diff --git a/daemons/based/based_io.c b/daemons/based/based_io.c index 7f2a959b1d..d1dad30d5e 100644 --- a/daemons/based/based_io.c +++ b/daemons/based/based_io.c @@ -1,422 +1,422 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include crm_trigger_t *cib_writer = NULL; int write_cib_contents(gpointer p); static void cib_rename(const char *old) { int new_fd; char *new = crm_strdup_printf("%s/cib.auto.XXXXXX", cib_root); umask(S_IWGRP | S_IWOTH | S_IROTH); new_fd = mkstemp(new); if ((new_fd < 0) || (rename(old, new) < 0)) { crm_err("Couldn't archive unusable file %s (disabling disk writes and continuing)", old); cib_writes_enabled = FALSE; } else { crm_err("Archived unusable file %s as %s", old, new); } if (new_fd > 0) { close(new_fd); } free(new); } /* * It is the callers responsibility to free the output of this function */ static xmlNode * retrieveCib(const char *filename, const char *sigfile) { xmlNode *root = NULL; int rc = cib_file_read_and_verify(filename, sigfile, &root); if (rc == pcmk_ok) { crm_info("Loaded CIB from %s (with digest %s)", filename, sigfile); } else { crm_warn("Continuing but NOT using CIB from %s (with digest %s): %s", filename, sigfile, pcmk_strerror(rc)); if (rc == -pcmk_err_cib_modified) { // Archive the original files so the contents are not lost cib_rename(filename); cib_rename(sigfile); } } return root; } static int cib_archive_filter(const struct dirent * a) { int rc = 0; // Looking for regular files starting with "cib-" and not ending in .sig struct stat s; char *a_path = crm_strdup_printf("%s/%s", cib_root, a->d_name); if(stat(a_path, &s) != 0) { rc = errno; crm_trace("%s - stat failed: %s (%d)", a->d_name, pcmk_rc_str(rc), rc); rc = 0; } else if (!S_ISREG(s.st_mode)) { crm_trace("%s - wrong type (%#o)", a->d_name, (unsigned int) (s.st_mode & S_IFMT)); } else if(strstr(a->d_name, "cib-") != a->d_name) { crm_trace("%s - wrong prefix", a->d_name); - } else if (pcmk__ends_with_ext(a->d_name, ".sig")) { + } else if (g_str_has_suffix(a->d_name, ".sig")) { crm_trace("%s - wrong suffix", a->d_name); } else { crm_debug("%s - candidate", a->d_name); rc = 1; } free(a_path); return rc; } static int cib_archive_sort(const struct dirent ** a, const struct dirent **b) { /* Order by creation date - most recently created file first */ int rc = 0; struct stat buf; time_t a_age = 0; time_t b_age = 0; char *a_path = crm_strdup_printf("%s/%s", cib_root, a[0]->d_name); char *b_path = crm_strdup_printf("%s/%s", cib_root, b[0]->d_name); if(stat(a_path, &buf) == 0) { a_age = buf.st_ctime; } if(stat(b_path, &buf) == 0) { b_age = buf.st_ctime; } free(a_path); free(b_path); if(a_age > b_age) { rc = 1; } else if(a_age < b_age) { rc = -1; } crm_trace("%s (%lu) vs. %s (%lu) : %d", a[0]->d_name, (unsigned long)a_age, b[0]->d_name, (unsigned long)b_age, rc); return rc; } xmlNode * readCibXmlFile(const char *dir, const char *file, gboolean discard_status) { struct dirent **namelist = NULL; int lpc = 0; char *sigfile = NULL; char *sigfilepath = NULL; char *filename = NULL; const char *name = NULL; const char *value = NULL; const char *use_valgrind = pcmk__env_option(PCMK__ENV_VALGRIND_ENABLED); xmlNode *root = NULL; xmlNode *status = NULL; sigfile = crm_strdup_printf("%s.sig", file); if (pcmk__daemon_can_write(dir, file) == FALSE || pcmk__daemon_can_write(dir, sigfile) == FALSE) { cib_status = -EACCES; return NULL; } filename = crm_strdup_printf("%s/%s", dir, file); sigfilepath = crm_strdup_printf("%s/%s", dir, sigfile); free(sigfile); cib_status = pcmk_ok; root = retrieveCib(filename, sigfilepath); free(filename); free(sigfilepath); if (root == NULL) { lpc = scandir(cib_root, &namelist, cib_archive_filter, cib_archive_sort); if (lpc < 0) { crm_err("Could not check for CIB backups in %s: %s", cib_root, pcmk_rc_str(errno)); } } while (root == NULL && lpc > 1) { int rc = pcmk_ok; lpc--; filename = crm_strdup_printf("%s/%s", cib_root, namelist[lpc]->d_name); sigfile = crm_strdup_printf("%s.sig", filename); rc = cib_file_read_and_verify(filename, sigfile, &root); if (rc == pcmk_ok) { crm_notice("Loaded CIB from last valid backup %s (with digest %s)", filename, sigfile); } else { crm_warn("Not using next most recent CIB backup from %s " "(with digest %s): %s", filename, sigfile, pcmk_strerror(rc)); } free(namelist[lpc]); free(filename); free(sigfile); } free(namelist); if (root == NULL) { root = createEmptyCib(0); crm_warn("Continuing with an empty configuration"); } if (cib_writes_enabled && (use_valgrind != NULL) && (crm_is_true(use_valgrind) || (strstr(use_valgrind, PCMK__SERVER_BASED) != NULL))) { cib_writes_enabled = FALSE; crm_err("*** Disabling disk writes to avoid confusing Valgrind ***"); } status = pcmk__xe_first_child(root, PCMK_XE_STATUS, NULL, NULL); if (discard_status && status != NULL) { // Strip out the PCMK_XE_STATUS section if there is one pcmk__xml_free(status); status = NULL; } if (status == NULL) { pcmk__xe_create(root, PCMK_XE_STATUS); } /* Do this before schema validation happens */ /* fill in some defaults */ value = pcmk__xe_get(root, PCMK_XA_ADMIN_EPOCH); if (value == NULL) { // Not possible with schema validation enabled crm_warn("Defaulting missing " PCMK_XA_ADMIN_EPOCH " to 0, but " "cluster may get confused about which node's configuration " "is most recent"); crm_xml_add_int(root, PCMK_XA_ADMIN_EPOCH, 0); } name = PCMK_XA_EPOCH; value = pcmk__xe_get(root, name); if (value == NULL) { crm_xml_add_int(root, name, 0); } name = PCMK_XA_NUM_UPDATES; value = pcmk__xe_get(root, name); if (value == NULL) { crm_xml_add_int(root, name, 0); } // Unset (DC should set appropriate value) pcmk__xe_remove_attr(root, PCMK_XA_DC_UUID); if (discard_status) { crm_log_xml_trace(root, "[on-disk]"); } if (!pcmk__configured_schema_validates(root)) { cib_status = -pcmk_err_schema_validation; } return root; } gboolean uninitializeCib(void) { xmlNode *tmp_cib = the_cib; if (tmp_cib == NULL) { return FALSE; } the_cib = NULL; pcmk__xml_free(tmp_cib); return TRUE; } /* * This method will free the old CIB pointer on success and the new one * on failure. */ int activateCibXml(xmlNode * new_cib, gboolean to_disk, const char *op) { if (new_cib) { xmlNode *saved_cib = the_cib; pcmk__assert(new_cib != saved_cib); the_cib = new_cib; pcmk__xml_free(saved_cib); if (cib_writes_enabled && cib_status == pcmk_ok && to_disk) { crm_debug("Triggering CIB write for %s op", op); mainloop_set_trigger(cib_writer); } return pcmk_ok; } crm_err("Ignoring invalid CIB"); if (the_cib) { crm_warn("Reverting to last known CIB"); } else { crm_crit("Could not write out new CIB and no saved version to revert to"); } return -ENODATA; } static void cib_diskwrite_complete(mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode) { const char *errmsg = "Could not write CIB to disk"; if ((exitcode != 0) && cib_writes_enabled) { cib_writes_enabled = FALSE; errmsg = "Disabling CIB disk writes after failure"; } if ((signo == 0) && (exitcode == 0)) { crm_trace("Disk write [%d] succeeded", (int) pid); } else if (signo == 0) { crm_err("%s: process %d exited %d", errmsg, (int) pid, exitcode); } else { crm_err("%s: process %d terminated with signal %d (%s)%s", errmsg, (int) pid, signo, strsignal(signo), (core? " and dumped core" : "")); } mainloop_trigger_complete(cib_writer); } int write_cib_contents(gpointer p) { int exit_rc = pcmk_ok; xmlNode *cib_local = NULL; /* Make a copy of the CIB to write (possibly in a forked child) */ if (p) { /* Synchronous write out */ cib_local = pcmk__xml_copy(NULL, p); } else { int pid = 0; int bb_state = qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_STATE_GET, 0); /* Turn it off before the fork() to avoid: * - 2 processes writing to the same shared mem * - the child needing to disable it * (which would close it from underneath the parent) * This way, the shared mem files are already closed */ qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_FALSE); pid = fork(); if (pid < 0) { crm_err("Disabling disk writes after fork failure: %s", pcmk_rc_str(errno)); cib_writes_enabled = FALSE; return FALSE; } if (pid) { /* Parent */ mainloop_child_add(pid, 0, "disk-writer", NULL, cib_diskwrite_complete); if (bb_state == QB_LOG_STATE_ENABLED) { /* Re-enable now that it it safe */ qb_log_ctl(QB_LOG_BLACKBOX, QB_LOG_CONF_ENABLED, QB_TRUE); } return -1; /* -1 means 'still work to do' */ } /* Asynchronous write-out after a fork() */ /* In theory, we can scribble on the_cib here and not affect the parent, * but let's be safe anyway. */ cib_local = pcmk__xml_copy(NULL, the_cib); } /* Write the CIB */ exit_rc = cib_file_write_with_digest(cib_local, cib_root, "cib.xml"); /* A nonzero exit code will cause further writes to be disabled */ pcmk__xml_free(cib_local); if (p == NULL) { crm_exit_t exit_code = CRM_EX_OK; switch (exit_rc) { case pcmk_ok: exit_code = CRM_EX_OK; break; case pcmk_err_cib_modified: exit_code = CRM_EX_DIGEST; // Existing CIB doesn't match digest break; case pcmk_err_cib_backup: // Existing CIB couldn't be backed up case pcmk_err_cib_save: // New CIB couldn't be saved exit_code = CRM_EX_CANTCREAT; break; default: exit_code = CRM_EX_ERROR; break; } /* Use _exit() because exit() could affect the parent adversely */ pcmk_common_cleanup(); _exit(exit_code); } return exit_rc; } diff --git a/daemons/execd/execd_commands.c b/daemons/execd/execd_commands.c index b011403719..a8b4a74e2e 100644 --- a/daemons/execd/execd_commands.c +++ b/daemons/execd/execd_commands.c @@ -1,1872 +1,1873 @@ /* * Copyright 2012-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include // xmlNode // Check whether we have a high-resolution monotonic clock #undef PCMK__TIME_USE_CGT #if HAVE_DECL_CLOCK_MONOTONIC && defined(CLOCK_MONOTONIC) # define PCMK__TIME_USE_CGT # include /* clock_gettime */ #endif #include #include #include #include #include #include #include #include #include #include "pacemaker-execd.h" GHashTable *rsc_list = NULL; typedef struct lrmd_cmd_s { int timeout; guint interval_ms; int start_delay; int timeout_orig; int call_id; int call_opts; /* Timer ids, must be removed on cmd destruction. */ int delay_id; int stonith_recurring_id; int rsc_deleted; int service_flags; char *client_id; char *origin; char *rsc_id; char *action; char *real_action; char *userdata_str; pcmk__action_result_t result; /* We can track operation queue time and run time, to be saved with the CIB * resource history (and displayed in cluster status). We need * high-resolution monotonic time for this purpose, so we use * clock_gettime(CLOCK_MONOTONIC, ...) (if available, otherwise this feature * is disabled). * * However, we also need epoch timestamps for recording the time the command * last ran and the time its return value last changed, for use in time * displays (as opposed to interval calculations). We keep time_t values for * this purpose. * * The last run time is used for both purposes, so we keep redundant * monotonic and epoch values for this. Technically the two could represent * different times, but since time_t has only second resolution and the * values are used for distinct purposes, that is not significant. */ #ifdef PCMK__TIME_USE_CGT /* Recurring and systemd operations may involve more than one executor * command per operation, so they need info about the original and the most * recent. */ struct timespec t_first_run; // When op first ran struct timespec t_run; // When op most recently ran struct timespec t_first_queue; // When op was first queued struct timespec t_queue; // When op was most recently queued #endif time_t epoch_last_run; // Epoch timestamp of when op last ran time_t epoch_rcchange; // Epoch timestamp of when rc last changed bool first_notify_sent; int last_notify_rc; int last_notify_op_status; int last_pid; GHashTable *params; } lrmd_cmd_t; static void cmd_finalize(lrmd_cmd_t * cmd, lrmd_rsc_t * rsc); static gboolean execute_resource_action(gpointer user_data); static void cancel_all_recurring(lrmd_rsc_t * rsc, const char *client_id); #ifdef PCMK__TIME_USE_CGT /*! * \internal * \brief Check whether a struct timespec has been set * * \param[in] timespec Time to check * * \return true if timespec has been set (i.e. is nonzero), false otherwise */ static inline bool time_is_set(const struct timespec *timespec) { return (timespec != NULL) && ((timespec->tv_sec != 0) || (timespec->tv_nsec != 0)); } /* * \internal * \brief Set a timespec (and its original if unset) to the current time * * \param[out] t_current Where to store current time * \param[out] t_orig Where to copy t_current if unset */ static void get_current_time(struct timespec *t_current, struct timespec *t_orig) { clock_gettime(CLOCK_MONOTONIC, t_current); if ((t_orig != NULL) && !time_is_set(t_orig)) { *t_orig = *t_current; } } /*! * \internal * \brief Return difference between two times in milliseconds * * \param[in] now More recent time (or NULL to use current time) * \param[in] old Earlier time * * \return milliseconds difference (or 0 if old is NULL or unset) * * \note Can overflow on 32bit machines when the differences is around * 24 days or more. */ static int time_diff_ms(const struct timespec *now, const struct timespec *old) { int diff_ms = 0; if (time_is_set(old)) { struct timespec local_now = { 0, }; if (now == NULL) { clock_gettime(CLOCK_MONOTONIC, &local_now); now = &local_now; } diff_ms = (now->tv_sec - old->tv_sec) * 1000 + (now->tv_nsec - old->tv_nsec) / 1000000; } return diff_ms; } /*! * \internal * \brief Reset a command's operation times to their original values. * * Reset a command's run and queued timestamps to the timestamps of the original * command, so we report the entire time since then and not just the time since * the most recent command (for recurring and systemd operations). * * \param[in,out] cmd Executor command object to reset * * \note It's not obvious what the queued time should be for a systemd * start/stop operation, which might go like this: * initial command queued 5ms, runs 3s * monitor command queued 10ms, runs 10s * monitor command queued 10ms, runs 10s * Is the queued time for that operation 5ms, 10ms or 25ms? The current * implementation will report 5ms. If it's 25ms, then we need to * subtract 20ms from the total exec time so as not to count it twice. * We can implement that later if it matters to anyone ... */ static void cmd_original_times(lrmd_cmd_t * cmd) { cmd->t_run = cmd->t_first_run; cmd->t_queue = cmd->t_first_queue; } #endif static inline bool action_matches(const lrmd_cmd_t *cmd, const char *action, guint interval_ms) { return (cmd->interval_ms == interval_ms) && pcmk__str_eq(cmd->action, action, pcmk__str_casei); } /*! * \internal * \brief Log the result of an asynchronous command * * \param[in] cmd Command to log result for * \param[in] exec_time_ms Execution time in milliseconds, if known * \param[in] queue_time_ms Queue time in milliseconds, if known */ static void log_finished(const lrmd_cmd_t *cmd, int exec_time_ms, int queue_time_ms) { int log_level = LOG_INFO; GString *str = g_string_sized_new(100); // reasonable starting size if (pcmk__str_eq(cmd->action, PCMK_ACTION_MONITOR, pcmk__str_casei)) { log_level = LOG_DEBUG; } g_string_append_printf(str, "%s %s (call %d", cmd->rsc_id, cmd->action, cmd->call_id); if (cmd->last_pid != 0) { g_string_append_printf(str, ", PID %d", cmd->last_pid); } switch (cmd->result.execution_status) { case PCMK_EXEC_DONE: g_string_append_printf(str, ") exited with status %d", cmd->result.exit_status); break; case PCMK_EXEC_CANCELLED: g_string_append_printf(str, ") cancelled"); break; default: pcmk__g_strcat(str, ") could not be executed: ", pcmk_exec_status_str(cmd->result.execution_status), NULL); break; } if (cmd->result.exit_reason != NULL) { pcmk__g_strcat(str, " (", cmd->result.exit_reason, ")", NULL); } #ifdef PCMK__TIME_USE_CGT pcmk__g_strcat(str, " (execution time ", pcmk__readable_interval(exec_time_ms), NULL); if (queue_time_ms > 0) { pcmk__g_strcat(str, " after being queued ", pcmk__readable_interval(queue_time_ms), NULL); } g_string_append_c(str, ')'); #endif do_crm_log(log_level, "%s", str->str); g_string_free(str, TRUE); } static void log_execute(lrmd_cmd_t * cmd) { int log_level = LOG_INFO; if (pcmk__str_eq(cmd->action, PCMK_ACTION_MONITOR, pcmk__str_casei)) { log_level = LOG_DEBUG; } do_crm_log(log_level, "executing - rsc:%s action:%s call_id:%d", cmd->rsc_id, cmd->action, cmd->call_id); } static const char * normalize_action_name(lrmd_rsc_t * rsc, const char *action) { if (pcmk__str_eq(action, PCMK_ACTION_MONITOR, pcmk__str_casei) && pcmk_is_set(pcmk_get_ra_caps(rsc->class), pcmk_ra_cap_status)) { return PCMK_ACTION_STATUS; } return action; } static lrmd_rsc_t * build_rsc_from_xml(xmlNode * msg) { xmlNode *rsc_xml = pcmk__xpath_find_one(msg->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); lrmd_rsc_t *rsc = NULL; rsc = pcmk__assert_alloc(1, sizeof(lrmd_rsc_t)); pcmk__xe_get_int(msg, PCMK__XA_LRMD_CALLOPT, &rsc->call_opts); rsc->rsc_id = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_RSC_ID); rsc->class = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_CLASS); rsc->provider = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_PROVIDER); rsc->type = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_TYPE); rsc->work = mainloop_add_trigger(G_PRIORITY_HIGH, execute_resource_action, rsc); // Initialize fence device probes (to return "not running") pcmk__set_result(&rsc->fence_probe_result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, NULL); return rsc; } static lrmd_cmd_t * create_lrmd_cmd(xmlNode *msg, pcmk__client_t *client) { int call_options = 0; xmlNode *rsc_xml = pcmk__xpath_find_one(msg->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); lrmd_cmd_t *cmd = NULL; cmd = pcmk__assert_alloc(1, sizeof(lrmd_cmd_t)); pcmk__xe_get_int(msg, PCMK__XA_LRMD_CALLOPT, &call_options); cmd->call_opts = call_options; cmd->client_id = pcmk__str_copy(client->id); pcmk__xe_get_int(msg, PCMK__XA_LRMD_CALLID, &cmd->call_id); pcmk__xe_get_guint(rsc_xml, PCMK__XA_LRMD_RSC_INTERVAL, &cmd->interval_ms); pcmk__xe_get_int(rsc_xml, PCMK__XA_LRMD_TIMEOUT, &cmd->timeout); pcmk__xe_get_int(rsc_xml, PCMK__XA_LRMD_RSC_START_DELAY, &cmd->start_delay); cmd->timeout_orig = cmd->timeout; cmd->origin = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_ORIGIN); cmd->action = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_RSC_ACTION); cmd->userdata_str = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_RSC_USERDATA_STR); cmd->rsc_id = pcmk__xe_get_copy(rsc_xml, PCMK__XA_LRMD_RSC_ID); cmd->params = xml2list(rsc_xml); if (pcmk__str_eq(g_hash_table_lookup(cmd->params, "CRM_meta_on_fail"), PCMK_VALUE_BLOCK, pcmk__str_casei)) { crm_debug("Setting flag to leave pid group on timeout and " "only kill action pid for " PCMK__OP_FMT, cmd->rsc_id, cmd->action, cmd->interval_ms); cmd->service_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, "Action", cmd->action, 0, SVC_ACTION_LEAVE_GROUP, "SVC_ACTION_LEAVE_GROUP"); } return cmd; } static void stop_recurring_timer(lrmd_cmd_t *cmd) { if (cmd) { if (cmd->stonith_recurring_id) { g_source_remove(cmd->stonith_recurring_id); } cmd->stonith_recurring_id = 0; } } static void free_lrmd_cmd(lrmd_cmd_t * cmd) { stop_recurring_timer(cmd); if (cmd->delay_id) { g_source_remove(cmd->delay_id); } if (cmd->params) { g_hash_table_destroy(cmd->params); } pcmk__reset_result(&(cmd->result)); free(cmd->origin); free(cmd->action); free(cmd->real_action); free(cmd->userdata_str); free(cmd->rsc_id); free(cmd->client_id); free(cmd); } static gboolean stonith_recurring_op_helper(gpointer data) { lrmd_cmd_t *cmd = data; lrmd_rsc_t *rsc; cmd->stonith_recurring_id = 0; if (!cmd->rsc_id) { return FALSE; } rsc = g_hash_table_lookup(rsc_list, cmd->rsc_id); pcmk__assert(rsc != NULL); /* take it out of recurring_ops list, and put it in the pending ops * to be executed */ rsc->recurring_ops = g_list_remove(rsc->recurring_ops, cmd); rsc->pending_ops = g_list_append(rsc->pending_ops, cmd); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_queue), &(cmd->t_first_queue)); #endif mainloop_set_trigger(rsc->work); return FALSE; } static inline void start_recurring_timer(lrmd_cmd_t *cmd) { if (!cmd || (cmd->interval_ms <= 0)) { return; } cmd->stonith_recurring_id = pcmk__create_timer(cmd->interval_ms, stonith_recurring_op_helper, cmd); } static gboolean start_delay_helper(gpointer data) { lrmd_cmd_t *cmd = data; lrmd_rsc_t *rsc = NULL; cmd->delay_id = 0; rsc = cmd->rsc_id ? g_hash_table_lookup(rsc_list, cmd->rsc_id) : NULL; if (rsc) { mainloop_set_trigger(rsc->work); } return FALSE; } /*! * \internal * \brief Check whether a list already contains the equivalent of a given action * * \param[in] action_list List to search * \param[in] cmd Action to search for */ static lrmd_cmd_t * find_duplicate_action(const GList *action_list, const lrmd_cmd_t *cmd) { for (const GList *item = action_list; item != NULL; item = item->next) { lrmd_cmd_t *dup = item->data; if (action_matches(cmd, dup->action, dup->interval_ms)) { return dup; } } return NULL; } static bool merge_recurring_duplicate(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { lrmd_cmd_t * dup = NULL; bool dup_pending = true; if (cmd->interval_ms == 0) { return false; } // Search for a duplicate of this action (in-flight or not) dup = find_duplicate_action(rsc->pending_ops, cmd); if (dup == NULL) { dup_pending = false; dup = find_duplicate_action(rsc->recurring_ops, cmd); if (dup == NULL) { return false; } } /* Do not merge fencing monitors marked for cancellation, so we can reply to * the cancellation separately. */ if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei) && (dup->result.execution_status == PCMK_EXEC_CANCELLED)) { return false; } /* This should not occur. If it does, we need to investigate how something * like this is possible in the controller. */ crm_warn("Duplicate recurring op entry detected (" PCMK__OP_FMT "), merging with previous op entry", rsc->rsc_id, normalize_action_name(rsc, dup->action), dup->interval_ms); // Merge new action's call ID and user data into existing action dup->first_notify_sent = false; free(dup->userdata_str); dup->userdata_str = cmd->userdata_str; cmd->userdata_str = NULL; dup->call_id = cmd->call_id; free_lrmd_cmd(cmd); cmd = NULL; /* If dup is not pending, that means it has already executed at least once * and is waiting in the interval. In that case, stop waiting and initiate * a new instance now. */ if (!dup_pending) { if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { stop_recurring_timer(dup); stonith_recurring_op_helper(dup); } else { services_action_kick(rsc->rsc_id, normalize_action_name(rsc, dup->action), dup->interval_ms); } } return true; } static void schedule_lrmd_cmd(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { CRM_CHECK(cmd != NULL, return); CRM_CHECK(rsc != NULL, return); crm_trace("Scheduling %s on %s", cmd->action, rsc->rsc_id); if (merge_recurring_duplicate(rsc, cmd)) { // Equivalent of cmd has already been scheduled return; } /* The controller expects the executor to automatically cancel * recurring operations before a resource stops. */ if (pcmk__str_eq(cmd->action, PCMK_ACTION_STOP, pcmk__str_casei)) { cancel_all_recurring(rsc, NULL); } rsc->pending_ops = g_list_append(rsc->pending_ops, cmd); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_queue), &(cmd->t_first_queue)); #endif mainloop_set_trigger(rsc->work); if (cmd->start_delay) { cmd->delay_id = pcmk__create_timer(cmd->start_delay, start_delay_helper, cmd); } } xmlNode * execd_create_reply_as(const char *origin, int rc, int call_id) { xmlNode *reply = pcmk__xe_create(NULL, PCMK__XE_LRMD_REPLY); crm_xml_add(reply, PCMK__XA_LRMD_ORIGIN, origin); crm_xml_add_int(reply, PCMK__XA_LRMD_RC, rc); crm_xml_add_int(reply, PCMK__XA_LRMD_CALLID, call_id); return reply; } static void send_client_notify(gpointer key, gpointer value, gpointer user_data) { xmlNode *update_msg = user_data; pcmk__client_t *client = value; int rc; int log_level = LOG_WARNING; const char *msg = NULL; CRM_CHECK(client != NULL, return); if (client->name == NULL) { crm_trace("Skipping notification to client without name"); return; } if (pcmk_is_set(client->flags, pcmk__client_to_proxy)) { /* We only want to notify clients of the executor IPC API. If we are * running as Pacemaker Remote, we may have clients proxied to other * IPC services in the cluster, so skip those. */ crm_trace("Skipping executor API notification to client %s", pcmk__client_name(client)); return; } rc = lrmd_server_send_notify(client, update_msg); if (rc == pcmk_rc_ok) { return; } switch (rc) { case ENOTCONN: case EPIPE: // Client exited without waiting for notification log_level = LOG_INFO; msg = "Disconnected"; break; default: msg = pcmk_rc_str(rc); break; } do_crm_log(log_level, "Could not notify client %s: %s " QB_XS " rc=%d", pcmk__client_name(client), msg, rc); } static void send_cmd_complete_notify(lrmd_cmd_t * cmd) { xmlNode *notify = NULL; int exec_time = 0; int queue_time = 0; #ifdef PCMK__TIME_USE_CGT exec_time = time_diff_ms(NULL, &(cmd->t_run)); queue_time = time_diff_ms(&cmd->t_run, &(cmd->t_queue)); #endif log_finished(cmd, exec_time, queue_time); /* If the originator requested to be notified only for changes in recurring * operation results, skip the notification if the result hasn't changed. */ if (cmd->first_notify_sent && pcmk_is_set(cmd->call_opts, lrmd_opt_notify_changes_only) && (cmd->last_notify_rc == cmd->result.exit_status) && (cmd->last_notify_op_status == cmd->result.execution_status)) { return; } cmd->first_notify_sent = true; cmd->last_notify_rc = cmd->result.exit_status; cmd->last_notify_op_status = cmd->result.execution_status; notify = pcmk__xe_create(NULL, PCMK__XE_LRMD_NOTIFY); crm_xml_add(notify, PCMK__XA_LRMD_ORIGIN, __func__); crm_xml_add_int(notify, PCMK__XA_LRMD_TIMEOUT, cmd->timeout); crm_xml_add_ms(notify, PCMK__XA_LRMD_RSC_INTERVAL, cmd->interval_ms); crm_xml_add_int(notify, PCMK__XA_LRMD_RSC_START_DELAY, cmd->start_delay); crm_xml_add_int(notify, PCMK__XA_LRMD_EXEC_RC, cmd->result.exit_status); crm_xml_add_int(notify, PCMK__XA_LRMD_EXEC_OP_STATUS, cmd->result.execution_status); crm_xml_add_int(notify, PCMK__XA_LRMD_CALLID, cmd->call_id); crm_xml_add_int(notify, PCMK__XA_LRMD_RSC_DELETED, cmd->rsc_deleted); crm_xml_add_ll(notify, PCMK__XA_LRMD_RUN_TIME, (long long) cmd->epoch_last_run); crm_xml_add_ll(notify, PCMK__XA_LRMD_RCCHANGE_TIME, (long long) cmd->epoch_rcchange); #ifdef PCMK__TIME_USE_CGT crm_xml_add_int(notify, PCMK__XA_LRMD_EXEC_TIME, exec_time); crm_xml_add_int(notify, PCMK__XA_LRMD_QUEUE_TIME, queue_time); #endif crm_xml_add(notify, PCMK__XA_LRMD_OP, LRMD_OP_RSC_EXEC); crm_xml_add(notify, PCMK__XA_LRMD_RSC_ID, cmd->rsc_id); if(cmd->real_action) { crm_xml_add(notify, PCMK__XA_LRMD_RSC_ACTION, cmd->real_action); } else { crm_xml_add(notify, PCMK__XA_LRMD_RSC_ACTION, cmd->action); } crm_xml_add(notify, PCMK__XA_LRMD_RSC_USERDATA_STR, cmd->userdata_str); crm_xml_add(notify, PCMK__XA_LRMD_RSC_EXIT_REASON, cmd->result.exit_reason); if (cmd->result.action_stderr != NULL) { crm_xml_add(notify, PCMK__XA_LRMD_RSC_OUTPUT, cmd->result.action_stderr); } else if (cmd->result.action_stdout != NULL) { crm_xml_add(notify, PCMK__XA_LRMD_RSC_OUTPUT, cmd->result.action_stdout); } if (cmd->params) { char *key = NULL; char *value = NULL; GHashTableIter iter; xmlNode *args = pcmk__xe_create(notify, PCMK__XE_ATTRIBUTES); g_hash_table_iter_init(&iter, cmd->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { hash2smartfield((gpointer) key, (gpointer) value, args); } } if ((cmd->client_id != NULL) && pcmk_is_set(cmd->call_opts, lrmd_opt_notify_orig_only)) { pcmk__client_t *client = pcmk__find_client_by_id(cmd->client_id); if (client != NULL) { send_client_notify(client->id, client, notify); } } else { pcmk__foreach_ipc_client(send_client_notify, notify); } pcmk__xml_free(notify); } void execd_send_generic_notify(int rc, xmlNode *request) { if (pcmk__ipc_client_count() != 0) { int call_id = 0; xmlNode *notify = NULL; xmlNode *rsc_xml = pcmk__xpath_find_one(request->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); const char *rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); const char *op = pcmk__xe_get(request, PCMK__XA_LRMD_OP); pcmk__xe_get_int(request, PCMK__XA_LRMD_CALLID, &call_id); notify = pcmk__xe_create(NULL, PCMK__XE_LRMD_NOTIFY); crm_xml_add(notify, PCMK__XA_LRMD_ORIGIN, __func__); crm_xml_add_int(notify, PCMK__XA_LRMD_RC, rc); crm_xml_add_int(notify, PCMK__XA_LRMD_CALLID, call_id); crm_xml_add(notify, PCMK__XA_LRMD_OP, op); crm_xml_add(notify, PCMK__XA_LRMD_RSC_ID, rsc_id); pcmk__foreach_ipc_client(send_client_notify, notify); pcmk__xml_free(notify); } } static void cmd_reset(lrmd_cmd_t * cmd) { cmd->last_pid = 0; #ifdef PCMK__TIME_USE_CGT memset(&cmd->t_run, 0, sizeof(cmd->t_run)); memset(&cmd->t_queue, 0, sizeof(cmd->t_queue)); #endif cmd->epoch_last_run = 0; pcmk__reset_result(&(cmd->result)); cmd->result.execution_status = PCMK_EXEC_DONE; } static void cmd_finalize(lrmd_cmd_t * cmd, lrmd_rsc_t * rsc) { crm_trace("Resource operation rsc:%s action:%s completed (%p %p)", cmd->rsc_id, cmd->action, rsc ? rsc->active : NULL, cmd); if (rsc && (rsc->active == cmd)) { rsc->active = NULL; mainloop_set_trigger(rsc->work); } if (!rsc) { cmd->rsc_deleted = 1; } /* reset original timeout so client notification has correct information */ cmd->timeout = cmd->timeout_orig; send_cmd_complete_notify(cmd); if ((cmd->interval_ms != 0) && (cmd->result.execution_status == PCMK_EXEC_CANCELLED)) { if (rsc) { rsc->recurring_ops = g_list_remove(rsc->recurring_ops, cmd); rsc->pending_ops = g_list_remove(rsc->pending_ops, cmd); } free_lrmd_cmd(cmd); } else if (cmd->interval_ms == 0) { if (rsc) { rsc->pending_ops = g_list_remove(rsc->pending_ops, cmd); } free_lrmd_cmd(cmd); } else { /* Clear all the values pertaining just to the last iteration of a recurring op. */ cmd_reset(cmd); } } struct notify_new_client_data { xmlNode *notify; pcmk__client_t *new_client; }; static void notify_one_client(gpointer key, gpointer value, gpointer user_data) { pcmk__client_t *client = value; struct notify_new_client_data *data = user_data; if (!pcmk__str_eq(client->id, data->new_client->id, pcmk__str_casei)) { send_client_notify(key, (gpointer) client, (gpointer) data->notify); } } void notify_of_new_client(pcmk__client_t *new_client) { struct notify_new_client_data data; data.new_client = new_client; data.notify = pcmk__xe_create(NULL, PCMK__XE_LRMD_NOTIFY); crm_xml_add(data.notify, PCMK__XA_LRMD_ORIGIN, __func__); crm_xml_add(data.notify, PCMK__XA_LRMD_OP, LRMD_OP_NEW_CLIENT); pcmk__foreach_ipc_client(notify_one_client, &data); pcmk__xml_free(data.notify); } void client_disconnect_cleanup(const char *client_id) { GHashTableIter iter; lrmd_rsc_t *rsc = NULL; char *key = NULL; g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & rsc)) { if (pcmk_all_flags_set(rsc->call_opts, lrmd_opt_drop_recurring)) { /* This client is disconnecting, drop any recurring operations * it may have initiated on the resource */ cancel_all_recurring(rsc, client_id); } } } static void action_complete(svc_action_t * action) { lrmd_rsc_t *rsc; lrmd_cmd_t *cmd = action->cb_data; enum ocf_exitcode code; #ifdef PCMK__TIME_USE_CGT const char *rclass = NULL; bool goagain = false; int time_sum = 0; int timeout_left = 0; int delay = 0; #endif if (!cmd) { crm_err("Completed executor action (%s) does not match any known operations", action->id); return; } #ifdef PCMK__TIME_USE_CGT if (cmd->result.exit_status != action->rc) { cmd->epoch_rcchange = time(NULL); } #endif cmd->last_pid = action->pid; // Cast variable instead of function return to keep compilers happy code = services_result2ocf(action->standard, cmd->action, action->rc); pcmk__set_result(&(cmd->result), (int) code, action->status, services__exit_reason(action)); rsc = cmd->rsc_id ? g_hash_table_lookup(rsc_list, cmd->rsc_id) : NULL; #ifdef PCMK__TIME_USE_CGT if (rsc != NULL) { rclass = rsc->class; #if PCMK__ENABLE_SERVICE if (pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_SERVICE, pcmk__str_casei)) { rclass = resources_find_service_class(rsc->type); } #endif } if (!pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { goto finalize; } if (pcmk__result_ok(&(cmd->result)) && pcmk__strcase_any_of(cmd->action, PCMK_ACTION_START, PCMK_ACTION_STOP, NULL)) { /* Getting results for when a start or stop action completes is now * handled by watching for JobRemoved() signals from systemd and * reacting to them. So, we can bypass the rest of the code in this * function for those actions, and simply finalize cmd. * * @TODO When monitors are handled in the same way, this function * can either be drastically simplified or done away with entirely. */ services__copy_result(action, &(cmd->result)); goto finalize; } else if (cmd->result.execution_status == PCMK_EXEC_PENDING && pcmk__str_any_of(cmd->action, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL) && cmd->interval_ms == 0 && cmd->real_action == NULL) { /* If the state is Pending at the time of probe, execute follow-up monitor. */ goagain = true; cmd->real_action = cmd->action; cmd->action = pcmk__str_copy(PCMK_ACTION_MONITOR); } else if (cmd->real_action != NULL) { // This is follow-up monitor to check whether start/stop/probe(monitor) completed if (cmd->result.execution_status == PCMK_EXEC_PENDING) { goagain = true; } else if (pcmk__result_ok(&(cmd->result)) && pcmk__str_eq(cmd->real_action, PCMK_ACTION_STOP, pcmk__str_casei)) { goagain = true; } else { int time_sum = time_diff_ms(NULL, &(cmd->t_first_run)); int timeout_left = cmd->timeout_orig - time_sum; crm_debug("%s systemd %s is now complete (elapsed=%dms, " "remaining=%dms): %s (%d)", cmd->rsc_id, cmd->real_action, time_sum, timeout_left, crm_exit_str(cmd->result.exit_status), cmd->result.exit_status); cmd_original_times(cmd); // Monitors may return "not running", but start/stop shouldn't if ((cmd->result.execution_status == PCMK_EXEC_DONE) && (cmd->result.exit_status == PCMK_OCF_NOT_RUNNING)) { if (pcmk__str_eq(cmd->real_action, PCMK_ACTION_START, pcmk__str_casei)) { cmd->result.exit_status = PCMK_OCF_UNKNOWN_ERROR; } else if (pcmk__str_eq(cmd->real_action, PCMK_ACTION_STOP, pcmk__str_casei)) { cmd->result.exit_status = PCMK_OCF_OK; } } } } else if (pcmk__str_any_of(cmd->action, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL) && (cmd->interval_ms > 0)) { /* For monitors, excluding follow-up monitors, */ /* if the pending state persists from the first notification until its timeout, */ /* it will be treated as a timeout. */ if ((cmd->result.execution_status == PCMK_EXEC_PENDING) && (cmd->last_notify_op_status == PCMK_EXEC_PENDING)) { int time_left = time(NULL) - (cmd->epoch_rcchange + (cmd->timeout_orig/1000)); if (time_left >= 0) { crm_notice("Giving up on %s %s (rc=%d): monitor pending timeout " "(first pending notification=%s timeout=%ds)", cmd->rsc_id, cmd->action, cmd->result.exit_status, - pcmk__trim(ctime(&cmd->epoch_rcchange)), cmd->timeout_orig); + g_strchomp(ctime(&cmd->epoch_rcchange)), + cmd->timeout_orig); pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, "Investigate reason for timeout, and adjust " "configured operation timeout if necessary"); cmd_original_times(cmd); } } } if (!goagain) { goto finalize; } time_sum = time_diff_ms(NULL, &(cmd->t_first_run)); timeout_left = cmd->timeout_orig - time_sum; delay = cmd->timeout_orig / 10; if (delay >= timeout_left && timeout_left > 20) { delay = timeout_left/2; } delay = QB_MIN(2000, delay); if (delay < timeout_left) { cmd->start_delay = delay; cmd->timeout = timeout_left; if (pcmk__result_ok(&(cmd->result))) { crm_debug("%s %s may still be in progress: re-scheduling (elapsed=%dms, remaining=%dms, start_delay=%dms)", cmd->rsc_id, cmd->real_action, time_sum, timeout_left, delay); } else if (cmd->result.execution_status == PCMK_EXEC_PENDING) { crm_info("%s %s is still in progress: re-scheduling (elapsed=%dms, remaining=%dms, start_delay=%dms)", cmd->rsc_id, cmd->action, time_sum, timeout_left, delay); } else { crm_notice("%s %s failed: %s: Re-scheduling (remaining " "timeout %s) " QB_XS " exitstatus=%d elapsed=%dms start_delay=%dms)", cmd->rsc_id, cmd->action, crm_exit_str(cmd->result.exit_status), pcmk__readable_interval(timeout_left), cmd->result.exit_status, time_sum, delay); } cmd_reset(cmd); if (rsc) { rsc->active = NULL; } schedule_lrmd_cmd(rsc, cmd); /* Don't finalize cmd, we're not done with it yet */ return; } else { crm_notice("Giving up on %s %s (rc=%d): timeout (elapsed=%dms, remaining=%dms)", cmd->rsc_id, (cmd->real_action? cmd->real_action : cmd->action), cmd->result.exit_status, time_sum, timeout_left); pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, "Investigate reason for timeout, and adjust " "configured operation timeout if necessary"); cmd_original_times(cmd); } #endif finalize: pcmk__set_result_output(&(cmd->result), services__grab_stdout(action), services__grab_stderr(action)); cmd_finalize(cmd, rsc); } /*! * \internal * \brief Process the result of a fence device action (start, stop, or monitor) * * \param[in,out] cmd Fence device action that completed * \param[in] exit_status Fencer API exit status for action * \param[in] execution_status Fencer API execution status for action * \param[in] exit_reason Human-friendly detail, if action failed */ static void stonith_action_complete(lrmd_cmd_t *cmd, int exit_status, enum pcmk_exec_status execution_status, const char *exit_reason) { // This can be NULL if resource was removed before command completed lrmd_rsc_t *rsc = g_hash_table_lookup(rsc_list, cmd->rsc_id); // Simplify fencer exit status to uniform exit status if (exit_status != CRM_EX_OK) { exit_status = PCMK_OCF_UNKNOWN_ERROR; } if (cmd->result.execution_status == PCMK_EXEC_CANCELLED) { /* An in-flight fence action was cancelled. The execution status is * already correct, so don't overwrite it. */ execution_status = PCMK_EXEC_CANCELLED; } else { /* Some execution status codes have specific meanings for the fencer * that executor clients may not expect, so map them to a simple error * status. */ switch (execution_status) { case PCMK_EXEC_NOT_CONNECTED: case PCMK_EXEC_INVALID: execution_status = PCMK_EXEC_ERROR; break; case PCMK_EXEC_NO_FENCE_DEVICE: /* This should be possible only for probes in practice, but * interpret for all actions to be safe. */ if (pcmk__str_eq(cmd->action, PCMK_ACTION_MONITOR, pcmk__str_none)) { exit_status = PCMK_OCF_NOT_RUNNING; } else if (pcmk__str_eq(cmd->action, PCMK_ACTION_STOP, pcmk__str_none)) { exit_status = PCMK_OCF_OK; } else { exit_status = PCMK_OCF_NOT_INSTALLED; } execution_status = PCMK_EXEC_ERROR; break; case PCMK_EXEC_NOT_SUPPORTED: exit_status = PCMK_OCF_UNIMPLEMENT_FEATURE; break; default: break; } } pcmk__set_result(&cmd->result, exit_status, execution_status, exit_reason); // Certain successful actions change the known state of the resource if ((rsc != NULL) && pcmk__result_ok(&(cmd->result))) { if (pcmk__str_eq(cmd->action, PCMK_ACTION_START, pcmk__str_casei)) { pcmk__set_result(&rsc->fence_probe_result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); // "running" } else if (pcmk__str_eq(cmd->action, PCMK_ACTION_STOP, pcmk__str_casei)) { pcmk__set_result(&rsc->fence_probe_result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, NULL); // "not running" } } /* The recurring timer should not be running at this point in any case, but * as a failsafe, stop it if it is. */ stop_recurring_timer(cmd); /* Reschedule this command if appropriate. If a recurring command is *not* * rescheduled, its status must be PCMK_EXEC_CANCELLED, otherwise it will * not be removed from recurring_ops by cmd_finalize(). */ if (rsc && (cmd->interval_ms > 0) && (cmd->result.execution_status != PCMK_EXEC_CANCELLED)) { start_recurring_timer(cmd); } cmd_finalize(cmd, rsc); } static void lrmd_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data) { if ((data == NULL) || (data->userdata == NULL)) { crm_err("Ignoring fence action result: " "Invalid callback arguments (bug?)"); } else { stonith_action_complete((lrmd_cmd_t *) data->userdata, stonith__exit_status(data), stonith__execution_status(data), stonith__exit_reason(data)); } } void stonith_connection_failed(void) { GHashTableIter iter; lrmd_rsc_t *rsc = NULL; crm_warn("Connection to fencer lost (any pending operations for " "fence devices will be considered failed)"); g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &rsc)) { if (!pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_none)) { continue; } /* If we registered this fence device, we don't know whether the * fencer still has the registration or not. Cause future probes to * return an error until the resource is stopped or started * successfully. This is especially important if the controller also * went away (possibly due to a cluster layer restart) and won't * receive our client notification of any monitors finalized below. */ if (rsc->fence_probe_result.execution_status == PCMK_EXEC_DONE) { pcmk__set_result(&rsc->fence_probe_result, CRM_EX_ERROR, PCMK_EXEC_NOT_CONNECTED, "Lost connection to fencer"); } // Consider any active, pending, or recurring operations as failed for (GList *op = rsc->recurring_ops; op != NULL; op = op->next) { lrmd_cmd_t *cmd = op->data; /* This won't free a recurring op but instead restart its timer. * If cmd is rsc->active, this will set rsc->active to NULL, so we * don't have to worry about finalizing it a second time below. */ stonith_action_complete(cmd, CRM_EX_ERROR, PCMK_EXEC_NOT_CONNECTED, "Lost connection to fencer"); } if (rsc->active != NULL) { rsc->pending_ops = g_list_prepend(rsc->pending_ops, rsc->active); } while (rsc->pending_ops != NULL) { // This will free the op and remove it from rsc->pending_ops stonith_action_complete((lrmd_cmd_t *) rsc->pending_ops->data, CRM_EX_ERROR, PCMK_EXEC_NOT_CONNECTED, "Lost connection to fencer"); } } } /*! * \internal * \brief Execute a stonith resource "start" action * * Start a stonith resource by registering it with the fencer. * (Stonith agents don't have a start command.) * * \param[in,out] stonith_api Connection to fencer * \param[in] rsc Stonith resource to start * \param[in] cmd Start command to execute * * \return pcmk_ok on success, -errno otherwise */ static int execd_stonith_start(stonith_t *stonith_api, const lrmd_rsc_t *rsc, const lrmd_cmd_t *cmd) { char *key = NULL; char *value = NULL; stonith_key_value_t *device_params = NULL; int rc = pcmk_ok; // Convert command parameters to stonith API key/values if (cmd->params) { GHashTableIter iter; g_hash_table_iter_init(&iter, cmd->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { device_params = stonith__key_value_add(device_params, key, value); } } /* The fencer will automatically register devices via CIB notifications * when the CIB changes, but to avoid a possible race condition between * the fencer receiving the notification and the executor requesting that * resource, the executor registers the device as well. The fencer knows how * to handle duplicate registrations. */ rc = stonith_api->cmds->register_device(stonith_api, st_opt_sync_call, cmd->rsc_id, rsc->provider, rsc->type, device_params); stonith__key_value_freeall(device_params, true, true); return rc; } /*! * \internal * \brief Execute a stonith resource "stop" action * * Stop a stonith resource by unregistering it with the fencer. * (Stonith agents don't have a stop command.) * * \param[in,out] stonith_api Connection to fencer * \param[in] rsc Stonith resource to stop * * \return pcmk_ok on success, -errno otherwise */ static inline int execd_stonith_stop(stonith_t *stonith_api, const lrmd_rsc_t *rsc) { /* @TODO Failure would indicate a problem communicating with fencer; * perhaps we should try reconnecting and retrying a few times? */ return stonith_api->cmds->remove_device(stonith_api, st_opt_sync_call, rsc->rsc_id); } /*! * \internal * \brief Initiate a stonith resource agent recurring "monitor" action * * \param[in,out] stonith_api Connection to fencer * \param[in,out] rsc Stonith resource to monitor * \param[in] cmd Monitor command being executed * * \return pcmk_ok if monitor was successfully initiated, -errno otherwise */ static inline int execd_stonith_monitor(stonith_t *stonith_api, lrmd_rsc_t *rsc, lrmd_cmd_t *cmd) { int rc = stonith_api->cmds->monitor(stonith_api, 0, cmd->rsc_id, pcmk__timeout_ms2s(cmd->timeout)); rc = stonith_api->cmds->register_callback(stonith_api, rc, 0, 0, cmd, "lrmd_stonith_callback", lrmd_stonith_callback); if (rc == TRUE) { rsc->active = cmd; rc = pcmk_ok; } else { rc = -pcmk_err_generic; } return rc; } static void execute_stonith_action(lrmd_rsc_t *rsc, lrmd_cmd_t *cmd) { int rc = pcmk_ok; const char *rc_s = NULL; bool do_monitor = false; // Don't free; belongs to pacemaker-execd.c stonith_t *stonith_api = get_stonith_connection(); if (pcmk__str_eq(cmd->action, PCMK_ACTION_MONITOR, pcmk__str_casei) && (cmd->interval_ms == 0)) { // Probes don't require a fencer connection stonith_action_complete(cmd, rsc->fence_probe_result.exit_status, rsc->fence_probe_result.execution_status, rsc->fence_probe_result.exit_reason); return; } if (stonith_api == NULL) { stonith_action_complete(cmd, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_NOT_CONNECTED, "No connection to fencer"); return; } if (pcmk__str_eq(cmd->action, PCMK_ACTION_START, pcmk__str_casei)) { rc = execd_stonith_start(stonith_api, rsc, cmd); if (rc == pcmk_ok) { do_monitor = true; } } else if (pcmk__str_eq(cmd->action, PCMK_ACTION_STOP, pcmk__str_casei)) { rc = execd_stonith_stop(stonith_api, rsc); } else if (pcmk__str_eq(cmd->action, PCMK_ACTION_MONITOR, pcmk__str_casei)) { do_monitor = true; } else { stonith_action_complete(cmd, PCMK_OCF_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR, "Invalid fence device action (bug?)"); return; } if (do_monitor) { rc = execd_stonith_monitor(stonith_api, rsc, cmd); if (rc == pcmk_ok) { // Don't clean up yet. We will get the result of the monitor later. return; } } if (rc != -pcmk_err_generic) { rc_s = pcmk_strerror(rc); } stonith_action_complete(cmd, ((rc == pcmk_rc_ok)? CRM_EX_OK : CRM_EX_ERROR), stonith__legacy2status(rc), rc_s); } static void execute_nonstonith_action(lrmd_rsc_t *rsc, lrmd_cmd_t *cmd) { svc_action_t *action = NULL; GHashTable *params_copy = NULL; pcmk__assert((rsc != NULL) && (cmd != NULL)); crm_trace("Creating action, resource:%s action:%s class:%s provider:%s agent:%s", rsc->rsc_id, cmd->action, rsc->class, rsc->provider, rsc->type); params_copy = pcmk__str_table_dup(cmd->params); action = services__create_resource_action(rsc->rsc_id, rsc->class, rsc->provider, rsc->type, normalize_action_name(rsc, cmd->action), cmd->interval_ms, cmd->timeout, params_copy, cmd->service_flags); if (action == NULL) { pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, strerror(ENOMEM)); cmd_finalize(cmd, rsc); return; } if (action->rc != PCMK_OCF_UNKNOWN) { services__copy_result(action, &(cmd->result)); services_action_free(action); cmd_finalize(cmd, rsc); return; } action->cb_data = cmd; if (services_action_async(action, action_complete)) { /* The services library has taken responsibility for the action. It * could be pending, blocked, or merged into a duplicate recurring * action, in which case the action callback (action_complete()) * will be called when the action completes, otherwise the callback has * already been called. * * action_complete() calls cmd_finalize() which can free cmd, so cmd * cannot be used here. */ } else { /* This is a recurring action that is not being cancelled and could not * be initiated. It has been rescheduled, and the action callback * (action_complete()) has been called, which in this case has already * called cmd_finalize(), which in this case should only reset (not * free) cmd. */ services__copy_result(action, &(cmd->result)); services_action_free(action); } } static gboolean execute_resource_action(gpointer user_data) { lrmd_rsc_t *rsc = (lrmd_rsc_t *) user_data; lrmd_cmd_t *cmd = NULL; CRM_CHECK(rsc != NULL, return FALSE); if (rsc->active) { crm_trace("%s is still active", rsc->rsc_id); return TRUE; } if (rsc->pending_ops) { GList *first = rsc->pending_ops; cmd = first->data; if (cmd->delay_id) { crm_trace ("Command %s %s was asked to run too early, waiting for start_delay timeout of %dms", cmd->rsc_id, cmd->action, cmd->start_delay); return TRUE; } rsc->pending_ops = g_list_remove_link(rsc->pending_ops, first); g_list_free_1(first); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_run), &(cmd->t_first_run)); #endif cmd->epoch_last_run = time(NULL); } if (!cmd) { crm_trace("Nothing further to do for %s", rsc->rsc_id); return TRUE; } rsc->active = cmd; /* only one op at a time for a rsc */ if (cmd->interval_ms) { rsc->recurring_ops = g_list_append(rsc->recurring_ops, cmd); } log_execute(cmd); if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { execute_stonith_action(rsc, cmd); } else { execute_nonstonith_action(rsc, cmd); } return TRUE; } void free_rsc(gpointer data) { GList *gIter = NULL; lrmd_rsc_t *rsc = data; int is_stonith = pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei); gIter = rsc->pending_ops; while (gIter != NULL) { GList *next = gIter->next; lrmd_cmd_t *cmd = gIter->data; /* command was never executed */ cmd->result.execution_status = PCMK_EXEC_CANCELLED; cmd_finalize(cmd, NULL); gIter = next; } /* frees list, but not list elements. */ g_list_free(rsc->pending_ops); gIter = rsc->recurring_ops; while (gIter != NULL) { GList *next = gIter->next; lrmd_cmd_t *cmd = gIter->data; if (is_stonith) { cmd->result.execution_status = PCMK_EXEC_CANCELLED; /* If a stonith command is in-flight, just mark it as cancelled; * it is not safe to finalize/free the cmd until the stonith api * says it has either completed or timed out. */ if (rsc->active != cmd) { cmd_finalize(cmd, NULL); } } else { /* This command is already handed off to service library, * let service library cancel it and tell us via the callback * when it is cancelled. The rsc can be safely destroyed * even if we are waiting for the cancel result */ services_action_cancel(rsc->rsc_id, normalize_action_name(rsc, cmd->action), cmd->interval_ms); } gIter = next; } /* frees list, but not list elements. */ g_list_free(rsc->recurring_ops); free(rsc->rsc_id); free(rsc->class); free(rsc->provider); free(rsc->type); mainloop_destroy_trigger(rsc->work); free(rsc); } int execd_process_signon(pcmk__client_t *client, xmlNode *request, int call_id, xmlNode **reply) { int rc = pcmk_rc_ok; time_t now = time(NULL); const char *protocol_version = pcmk__xe_get(request, PCMK__XA_LRMD_PROTOCOL_VERSION); const char *start_state = pcmk__env_option(PCMK__ENV_NODE_START_STATE); if (compare_version(protocol_version, LRMD_COMPATIBLE_PROTOCOL) < 0) { crm_err("Cluster API version must be greater than or equal to %s, not %s", LRMD_COMPATIBLE_PROTOCOL, protocol_version); rc = EPROTO; } if (pcmk__xe_attr_is_true(request, PCMK__XA_LRMD_IS_IPC_PROVIDER)) { #ifdef PCMK__COMPILE_REMOTE if ((client->remote != NULL) && pcmk_is_set(client->flags, pcmk__client_tls_handshake_complete)) { const char *op = pcmk__xe_get(request, PCMK__XA_LRMD_OP); // This is a remote connection from a cluster node's controller ipc_proxy_add_provider(client); /* @TODO Allowing multiple proxies makes no sense given that clients * have no way to choose between them. Maybe always use the most * recent one and switch any existing IPC connections to use it, * by iterating over ipc_clients here, and if client->id doesn't * match the client's userdata, replace the userdata with the new * ID. After the iteration, call lrmd_remote_client_destroy() on any * of the replaced values in ipc_providers. */ /* If this was a register operation, also ask for new schema files but * only if it's supported by the protocol version. */ if (pcmk__str_eq(op, CRM_OP_REGISTER, pcmk__str_none) && LRMD_SUPPORTS_SCHEMA_XFER(protocol_version)) { remoted_request_cib_schema_files(); } } else { rc = EACCES; } #else rc = EPROTONOSUPPORT; #endif } pcmk__assert(reply != NULL); *reply = execd_create_reply(pcmk_rc2legacy(rc), call_id); crm_xml_add(*reply, PCMK__XA_LRMD_OP, CRM_OP_REGISTER); crm_xml_add(*reply, PCMK__XA_LRMD_CLIENTID, client->id); crm_xml_add(*reply, PCMK__XA_LRMD_PROTOCOL_VERSION, LRMD_PROTOCOL_VERSION); crm_xml_add_ll(*reply, PCMK__XA_UPTIME, now - start_time); if (start_state) { crm_xml_add(*reply, PCMK__XA_NODE_START_STATE, start_state); } return rc; } void execd_process_rsc_register(pcmk__client_t *client, uint32_t id, xmlNode *request) { lrmd_rsc_t *rsc = build_rsc_from_xml(request); lrmd_rsc_t *dup = g_hash_table_lookup(rsc_list, rsc->rsc_id); if (dup && pcmk__str_eq(rsc->class, dup->class, pcmk__str_casei) && pcmk__str_eq(rsc->provider, dup->provider, pcmk__str_casei) && pcmk__str_eq(rsc->type, dup->type, pcmk__str_casei)) { crm_notice("Ignoring duplicate registration of '%s'", rsc->rsc_id); free_rsc(rsc); return; } g_hash_table_replace(rsc_list, rsc->rsc_id, rsc); crm_info("Cached agent information for '%s'", rsc->rsc_id); } int execd_process_get_rsc_info(xmlNode *request, int call_id, xmlNode **reply) { int rc = pcmk_rc_ok; xmlNode *rsc_xml = pcmk__xpath_find_one(request->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); const char *rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); lrmd_rsc_t *rsc = NULL; if (rsc_id == NULL) { rc = ENODEV; } else { rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Agent information for '%s' not in cache", rsc_id); rc = ENODEV; } } CRM_LOG_ASSERT(reply != NULL); *reply = execd_create_reply(pcmk_rc2legacy(rc), call_id); if (rsc) { crm_xml_add(*reply, PCMK__XA_LRMD_RSC_ID, rsc->rsc_id); crm_xml_add(*reply, PCMK__XA_LRMD_CLASS, rsc->class); crm_xml_add(*reply, PCMK__XA_LRMD_PROVIDER, rsc->provider); crm_xml_add(*reply, PCMK__XA_LRMD_TYPE, rsc->type); } return rc; } int execd_process_rsc_unregister(pcmk__client_t *client, xmlNode *request) { int rc = pcmk_rc_ok; lrmd_rsc_t *rsc = NULL; xmlNode *rsc_xml = pcmk__xpath_find_one(request->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); const char *rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); if (!rsc_id) { return ENODEV; } rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Ignoring unregistration of resource '%s', which is not registered", rsc_id); return pcmk_rc_ok; } if (rsc->active) { /* let the caller know there are still active ops on this rsc to watch for */ crm_trace("Operation (%p) still in progress for unregistered resource %s", rsc->active, rsc_id); rc = EINPROGRESS; } g_hash_table_remove(rsc_list, rsc_id); return rc; } int execd_process_rsc_exec(pcmk__client_t *client, xmlNode *request) { lrmd_rsc_t *rsc = NULL; lrmd_cmd_t *cmd = NULL; xmlNode *rsc_xml = pcmk__xpath_find_one(request->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); const char *rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); if (!rsc_id) { return EINVAL; } if (!(rsc = g_hash_table_lookup(rsc_list, rsc_id))) { crm_info("Resource '%s' not found (%d active resources)", rsc_id, g_hash_table_size(rsc_list)); return ENODEV; } cmd = create_lrmd_cmd(request, client); /* Don't reference cmd after handing it off to be scheduled. * The cmd could get merged and freed. */ schedule_lrmd_cmd(rsc, cmd); return pcmk_rc_ok; } static int cancel_op(const char *rsc_id, const char *action, guint interval_ms) { GList *gIter = NULL; lrmd_rsc_t *rsc = g_hash_table_lookup(rsc_list, rsc_id); /* How to cancel an action. * 1. Check pending ops list, if it hasn't been handed off * to the service library or stonith recurring list remove * it there and that will stop it. * 2. If it isn't in the pending ops list, then it's either a * recurring op in the stonith recurring list, or the service * library's recurring list. Stop it there * 3. If not found in any lists, then this operation has either * been executed already and is not a recurring operation, or * never existed. */ if (!rsc) { return ENODEV; } for (gIter = rsc->pending_ops; gIter != NULL; gIter = gIter->next) { lrmd_cmd_t *cmd = gIter->data; if (action_matches(cmd, action, interval_ms)) { cmd->result.execution_status = PCMK_EXEC_CANCELLED; cmd_finalize(cmd, rsc); return pcmk_rc_ok; } } if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { /* The service library does not handle stonith operations. * We have to handle recurring stonith operations ourselves. */ for (gIter = rsc->recurring_ops; gIter != NULL; gIter = gIter->next) { lrmd_cmd_t *cmd = gIter->data; if (action_matches(cmd, action, interval_ms)) { cmd->result.execution_status = PCMK_EXEC_CANCELLED; if (rsc->active != cmd) { cmd_finalize(cmd, rsc); } return pcmk_rc_ok; } } } else if (services_action_cancel(rsc_id, normalize_action_name(rsc, action), interval_ms) == TRUE) { /* The service library will tell the action_complete callback function * this action was cancelled, which will destroy the cmd and remove * it from the recurring_op list. Do not do that in this function * if the service library says it cancelled it. */ return pcmk_rc_ok; } return EOPNOTSUPP; } static void cancel_all_recurring(lrmd_rsc_t * rsc, const char *client_id) { GList *cmd_list = NULL; GList *cmd_iter = NULL; /* Notice a copy of each list is created when concat is called. * This prevents odd behavior from occurring when the cmd_list * is iterated through later on. It is possible the cancel_op * function may end up modifying the recurring_ops and pending_ops * lists. If we did not copy those lists, our cmd_list iteration * could get messed up.*/ if (rsc->recurring_ops) { cmd_list = g_list_concat(cmd_list, g_list_copy(rsc->recurring_ops)); } if (rsc->pending_ops) { cmd_list = g_list_concat(cmd_list, g_list_copy(rsc->pending_ops)); } if (!cmd_list) { return; } for (cmd_iter = cmd_list; cmd_iter; cmd_iter = cmd_iter->next) { lrmd_cmd_t *cmd = cmd_iter->data; if (cmd->interval_ms == 0) { continue; } if (client_id && !pcmk__str_eq(cmd->client_id, client_id, pcmk__str_casei)) { continue; } cancel_op(rsc->rsc_id, cmd->action, cmd->interval_ms); } /* frees only the copied list data, not the cmds */ g_list_free(cmd_list); } int execd_process_rsc_cancel(pcmk__client_t *client, xmlNode *request) { xmlNode *rsc_xml = pcmk__xpath_find_one(request->doc, "//" PCMK__XE_LRMD_RSC, LOG_ERR); const char *rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); const char *action = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ACTION); guint interval_ms = 0; pcmk__xe_get_guint(rsc_xml, PCMK__XA_LRMD_RSC_INTERVAL, &interval_ms); if (!rsc_id || !action) { return EINVAL; } return cancel_op(rsc_id, action, interval_ms); } static void add_recurring_op_xml(xmlNode *reply, lrmd_rsc_t *rsc) { xmlNode *rsc_xml = pcmk__xe_create(reply, PCMK__XE_LRMD_RSC); crm_xml_add(rsc_xml, PCMK__XA_LRMD_RSC_ID, rsc->rsc_id); for (GList *item = rsc->recurring_ops; item != NULL; item = item->next) { lrmd_cmd_t *cmd = item->data; xmlNode *op_xml = pcmk__xe_create(rsc_xml, PCMK__XE_LRMD_RSC_OP); crm_xml_add(op_xml, PCMK__XA_LRMD_RSC_ACTION, pcmk__s(cmd->real_action, cmd->action)); crm_xml_add_ms(op_xml, PCMK__XA_LRMD_RSC_INTERVAL, cmd->interval_ms); crm_xml_add_int(op_xml, PCMK__XA_LRMD_TIMEOUT, cmd->timeout_orig); } } int execd_process_get_recurring(xmlNode *request, int call_id, xmlNode **reply) { int rc = pcmk_rc_ok; const char *rsc_id = NULL; lrmd_rsc_t *rsc = NULL; xmlNode *rsc_xml = NULL; // Resource ID is optional rsc_xml = pcmk__xe_first_child(request, PCMK__XE_LRMD_CALLDATA, NULL, NULL); if (rsc_xml) { rsc_xml = pcmk__xe_first_child(rsc_xml, PCMK__XE_LRMD_RSC, NULL, NULL); } if (rsc_xml) { rsc_id = pcmk__xe_get(rsc_xml, PCMK__XA_LRMD_RSC_ID); } // If resource ID is specified, resource must exist if (rsc_id != NULL) { rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Resource '%s' not found (%d active resources)", rsc_id, g_hash_table_size(rsc_list)); rc = ENODEV; } } CRM_LOG_ASSERT(reply != NULL); *reply = execd_create_reply(pcmk_rc2legacy(rc), call_id); // If resource ID is not specified, check all resources if (rsc_id == NULL) { GHashTableIter iter; char *key = NULL; g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &rsc)) { add_recurring_op_xml(*reply, rsc); } } else if (rsc) { add_recurring_op_xml(*reply, rsc); } return rc; } diff --git a/include/crm/common/strings_internal.h b/include/crm/common/strings_internal.h index 5d4c0de605..2bbbc435b8 100644 --- a/include/crm/common/strings_internal.h +++ b/include/crm/common/strings_internal.h @@ -1,244 +1,240 @@ /* - * Copyright 2015-2024 the Pacemaker project contributors + * Copyright 2015-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__CRM_COMMON_STRINGS_INTERNAL__H #define PCMK__CRM_COMMON_STRINGS_INTERNAL__H #include // bool #include // uint32_t, etc. #include // guint, GList, GHashTable #include // PCMK_VALUE_TRUE, PCMK_VALUE_FALSE #include // crm_strdup_printf() #ifdef __cplusplus extern "C" { #endif /* internal constants for generic string functions (from strings.c) */ #define PCMK__PARSE_INT_DEFAULT -1 #define PCMK__PARSE_DBL_DEFAULT -1.0 /* internal generic string functions (from strings.c) */ enum pcmk__str_flags { pcmk__str_none = 0, pcmk__str_casei = 1 << 0, pcmk__str_null_matches = 1 << 1, pcmk__str_regex = 1 << 2, pcmk__str_star_matches = 1 << 3, }; int pcmk__scan_double(const char *text, double *result, const char *default_text, char **end_text); int pcmk__guint_from_hash(GHashTable *table, const char *key, guint default_val, guint *result); -bool pcmk__starts_with(const char *str, const char *prefix); -bool pcmk__ends_with(const char *s, const char *match); -bool pcmk__ends_with_ext(const char *s, const char *match); -char *pcmk__trim(char *str); void pcmk__add_separated_word(GString **list, size_t init_size, const char *word, const char *separator); int pcmk__compress(const char *data, unsigned int length, unsigned int max, char **result, unsigned int *result_len); int pcmk__scan_ll(const char *text, long long *result, long long default_value); int pcmk__scan_min_int(const char *text, int *result, int minimum); int pcmk__scan_port(const char *text, int *port); -int pcmk__parse_ll_range(const char *srcstring, long long *start, long long *end); +int pcmk__parse_ll_range(const char *text, long long *start, long long *end); GHashTable *pcmk__strkey_table(GDestroyNotify key_destroy_func, GDestroyNotify value_destroy_func); GHashTable *pcmk__strikey_table(GDestroyNotify key_destroy_func, GDestroyNotify value_destroy_func); GHashTable *pcmk__str_table_dup(GHashTable *old_table); void pcmk__insert_dup(GHashTable *table, const char *name, const char *value); /*! * \internal * \brief Get a string value with a default if NULL * * \param[in] s String to return if non-NULL * \param[in] default_value String (or NULL) to return if \p s is NULL * * \return \p s if \p s is non-NULL, otherwise \p default_value */ static inline const char * pcmk__s(const char *s, const char *default_value) { return (s == NULL)? default_value : s; } /*! * \internal * \brief Create a hash table with integer keys * * \param[in] value_destroy_func Function to free a value * * \return Newly allocated hash table * \note It is the caller's responsibility to free the result, using * g_hash_table_destroy(). */ static inline GHashTable * pcmk__intkey_table(GDestroyNotify value_destroy_func) { return g_hash_table_new_full(g_direct_hash, g_direct_equal, NULL, value_destroy_func); } /*! * \internal * \brief Insert a value into a hash table with integer keys * * \param[in,out] hash_table Table to insert into * \param[in] key Integer key to insert * \param[in] value Value to insert * * \return Whether the key/value was already in the table * \note This has the same semantics as g_hash_table_insert(). If the key * already exists in the table, the old value is freed and replaced. */ static inline gboolean pcmk__intkey_table_insert(GHashTable *hash_table, int key, gpointer value) { return g_hash_table_insert(hash_table, GINT_TO_POINTER(key), value); } /*! * \internal * \brief Look up a value in a hash table with integer keys * * \param[in] hash_table Table to check * \param[in] key Integer key to look for * * \return Value in table for \key (or NULL if not found) */ static inline gpointer pcmk__intkey_table_lookup(GHashTable *hash_table, int key) { return g_hash_table_lookup(hash_table, GINT_TO_POINTER(key)); } /*! * \internal * \brief Remove a key/value from a hash table with integer keys * * \param[in,out] hash_table Table to modify * \param[in] key Integer key of entry to remove * * \return Whether \p key was found and removed from \p hash_table */ static inline gboolean pcmk__intkey_table_remove(GHashTable *hash_table, int key) { return g_hash_table_remove(hash_table, GINT_TO_POINTER(key)); } gboolean pcmk__str_in_list(const gchar *s, const GList *lst, uint32_t flags); bool pcmk__strcase_any_of(const char *s, ...) G_GNUC_NULL_TERMINATED; bool pcmk__str_any_of(const char *s, ...) G_GNUC_NULL_TERMINATED; bool pcmk__char_in_any_str(int ch, ...) G_GNUC_NULL_TERMINATED; int pcmk__strcmp(const char *s1, const char *s2, uint32_t flags); int pcmk__numeric_strcasecmp(const char *s1, const char *s2); char *pcmk__str_copy_as(const char *file, const char *function, uint32_t line, const char *str); /*! * \internal * \brief Copy a string, asserting on failure * * \param[in] str String to copy (can be \c NULL) * * \return Newly allocated copy of \p str, or \c NULL if \p str is \c NULL * * \note The caller is responsible for freeing the return value using \c free(). */ #define pcmk__str_copy(str) pcmk__str_copy_as(__FILE__, __func__, __LINE__, str) void pcmk__str_update(char **str, const char *value); void pcmk__g_strcat(GString *buffer, ...) G_GNUC_NULL_TERMINATED; static inline bool pcmk__str_eq(const char *s1, const char *s2, uint32_t flags) { return pcmk__strcmp(s1, s2, flags) == 0; } // Like pcmk__add_separated_word() but using a space as separator static inline void pcmk__add_word(GString **list, size_t init_size, const char *word) { return pcmk__add_separated_word(list, init_size, word, " "); } /* Correctly displaying singular or plural is complicated; consider "1 node has" * vs. "2 nodes have". A flexible solution is to pluralize entire strings, e.g. * * if (a == 1) { * crm_info("singular message"): * } else { * crm_info("plural message"); * } * * though even that's not sufficient for all languages besides English (if we * ever desire to do translations of output and log messages). But the following * convenience macros are "good enough" and more concise for many cases. */ /* Example: * crm_info("Found %d %s", nentries, * pcmk__plural_alt(nentries, "entry", "entries")); */ #define pcmk__plural_alt(i, s1, s2) (((i) == 1)? (s1) : (s2)) // Example: crm_info("Found %d node%s", nnodes, pcmk__plural_s(nnodes)); #define pcmk__plural_s(i) pcmk__plural_alt(i, "", "s") static inline int pcmk__str_empty(const char *s) { return (s == NULL) || (s[0] == '\0'); } static inline char * pcmk__itoa(int an_int) { return crm_strdup_printf("%d", an_int); } static inline char * pcmk__ftoa(double a_float) { return crm_strdup_printf("%f", a_float); } static inline char * pcmk__ttoa(time_t epoch_time) { return crm_strdup_printf("%lld", (long long) epoch_time); } // note this returns const not allocated static inline const char * pcmk__btoa(bool condition) { return condition? PCMK_VALUE_TRUE : PCMK_VALUE_FALSE; } #ifdef __cplusplus } #endif #endif // PCMK__CRM_COMMON_STRINGS_INTERNAL__H diff --git a/lib/cib/cib_file.c b/lib/cib/cib_file.c index c0a9505b3d..7348ce5f06 100644 --- a/lib/cib/cib_file.c +++ b/lib/cib/cib_file.c @@ -1,1178 +1,1178 @@ /* * Original copyright 2004 International Business Machines * Later changes copyright 2008-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define CIB_SERIES "cib" #define CIB_SERIES_MAX 100 #define CIB_SERIES_BZIP FALSE /* Must be false because archived copies are created with hard links */ #define CIB_LIVE_NAME CIB_SERIES ".xml" // key: client ID (const char *) -> value: client (cib_t *) static GHashTable *client_table = NULL; enum cib_file_flags { cib_file_flag_dirty = (1 << 0), cib_file_flag_live = (1 << 1), }; typedef struct cib_file_opaque_s { char *id; char *filename; uint32_t flags; // Group of enum cib_file_flags xmlNode *cib_xml; } cib_file_opaque_t; static int cib_file_process_commit_transaction(const char *op, int options, const char *section, xmlNode *req, xmlNode *input, xmlNode *existing_cib, xmlNode **result_cib, xmlNode **answer); /*! * \internal * \brief Add a CIB file client to client table * * \param[in] cib CIB client */ static void register_client(const cib_t *cib) { cib_file_opaque_t *private = cib->variant_opaque; if (client_table == NULL) { client_table = pcmk__strkey_table(NULL, NULL); } g_hash_table_insert(client_table, private->id, (gpointer) cib); } /*! * \internal * \brief Remove a CIB file client from client table * * \param[in] cib CIB client */ static void unregister_client(const cib_t *cib) { cib_file_opaque_t *private = cib->variant_opaque; if (client_table == NULL) { return; } g_hash_table_remove(client_table, private->id); /* @COMPAT: Add to crm_exit() when libcib and libcrmcommon are merged, * instead of destroying the client table when there are no more clients. */ if (g_hash_table_size(client_table) == 0) { g_hash_table_destroy(client_table); client_table = NULL; } } /*! * \internal * \brief Look up a CIB file client by its ID * * \param[in] client_id CIB client ID * * \return CIB client with matching ID if found, or \p NULL otherwise */ static cib_t * get_client(const char *client_id) { if (client_table == NULL) { return NULL; } return g_hash_table_lookup(client_table, (gpointer) client_id); } static const cib__op_fn_t cib_op_functions[] = { [cib__op_apply_patch] = cib_process_diff, [cib__op_bump] = cib_process_bump, [cib__op_commit_transact] = cib_file_process_commit_transaction, [cib__op_create] = cib_process_create, [cib__op_delete] = cib_process_delete, [cib__op_erase] = cib_process_erase, [cib__op_modify] = cib_process_modify, [cib__op_query] = cib_process_query, [cib__op_replace] = cib_process_replace, [cib__op_upgrade] = cib_process_upgrade, }; /* cib_file_backup() and cib_file_write_with_digest() need to chown the * written files only in limited circumstances, so these variables allow * that to be indicated without affecting external callers */ static uid_t cib_file_owner = 0; static uid_t cib_file_group = 0; static gboolean cib_do_chown = FALSE; #define cib_set_file_flags(cibfile, flags_to_set) do { \ (cibfile)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "CIB file", \ cibfile->filename, \ (cibfile)->flags, \ (flags_to_set), \ #flags_to_set); \ } while (0) #define cib_clear_file_flags(cibfile, flags_to_clear) do { \ (cibfile)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "CIB file", \ cibfile->filename, \ (cibfile)->flags, \ (flags_to_clear), \ #flags_to_clear); \ } while (0) /*! * \internal * \brief Get the function that performs a given CIB file operation * * \param[in] operation Operation whose function to look up * * \return Function that performs \p operation for a CIB file client */ static cib__op_fn_t file_get_op_function(const cib__operation_t *operation) { enum cib__op_type type = operation->type; pcmk__assert(type >= 0); if (type >= PCMK__NELEM(cib_op_functions)) { return NULL; } return cib_op_functions[type]; } /*! * \internal * \brief Check whether a file is the live CIB * * \param[in] filename Name of file to check * * \return TRUE if file exists and its real path is same as live CIB's */ static gboolean cib_file_is_live(const char *filename) { gboolean same = FALSE; if (filename != NULL) { // Canonicalize file names for true comparison char *real_filename = NULL; if (pcmk__real_path(filename, &real_filename) == pcmk_rc_ok) { char *real_livename = NULL; if (pcmk__real_path(CRM_CONFIG_DIR "/" CIB_LIVE_NAME, &real_livename) == pcmk_rc_ok) { same = !strcmp(real_filename, real_livename); free(real_livename); } free(real_filename); } } return same; } static int cib_file_process_request(cib_t *cib, xmlNode *request, xmlNode **output) { int rc = pcmk_ok; const cib__operation_t *operation = NULL; cib__op_fn_t op_function = NULL; int call_id = 0; uint32_t call_options = cib_none; const char *op = pcmk__xe_get(request, PCMK__XA_CIB_OP); const char *section = pcmk__xe_get(request, PCMK__XA_CIB_SECTION); xmlNode *wrapper = pcmk__xe_first_child(request, PCMK__XE_CIB_CALLDATA, NULL, NULL); xmlNode *data = pcmk__xe_first_child(wrapper, NULL, NULL, NULL); bool changed = false; bool read_only = false; xmlNode *result_cib = NULL; xmlNode *cib_diff = NULL; cib_file_opaque_t *private = cib->variant_opaque; // We error checked these in callers cib__get_operation(op, &operation); op_function = file_get_op_function(operation); pcmk__xe_get_int(request, PCMK__XA_CIB_CALLID, &call_id); rc = pcmk__xe_get_flags(request, PCMK__XA_CIB_CALLOPT, &call_options, cib_none); if (rc != pcmk_rc_ok) { crm_warn("Couldn't parse options from request: %s", pcmk_rc_str(rc)); } read_only = !pcmk_is_set(operation->flags, cib__op_attr_modifies); // Mirror the logic in prepare_input() in the CIB manager if ((section != NULL) && pcmk__xe_is(data, PCMK_XE_CIB)) { data = pcmk_find_cib_element(data, section); } rc = cib_perform_op(cib, op, call_options, op_function, read_only, section, request, data, true, &changed, &private->cib_xml, &result_cib, &cib_diff, output); if (pcmk_is_set(call_options, cib_transaction)) { /* The rest of the logic applies only to the transaction as a whole, not * to individual requests. */ goto done; } if (rc == -pcmk_err_schema_validation) { // Show validation errors to stderr pcmk__validate_xml(result_cib, NULL, NULL, NULL); } else if ((rc == pcmk_ok) && !read_only) { pcmk__log_xml_patchset(LOG_DEBUG, cib_diff); if (result_cib != private->cib_xml) { pcmk__xml_free(private->cib_xml); private->cib_xml = result_cib; } cib_set_file_flags(private, cib_file_flag_dirty); } done: if ((result_cib != private->cib_xml) && (result_cib != *output)) { pcmk__xml_free(result_cib); } pcmk__xml_free(cib_diff); return rc; } static int cib_file_perform_op_delegate(cib_t *cib, const char *op, const char *host, const char *section, xmlNode *data, xmlNode **output_data, int call_options, const char *user_name) { int rc = pcmk_ok; xmlNode *request = NULL; xmlNode *output = NULL; cib_file_opaque_t *private = cib->variant_opaque; const cib__operation_t *operation = NULL; crm_info("Handling %s operation for %s as %s", pcmk__s(op, "invalid"), pcmk__s(section, "entire CIB"), pcmk__s(user_name, "default user")); if (output_data != NULL) { *output_data = NULL; } if (cib->state == cib_disconnected) { return -ENOTCONN; } rc = cib__get_operation(op, &operation); rc = pcmk_rc2legacy(rc); if (rc != pcmk_ok) { // @COMPAT: At compatibility break, use rc directly return -EPROTONOSUPPORT; } if (file_get_op_function(operation) == NULL) { // @COMPAT: At compatibility break, use EOPNOTSUPP crm_err("Operation %s is not supported by CIB file clients", op); return -EPROTONOSUPPORT; } cib__set_call_options(call_options, "file operation", cib_no_mtime); rc = cib__create_op(cib, op, host, section, data, call_options, user_name, NULL, &request); if (rc != pcmk_ok) { return rc; } crm_xml_add(request, PCMK__XA_ACL_TARGET, user_name); crm_xml_add(request, PCMK__XA_CIB_CLIENTID, private->id); if (pcmk_is_set(call_options, cib_transaction)) { rc = cib__extend_transaction(cib, request); goto done; } rc = cib_file_process_request(cib, request, &output); if ((output_data != NULL) && (output != NULL)) { if (output->doc == private->cib_xml->doc) { *output_data = pcmk__xml_copy(NULL, output); } else { *output_data = output; } } done: if ((output != NULL) && (output->doc != private->cib_xml->doc) && ((output_data == NULL) || (output != *output_data))) { pcmk__xml_free(output); } pcmk__xml_free(request); return rc; } /*! * \internal * \brief Read CIB from disk and validate it against XML schema * * \param[in] filename Name of file to read CIB from * \param[out] output Where to store the read CIB XML * * \return pcmk_ok on success, * -ENXIO if file does not exist (or stat() otherwise fails), or * -pcmk_err_schema_validation if XML doesn't parse or validate * \note If filename is the live CIB, this will *not* verify its digest, * though that functionality would be trivial to add here. * Also, this will *not* verify that the file is writable, * because some callers might not need to write. */ static int load_file_cib(const char *filename, xmlNode **output) { struct stat buf; xmlNode *root = NULL; /* Ensure file is readable */ if (strcmp(filename, "-") && (stat(filename, &buf) < 0)) { return -ENXIO; } /* Parse XML from file */ root = pcmk__xml_read(filename); if (root == NULL) { return -pcmk_err_schema_validation; } /* Add a status section if not already present */ if (pcmk__xe_first_child(root, PCMK_XE_STATUS, NULL, NULL) == NULL) { pcmk__xe_create(root, PCMK_XE_STATUS); } /* Validate XML against its specified schema */ if (!pcmk__configured_schema_validates(root)) { pcmk__xml_free(root); return -pcmk_err_schema_validation; } /* Remember the parsed XML for later use */ *output = root; return pcmk_ok; } static int cib_file_signon(cib_t *cib, const char *name, enum cib_conn_type type) { int rc = pcmk_ok; cib_file_opaque_t *private = cib->variant_opaque; if (private->filename == NULL) { rc = -EINVAL; } else { rc = load_file_cib(private->filename, &private->cib_xml); } if (rc == pcmk_ok) { crm_debug("Opened connection to local file '%s' for %s", private->filename, pcmk__s(name, "client")); cib->state = cib_connected_command; cib->type = cib_command; register_client(cib); } else { crm_info("Connection to local file '%s' for %s (client %s) failed: %s", private->filename, pcmk__s(name, "client"), private->id, pcmk_strerror(rc)); } return rc; } /*! * \internal * \brief Write out the in-memory CIB to a live CIB file * * \param[in] cib_root Root of XML tree to write * \param[in,out] path Full path to file to write * * \return 0 on success, -1 on failure */ static int cib_file_write_live(xmlNode *cib_root, char *path) { uid_t uid = geteuid(); struct passwd *daemon_pwent; char *sep = strrchr(path, '/'); const char *cib_dirname, *cib_filename; int rc = 0; /* Get the desired uid/gid */ errno = 0; daemon_pwent = getpwnam(CRM_DAEMON_USER); if (daemon_pwent == NULL) { crm_perror(LOG_ERR, "Could not find %s user", CRM_DAEMON_USER); return -1; } /* If we're root, we can change the ownership; * if we're daemon, anything we create will be OK; * otherwise, block access so we don't create wrong owner */ if ((uid != 0) && (uid != daemon_pwent->pw_uid)) { crm_perror(LOG_ERR, "Must be root or %s to modify live CIB", CRM_DAEMON_USER); return 0; } /* fancy footwork to separate dirname from filename * (we know the canonical name maps to the live CIB, * but the given name might be relative, or symlinked) */ if (sep == NULL) { /* no directory component specified */ cib_dirname = "./"; cib_filename = path; } else if (sep == path) { /* given name is in / */ cib_dirname = "/"; cib_filename = path + 1; } else { /* typical case; split given name into parts */ *sep = '\0'; cib_dirname = path; cib_filename = sep + 1; } /* if we're root, we want to update the file ownership */ if (uid == 0) { cib_file_owner = daemon_pwent->pw_uid; cib_file_group = daemon_pwent->pw_gid; cib_do_chown = TRUE; } /* write the file */ if (cib_file_write_with_digest(cib_root, cib_dirname, cib_filename) != pcmk_ok) { rc = -1; } /* turn off file ownership changes, for other callers */ if (uid == 0) { cib_do_chown = FALSE; } /* undo fancy stuff */ if ((sep != NULL) && (*sep == '\0')) { *sep = '/'; } return rc; } /*! * \internal * \brief Sign-off method for CIB file variants * * This will write the file to disk if needed, and free the in-memory CIB. If * the file is the live CIB, it will compute and write a signature as well. * * \param[in,out] cib CIB object to sign off * * \return pcmk_ok on success, pcmk_err_generic on failure * \todo This method should refuse to write the live CIB if the CIB manager is * running. */ static int cib_file_signoff(cib_t *cib) { int rc = pcmk_ok; cib_file_opaque_t *private = cib->variant_opaque; crm_debug("Disconnecting from the CIB manager"); cib->state = cib_disconnected; cib->type = cib_no_connection; unregister_client(cib); cib->cmds->end_transaction(cib, false, cib_none); /* If the in-memory CIB has been changed, write it to disk */ if (pcmk_is_set(private->flags, cib_file_flag_dirty)) { /* If this is the live CIB, write it out with a digest */ if (pcmk_is_set(private->flags, cib_file_flag_live)) { if (cib_file_write_live(private->cib_xml, private->filename) < 0) { rc = pcmk_err_generic; } /* Otherwise, it's a simple write */ } else { - bool compress = pcmk__ends_with_ext(private->filename, ".bz2"); + bool compress = g_str_has_suffix(private->filename, ".bz2"); if (pcmk__xml_write_file(private->cib_xml, private->filename, compress) != pcmk_rc_ok) { rc = pcmk_err_generic; } } if (rc == pcmk_ok) { crm_info("Wrote CIB to %s", private->filename); cib_clear_file_flags(private, cib_file_flag_dirty); } else { crm_err("Could not write CIB to %s", private->filename); } } /* Free the in-memory CIB */ pcmk__xml_free(private->cib_xml); private->cib_xml = NULL; return rc; } static int cib_file_free(cib_t *cib) { int rc = pcmk_ok; if (cib->state != cib_disconnected) { rc = cib_file_signoff(cib); } if (rc == pcmk_ok) { cib_file_opaque_t *private = cib->variant_opaque; free(private->id); free(private->filename); free(private); free(cib->cmds); free(cib->user); free(cib); } else { fprintf(stderr, "Couldn't sign off: %d\n", rc); } return rc; } static int cib_file_register_notification(cib_t *cib, const char *callback, int enabled) { return -EPROTONOSUPPORT; } static int cib_file_set_connection_dnotify(cib_t *cib, void (*dnotify) (gpointer user_data)) { return -EPROTONOSUPPORT; } /*! * \internal * \brief Get the given CIB connection's unique client identifier * * \param[in] cib CIB connection * \param[out] async_id If not \p NULL, where to store asynchronous client ID * \param[out] sync_id If not \p NULL, where to store synchronous client ID * * \return Legacy Pacemaker return code * * \note This is the \p cib_file variant implementation of * \p cib_api_operations_t:client_id(). */ static int cib_file_client_id(const cib_t *cib, const char **async_id, const char **sync_id) { cib_file_opaque_t *private = cib->variant_opaque; if (async_id != NULL) { *async_id = private->id; } if (sync_id != NULL) { *sync_id = private->id; } return pcmk_ok; } cib_t * cib_file_new(const char *cib_location) { cib_t *cib = NULL; cib_file_opaque_t *private = NULL; char *filename = NULL; if (cib_location == NULL) { cib_location = getenv("CIB_file"); if (cib_location == NULL) { return NULL; // Shouldn't be possible if we were called internally } } cib = cib_new_variant(); if (cib == NULL) { return NULL; } filename = strdup(cib_location); if (filename == NULL) { free(cib); return NULL; } private = calloc(1, sizeof(cib_file_opaque_t)); if (private == NULL) { free(cib); free(filename); return NULL; } private->id = crm_generate_uuid(); private->filename = filename; cib->variant = cib_file; cib->variant_opaque = private; private->flags = 0; if (cib_file_is_live(cib_location)) { cib_set_file_flags(private, cib_file_flag_live); crm_trace("File %s detected as live CIB", cib_location); } /* assign variant specific ops */ cib->delegate_fn = cib_file_perform_op_delegate; cib->cmds->signon = cib_file_signon; cib->cmds->signoff = cib_file_signoff; cib->cmds->free = cib_file_free; cib->cmds->register_notification = cib_file_register_notification; cib->cmds->set_connection_dnotify = cib_file_set_connection_dnotify; cib->cmds->client_id = cib_file_client_id; return cib; } /*! * \internal * \brief Compare the calculated digest of an XML tree against a signature file * * \param[in] root Root of XML tree to compare * \param[in] sigfile Name of signature file containing digest to compare * * \return TRUE if digests match or signature file does not exist, else FALSE */ static gboolean cib_file_verify_digest(xmlNode *root, const char *sigfile) { gboolean passed = FALSE; char *expected; int rc = pcmk__file_contents(sigfile, &expected); switch (rc) { case pcmk_rc_ok: if (expected == NULL) { crm_err("On-disk digest at %s is empty", sigfile); return FALSE; } break; case ENOENT: crm_warn("No on-disk digest present at %s", sigfile); return TRUE; default: crm_err("Could not read on-disk digest from %s: %s", sigfile, pcmk_rc_str(rc)); return FALSE; } passed = pcmk__verify_digest(root, expected); free(expected); return passed; } /*! * \internal * \brief Read an XML tree from a file and verify its digest * * \param[in] filename Name of XML file to read * \param[in] sigfile Name of signature file containing digest to compare * \param[out] root If non-NULL, will be set to pointer to parsed XML tree * * \return 0 if file was successfully read, parsed and verified, otherwise: * -errno on stat() failure, * -pcmk_err_cib_corrupt if file size is 0 or XML is not parseable, or * -pcmk_err_cib_modified if digests do not match * \note If root is non-NULL, it is the caller's responsibility to free *root on * successful return. */ int cib_file_read_and_verify(const char *filename, const char *sigfile, xmlNode **root) { int s_res; struct stat buf; char *local_sigfile = NULL; xmlNode *local_root = NULL; pcmk__assert(filename != NULL); if (root) { *root = NULL; } /* Verify that file exists and its size is nonzero */ s_res = stat(filename, &buf); if (s_res < 0) { crm_perror(LOG_WARNING, "Could not verify cluster configuration file %s", filename); return -errno; } else if (buf.st_size == 0) { crm_warn("Cluster configuration file %s is corrupt (size is zero)", filename); return -pcmk_err_cib_corrupt; } /* Parse XML */ local_root = pcmk__xml_read(filename); if (local_root == NULL) { crm_warn("Cluster configuration file %s is corrupt (unparseable as XML)", filename); return -pcmk_err_cib_corrupt; } /* If sigfile is not specified, use original file name plus .sig */ if (sigfile == NULL) { sigfile = local_sigfile = crm_strdup_printf("%s.sig", filename); } /* Verify that digests match */ if (cib_file_verify_digest(local_root, sigfile) == FALSE) { free(local_sigfile); pcmk__xml_free(local_root); return -pcmk_err_cib_modified; } free(local_sigfile); if (root) { *root = local_root; } else { pcmk__xml_free(local_root); } return pcmk_ok; } /*! * \internal * \brief Back up a CIB * * \param[in] cib_dirname Directory containing CIB file and backups * \param[in] cib_filename Name (relative to cib_dirname) of CIB file to back up * * \return 0 on success, -1 on error */ static int cib_file_backup(const char *cib_dirname, const char *cib_filename) { int rc = 0; unsigned int seq = 0U; char *cib_path = crm_strdup_printf("%s/%s", cib_dirname, cib_filename); char *cib_digest = crm_strdup_printf("%s.sig", cib_path); char *backup_path; char *backup_digest; // Determine backup and digest file names if (pcmk__read_series_sequence(cib_dirname, CIB_SERIES, &seq) != pcmk_rc_ok) { // @TODO maybe handle errors better ... seq = 0U; } backup_path = pcmk__series_filename(cib_dirname, CIB_SERIES, seq, CIB_SERIES_BZIP); backup_digest = crm_strdup_printf("%s.sig", backup_path); /* Remove the old backups if they exist */ unlink(backup_path); unlink(backup_digest); /* Back up the CIB, by hard-linking it to the backup name */ if ((link(cib_path, backup_path) < 0) && (errno != ENOENT)) { crm_perror(LOG_ERR, "Could not archive %s by linking to %s", cib_path, backup_path); rc = -1; /* Back up the CIB signature similarly */ } else if ((link(cib_digest, backup_digest) < 0) && (errno != ENOENT)) { crm_perror(LOG_ERR, "Could not archive %s by linking to %s", cib_digest, backup_digest); rc = -1; /* Update the last counter and ensure everything is sync'd to media */ } else { pcmk__write_series_sequence(cib_dirname, CIB_SERIES, ++seq, CIB_SERIES_MAX); if (cib_do_chown) { int rc2; if ((chown(backup_path, cib_file_owner, cib_file_group) < 0) && (errno != ENOENT)) { crm_perror(LOG_ERR, "Could not set owner of %s", backup_path); rc = -1; } if ((chown(backup_digest, cib_file_owner, cib_file_group) < 0) && (errno != ENOENT)) { crm_perror(LOG_ERR, "Could not set owner of %s", backup_digest); rc = -1; } rc2 = pcmk__chown_series_sequence(cib_dirname, CIB_SERIES, cib_file_owner, cib_file_group); if (rc2 != pcmk_rc_ok) { crm_err("Could not set owner of sequence file in %s: %s", cib_dirname, pcmk_rc_str(rc2)); rc = -1; } } pcmk__sync_directory(cib_dirname); crm_info("Archived previous version as %s", backup_path); } free(cib_path); free(cib_digest); free(backup_path); free(backup_digest); return rc; } /*! * \internal * \brief Prepare CIB XML to be written to disk * * Set \c PCMK_XA_NUM_UPDATES to 0, set \c PCMK_XA_CIB_LAST_WRITTEN to the * current timestamp, and strip out the status section. * * \param[in,out] root Root of CIB XML tree * * \return void */ static void cib_file_prepare_xml(xmlNode *root) { xmlNode *cib_status_root = NULL; /* Always write out with num_updates=0 and current last-written timestamp */ crm_xml_add(root, PCMK_XA_NUM_UPDATES, "0"); pcmk__xe_add_last_written(root); /* Delete status section before writing to file, because * we discard it on startup anyway, and users get confused by it */ cib_status_root = pcmk__xe_first_child(root, PCMK_XE_STATUS, NULL, NULL); CRM_CHECK(cib_status_root != NULL, return); pcmk__xml_free(cib_status_root); } /*! * \internal * \brief Write CIB to disk, along with a signature file containing its digest * * \param[in,out] cib_root Root of XML tree to write * \param[in] cib_dirname Directory containing CIB and signature files * \param[in] cib_filename Name (relative to cib_dirname) of file to write * * \return pcmk_ok on success, * pcmk_err_cib_modified if existing cib_filename doesn't match digest, * pcmk_err_cib_backup if existing cib_filename couldn't be backed up, * or pcmk_err_cib_save if new cib_filename couldn't be saved */ int cib_file_write_with_digest(xmlNode *cib_root, const char *cib_dirname, const char *cib_filename) { int exit_rc = pcmk_ok; int rc, fd; char *digest = NULL; /* Detect CIB version for diagnostic purposes */ const char *epoch = pcmk__xe_get(cib_root, PCMK_XA_EPOCH); const char *admin_epoch = pcmk__xe_get(cib_root, PCMK_XA_ADMIN_EPOCH); /* Determine full CIB and signature pathnames */ char *cib_path = crm_strdup_printf("%s/%s", cib_dirname, cib_filename); char *digest_path = crm_strdup_printf("%s.sig", cib_path); /* Create temporary file name patterns for writing out CIB and signature */ char *tmp_cib = crm_strdup_printf("%s/cib.XXXXXX", cib_dirname); char *tmp_digest = crm_strdup_printf("%s/cib.XXXXXX", cib_dirname); /* Ensure the admin didn't modify the existing CIB underneath us */ crm_trace("Reading cluster configuration file %s", cib_path); rc = cib_file_read_and_verify(cib_path, NULL, NULL); if ((rc != pcmk_ok) && (rc != -ENOENT)) { crm_err("%s was manually modified while the cluster was active!", cib_path); exit_rc = pcmk_err_cib_modified; goto cleanup; } /* Back up the existing CIB */ if (cib_file_backup(cib_dirname, cib_filename) < 0) { exit_rc = pcmk_err_cib_backup; goto cleanup; } crm_debug("Writing CIB to disk"); umask(S_IWGRP | S_IWOTH | S_IROTH); cib_file_prepare_xml(cib_root); /* Write the CIB to a temporary file, so we can deploy (near) atomically */ fd = mkstemp(tmp_cib); if (fd < 0) { crm_perror(LOG_ERR, "Couldn't open temporary file %s for writing CIB", tmp_cib); exit_rc = pcmk_err_cib_save; goto cleanup; } /* Protect the temporary file */ if (fchmod(fd, S_IRUSR | S_IWUSR) < 0) { crm_perror(LOG_ERR, "Couldn't protect temporary file %s for writing CIB", tmp_cib); exit_rc = pcmk_err_cib_save; goto cleanup; } if (cib_do_chown && (fchown(fd, cib_file_owner, cib_file_group) < 0)) { crm_perror(LOG_ERR, "Couldn't protect temporary file %s for writing CIB", tmp_cib); exit_rc = pcmk_err_cib_save; goto cleanup; } /* Write out the CIB */ if (pcmk__xml_write_fd(cib_root, tmp_cib, fd) != pcmk_rc_ok) { crm_err("Changes couldn't be written to %s", tmp_cib); exit_rc = pcmk_err_cib_save; goto cleanup; } /* Calculate CIB digest */ digest = pcmk__digest_on_disk_cib(cib_root); pcmk__assert(digest != NULL); crm_info("Wrote version %s.%s.0 of the CIB to disk (digest: %s)", (admin_epoch ? admin_epoch : "0"), (epoch ? epoch : "0"), digest); /* Write the CIB digest to a temporary file */ fd = mkstemp(tmp_digest); if (fd < 0) { crm_perror(LOG_ERR, "Could not create temporary file for CIB digest"); exit_rc = pcmk_err_cib_save; goto cleanup; } if (cib_do_chown && (fchown(fd, cib_file_owner, cib_file_group) < 0)) { crm_perror(LOG_ERR, "Couldn't protect temporary file %s for writing CIB", tmp_cib); exit_rc = pcmk_err_cib_save; close(fd); goto cleanup; } rc = pcmk__write_sync(fd, digest); if (rc != pcmk_rc_ok) { crm_err("Could not write digest to %s: %s", tmp_digest, pcmk_rc_str(rc)); exit_rc = pcmk_err_cib_save; close(fd); goto cleanup; } close(fd); crm_debug("Wrote digest %s to disk", digest); /* Verify that what we wrote is sane */ crm_info("Reading cluster configuration file %s (digest: %s)", tmp_cib, tmp_digest); rc = cib_file_read_and_verify(tmp_cib, tmp_digest, NULL); pcmk__assert(rc == 0); /* Rename temporary files to live, and sync directory changes to media */ crm_debug("Activating %s", tmp_cib); if (rename(tmp_cib, cib_path) < 0) { crm_perror(LOG_ERR, "Couldn't rename %s as %s", tmp_cib, cib_path); exit_rc = pcmk_err_cib_save; } if (rename(tmp_digest, digest_path) < 0) { crm_perror(LOG_ERR, "Couldn't rename %s as %s", tmp_digest, digest_path); exit_rc = pcmk_err_cib_save; } pcmk__sync_directory(cib_dirname); cleanup: free(cib_path); free(digest_path); free(digest); free(tmp_digest); free(tmp_cib); return exit_rc; } /*! * \internal * \brief Process requests in a CIB transaction * * Stop when a request fails or when all requests have been processed. * * \param[in,out] cib CIB client * \param[in,out] transaction CIB transaction * * \return Standard Pacemaker return code */ static int cib_file_process_transaction_requests(cib_t *cib, xmlNode *transaction) { cib_file_opaque_t *private = cib->variant_opaque; for (xmlNode *request = pcmk__xe_first_child(transaction, PCMK__XE_CIB_COMMAND, NULL, NULL); request != NULL; request = pcmk__xe_next(request, PCMK__XE_CIB_COMMAND)) { xmlNode *output = NULL; const char *op = pcmk__xe_get(request, PCMK__XA_CIB_OP); int rc = cib_file_process_request(cib, request, &output); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { crm_err("Aborting transaction for CIB file client (%s) on file " "'%s' due to failed %s request: %s", private->id, private->filename, op, pcmk_rc_str(rc)); crm_log_xml_info(request, "Failed request"); return rc; } crm_trace("Applied %s request to transaction working CIB for CIB file " "client (%s) on file '%s'", op, private->id, private->filename); crm_log_xml_trace(request, "Successful request"); } return pcmk_rc_ok; } /*! * \internal * \brief Commit a given CIB file client's transaction to a working CIB copy * * \param[in,out] cib CIB file client * \param[in] transaction CIB transaction * \param[in,out] result_cib Where to store result CIB * * \return Standard Pacemaker return code * * \note The caller is responsible for replacing the \p cib argument's * \p private->cib_xml with \p result_cib on success, and for freeing * \p result_cib using \p pcmk__xml_free() on failure. */ static int cib_file_commit_transaction(cib_t *cib, xmlNode *transaction, xmlNode **result_cib) { int rc = pcmk_rc_ok; cib_file_opaque_t *private = cib->variant_opaque; xmlNode *saved_cib = private->cib_xml; CRM_CHECK(pcmk__xe_is(transaction, PCMK__XE_CIB_TRANSACTION), return pcmk_rc_no_transaction); /* *result_cib should be a copy of private->cib_xml (created by * cib_perform_op()). If not, make a copy now. Change tracking isn't * strictly required here because: * * Each request in the transaction will have changes tracked and ACLs * checked if appropriate. * * cib_perform_op() will infer changes for the commit request at the end. */ CRM_CHECK((*result_cib != NULL) && (*result_cib != private->cib_xml), *result_cib = pcmk__xml_copy(NULL, private->cib_xml)); crm_trace("Committing transaction for CIB file client (%s) on file '%s' to " "working CIB", private->id, private->filename); // Apply all changes to a working copy of the CIB private->cib_xml = *result_cib; rc = cib_file_process_transaction_requests(cib, transaction); crm_trace("Transaction commit %s for CIB file client (%s) on file '%s'", ((rc == pcmk_rc_ok)? "succeeded" : "failed"), private->id, private->filename); /* Some request types (for example, erase) may have freed private->cib_xml * (the working copy) and pointed it at a new XML object. In that case, it * follows that *result_cib (the working copy) was freed. * * Point *result_cib at the updated working copy stored in private->cib_xml. */ *result_cib = private->cib_xml; // Point private->cib_xml back to the unchanged original copy private->cib_xml = saved_cib; return rc; } static int cib_file_process_commit_transaction(const char *op, int options, const char *section, xmlNode *req, xmlNode *input, xmlNode *existing_cib, xmlNode **result_cib, xmlNode **answer) { int rc = pcmk_rc_ok; const char *client_id = pcmk__xe_get(req, PCMK__XA_CIB_CLIENTID); cib_t *cib = NULL; CRM_CHECK(client_id != NULL, return -EINVAL); cib = get_client(client_id); CRM_CHECK(cib != NULL, return -EINVAL); rc = cib_file_commit_transaction(cib, input, result_cib); if (rc != pcmk_rc_ok) { cib_file_opaque_t *private = cib->variant_opaque; crm_err("Could not commit transaction for CIB file client (%s) on " "file '%s': %s", private->id, private->filename, pcmk_rc_str(rc)); } return pcmk_rc2legacy(rc); } diff --git a/lib/common/agents.c b/lib/common/agents.c index 34acbb0b45..4448d0ba8d 100644 --- a/lib/common/agents.c +++ b/lib/common/agents.c @@ -1,191 +1,193 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include +#include // g_str_has_prefix() + #include #include /*! * \brief Get capabilities of a resource agent standard * * \param[in] standard Standard name * * \return Bitmask of enum pcmk_ra_caps values */ uint32_t pcmk_get_ra_caps(const char *standard) { /* @COMPAT This should probably be case-sensitive, but isn't, * for backward compatibility. */ if (standard == NULL) { return pcmk_ra_cap_none; } else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_OCF)) { return pcmk_ra_cap_provider | pcmk_ra_cap_params | pcmk_ra_cap_unique | pcmk_ra_cap_promotable | pcmk_ra_cap_cli_exec; } else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_STONITH)) { /* @COMPAT Stonith resources can't really be unique clones, but we've * allowed it in the past and have it in some scheduler regression tests * (which were likely never used as real configurations). * * @TODO Remove pcmk_ra_cap_unique at the next major schema version * bump, with a transform to remove PCMK_META_GLOBALLY_UNIQUE from the * config. */ return pcmk_ra_cap_params | pcmk_ra_cap_unique | pcmk_ra_cap_stdin | pcmk_ra_cap_fence_params; } else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_LSB)) { return pcmk_ra_cap_status | pcmk_ra_cap_cli_exec; } else if (!strcasecmp(standard, PCMK_RESOURCE_CLASS_SYSTEMD) || !strcasecmp(standard, PCMK_RESOURCE_CLASS_SERVICE)) { return pcmk_ra_cap_status; } return pcmk_ra_cap_none; } int pcmk__effective_rc(int rc) { int remapped_rc = rc; switch (rc) { case PCMK_OCF_DEGRADED: remapped_rc = PCMK_OCF_OK; break; case PCMK_OCF_DEGRADED_PROMOTED: remapped_rc = PCMK_OCF_RUNNING_PROMOTED; break; default: break; } return remapped_rc; } char * crm_generate_ra_key(const char *standard, const char *provider, const char *type) { bool std_empty = pcmk__str_empty(standard); bool prov_empty = pcmk__str_empty(provider); bool ty_empty = pcmk__str_empty(type); if (std_empty || ty_empty) { return NULL; } return crm_strdup_printf("%s%s%s:%s", standard, (prov_empty ? "" : ":"), (prov_empty ? "" : provider), type); } /*! * \brief Parse a "standard[:provider]:type" agent specification * * \param[in] spec Agent specification * \param[out] standard Newly allocated memory containing agent standard (or NULL) * \param[out] provider Newly allocated memory containing agent provider (or NULL) * \param[put] type Newly allocated memory containing agent type (or NULL) * * \return pcmk_ok if the string could be parsed, -EINVAL otherwise * * \note It is acceptable for the type to contain a ':' if the standard supports * that. For example, systemd supports the form "systemd:UNIT@A:B". * \note It is the caller's responsibility to free the returned values. */ int crm_parse_agent_spec(const char *spec, char **standard, char **provider, char **type) { char *colon; CRM_CHECK(spec && standard && provider && type, return -EINVAL); *standard = NULL; *provider = NULL; *type = NULL; colon = strchr(spec, ':'); if ((colon == NULL) || (colon == spec)) { return -EINVAL; } *standard = strndup(spec, colon - spec); spec = colon + 1; if (pcmk_is_set(pcmk_get_ra_caps(*standard), pcmk_ra_cap_provider)) { colon = strchr(spec, ':'); if ((colon == NULL) || (colon == spec)) { free(*standard); return -EINVAL; } *provider = strndup(spec, colon - spec); spec = colon + 1; } if (*spec == '\0') { free(*standard); free(*provider); return -EINVAL; } *type = strdup(spec); return pcmk_ok; } /*! * \brief Check whether a given stonith parameter is handled by Pacemaker * * Return true if a given string is the name of one of the special resource * instance attributes interpreted directly by Pacemaker for stonith-class * resources. * * \param[in] param Parameter name to check * * \return true if \p param is a special fencing parameter */ bool pcmk_stonith_param(const char *param) { if (param == NULL) { return false; } if (pcmk__str_any_of(param, PCMK_STONITH_PROVIDES, PCMK_STONITH_STONITH_TIMEOUT, NULL)) { return true; } - if (!pcmk__starts_with(param, "pcmk_")) { // Short-circuit common case + if (!g_str_has_prefix(param, "pcmk_")) { // Short-circuit common case return false; } if (pcmk__str_any_of(param, PCMK_STONITH_ACTION_LIMIT, PCMK_STONITH_DELAY_BASE, PCMK_STONITH_DELAY_MAX, PCMK_STONITH_HOST_ARGUMENT, PCMK_STONITH_HOST_CHECK, PCMK_STONITH_HOST_LIST, PCMK_STONITH_HOST_MAP, NULL)) { return true; } param = strchr(param + 5, '_'); // Skip past "pcmk_ACTION" return pcmk__str_any_of(param, "_action", "_timeout", "_retries", NULL); } diff --git a/lib/common/fuzzers/strings_fuzzer.c b/lib/common/fuzzers/strings_fuzzer.c index f8dfc8e39f..2c6d36574f 100644 --- a/lib/common/fuzzers/strings_fuzzer.c +++ b/lib/common/fuzzers/strings_fuzzer.c @@ -1,40 +1,40 @@ /* - * Copyright 2024 the Pacemaker project contributors + * Copyright 2024-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ +#include + #include #include #include #include #include #include #include -#include int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) { char *ns = NULL; guint res = 0U; if (size < 10) { return -1; // Do not add input to testing corpus } ns = pcmk__assert_alloc(1, size + 1); memcpy(ns, data, size); ns[size] = '\0'; pcmk__numeric_strcasecmp(ns, ns); - pcmk__trim(ns); pcmk_parse_interval_spec(ns, &res); crm_get_msec(ns); free(ns); return 0; } diff --git a/lib/common/iso8601.c b/lib/common/iso8601.c index 3ecad9a8e5..2a9e3f60d2 100644 --- a/lib/common/iso8601.c +++ b/lib/common/iso8601.c @@ -1,2353 +1,2356 @@ /* * Copyright 2005-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ /* * References: * https://en.wikipedia.org/wiki/ISO_8601 * http://www.staff.science.uu.nl/~gent0113/calendar/isocalendar.htm */ #include #include #include #include #include #include // INT_MIN, INT_MAX #include #include + +#include // g_strchomp() + #include #include #include "crmcommon_private.h" /* * Andrew's code was originally written for OSes whose "struct tm" contains: * long tm_gmtoff; :: Seconds east of UTC * const char *tm_zone; :: Timezone abbreviation * Some OSes lack these, instead having: * time_t (or long) timezone; :: "difference between UTC and local standard time" * char *tzname[2] = { "...", "..." }; * I (David Lee) confess to not understanding the details. So my attempted * generalisations for where their use is necessary may be flawed. * * 1. Does "difference between ..." subtract the same or opposite way? * 2. Should it use "altzone" instead of "timezone"? * 3. Should it use tzname[0] or tzname[1]? Interaction with timezone/altzone? */ #if defined(HAVE_STRUCT_TM_TM_GMTOFF) # define GMTOFF(tm) ((tm)->tm_gmtoff) #else /* Note: extern variable; macro argument not actually used. */ # define GMTOFF(tm) (-timezone+daylight) #endif #define HOUR_SECONDS (60 * 60) #define DAY_SECONDS (HOUR_SECONDS * 24) /*! * \internal * \brief Validate a seconds/microseconds tuple * * The microseconds value must be in the correct range, and if both are nonzero * they must have the same sign. * * \param[in] sec Seconds * \param[in] usec Microseconds * * \return true if the seconds/microseconds tuple is valid, or false otherwise */ #define valid_sec_usec(sec, usec) \ ((QB_ABS(usec) < QB_TIME_US_IN_SEC) \ && (((sec) == 0) || ((usec) == 0) || (((sec) < 0) == ((usec) < 0)))) // A date/time or duration struct crm_time_s { int years; // Calendar year (date/time) or number of years (duration) int months; // Number of months (duration only) int days; // Ordinal day of year (date/time) or number of days (duration) int seconds; // Seconds of day (date/time) or number of seconds (duration) int offset; // Seconds offset from UTC (date/time only) bool duration; // True if duration }; static crm_time_t *parse_date(const char *date_str); static crm_time_t * crm_get_utc_time(const crm_time_t *dt) { crm_time_t *utc = NULL; if (dt == NULL) { errno = EINVAL; return NULL; } utc = crm_time_new_undefined(); utc->years = dt->years; utc->days = dt->days; utc->seconds = dt->seconds; utc->offset = 0; if (dt->offset) { crm_time_add_seconds(utc, -dt->offset); } else { /* Durations (which are the only things that can include months, never have a timezone */ utc->months = dt->months; } crm_time_log(LOG_TRACE, "utc-source", dt, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_TRACE, "utc-target", utc, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); return utc; } crm_time_t * crm_time_new(const char *date_time) { tzset(); if (date_time == NULL) { return pcmk__copy_timet(time(NULL)); } return parse_date(date_time); } /*! * \brief Allocate memory for an uninitialized time object * * \return Newly allocated time object * \note The caller is responsible for freeing the return value using * crm_time_free(). */ crm_time_t * crm_time_new_undefined(void) { return (crm_time_t *) pcmk__assert_alloc(1, sizeof(crm_time_t)); } /*! * \brief Check whether a time object has been initialized yet * * \param[in] t Time object to check * * \return TRUE if time object has been initialized, FALSE otherwise */ bool crm_time_is_defined(const crm_time_t *t) { // Any nonzero member indicates something has been done to t return (t != NULL) && (t->years || t->months || t->days || t->seconds || t->offset || t->duration); } void crm_time_free(crm_time_t * dt) { if (dt == NULL) { return; } free(dt); } static int year_days(int year) { int d = 365; if (crm_time_leapyear(year)) { d++; } return d; } /* From http://myweb.ecu.edu/mccartyr/ISOwdALG.txt : * * 5. Find the Jan1Weekday for Y (Monday=1, Sunday=7) * YY = (Y-1) % 100 * C = (Y-1) - YY * G = YY + YY/4 * Jan1Weekday = 1 + (((((C / 100) % 4) x 5) + G) % 7) */ int crm_time_january1_weekday(int year) { int YY = (year - 1) % 100; int C = (year - 1) - YY; int G = YY + YY / 4; int jan1 = 1 + (((((C / 100) % 4) * 5) + G) % 7); crm_trace("YY=%d, C=%d, G=%d", YY, C, G); crm_trace("January 1 %.4d: %d", year, jan1); return jan1; } int crm_time_weeks_in_year(int year) { int weeks = 52; int jan1 = crm_time_january1_weekday(year); /* if jan1 == thursday */ if (jan1 == 4) { weeks++; } else { jan1 = crm_time_january1_weekday(year + 1); /* if dec31 == thursday aka. jan1 of next year is a friday */ if (jan1 == 5) { weeks++; } } return weeks; } // Jan-Dec plus Feb of leap years static int month_days[13] = { 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 29 }; /*! * \brief Return number of days in given month of given year * * \param[in] month Ordinal month (1-12) * \param[in] year Gregorian year * * \return Number of days in given month (0 if given month or year is invalid) */ int crm_time_days_in_month(int month, int year) { if ((month < 1) || (month > 12) || (year < 1)) { return 0; } if ((month == 2) && crm_time_leapyear(year)) { month = 13; } return month_days[month - 1]; } bool crm_time_leapyear(int year) { gboolean is_leap = FALSE; if (year % 4 == 0) { is_leap = TRUE; } if (year % 100 == 0 && year % 400 != 0) { is_leap = FALSE; } return is_leap; } /*! * \internal * \brief Get ordinal day number of year corresponding to given date * * \param[in] y Year * \param[in] m Month (1-12) * \param[in] d Day of month (1-31) * * \return Day number of year \p y corresponding to month \p m and day \p d, * or 0 for invalid arguments */ static int get_ordinal_days(uint32_t y, uint32_t m, uint32_t d) { int result = 0; CRM_CHECK((y > 0) && (y <= INT_MAX) && (m >= 1) && (m <= 12) && (d >= 1) && (d <= 31), return 0); result = d; for (int lpc = 1; lpc < m; lpc++) { result += crm_time_days_in_month(lpc, y); } return result; } void crm_time_log_alias(int log_level, const char *file, const char *function, int line, const char *prefix, const crm_time_t *date_time, int flags) { char *date_s = crm_time_as_string(date_time, flags); if (log_level == LOG_STDOUT) { printf("%s%s%s\n", (prefix? prefix : ""), (prefix? ": " : ""), date_s); } else { do_crm_log_alias(log_level, file, function, line, "%s%s%s", (prefix? prefix : ""), (prefix? ": " : ""), date_s); } free(date_s); } static void crm_time_get_sec(int sec, uint32_t *h, uint32_t *m, uint32_t *s) { uint32_t hours, minutes, seconds; seconds = QB_ABS(sec); hours = seconds / HOUR_SECONDS; seconds -= HOUR_SECONDS * hours; minutes = seconds / 60; seconds -= 60 * minutes; crm_trace("%d == %.2" PRIu32 ":%.2" PRIu32 ":%.2" PRIu32, sec, hours, minutes, seconds); *h = hours; *m = minutes; *s = seconds; } int crm_time_get_timeofday(const crm_time_t *dt, uint32_t *h, uint32_t *m, uint32_t *s) { crm_time_get_sec(dt->seconds, h, m, s); return TRUE; } int crm_time_get_timezone(const crm_time_t *dt, uint32_t *h, uint32_t *m) { uint32_t s; crm_time_get_sec(dt->seconds, h, m, &s); return TRUE; } long long crm_time_get_seconds(const crm_time_t *dt) { int lpc; crm_time_t *utc = NULL; long long in_seconds = 0; if (dt == NULL) { return 0; } // @TODO This is inefficient if dt is already in UTC utc = crm_get_utc_time(dt); if (utc == NULL) { return 0; } // @TODO We should probably use <= if dt is a duration for (lpc = 1; lpc < utc->years; lpc++) { long long dmax = year_days(lpc); in_seconds += DAY_SECONDS * dmax; } /* utc->months can be set only for durations. By definition, the value * varies depending on the (unknown) start date to which the duration will * be applied. Assume 30-day months so that something vaguely sane happens * in this case. */ if (utc->months > 0) { in_seconds += DAY_SECONDS * 30 * (long long) (utc->months); } if (utc->days > 0) { in_seconds += DAY_SECONDS * (long long) (utc->days - 1); } in_seconds += utc->seconds; crm_time_free(utc); return in_seconds; } #define EPOCH_SECONDS 62135596800ULL /* Calculated using crm_time_get_seconds() */ long long crm_time_get_seconds_since_epoch(const crm_time_t *dt) { return (dt == NULL)? 0 : (crm_time_get_seconds(dt) - EPOCH_SECONDS); } int crm_time_get_gregorian(const crm_time_t *dt, uint32_t *y, uint32_t *m, uint32_t *d) { int months = 0; int days = dt->days; if(dt->years != 0) { for (months = 1; months <= 12 && days > 0; months++) { int mdays = crm_time_days_in_month(months, dt->years); if (mdays >= days) { break; } else { days -= mdays; } } } else if (dt->months) { /* This is a duration including months, don't convert the days field */ months = dt->months; } else { /* This is a duration not including months, still don't convert the days field */ } *y = dt->years; *m = months; *d = days; crm_trace("%.4d-%.3d -> %.4d-%.2d-%.2d", dt->years, dt->days, dt->years, months, days); return TRUE; } int crm_time_get_ordinal(const crm_time_t *dt, uint32_t *y, uint32_t *d) { *y = dt->years; *d = dt->days; return TRUE; } int crm_time_get_isoweek(const crm_time_t *dt, uint32_t *y, uint32_t *w, uint32_t *d) { /* * Monday 29 December 2008 is written "2009-W01-1" * Sunday 3 January 2010 is written "2009-W53-7" */ int year_num = 0; int jan1 = crm_time_january1_weekday(dt->years); int h = -1; CRM_CHECK(dt->days > 0, return FALSE); /* 6. Find the Weekday for Y M D */ h = dt->days + jan1 - 1; *d = 1 + ((h - 1) % 7); /* 7. Find if Y M D falls in YearNumber Y-1, WeekNumber 52 or 53 */ if (dt->days <= (8 - jan1) && jan1 > 4) { crm_trace("year--, jan1=%d", jan1); year_num = dt->years - 1; *w = crm_time_weeks_in_year(year_num); } else { year_num = dt->years; } /* 8. Find if Y M D falls in YearNumber Y+1, WeekNumber 1 */ if (year_num == dt->years) { int dmax = year_days(year_num); int correction = 4 - *d; if ((dmax - dt->days) < correction) { crm_trace("year++, jan1=%d, i=%d vs. %d", jan1, dmax - dt->days, correction); year_num = dt->years + 1; *w = 1; } } /* 9. Find if Y M D falls in YearNumber Y, WeekNumber 1 through 53 */ if (year_num == dt->years) { int j = dt->days + (7 - *d) + (jan1 - 1); *w = j / 7; if (jan1 > 4) { *w -= 1; } } *y = year_num; crm_trace("Converted %.4d-%.3d to %.4" PRIu32 "-W%.2" PRIu32 "-%" PRIu32, dt->years, dt->days, *y, *w, *d); return TRUE; } #define DATE_MAX 128 /*! * \internal * \brief Print "." to a buffer * * \param[in] sec Seconds * \param[in] usec Microseconds (must be of same sign as \p sec and of * absolute value less than \c QB_TIME_US_IN_SEC) * \param[in,out] buf Result buffer */ static inline void sec_usec_as_string(long long sec, int usec, GString *buf) { /* A negative value smaller than -1 second should have the negative sign * before the 0, not before the usec part */ if ((sec == 0) && (usec < 0)) { g_string_append_c(buf, '-'); } g_string_append_printf(buf, "%lld.%06d", sec, QB_ABS(usec)); } /*! * \internal * \brief Get a string representation of a duration * * \param[in] dt Time object to interpret as a duration * \param[in] usec Microseconds to add to \p dt * \param[in] show_usec Whether to include microseconds in \p buf * \param[in,out] buf Result buffer */ static void duration_as_string(const crm_time_t *dt, int usec, bool show_usec, GString *buf) { pcmk__assert(valid_sec_usec(dt->seconds, usec)); if (dt->years) { g_string_append_printf(buf, "%4d year%s ", dt->years, pcmk__plural_s(dt->years)); } if (dt->months) { g_string_append_printf(buf, "%2d month%s ", dt->months, pcmk__plural_s(dt->months)); } if (dt->days) { g_string_append_printf(buf, "%2d day%s ", dt->days, pcmk__plural_s(dt->days)); } // At least print seconds (and optionally usecs) if ((buf->len == 0) || (dt->seconds != 0) || (show_usec && (usec != 0))) { if (show_usec) { sec_usec_as_string(dt->seconds, usec, buf); } else { g_string_append_printf(buf, "%d", dt->seconds); } g_string_append_printf(buf, " second%s", pcmk__plural_s(dt->seconds)); } // More than one minute, so provide a more readable breakdown into units if (QB_ABS(dt->seconds) >= 60) { uint32_t h = 0; uint32_t m = 0; uint32_t s = 0; uint32_t u = QB_ABS(usec); bool print_sec_component = false; crm_time_get_sec(dt->seconds, &h, &m, &s); print_sec_component = ((s != 0) || (show_usec && (u != 0))); g_string_append(buf, " ("); if (h) { g_string_append_printf(buf, "%" PRIu32 " hour%s", h, pcmk__plural_s(h)); if ((m != 0) || print_sec_component) { g_string_append_c(buf, ' '); } } if (m) { g_string_append_printf(buf, "%" PRIu32 " minute%s", m, pcmk__plural_s(m)); if (print_sec_component) { g_string_append_c(buf, ' '); } } if (print_sec_component) { if (show_usec) { sec_usec_as_string(s, u, buf); } else { g_string_append_printf(buf, "%" PRIu32, s); } g_string_append_printf(buf, " second%s", pcmk__plural_s(dt->seconds)); } g_string_append_c(buf, ')'); } } /*! * \internal * \brief Get a string representation of a time object * * \param[in] dt Time to convert to string * \param[in] usec Microseconds to add to \p dt * \param[in] flags Group of \c crm_time_* string format options * * \return Newly allocated string representation of \p dt plus \p usec * * \note The caller is responsible for freeing the return value using \c free(). */ static char * time_as_string_common(const crm_time_t *dt, int usec, uint32_t flags) { crm_time_t *utc = NULL; GString *buf = NULL; char *result = NULL; if (!crm_time_is_defined(dt)) { return pcmk__str_copy(""); } pcmk__assert(valid_sec_usec(dt->seconds, usec)); buf = g_string_sized_new(128); /* Simple cases: as duration, seconds, or seconds since epoch. * These never depend on time zone. */ if (pcmk_is_set(flags, crm_time_log_duration)) { duration_as_string(dt, usec, pcmk_is_set(flags, crm_time_usecs), buf); goto done; } if (pcmk_any_flags_set(flags, crm_time_seconds|crm_time_epoch)) { long long seconds = 0; if (pcmk_is_set(flags, crm_time_seconds)) { seconds = crm_time_get_seconds(dt); } else { seconds = crm_time_get_seconds_since_epoch(dt); } if (pcmk_is_set(flags, crm_time_usecs)) { sec_usec_as_string(seconds, usec, buf); } else { g_string_append_printf(buf, "%lld", seconds); } goto done; } // Convert to UTC if local timezone was not requested if ((dt->offset != 0) && !pcmk_is_set(flags, crm_time_log_with_timezone)) { crm_trace("UTC conversion"); utc = crm_get_utc_time(dt); dt = utc; } // As readable string if (pcmk_is_set(flags, crm_time_log_date)) { if (pcmk_is_set(flags, crm_time_weeks)) { // YYYY-WW-D uint32_t y = 0; uint32_t w = 0; uint32_t d = 0; if (crm_time_get_isoweek(dt, &y, &w, &d)) { g_string_append_printf(buf, "%" PRIu32 "-W%.2" PRIu32 "-%" PRIu32, y, w, d); } } else if (pcmk_is_set(flags, crm_time_ordinal)) { // YYYY-DDD uint32_t y = 0; uint32_t d = 0; if (crm_time_get_ordinal(dt, &y, &d)) { g_string_append_printf(buf, "%" PRIu32 "-%.3" PRIu32, y, d); } } else { // YYYY-MM-DD uint32_t y = 0; uint32_t m = 0; uint32_t d = 0; if (crm_time_get_gregorian(dt, &y, &m, &d)) { g_string_append_printf(buf, "%.4" PRIu32 "-%.2" PRIu32 "-%.2" PRIu32, y, m, d); } } } if (pcmk_is_set(flags, crm_time_log_timeofday)) { uint32_t h = 0, m = 0, s = 0; if (buf->len > 0) { g_string_append_c(buf, ' '); } if (crm_time_get_timeofday(dt, &h, &m, &s)) { g_string_append_printf(buf, "%.2" PRIu32 ":%.2" PRIu32 ":%.2" PRIu32, h, m, s); if (pcmk_is_set(flags, crm_time_usecs)) { g_string_append_printf(buf, ".%06" PRIu32, QB_ABS(usec)); } } if (pcmk_is_set(flags, crm_time_log_with_timezone) && (dt->offset != 0)) { crm_time_get_sec(dt->offset, &h, &m, &s); g_string_append_printf(buf, " %c%.2" PRIu32 ":%.2" PRIu32, ((dt->offset < 0)? '-' : '+'), h, m); } else { g_string_append_c(buf, 'Z'); } } done: crm_time_free(utc); result = pcmk__str_copy(buf->str); g_string_free(buf, TRUE); return result; } /*! * \brief Get a string representation of a \p crm_time_t object * * \param[in] dt Time to convert to string * \param[in] flags Group of \p crm_time_* string format options * * \note The caller is responsible for freeing the return value using \p free(). */ char * crm_time_as_string(const crm_time_t *dt, int flags) { return time_as_string_common(dt, 0, flags); } /*! * \internal * \brief Determine number of seconds from an hour:minute:second string * * \param[in] time_str Time specification string * \param[out] result Number of seconds equivalent to time_str * * \return TRUE if specification was valid, FALSE (and set errno) otherwise * \note This may return the number of seconds in a day (which is out of bounds * for a time object) if given 24:00:00. */ static bool crm_time_parse_sec(const char *time_str, int *result) { int rc; uint32_t hour = 0; uint32_t minute = 0; uint32_t second = 0; *result = 0; // Must have at least hour, but minutes and seconds are optional rc = sscanf(time_str, "%" SCNu32 ":%" SCNu32 ":%" SCNu32, &hour, &minute, &second); if (rc == 1) { rc = sscanf(time_str, "%2" SCNu32 "%2" SCNu32 "%2" SCNu32, &hour, &minute, &second); } if (rc == 0) { crm_err("%s is not a valid ISO 8601 time specification", time_str); errno = EINVAL; return FALSE; } crm_trace("Got valid time: %.2" PRIu32 ":%.2" PRIu32 ":%.2" PRIu32, hour, minute, second); if ((hour == 24) && (minute == 0) && (second == 0)) { // Equivalent to 00:00:00 of next day, return number of seconds in day } else if (hour >= 24) { crm_err("%s is not a valid ISO 8601 time specification " "because %" PRIu32 " is not a valid hour", time_str, hour); errno = EINVAL; return FALSE; } if (minute >= 60) { crm_err("%s is not a valid ISO 8601 time specification " "because %" PRIu32 " is not a valid minute", time_str, minute); errno = EINVAL; return FALSE; } if (second >= 60) { crm_err("%s is not a valid ISO 8601 time specification " "because %" PRIu32 " is not a valid second", time_str, second); errno = EINVAL; return FALSE; } *result = (hour * HOUR_SECONDS) + (minute * 60) + second; return TRUE; } static bool crm_time_parse_offset(const char *offset_str, int *offset) { tzset(); if (offset_str == NULL) { // Use local offset #if defined(HAVE_STRUCT_TM_TM_GMTOFF) time_t now = time(NULL); struct tm *now_tm = localtime(&now); #endif int h_offset = GMTOFF(now_tm) / HOUR_SECONDS; int m_offset = (GMTOFF(now_tm) - (HOUR_SECONDS * h_offset)) / 60; if (h_offset < 0 && m_offset < 0) { m_offset = 0 - m_offset; } *offset = (HOUR_SECONDS * h_offset) + (60 * m_offset); return TRUE; } if (offset_str[0] == 'Z') { // @TODO invalid if anything after? *offset = 0; return TRUE; } *offset = 0; if ((offset_str[0] == '+') || (offset_str[0] == '-') || isdigit((int)offset_str[0])) { gboolean negate = FALSE; if (offset_str[0] == '+') { offset_str++; } else if (offset_str[0] == '-') { negate = TRUE; offset_str++; } if (crm_time_parse_sec(offset_str, offset) == FALSE) { return FALSE; } if (negate) { *offset = 0 - *offset; } } // @TODO else invalid? return TRUE; } /*! * \internal * \brief Parse the time portion of an ISO 8601 date/time string * * \param[in] time_str Time portion of specification (after any 'T') * \param[in,out] a_time Time object to parse into * * \return TRUE if valid time was parsed, FALSE (and set errno) otherwise * \note This may add a day to a_time (if the time is 24:00:00). */ static bool crm_time_parse(const char *time_str, crm_time_t *a_time) { uint32_t h, m, s; char *offset_s = NULL; tzset(); if (time_str) { if (crm_time_parse_sec(time_str, &(a_time->seconds)) == FALSE) { return FALSE; } offset_s = strstr(time_str, "Z"); if (offset_s == NULL) { offset_s = strstr(time_str, " "); if (offset_s) { while (isspace(offset_s[0])) { offset_s++; } } } } if (crm_time_parse_offset(offset_s, &(a_time->offset)) == FALSE) { return FALSE; } crm_time_get_sec(a_time->offset, &h, &m, &s); crm_trace("Got tz: %c%2." PRIu32 ":%.2" PRIu32, (a_time->offset < 0)? '-' : '+', h, m); if (a_time->seconds == DAY_SECONDS) { // 24:00:00 == 00:00:00 of next day a_time->seconds = 0; crm_time_add_days(a_time, 1); } return TRUE; } /* * \internal * \brief Parse a time object from an ISO 8601 date/time specification * * \param[in] date_str ISO 8601 date/time specification (or * \c PCMK__VALUE_EPOCH) * * \return New time object on success, NULL (and set errno) otherwise */ static crm_time_t * parse_date(const char *date_str) { const char *time_s = NULL; crm_time_t *dt = NULL; uint32_t year = 0U; uint32_t month = 0U; uint32_t day = 0U; uint32_t week = 0U; int rc = 0; if (pcmk__str_empty(date_str)) { crm_err("No ISO 8601 date/time specification given"); goto invalid; } if ((date_str[0] == 'T') || ((strlen(date_str) > 2) && (date_str[2] == ':'))) { /* Just a time supplied - Infer current date */ dt = crm_time_new(NULL); if (date_str[0] == 'T') { time_s = date_str + 1; } else { time_s = date_str; } goto parse_time; } dt = crm_time_new_undefined(); if ((strncasecmp(PCMK__VALUE_EPOCH, date_str, 5) == 0) && ((date_str[5] == '\0') || (date_str[5] == '/') || isspace(date_str[5]))) { dt->days = 1; dt->years = 1970; crm_time_log(LOG_TRACE, "Unpacked", dt, crm_time_log_date | crm_time_log_timeofday); return dt; } /* YYYY-MM-DD */ rc = sscanf(date_str, "%" SCNu32 "-%" SCNu32 "-%" SCNu32 "", &year, &month, &day); if (rc == 1) { /* YYYYMMDD */ rc = sscanf(date_str, "%4" SCNu32 "%2" SCNu32 "%2" SCNu32 "", &year, &month, &day); } if (rc == 3) { if ((month < 1U) || (month > 12U)) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid month", date_str, month); goto invalid; } else if ((year < 1U) || (year > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid year", date_str, year); goto invalid; } else if ((day < 1) || (day > INT_MAX) || (day > crm_time_days_in_month(month, year))) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid day of the month", date_str, day); goto invalid; } else { dt->years = year; dt->days = get_ordinal_days(year, month, day); crm_trace("Parsed Gregorian date '%.4" PRIu32 "-%.3d' " "from date string '%s'", year, dt->days, date_str); } goto parse_time; } /* YYYY-DDD */ rc = sscanf(date_str, "%" SCNu32 "-%" SCNu32, &year, &day); if (rc == 2) { if ((year < 1U) || (year > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid year", date_str, year); goto invalid; } else if ((day < 1U) || (day > INT_MAX) || (day > year_days(year))) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid day of year %" PRIu32 " (1-%d)", date_str, day, year, year_days(year)); goto invalid; } crm_trace("Parsed ordinal year %d and days %d from date string '%s'", year, day, date_str); dt->days = day; dt->years = year; goto parse_time; } /* YYYY-Www-D */ rc = sscanf(date_str, "%" SCNu32 "-W%" SCNu32 "-%" SCNu32, &year, &week, &day); if (rc == 3) { if ((week < 1U) || (week > crm_time_weeks_in_year(year))) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid week of year %" PRIu32 " (1-%d)", date_str, week, year, crm_time_weeks_in_year(year)); goto invalid; } else if ((day < 1U) || (day > 7U)) { crm_err("'%s' is not a valid ISO 8601 date/time specification " "because '%" PRIu32 "' is not a valid day of the week", date_str, day); goto invalid; } else { /* * See https://en.wikipedia.org/wiki/ISO_week_date * * Monday 29 December 2008 is written "2009-W01-1" * Sunday 3 January 2010 is written "2009-W53-7" * Saturday 27 September 2008 is written "2008-W37-6" * * If 1 January is on a Monday, Tuesday, Wednesday or Thursday, it * is in week 1. If 1 January is on a Friday, Saturday or Sunday, * it is in week 52 or 53 of the previous year. */ int jan1 = crm_time_january1_weekday(year); crm_trace("Parsed year %" PRIu32 " (Jan 1 = %d), week %" PRIu32 ", and day %" PRIu32 " from date string '%s'", year, jan1, week, day, date_str); dt->years = year; crm_time_add_days(dt, (week - 1) * 7); if (jan1 <= 4) { crm_time_add_days(dt, 1 - jan1); } else { crm_time_add_days(dt, 8 - jan1); } crm_time_add_days(dt, day); } goto parse_time; } crm_err("'%s' is not a valid ISO 8601 date/time specification", date_str); goto invalid; parse_time: if (time_s == NULL) { time_s = date_str + strspn(date_str, "0123456789-W"); if ((time_s[0] == ' ') || (time_s[0] == 'T')) { ++time_s; } else { time_s = NULL; } } if ((time_s != NULL) && (crm_time_parse(time_s, dt) == FALSE)) { goto invalid; } crm_time_log(LOG_TRACE, "Unpacked", dt, crm_time_log_date | crm_time_log_timeofday); if (crm_time_check(dt) == FALSE) { crm_err("'%s' is not a valid ISO 8601 date/time specification", date_str); goto invalid; } return dt; invalid: crm_time_free(dt); errno = EINVAL; return NULL; } // Parse an ISO 8601 numeric value and return number of characters consumed static int parse_int(const char *str, int *result) { unsigned int lpc; int offset = (str[0] == 'T')? 1 : 0; bool negate = false; *result = 0; // @TODO This cannot handle combinations of these characters switch (str[offset]) { case '.': case ',': return 0; // Fractions are not supported case '-': negate = true; offset++; break; case '+': case ':': offset++; break; default: break; } for (lpc = 0; (lpc < 10) && isdigit(str[offset]); lpc++) { const int digit = str[offset++] - '0'; if ((*result * 10LL + digit) > INT_MAX) { return 0; // Overflow } *result = *result * 10 + digit; } if (negate) { *result = 0 - *result; } return (lpc > 0)? offset : 0; } /*! * \brief Parse a time duration from an ISO 8601 duration specification * * \param[in] period_s ISO 8601 duration specification (optionally followed by * whitespace, after which the rest of the string will be * ignored) * * \return New time object on success, NULL (and set errno) otherwise * \note It is the caller's responsibility to return the result using * crm_time_free(). */ crm_time_t * crm_time_parse_duration(const char *period_s) { gboolean is_time = FALSE; crm_time_t *diff = NULL; if (pcmk__str_empty(period_s)) { crm_err("No ISO 8601 time duration given"); goto invalid; } if (period_s[0] != 'P') { crm_err("'%s' is not a valid ISO 8601 time duration " "because it does not start with a 'P'", period_s); goto invalid; } if ((period_s[1] == '\0') || isspace(period_s[1])) { crm_err("'%s' is not a valid ISO 8601 time duration " "because nothing follows 'P'", period_s); goto invalid; } diff = crm_time_new_undefined(); for (const char *current = period_s + 1; current[0] && (current[0] != '/') && !isspace(current[0]); ++current) { int an_int = 0, rc; long long result = 0LL; if (current[0] == 'T') { /* A 'T' separates year/month/day from hour/minute/seconds. We don't * require it strictly, but just use it to differentiate month from * minutes. */ is_time = TRUE; continue; } // An integer must be next rc = parse_int(current, &an_int); if (rc == 0) { crm_err("'%s' is not a valid ISO 8601 time duration " "because no valid integer at '%s'", period_s, current); goto invalid; } current += rc; // A time unit must be next (we're not strict about the order) switch (current[0]) { case 'Y': diff->years = an_int; break; case 'M': if (!is_time) { // Months diff->months = an_int; } else { // Minutes result = diff->seconds + an_int * 60LL; if ((result < INT_MIN) || (result > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because integer at '%s' is too %s", period_s, current - rc, ((result > 0)? "large" : "small")); goto invalid; } else { diff->seconds = (int) result; } } break; case 'W': result = diff->days + an_int * 7LL; if ((result < INT_MIN) || (result > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because integer at '%s' is too %s", period_s, current - rc, ((result > 0)? "large" : "small")); goto invalid; } else { diff->days = (int) result; } break; case 'D': result = diff->days + (long long) an_int; if ((result < INT_MIN) || (result > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because integer at '%s' is too %s", period_s, current - rc, ((result > 0)? "large" : "small")); goto invalid; } else { diff->days = (int) result; } break; case 'H': result = diff->seconds + (long long) an_int * HOUR_SECONDS; if ((result < INT_MIN) || (result > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because integer at '%s' is too %s", period_s, current - rc, ((result > 0)? "large" : "small")); goto invalid; } else { diff->seconds = (int) result; } break; case 'S': result = diff->seconds + (long long) an_int; if ((result < INT_MIN) || (result > INT_MAX)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because integer at '%s' is too %s", period_s, current - rc, ((result > 0)? "large" : "small")); goto invalid; } else { diff->seconds = (int) result; } break; case '\0': crm_err("'%s' is not a valid ISO 8601 time duration " "because no units after %d", period_s, an_int); goto invalid; default: crm_err("'%s' is not a valid ISO 8601 time duration " "because '%c' is not a valid time unit", period_s, current[0]); goto invalid; } } if (!crm_time_is_defined(diff)) { crm_err("'%s' is not a valid ISO 8601 time duration " "because no amounts and units given", period_s); goto invalid; } diff->duration = TRUE; return diff; invalid: crm_time_free(diff); errno = EINVAL; return NULL; } /*! * \brief Parse a time period from an ISO 8601 interval specification * * \param[in] period_str ISO 8601 interval specification (start/end, * start/duration, or duration/end) * * \return New time period object on success, NULL (and set errno) otherwise * \note The caller is responsible for freeing the result using * crm_time_free_period(). */ crm_time_period_t * crm_time_parse_period(const char *period_str) { const char *original = period_str; crm_time_period_t *period = NULL; if (pcmk__str_empty(period_str)) { crm_err("No ISO 8601 time period given"); goto invalid; } tzset(); period = pcmk__assert_alloc(1, sizeof(crm_time_period_t)); if (period_str[0] == 'P') { period->diff = crm_time_parse_duration(period_str); if (period->diff == NULL) { goto error; } } else { period->start = parse_date(period_str); if (period->start == NULL) { goto error; } } period_str = strstr(original, "/"); if (period_str) { ++period_str; if (period_str[0] == 'P') { if (period->diff != NULL) { crm_err("'%s' is not a valid ISO 8601 time period " "because it has two durations", original); goto invalid; } period->diff = crm_time_parse_duration(period_str); if (period->diff == NULL) { goto error; } } else { period->end = parse_date(period_str); if (period->end == NULL) { goto error; } } } else if (period->diff != NULL) { // Only duration given, assume start is now period->start = crm_time_new(NULL); } else { // Only start given crm_err("'%s' is not a valid ISO 8601 time period " "because it has no duration or ending time", original); goto invalid; } if (period->start == NULL) { period->start = crm_time_subtract(period->end, period->diff); } else if (period->end == NULL) { period->end = crm_time_add(period->start, period->diff); } if (crm_time_check(period->start) == FALSE) { crm_err("'%s' is not a valid ISO 8601 time period " "because the start is invalid", period_str); goto invalid; } if (crm_time_check(period->end) == FALSE) { crm_err("'%s' is not a valid ISO 8601 time period " "because the end is invalid", period_str); goto invalid; } return period; invalid: errno = EINVAL; error: crm_time_free_period(period); return NULL; } /*! * \brief Free a dynamically allocated time period object * * \param[in,out] period Time period to free */ void crm_time_free_period(crm_time_period_t *period) { if (period) { crm_time_free(period->start); crm_time_free(period->end); crm_time_free(period->diff); free(period); } } void crm_time_set(crm_time_t *target, const crm_time_t *source) { crm_trace("target=%p, source=%p", target, source); CRM_CHECK(target != NULL && source != NULL, return); target->years = source->years; target->days = source->days; target->months = source->months; /* Only for durations */ target->seconds = source->seconds; target->offset = source->offset; crm_time_log(LOG_TRACE, "source", source, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_TRACE, "target", target, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); } static void ha_set_tm_time(crm_time_t *target, const struct tm *source) { int h_offset = 0; int m_offset = 0; /* Ensure target is fully initialized */ target->years = 0; target->months = 0; target->days = 0; target->seconds = 0; target->offset = 0; target->duration = FALSE; if (source->tm_year > 0) { /* years since 1900 */ target->years = 1900; crm_time_add_years(target, source->tm_year); } if (source->tm_yday >= 0) { /* days since January 1 [0-365] */ target->days = 1 + source->tm_yday; } if (source->tm_hour >= 0) { target->seconds += HOUR_SECONDS * source->tm_hour; } if (source->tm_min >= 0) { target->seconds += 60 * source->tm_min; } if (source->tm_sec >= 0) { target->seconds += source->tm_sec; } /* tm_gmtoff == offset from UTC in seconds */ h_offset = GMTOFF(source) / HOUR_SECONDS; m_offset = (GMTOFF(source) - (HOUR_SECONDS * h_offset)) / 60; crm_trace("Time offset is %lds (%.2d:%.2d)", GMTOFF(source), h_offset, m_offset); target->offset += HOUR_SECONDS * h_offset; target->offset += 60 * m_offset; } void crm_time_set_timet(crm_time_t *target, const time_t *source) { ha_set_tm_time(target, localtime(source)); } /*! * \internal * \brief Set one time object to another if the other is earlier * * \param[in,out] target Time object to set * \param[in] source Time object to use if earlier */ void pcmk__set_time_if_earlier(crm_time_t *target, const crm_time_t *source) { if ((target != NULL) && (source != NULL) && (!crm_time_is_defined(target) || (crm_time_compare(source, target) < 0))) { crm_time_set(target, source); } } crm_time_t * pcmk_copy_time(const crm_time_t *source) { crm_time_t *target = crm_time_new_undefined(); crm_time_set(target, source); return target; } /*! * \internal * \brief Convert a \p time_t time to a \p crm_time_t time * * \param[in] source Time to convert * * \return A \p crm_time_t object representing \p source */ crm_time_t * pcmk__copy_timet(time_t source) { crm_time_t *target = crm_time_new_undefined(); crm_time_set_timet(target, &source); return target; } crm_time_t * crm_time_add(const crm_time_t *dt, const crm_time_t *value) { crm_time_t *utc = NULL; crm_time_t *answer = NULL; if ((dt == NULL) || (value == NULL)) { errno = EINVAL; return NULL; } answer = pcmk_copy_time(dt); utc = crm_get_utc_time(value); if (utc == NULL) { crm_time_free(answer); return NULL; } crm_time_add_years(answer, utc->years); crm_time_add_months(answer, utc->months); crm_time_add_days(answer, utc->days); crm_time_add_seconds(answer, utc->seconds); crm_time_free(utc); return answer; } /*! * \internal * \brief Return the XML attribute name corresponding to a time component * * \param[in] component Component to check * * \return XML attribute name corresponding to \p component, or NULL if * \p component is invalid */ const char * pcmk__time_component_attr(enum pcmk__time_component component) { switch (component) { case pcmk__time_years: return PCMK_XA_YEARS; case pcmk__time_months: return PCMK_XA_MONTHS; case pcmk__time_weeks: return PCMK_XA_WEEKS; case pcmk__time_days: return PCMK_XA_DAYS; case pcmk__time_hours: return PCMK_XA_HOURS; case pcmk__time_minutes: return PCMK_XA_MINUTES; case pcmk__time_seconds: return PCMK_XA_SECONDS; default: return NULL; } } typedef void (*component_fn_t)(crm_time_t *, int); /*! * \internal * \brief Get the addition function corresponding to a time component * \param[in] component Component to check * * \return Addition function corresponding to \p component, or NULL if * \p component is invalid */ static component_fn_t component_fn(enum pcmk__time_component component) { switch (component) { case pcmk__time_years: return crm_time_add_years; case pcmk__time_months: return crm_time_add_months; case pcmk__time_weeks: return crm_time_add_weeks; case pcmk__time_days: return crm_time_add_days; case pcmk__time_hours: return crm_time_add_hours; case pcmk__time_minutes: return crm_time_add_minutes; case pcmk__time_seconds: return crm_time_add_seconds; default: return NULL; } } /*! * \internal * \brief Add the value of an XML attribute to a time object * * \param[in,out] t Time object to add to * \param[in] component Component of \p t to add to * \param[in] xml XML with value to add * * \return Standard Pacemaker return code */ int pcmk__add_time_from_xml(crm_time_t *t, enum pcmk__time_component component, const xmlNode *xml) { long long value; const char *attr = pcmk__time_component_attr(component); component_fn_t add = component_fn(component); if ((t == NULL) || (attr == NULL) || (add == NULL)) { return EINVAL; } if (xml == NULL) { return pcmk_rc_ok; } if (pcmk__scan_ll(pcmk__xe_get(xml, attr), &value, 0LL) != pcmk_rc_ok) { return pcmk_rc_unpack_error; } if ((value < INT_MIN) || (value > INT_MAX)) { return ERANGE; } if (value != 0LL) { add(t, (int) value); } return pcmk_rc_ok; } crm_time_t * crm_time_calculate_duration(const crm_time_t *dt, const crm_time_t *value) { crm_time_t *utc = NULL; crm_time_t *answer = NULL; if ((dt == NULL) || (value == NULL)) { errno = EINVAL; return NULL; } utc = crm_get_utc_time(value); if (utc == NULL) { return NULL; } answer = crm_get_utc_time(dt); if (answer == NULL) { crm_time_free(utc); return NULL; } answer->duration = TRUE; crm_time_add_years(answer, -utc->years); if(utc->months != 0) { crm_time_add_months(answer, -utc->months); } crm_time_add_days(answer, -utc->days); crm_time_add_seconds(answer, -utc->seconds); crm_time_free(utc); return answer; } crm_time_t * crm_time_subtract(const crm_time_t *dt, const crm_time_t *value) { crm_time_t *utc = NULL; crm_time_t *answer = NULL; if ((dt == NULL) || (value == NULL)) { errno = EINVAL; return NULL; } utc = crm_get_utc_time(value); if (utc == NULL) { return NULL; } answer = pcmk_copy_time(dt); crm_time_add_years(answer, -utc->years); if(utc->months != 0) { crm_time_add_months(answer, -utc->months); } crm_time_add_days(answer, -utc->days); crm_time_add_seconds(answer, -utc->seconds); crm_time_free(utc); return answer; } /*! * \brief Check whether a time object represents a sensible date/time * * \param[in] dt Date/time object to check * * \return \c true if years, days, and seconds are sensible, \c false otherwise */ bool crm_time_check(const crm_time_t *dt) { return (dt != NULL) && (dt->days > 0) && (dt->days <= year_days(dt->years)) && (dt->seconds >= 0) && (dt->seconds < DAY_SECONDS); } #define do_cmp_field(l, r, field) \ if(rc == 0) { \ if(l->field > r->field) { \ crm_trace("%s: %d > %d", \ #field, l->field, r->field); \ rc = 1; \ } else if(l->field < r->field) { \ crm_trace("%s: %d < %d", \ #field, l->field, r->field); \ rc = -1; \ } \ } int crm_time_compare(const crm_time_t *a, const crm_time_t *b) { int rc = 0; crm_time_t *t1 = crm_get_utc_time(a); crm_time_t *t2 = crm_get_utc_time(b); if ((t1 == NULL) && (t2 == NULL)) { rc = 0; } else if (t1 == NULL) { rc = -1; } else if (t2 == NULL) { rc = 1; } else { do_cmp_field(t1, t2, years); do_cmp_field(t1, t2, days); do_cmp_field(t1, t2, seconds); } crm_time_free(t1); crm_time_free(t2); return rc; } /*! * \brief Add a given number of seconds to a date/time or duration * * \param[in,out] a_time Date/time or duration to add seconds to * \param[in] extra Number of seconds to add */ void crm_time_add_seconds(crm_time_t *a_time, int extra) { int days = extra / DAY_SECONDS; pcmk__assert(a_time != NULL); crm_trace("Adding %d seconds (including %d whole day%s) to %d", extra, days, pcmk__plural_s(days), a_time->seconds); a_time->seconds += extra % DAY_SECONDS; // Check whether the addition crossed a day boundary if (a_time->seconds > DAY_SECONDS) { ++days; a_time->seconds -= DAY_SECONDS; } else if (a_time->seconds < 0) { --days; a_time->seconds += DAY_SECONDS; } crm_time_add_days(a_time, days); } #define ydays(t) (crm_time_leapyear((t)->years)? 366 : 365) /*! * \brief Add days to a date/time * * \param[in,out] a_time Time to modify * \param[in] extra Number of days to add (may be negative to subtract) */ void crm_time_add_days(crm_time_t *a_time, int extra) { pcmk__assert(a_time != NULL); crm_trace("Adding %d days to %.4d-%.3d", extra, a_time->years, a_time->days); if (extra > 0) { while ((a_time->days + (long long) extra) > ydays(a_time)) { if ((a_time->years + 1LL) > INT_MAX) { a_time->days = ydays(a_time); // Clip to latest we can handle return; } extra -= ydays(a_time); a_time->years++; } } else if (extra < 0) { const int min_days = a_time->duration? 0 : 1; while ((a_time->days + (long long) extra) < min_days) { if ((a_time->years - 1) < 1) { a_time->days = 1; // Clip to earliest we can handle (no BCE) return; } a_time->years--; extra += ydays(a_time); } } a_time->days += extra; } void crm_time_add_months(crm_time_t * a_time, int extra) { int lpc; uint32_t y, m, d, dmax; crm_time_get_gregorian(a_time, &y, &m, &d); crm_trace("Adding %d months to %.4" PRIu32 "-%.2" PRIu32 "-%.2" PRIu32, extra, y, m, d); if (extra > 0) { for (lpc = extra; lpc > 0; lpc--) { m++; if (m == 13) { m = 1; y++; } } } else { for (lpc = -extra; lpc > 0; lpc--) { m--; if (m == 0) { m = 12; y--; } } } dmax = crm_time_days_in_month(m, y); if (dmax < d) { /* Preserve day-of-month unless the month doesn't have enough days */ d = dmax; } crm_trace("Calculated %.4" PRIu32 "-%.2" PRIu32 "-%.2" PRIu32, y, m, d); a_time->years = y; a_time->days = get_ordinal_days(y, m, d); crm_time_get_gregorian(a_time, &y, &m, &d); crm_trace("Got %.4" PRIu32 "-%.2" PRIu32 "-%.2" PRIu32, y, m, d); } void crm_time_add_minutes(crm_time_t * a_time, int extra) { crm_time_add_seconds(a_time, extra * 60); } void crm_time_add_hours(crm_time_t * a_time, int extra) { crm_time_add_seconds(a_time, extra * HOUR_SECONDS); } void crm_time_add_weeks(crm_time_t * a_time, int extra) { crm_time_add_days(a_time, extra * 7); } void crm_time_add_years(crm_time_t * a_time, int extra) { pcmk__assert(a_time != NULL); if ((extra > 0) && ((a_time->years + (long long) extra) > INT_MAX)) { a_time->years = INT_MAX; } else if ((extra < 0) && ((a_time->years + (long long) extra) < 1)) { a_time->years = 1; // Clip to earliest we can handle (no BCE) } else { a_time->years += extra; } } /* The high-resolution variant of time object was added to meet an immediate * need, and is kept internal API. * * @TODO The long-term goal is to come up with a clean, unified design for a * time type (or types) that meets all the various needs, to replace * crm_time_t, pcmk__time_hr_t, and struct timespec (in lrmd_cmd_t). */ static pcmk__time_hr_t * time_to_hr(const crm_time_t *dt) { pcmk__time_hr_t *hr_dt = NULL; pcmk__assert(dt != NULL); hr_dt = pcmk__assert_alloc(1, sizeof(pcmk__time_hr_t)); hr_dt->years = dt->years; hr_dt->months = dt->months; hr_dt->days = dt->days; hr_dt->seconds = dt->seconds; hr_dt->offset = dt->offset; hr_dt->duration = dt->duration; return hr_dt; } /*! * \internal * \brief Return the current time as a high-resolution time * * \param[out] epoch If not NULL, this will be set to seconds since epoch * * \return Newly allocated high-resolution time set to the current time */ pcmk__time_hr_t * pcmk__time_hr_now(time_t *epoch) { struct timespec tv; crm_time_t dt; pcmk__time_hr_t *hr; qb_util_timespec_from_epoch_get(&tv); if (epoch != NULL) { *epoch = tv.tv_sec; } crm_time_set_timet(&dt, &(tv.tv_sec)); hr = time_to_hr(&dt); hr->useconds = tv.tv_nsec / QB_TIME_NS_IN_USEC; return hr; } pcmk__time_hr_t * pcmk__time_hr_new(const char *date_time) { pcmk__time_hr_t *hr_dt = NULL; if (date_time == NULL) { hr_dt = pcmk__time_hr_now(NULL); } else { crm_time_t *dt = parse_date(date_time); hr_dt = time_to_hr(dt); crm_time_free(dt); } return hr_dt; } void pcmk__time_hr_free(pcmk__time_hr_t * hr_dt) { free(hr_dt); } static void ha_get_tm_time(struct tm *target, const pcmk__time_hr_t *source) { *target = (struct tm) { .tm_year = source->years - 1900, /* source->days is day of year, but we assign it to tm_mday instead of * tm_yday. mktime() fixes it. See the mktime(3) man page for details. */ .tm_mday = source->days, // mktime() converts this to hours/minutes/seconds appropriately .tm_sec = source->seconds, // Don't adjust DST here; let mktime() try to determine DST status .tm_isdst = -1, #if defined(HAVE_STRUCT_TM_TM_GMTOFF) .tm_gmtoff = source->offset #endif }; mktime(target); } /*! * \internal * \brief Convert a struct tm to a \c GDateTime * * \param[in] tm Time object to convert * \param[in] offset Offset from UTC (in seconds) * * \return Newly allocated \c GDateTime object corresponding to \p tm, or * \c NULL on error * * \note The caller is responsible for freeing the return value using * \c g_date_time_unref(). */ static GDateTime * get_g_date_time(const struct tm *tm, int offset) { // Accept an offset argument in case tm lacks a tm_gmtoff member char buf[sizeof("+hh:mm")] = { '\0', }; const char *offset_s = NULL; GTimeZone *tz = NULL; GDateTime *dt = NULL; if (QB_ABS(offset) <= DAY_SECONDS) { uint32_t hours = 0; uint32_t minutes = 0; uint32_t seconds = 0; int rc = 0; crm_time_get_sec(offset, &hours, &minutes, &seconds); rc = snprintf(buf, sizeof(buf), "%c%02" PRIu32 ":%02" PRIu32, ((offset >= 0)? '+' : '-'), hours, minutes); pcmk__assert(rc == (sizeof(buf) - 1)); offset_s = buf; } else { // offset out of range; use NULL as offset_s CRM_LOG_ASSERT(QB_ABS(offset) <= DAY_SECONDS); } /* @FIXME @COMPAT As of glib 2.68, g_time_zone_new() is deprecated in favor * of g_time_zone_new_identifier(). However, calling * g_time_zone_new_identifier() results in compiler warnings, even on a * system with glib 2.84 installed. It is unclear why. * * The *_new_identifier() function was added (and the *_new() function * deprecated) in version 2.68. They have the same signature. Ideally, we * would choose which function to call here and below based the installed * glib version using a CPP guard. */ tz = g_time_zone_new(offset_s); dt = g_date_time_new(tz, tm->tm_year + 1900, tm->tm_mon + 1, tm->tm_mday, tm->tm_hour, tm->tm_min, tm->tm_sec); g_time_zone_unref(tz); return dt; } /*! * \internal * \brief Expand a date/time format string, with support for fractional seconds * * \param[in] format Date/time format string compatible with * \c g_date_time_format(), with additional support for * \c "%N" for fractional seconds * \param[in] hr_dt Time value to format * * \return Newly allocated string with formatted string, or \c NULL on error * * \note This function falls back to trying \c strftime() with a fixed-size * buffer if \c g_date_time_format() fails. This fallback will be removed * in a future release. */ char * pcmk__time_format_hr(const char *format, const pcmk__time_hr_t *hr_dt) { int scanned_pos = 0; // How many characters of format have been parsed int printed_pos = 0; // How many characters of format have been processed GString *buf = NULL; char *result = NULL; struct tm tm = { 0, }; GDateTime *gdt = NULL; if (format == NULL) { return NULL; } buf = g_string_sized_new(128); ha_get_tm_time(&tm, hr_dt); gdt = get_g_date_time(&tm, hr_dt->offset); if (gdt == NULL) { goto done; } while (format[scanned_pos] != '\0') { int fmt_pos = 0; // Index after last character to pass as-is int frac_digits = 0; // %N specifier's width field value (if any) gchar *tmp_fmt_s = NULL; gchar *date_s = NULL; // Look for next format specifier const char *mark_s = strchr(&format[scanned_pos], '%'); if (mark_s == NULL) { // No more specifiers, so pass remaining string to strftime() as-is scanned_pos = strlen(format); fmt_pos = scanned_pos; } else { fmt_pos = mark_s - format; // Index of % // Skip % and any width field scanned_pos = fmt_pos + 1; while (isdigit(format[scanned_pos])) { scanned_pos++; } switch (format[scanned_pos]) { case '\0': // Literal % and possibly digits at end of string fmt_pos = scanned_pos; // Pass remaining string as-is break; case 'N': // %[width]N /* Fractional seconds. This was supposed to represent * nanoseconds. However, we only store times at microsecond * resolution, and the width field support makes this a * general fractional component specifier rather than a * nanoseconds specifier. * * Further, since we cap the width at 6 digits, a user * cannot display times at greater than microsecond * resolution. * * A leading zero in the width field is ignored, not treated * as "use zero-padding." For example, "%03N" and "%3N" * produce the same result. */ scanned_pos++; // Parse width field frac_digits = atoi(&format[fmt_pos + 1]); frac_digits = QB_MAX(frac_digits, 0); frac_digits = QB_MIN(frac_digits, 6); break; default: // Some other specifier if (format[++scanned_pos] != '\0') { // More to parse continue; } fmt_pos = scanned_pos; // Pass remaining string as-is break; } } tmp_fmt_s = g_strndup(&format[printed_pos], fmt_pos - printed_pos); date_s = g_date_time_format(gdt, tmp_fmt_s); if (date_s == NULL) { char compat_date_s[1024] = { '\0' }; size_t nbytes = 0; // @COMPAT Drop this fallback crm_warn("Could not format time using format string '%s' with " "g_date_time_format(); trying strftime(). In a future " "release, use of strftime() as a fallback will be removed", format); #ifdef HAVE_FORMAT_NONLITERAL #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wformat-nonliteral" #endif // HAVE_FORMAT_NONLITERAL nbytes = strftime(compat_date_s, sizeof(compat_date_s), tmp_fmt_s, &tm); #ifdef HAVE_FORMAT_NONLITERAL #pragma GCC diagnostic pop #endif // HAVE_FORMAT_NONLITERAL if (nbytes == 0) { // Truncation, empty string, or error; impossible to discern crm_err("Could not format time using format string '%s'", format); // Ensure we return NULL g_string_truncate(buf, 0); g_free(tmp_fmt_s); goto done; } date_s = g_strdup(compat_date_s); } g_string_append(buf, date_s); g_free(date_s); g_free(tmp_fmt_s); printed_pos = scanned_pos; if (frac_digits != 0) { // Descending powers of 10 (10^5 down to 10^0) static const int powers[6] = { 1e5, 1e4, 1e3, 1e2, 1e1, 1e0 }; // Sanity check to ensure array access is in bounds pcmk__assert((frac_digits > 0) && (frac_digits <= 6)); /* Append fractional seconds at the requested resolution, truncated * toward zero. We're basically converting from microseconds to * another unit here. For example, suppose the width field * (frac_digits) is 3. This means "use millisecond resolution." Then * we need to divide our microseconds value by 10^3, which is * powers[3 - 1]. * * If the width field is 6 (microsecond resolution), then we divide * our microseconds value by 10^0 == 1, which is powers[6 - 1]. */ g_string_append_printf(buf, "%0*d", frac_digits, hr_dt->useconds / powers[frac_digits - 1]); } } done: if (buf->len > 0) { result = pcmk__str_copy(buf->str); } g_string_free(buf, TRUE); if (gdt != NULL) { g_date_time_unref(gdt); } return result; } /*! * \internal * \brief Return a human-friendly string corresponding to an epoch time value * * \param[in] source Pointer to epoch time value (or \p NULL for current time) * \param[in] flags Group of \p crm_time_* flags controlling display format * (0 to use \p ctime() with newline removed) * * \return String representation of \p source on success (may be empty depending * on \p flags; guaranteed not to be \p NULL) * * \note The caller is responsible for freeing the return value using \p free(). */ char * pcmk__epoch2str(const time_t *source, uint32_t flags) { time_t epoch_time = (source == NULL)? time(NULL) : *source; if (flags == 0) { - return pcmk__str_copy(pcmk__trim(ctime(&epoch_time))); + return pcmk__str_copy(g_strchomp(ctime(&epoch_time))); } else { crm_time_t dt; crm_time_set_timet(&dt, &epoch_time); return crm_time_as_string(&dt, flags); } } /*! * \internal * \brief Return a human-friendly string corresponding to seconds-and- * nanoseconds value * * Time is shown with microsecond resolution if \p crm_time_usecs is in \p * flags. * * \param[in] ts Time in seconds and nanoseconds (or \p NULL for current * time) * \param[in] flags Group of \p crm_time_* flags controlling display format * * \return String representation of \p ts on success (may be empty depending on * \p flags; guaranteed not to be \p NULL) * * \note The caller is responsible for freeing the return value using \p free(). */ char * pcmk__timespec2str(const struct timespec *ts, uint32_t flags) { struct timespec tmp_ts; crm_time_t dt; if (ts == NULL) { qb_util_timespec_from_epoch_get(&tmp_ts); ts = &tmp_ts; } crm_time_set_timet(&dt, &ts->tv_sec); return time_as_string_common(&dt, ts->tv_nsec / QB_TIME_NS_IN_USEC, flags); } /*! * \internal * \brief Given a millisecond interval, return a log-friendly string * * \param[in] interval_ms Interval in milliseconds * * \return Readable version of \p interval_ms * * \note The return value is a pointer to static memory that may be overwritten * by later calls to this function. */ const char * pcmk__readable_interval(guint interval_ms) { #define MS_IN_S (1000) #define MS_IN_M (MS_IN_S * 60) #define MS_IN_H (MS_IN_M * 60) #define MS_IN_D (MS_IN_H * 24) #define MAXSTR sizeof("..d..h..m..s...ms") static char str[MAXSTR]; GString *buf = NULL; if (interval_ms == 0) { return "0s"; } buf = g_string_sized_new(128); if (interval_ms >= MS_IN_D) { g_string_append_printf(buf, "%ud", interval_ms / MS_IN_D); interval_ms -= (interval_ms / MS_IN_D) * MS_IN_D; } if (interval_ms >= MS_IN_H) { g_string_append_printf(buf, "%uh", interval_ms / MS_IN_H); interval_ms -= (interval_ms / MS_IN_H) * MS_IN_H; } if (interval_ms >= MS_IN_M) { g_string_append_printf(buf, "%um", interval_ms / MS_IN_M); interval_ms -= (interval_ms / MS_IN_M) * MS_IN_M; } // Ns, N.NNNs, or NNNms if (interval_ms >= MS_IN_S) { g_string_append_printf(buf, "%u", interval_ms / MS_IN_S); interval_ms -= (interval_ms / MS_IN_S) * MS_IN_S; if (interval_ms > 0) { g_string_append_printf(buf, ".%03u", interval_ms); } g_string_append_c(buf, 's'); } else if (interval_ms > 0) { g_string_append_printf(buf, "%ums", interval_ms); } pcmk__assert(buf->len < sizeof(str)); strncpy(str, buf->str, sizeof(str) - 1); g_string_free(buf, TRUE); return str; } diff --git a/lib/common/procfs.c b/lib/common/procfs.c index e6e32e3d5f..35ffd5c6de 100644 --- a/lib/common/procfs.c +++ b/lib/common/procfs.c @@ -1,476 +1,478 @@ /* * Copyright 2015-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include +#include // g_str_has_prefix() + #if HAVE_LINUX_PROCFS /*! * \internal * \brief Return name of /proc file containing the CIB daemon's load statistics * * \return Newly allocated memory with file name on success, NULL otherwise * * \note It is the caller's responsibility to free the return value. * This will return NULL if the daemon is being run via valgrind. * This should be called only on Linux systems. */ static char * find_cib_loadfile(const char *server) { pid_t pid = pcmk__procfs_pid_of(server); return pid? crm_strdup_printf("/proc/%lld/stat", (long long) pid) : NULL; } /*! * \internal * \brief Get process ID and name associated with a /proc directory entry * * \param[in] entry Directory entry (must be result of readdir() on /proc) * \param[out] name If not NULL, a char[16] to hold the process name * \param[out] pid If not NULL, will be set to process ID of entry * * \return Standard Pacemaker return code * \note This should be called only on Linux systems, as not all systems that * support /proc store process names and IDs in the same way. The kernel * limits the process name to the first 15 characters (plus terminator). * It would be nice if there were a public kernel API constant for that * limit, but there isn't. */ static int pcmk__procfs_process_info(const struct dirent *entry, char *name, pid_t *pid) { int fd, local_pid; FILE *file; struct stat statbuf; char procpath[128] = { 0 }; /* We're only interested in entries whose name is a PID, * so skip anything non-numeric or that is too long. * * 114 = 128 - strlen("/proc/") - strlen("/status") - 1 */ local_pid = atoi(entry->d_name); if ((local_pid <= 0) || (strlen(entry->d_name) > 114)) { return -1; } if (pid) { *pid = (pid_t) local_pid; } /* Get this entry's file information */ strcpy(procpath, "/proc/"); strcat(procpath, entry->d_name); fd = open(procpath, O_RDONLY); if (fd < 0 ) { return -1; } if (fstat(fd, &statbuf) < 0) { close(fd); return -1; } close(fd); /* We're only interested in subdirectories */ if (!S_ISDIR(statbuf.st_mode)) { return -1; } /* Read the first entry ("Name:") from the process's status file. * We could handle the valgrind case if we parsed the cmdline file * instead, but that's more of a pain than it's worth. */ if (name != NULL) { strcat(procpath, "/status"); file = fopen(procpath, "r"); if (!file) { return -1; } if (fscanf(file, "Name:\t%15[^\n]", name) != 1) { fclose(file); return -1; } name[15] = 0; fclose(file); } return 0; } #endif // HAVE_LINUX_PROCFS /*! * \internal * \brief Return process ID of a named process * * \param[in] name Process name (as used in /proc/.../status) * * \return Process ID of named process if running, 0 otherwise * * \note This will return 0 if the process is being run via valgrind. * This should be called only on Linux systems. */ pid_t pcmk__procfs_pid_of(const char *name) { #if HAVE_LINUX_PROCFS DIR *dp; struct dirent *entry; pid_t pid = 0; char entry_name[64] = { 0 }; dp = opendir("/proc"); if (dp == NULL) { crm_notice("Can not read /proc directory to track existing components"); return 0; } while ((entry = readdir(dp)) != NULL) { if ((pcmk__procfs_process_info(entry, entry_name, &pid) == pcmk_rc_ok) && pcmk__str_eq(entry_name, name, pcmk__str_casei) && (pcmk__pid_active(pid, NULL) == pcmk_rc_ok)) { crm_info("Found %s active as process %lld", name, (long long) pid); break; } pid = 0; } closedir(dp); return pid; #else return 0; #endif // HAVE_LINUX_PROCFS } /*! * \internal * \brief Calculate number of logical CPU cores from procfs * * \return Number of cores (or 1 if unable to determine) */ unsigned int pcmk__procfs_num_cores(void) { #if HAVE_LINUX_PROCFS int cores = 0; FILE *stream = NULL; /* Parse /proc/stat instead of /proc/cpuinfo because it's smaller */ stream = fopen("/proc/stat", "r"); if (stream == NULL) { crm_perror(LOG_INFO, "Could not open /proc/stat"); } else { char buffer[2048]; while (fgets(buffer, sizeof(buffer), stream)) { - if (pcmk__starts_with(buffer, "cpu") && isdigit(buffer[3])) { + if (g_str_has_prefix(buffer, "cpu") && isdigit(buffer[3])) { ++cores; } } fclose(stream); } return cores? cores : 1; #else return 1; #endif // HAVE_LINUX_PROCFS } /*! * \internal * \brief Get the executable path corresponding to a process ID * * \param[in] pid Process ID to check * \param[out] path Where to store executable path (can be \c NULL) * * \return Standard Pacemaker error code (as possible errno values from * readlink()) */ int pcmk__procfs_pid2path(pid_t pid, char **path) { #if HAVE_LINUX_PROCFS char *procfs_path = NULL; ssize_t link_rc; /* The readlink(2) man page recommends calling lstat() to get the required * buffer size, and then dynamically allocate the buffer. However, * st_size == 0 for symlinks under /proc. So we use PATH_MAX. */ char real_path[PATH_MAX] = { '\0', }; pcmk__assert((path == NULL) || (*path == NULL)); procfs_path = crm_strdup_printf("/proc/%lld/exe", (long long) pid); link_rc = readlink(procfs_path, real_path, sizeof(real_path)); free(procfs_path); if (link_rc < 0) { return errno; } else if (link_rc >= sizeof(real_path)) { return ENAMETOOLONG; } if (path != NULL) { /* Make Coverity happy; we already zero-initialized real_path, and we * returned ENAMETOOLONG if it's no longer null-terminated */ real_path[link_rc] = '\0'; *path = pcmk__str_copy(real_path); } return pcmk_rc_ok; #else return EOPNOTSUPP; #endif // HAVE_LINUX_PROCFS } /*! * \internal * \brief Check whether process ID information is available from procfs * * \return true if process ID information is available, otherwise false */ bool pcmk__procfs_has_pids(void) { #if HAVE_LINUX_PROCFS static bool have_pids = false; static bool checked = false; if (!checked) { have_pids = pcmk__procfs_pid2path(getpid(), NULL) == pcmk_rc_ok; checked = true; } return have_pids; #else return false; #endif // HAVE_LINUX_PROCFS } /*! * \internal * \brief Return an open handle on the directory containing links to open file * descriptors, or NULL on error */ DIR * pcmk__procfs_fd_dir(void) { DIR *dir = NULL; /* /proc/self/fd (on Linux) or /dev/fd (on most OSes) contains symlinks to * all open files for the current process, named as the file descriptor. * Use this if available, because it's more efficient than a shotgun * approach to closing descriptors. */ #if HAVE_LINUX_PROCFS dir = opendir("/proc/self/fd"); #endif // HAVE_LINUX_PROCFS return dir; } /*! * \internal * \brief Trigger a sysrq command if supported on current platform * * \param[in] t Sysrq command to trigger */ void pcmk__sysrq_trigger(char t) { #if HAVE_LINUX_PROCFS // Root can always write here, regardless of kernel.sysrq value FILE *procf = fopen("/proc/sysrq-trigger", "a"); if (procf == NULL) { crm_warn("Could not open sysrq-trigger: %s", strerror(errno)); } else { fprintf(procf, "%c\n", t); fclose(procf); } #endif // HAVE_LINUX_PROCFS } bool pcmk__throttle_cib_load(const char *server, float *load) { /* /proc/[pid]/stat * * Status information about the process. This is used by ps(1). It is defined * in /usr/src/linux/fs/proc/array.c. * * The fields, in order, with their proper scanf(3) format specifiers, are: * * pid %d (1) The process ID. * comm %s (2) The filename of the executable, in parentheses. This is * visible whether or not the executable is swapped out. * state %c (3) One character from the string "RSDZTW" where R is running, * S is sleeping in an interruptible wait, D is waiting in * uninterruptible disk sleep, Z is zombie, T is traced or * stopped (on a signal), and W is paging. * ppid %d (4) The PID of the parent. * pgrp %d (5) The process group ID of the process. * session %d (6) The session ID of the process. * tty_nr %d (7) The controlling terminal of the process. (The minor device * number is contained in the combination of bits 31 to 20 and * 7 to 0; the major device number is in bits 15 to 8.) * tpgid %d (8) The ID of the foreground process group of the controlling * terminal of the process. * flags %u (9) The kernel flags word of the process. For bit meanings, see * the PF_* defines in the Linux kernel source file include/linux/sched.h. * Details depend on the kernel version. * minflt %lu (10) The number of minor faults the process has made which have * not required loading a memory page from disk. * cminflt %lu (11) The number of minor faults that the process's waited-for * children have made. * majflt %lu (12) The number of major faults the process has made which have * required loading a memory page from disk. * cmajflt %lu (13) The number of major faults that the process's waited-for * children have made. * utime %lu (14) Amount of time that this process has been scheduled in user * mode, measured in clock ticks (divide by sysconf(_SC_CLK_TCK)). * This includes guest time, guest_time (time spent running a * virtual CPU, see below), so that applications that are not * aware of the guest time field do not lose that time from * their calculations. * stime %lu (15) Amount of time that this process has been scheduled in * kernel mode, measured in clock ticks (divide by sysconf(_SC_CLK_TCK)). */ #if HAVE_LINUX_PROCFS static char *loadfile = NULL; static time_t last_call = 0; static long ticks_per_s = 0; static unsigned long last_utime, last_stime; char buffer[64*1024]; FILE *stream = NULL; time_t now = time(NULL); if (load == NULL) { return false; } else { *load = 0.0; } if (loadfile == NULL) { last_call = 0; last_utime = 0; last_stime = 0; loadfile = find_cib_loadfile(server); if (loadfile == NULL) { crm_warn("Couldn't find CIB load file"); return false; } ticks_per_s = sysconf(_SC_CLK_TCK); crm_trace("Found %s", loadfile); } stream = fopen(loadfile, "r"); if (stream == NULL) { int rc = errno; crm_warn("Couldn't read %s: %s (%d)", loadfile, pcmk_rc_str(rc), rc); free(loadfile); loadfile = NULL; return false; } if (fgets(buffer, sizeof(buffer), stream) != NULL) { char *comm = pcmk__assert_alloc(1, 256); char state = 0; int rc = 0, pid = 0, ppid = 0, pgrp = 0, session = 0, tty_nr = 0, tpgid = 0; unsigned long flags = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0, utime = 0, stime = 0; rc = sscanf(buffer, "%d %[^ ] %c %d %d %d %d %d %lu %lu %lu %lu %lu %lu %lu", &pid, comm, &state, &ppid, &pgrp, &session, &tty_nr, &tpgid, &flags, &minflt, &cminflt, &majflt, &cmajflt, &utime, &stime); free(comm); if (rc != 15) { crm_err("Only %d of 15 fields found in %s", rc, loadfile); fclose(stream); return false; } else if ((last_call > 0) && (last_call < now) && (last_utime <= utime) && (last_stime <= stime)) { time_t elapsed = now - last_call; unsigned long delta_utime = utime - last_utime; unsigned long delta_stime = stime - last_stime; *load = delta_utime + delta_stime; /* Cast to a float before division */ *load /= ticks_per_s; *load /= elapsed; crm_debug("cib load: %f (%lu ticks in %lds)", *load, delta_utime + delta_stime, (long) elapsed); } else { crm_debug("Init %lu + %lu ticks at %ld (%lu tps)", utime, stime, (long) now, ticks_per_s); } last_call = now; last_utime = utime; last_stime = stime; fclose(stream); return true; } fclose(stream); #endif // HAVE_LINUX_PROCFS return false; } bool pcmk__throttle_load_avg(float *load) { #if HAVE_LINUX_PROCFS char buffer[256]; FILE *stream = NULL; const char *loadfile = "/proc/loadavg"; if (load == NULL) { return false; } stream = fopen(loadfile, "r"); if (stream == NULL) { int rc = errno; crm_warn("Couldn't read %s: %s (%d)", loadfile, pcmk_rc_str(rc), rc); return false; } if (fgets(buffer, sizeof(buffer), stream) != NULL) { char *nl = strstr(buffer, "\n"); /* Grab the 1-minute average, ignore the rest */ *load = strtof(buffer, NULL); if (nl != NULL) { nl[0] = 0; } fclose(stream); return true; } fclose(stream); #endif // HAVE_LINUX_PROCFS return false; } diff --git a/lib/common/scheduler.c b/lib/common/scheduler.c index ceaa570b0f..d31edf97d9 100644 --- a/lib/common/scheduler.c +++ b/lib/common/scheduler.c @@ -1,416 +1,415 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include // uint32_t #include // EINVAL #include // gboolean, FALSE, etc. #include // xmlNode #include uint32_t pcmk__warnings = 0; /*! * \brief Create a new object to hold scheduler data * * \return New, initialized scheduler data, or NULL on memory error * \note Only pcmk_scheduler_t objects created with this function (as opposed * to statically declared or directly allocated) should be used with the * functions in this library, to allow for future extensions to the * data type. The caller is responsible for freeing the memory with * pcmk_free_scheduler() when the instance is no longer needed. */ pcmk_scheduler_t * pcmk_new_scheduler(void) { pcmk_scheduler_t *scheduler = calloc(1, sizeof(pcmk_scheduler_t)); if (scheduler == NULL) { return NULL; } scheduler->priv = calloc(1, sizeof(pcmk__scheduler_private_t)); if (scheduler->priv == NULL) { free(scheduler); return NULL; } pcmk__set_scheduler_defaults(scheduler); return scheduler; } /*! * \internal * \brief Set non-zero default values in scheduler data * * \param[in,out] scheduler Scheduler data to modify * * \note Values that default to NULL or 0 will remain unchanged */ void pcmk__set_scheduler_defaults(pcmk_scheduler_t *scheduler) { pcmk__assert(scheduler != NULL); scheduler->flags = 0U; #if PCMK__CONCURRENT_FENCING_DEFAULT_TRUE pcmk__set_scheduler_flags(scheduler, pcmk__sched_symmetric_cluster |pcmk__sched_concurrent_fencing |pcmk__sched_stop_removed_resources |pcmk__sched_cancel_removed_actions); #else pcmk__set_scheduler_flags(scheduler, pcmk__sched_symmetric_cluster |pcmk__sched_stop_removed_resources |pcmk__sched_cancel_removed_actions); #endif scheduler->no_quorum_policy = pcmk_no_quorum_stop; scheduler->priv->next_action_id = 1; scheduler->priv->next_ordering_id = 1; } /*! * \brief Reset scheduler data to defaults * * Free scheduler data except the local node name and output object, and reset * all other values to defaults, so the data is suitable for rerunning status * * \param[in,out] scheduler Scheduler data to reset */ void pcmk_reset_scheduler(pcmk_scheduler_t *scheduler) { if (scheduler == NULL) { return; } /* Be careful about the order of freeing members. Many contain references to * other members that will become dangling if those members are freed first. * For example, the node name and ID of Pacemaker Remote nodes are pointers * into resource objects. Ensure that earlier-freed members are not needed * by any of the free functions for later-freed members. */ scheduler->dc_node = NULL; g_list_free_full(scheduler->nodes, pcmk__free_node); scheduler->nodes = NULL; // Do not reset local_node_name or out crm_time_free(scheduler->priv->now); scheduler->priv->now = NULL; if (scheduler->priv->options != NULL) { g_hash_table_destroy(scheduler->priv->options); scheduler->priv->options = NULL; } scheduler->priv->fence_action = NULL; scheduler->priv->fence_timeout_ms = 0U; scheduler->priv->priority_fencing_ms = 0U; scheduler->priv->shutdown_lock_ms = 0U; scheduler->priv->node_pending_ms = 0U; scheduler->priv->placement_strategy = NULL; scheduler->priv->rsc_defaults = NULL; scheduler->priv->op_defaults = NULL; g_list_free_full(scheduler->priv->resources, pcmk__free_resource); scheduler->priv->resources = NULL; if (scheduler->priv->templates != NULL) { g_hash_table_destroy(scheduler->priv->templates); scheduler->priv->templates = NULL; } if (scheduler->priv->tags != NULL) { g_hash_table_destroy(scheduler->priv->tags); scheduler->priv->tags = NULL; } g_list_free_full(scheduler->priv->actions, pcmk__free_action); scheduler->priv->actions = NULL; if (scheduler->priv->singletons != NULL) { g_hash_table_destroy(scheduler->priv->singletons); scheduler->priv->singletons = NULL; } pcmk__xml_free(scheduler->priv->failed); scheduler->priv->failed = NULL; pcmk__free_param_checks(scheduler); g_list_free(scheduler->priv->stop_needed); scheduler->priv->stop_needed = NULL; g_list_free_full(scheduler->priv->location_constraints, pcmk__free_location); scheduler->priv->location_constraints = NULL; g_list_free_full(scheduler->priv->colocation_constraints, free); scheduler->priv->colocation_constraints = NULL; g_list_free_full(scheduler->priv->ordering_constraints, pcmk__free_action_relation); scheduler->priv->ordering_constraints = NULL; if (scheduler->priv->ticket_constraints != NULL) { g_hash_table_destroy(scheduler->priv->ticket_constraints); scheduler->priv->ticket_constraints = NULL; } scheduler->priv->ninstances = 0; scheduler->priv->blocked_resources = 0; scheduler->priv->disabled_resources = 0; scheduler->priv->recheck_by = 0; pcmk__xml_free(scheduler->priv->graph); scheduler->priv->graph = NULL; scheduler->priv->synapse_count = 0; pcmk__xml_free(scheduler->input); scheduler->input = NULL; pcmk__set_scheduler_defaults(scheduler); pcmk__config_has_error = false; pcmk__config_has_warning = false; } /*! * \brief Free scheduler data * * \param[in,out] scheduler Scheduler data to free */ void pcmk_free_scheduler(pcmk_scheduler_t *scheduler) { if (scheduler != NULL) { pcmk_reset_scheduler(scheduler); free(scheduler->priv->local_node_name); free(scheduler->priv); free(scheduler); } } /*! * \internal * \brief Get the Designated Controller node from scheduler data * * \param[in] scheduler Scheduler data * * \return Designated Controller node from scheduler data, or NULL if none */ pcmk_node_t * pcmk_get_dc(const pcmk_scheduler_t *scheduler) { return (scheduler == NULL)? NULL : scheduler->dc_node; } /*! * \internal * \brief Get the no quorum policy from scheduler data * * \param[in] scheduler Scheduler data * * \return No quorum policy from scheduler data */ enum pe_quorum_policy pcmk_get_no_quorum_policy(const pcmk_scheduler_t *scheduler) { if (scheduler == NULL) { return pcmk_no_quorum_stop; // The default } return scheduler->no_quorum_policy; } /*! * \internal * \brief Set CIB XML as scheduler input in scheduler data * * \param[out] scheduler Scheduler data * \param[in] cib CIB XML to set as scheduler input * * \return Standard Pacemaker return code (EINVAL if \p scheduler is NULL, * otherwise pcmk_rc_ok) * \note This will not free any previously set scheduler CIB. */ int pcmk_set_scheduler_cib(pcmk_scheduler_t *scheduler, xmlNode *cib) { if (scheduler == NULL) { return EINVAL; } scheduler->input = cib; return pcmk_rc_ok; } /*! * \internal * \brief Check whether cluster has quorum * * \param[in] scheduler Scheduler data * * \return true if cluster has quorum, otherwise false */ bool pcmk_has_quorum(const pcmk_scheduler_t *scheduler) { if (scheduler == NULL) { return false; } return pcmk_is_set(scheduler->flags, pcmk__sched_quorate); } /*! * \brief Find a node by name in scheduler data * * \param[in] scheduler Scheduler data * \param[in] node_name Name of node to find * * \return Node from scheduler data that matches \p node_name if any, * otherwise NULL */ pcmk_node_t * pcmk_find_node(const pcmk_scheduler_t *scheduler, const char *node_name) { if ((scheduler == NULL) || (node_name == NULL)) { return NULL; } return pcmk__find_node_in_list(scheduler->nodes, node_name); } /*! * \internal * \brief Get scheduler data's "now" in epoch time * * \param[in,out] scheduler Scheduler data * * \return Scheduler data's "now" as seconds since epoch (defaulting to current * time) */ time_t pcmk__scheduler_epoch_time(pcmk_scheduler_t *scheduler) { if (scheduler == NULL) { return time(NULL); } if (scheduler->priv->now == NULL) { crm_trace("Scheduler 'now' set to current time"); scheduler->priv->now = crm_time_new(NULL); } return crm_time_get_seconds_since_epoch(scheduler->priv->now); } /*! * \internal * \brief Update "recheck by" time in scheduler data * * \param[in] recheck Epoch time when recheck should happen * \param[in,out] scheduler Scheduler data * \param[in] reason What time is being updated for (for logs) */ void pcmk__update_recheck_time(time_t recheck, pcmk_scheduler_t *scheduler, const char *reason) { pcmk__assert(scheduler != NULL); if ((recheck > pcmk__scheduler_epoch_time(scheduler)) && ((scheduler->priv->recheck_by == 0) || (scheduler->priv->recheck_by > recheck))) { scheduler->priv->recheck_by = recheck; crm_debug("Updated next scheduler recheck to %s for %s", - pcmk__trim(ctime(&recheck)), - pcmk__s(reason, "some reason")); + g_strchomp(ctime(&recheck)), pcmk__s(reason, "some reason")); } } /* Fail count clearing for parameter changes normally happens when unpacking * history, before resources are unpacked. However, for bundles using the * REMOTE_CONTAINER_HACK, we can't check the conditions until after unpacking * the bundle, so those parameter checks are deferred using the APIs below. */ // History entry to be checked later for fail count clearing struct param_check { const xmlNode *rsc_history; // History entry XML pcmk_resource_t *rsc; // Resource corresponding to history entry pcmk_node_t *node; // Node corresponding to history entry enum pcmk__check_parameters check_type; // What needs checking }; /*! * \internal * \brief Add a deferred parameter check * * \param[in] rsc_history Resource history XML to check later * \param[in,out] rsc Resource that history is for * \param[in] node Node that history is for * \param[in] flag What needs to be checked later */ void pcmk__add_param_check(const xmlNode *rsc_history, pcmk_resource_t *rsc, pcmk_node_t *node, enum pcmk__check_parameters flag) { struct param_check *param_check = NULL; CRM_CHECK((rsc_history != NULL) && (rsc != NULL) && (node != NULL), return); crm_trace("Deferring checks of %s until after assignment", pcmk__xe_id(rsc_history)); param_check = pcmk__assert_alloc(1, sizeof(struct param_check)); param_check->rsc_history = rsc_history; param_check->rsc = rsc; param_check->node = node; param_check->check_type = flag; rsc->priv->scheduler->priv->param_check = g_list_prepend(rsc->priv->scheduler->priv->param_check, param_check); } /*! * \internal * \brief Call a function for each deferred parameter check * * \param[in,out] scheduler Scheduler data * \param[in] cb Function to be called */ void pcmk__foreach_param_check(pcmk_scheduler_t *scheduler, void (*cb)(pcmk_resource_t*, pcmk_node_t*, const xmlNode*, enum pcmk__check_parameters)) { CRM_CHECK((scheduler != NULL) && (cb != NULL), return); for (GList *item = scheduler->priv->param_check; item != NULL; item = item->next) { struct param_check *param_check = item->data; cb(param_check->rsc, param_check->node, param_check->rsc_history, param_check->check_type); } } /*! * \internal * \brief Free all deferred parameter checks * * \param[in,out] scheduler Scheduler data */ void pcmk__free_param_checks(pcmk_scheduler_t *scheduler) { if ((scheduler != NULL) && (scheduler->priv->param_check != NULL)) { g_list_free_full(scheduler->priv->param_check, free); scheduler->priv->param_check = NULL; } } diff --git a/lib/common/schemas.c b/lib/common/schemas.c index 38b936d816..61b3300c87 100644 --- a/lib/common/schemas.c +++ b/lib/common/schemas.c @@ -1,1603 +1,1611 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include // UCHAR_MAX #include #include +#include // g_str_has_prefix() #include #include // xmlNode #include // xmlChar #include #include #include #include #include #include #include /* PCMK__XML_LOG_BASE */ #include "crmcommon_private.h" #define SCHEMA_ZERO { .v = { 0, 0 } } #define schema_strdup_printf(prefix, version, suffix) \ crm_strdup_printf(prefix "%u.%u" suffix, (version).v[0], (version).v[1]) typedef struct { xmlRelaxNGPtr rng; xmlRelaxNGValidCtxtPtr valid; xmlRelaxNGParserCtxtPtr parser; } relaxng_ctx_cache_t; static GList *known_schemas = NULL; static bool initialized = false; static bool silent_logging = FALSE; static void G_GNUC_PRINTF(2, 3) xml_log(int priority, const char *fmt, ...) { va_list ap; va_start(ap, fmt); if (silent_logging == FALSE) { /* XXX should not this enable dechunking as well? */ PCMK__XML_LOG_BASE(priority, FALSE, 0, NULL, fmt, ap); } va_end(ap); } static int xml_latest_schema_index(void) { /* This function assumes that pcmk__schema_init() has been called * beforehand, so we have at least two schemas (one real schema and the * "none" schema). * * @COMPAT: The "none" schema is deprecated since 2.1.8. * Update this when we drop that schema. */ return g_list_length(known_schemas) - 2; } /*! * \internal * \brief Return the schema entry of the highest-versioned schema * * \return Schema entry of highest-versioned schema */ static GList * get_highest_schema(void) { /* The highest numerically versioned schema is the one before none * * @COMPAT none is deprecated since 2.1.8 */ GList *entry = pcmk__get_schema("none"); pcmk__assert((entry != NULL) && (entry->prev != NULL)); return entry->prev; } /*! * \internal * \brief Return the name of the highest-versioned schema * * \return Name of highest-versioned schema (or NULL on error) */ const char * pcmk__highest_schema_name(void) { GList *entry = get_highest_schema(); return ((pcmk__schema_t *)(entry->data))->name; } /*! * \internal * \brief Find first entry of highest major schema version series * * \return Schema entry of first schema with highest major version */ GList * pcmk__find_x_0_schema(void) { #if defined(PCMK__UNIT_TESTING) /* If we're unit testing, this can't be static because it'll stick * around from one test run to the next. It needs to be cleared out * every time. */ GList *x_0_entry = NULL; #else static GList *x_0_entry = NULL; #endif pcmk__schema_t *highest_schema = NULL; if (x_0_entry != NULL) { return x_0_entry; } x_0_entry = get_highest_schema(); highest_schema = x_0_entry->data; for (GList *iter = x_0_entry->prev; iter != NULL; iter = iter->prev) { pcmk__schema_t *schema = iter->data; /* We've found a schema in an older major version series. Return * the index of the first one in the same major version series as * the highest schema. */ if (schema->version.v[0] < highest_schema->version.v[0]) { x_0_entry = iter->next; break; } /* We're out of list to examine. This probably means there was only * one major version series, so return the first schema entry. */ if (iter->prev == NULL) { x_0_entry = known_schemas->data; break; } } return x_0_entry; } static inline bool version_from_filename(const char *filename, pcmk__schema_version_t *version) { - if (pcmk__ends_with(filename, ".rng")) { + if (filename == NULL) { + return false; + } + + if (g_str_has_suffix(filename, ".rng")) { return sscanf(filename, "pacemaker-%hhu.%hhu.rng", &(version->v[0]), &(version->v[1])) == 2; } else { return sscanf(filename, "pacemaker-%hhu.%hhu", &(version->v[0]), &(version->v[1])) == 2; } } static int schema_filter(const struct dirent *a) { int rc = 0; pcmk__schema_version_t version = SCHEMA_ZERO; - if (strstr(a->d_name, "pacemaker-") != a->d_name) { + if (!g_str_has_prefix(a->d_name, "pacemaker-")) { /* crm_trace("%s - wrong prefix", a->d_name); */ - } else if (!pcmk__ends_with_ext(a->d_name, ".rng")) { + } else if (!g_str_has_suffix(a->d_name, ".rng")) { /* crm_trace("%s - wrong suffix", a->d_name); */ } else if (!version_from_filename(a->d_name, &version)) { /* crm_trace("%s - wrong format", a->d_name); */ } else { /* crm_debug("%s - candidate", a->d_name); */ rc = 1; } return rc; } static int schema_cmp(pcmk__schema_version_t a_version, pcmk__schema_version_t b_version) { for (int i = 0; i < 2; ++i) { if (a_version.v[i] < b_version.v[i]) { return -1; } else if (a_version.v[i] > b_version.v[i]) { return 1; } } return 0; } static int schema_cmp_directory(const struct dirent **a, const struct dirent **b) { pcmk__schema_version_t a_version = SCHEMA_ZERO; pcmk__schema_version_t b_version = SCHEMA_ZERO; if (!version_from_filename(a[0]->d_name, &a_version) || !version_from_filename(b[0]->d_name, &b_version)) { // Shouldn't be possible, but makes static analysis happy return 0; } return schema_cmp(a_version, b_version); } /*! * \internal * \brief Add given schema + auxiliary data to internal bookkeeping. */ static void add_schema(enum pcmk__schema_validator validator, const pcmk__schema_version_t *version, const char *name, GList *transforms) { pcmk__schema_t *schema = NULL; schema = pcmk__assert_alloc(1, sizeof(pcmk__schema_t)); schema->validator = validator; schema->version.v[0] = version->v[0]; schema->version.v[1] = version->v[1]; schema->transforms = transforms; // schema->schema_index is set after all schemas are loaded and sorted if (version->v[0] || version->v[1]) { schema->name = schema_strdup_printf("pacemaker-", *version, ""); } else { schema->name = pcmk__str_copy(name); } known_schemas = g_list_prepend(known_schemas, schema); } static void wrap_libxslt(bool finalize) { static xsltSecurityPrefsPtr secprefs; int ret = 0; /* security framework preferences */ if (!finalize) { pcmk__assert(secprefs == NULL); secprefs = xsltNewSecurityPrefs(); ret = xsltSetSecurityPrefs(secprefs, XSLT_SECPREF_WRITE_FILE, xsltSecurityForbid) | xsltSetSecurityPrefs(secprefs, XSLT_SECPREF_CREATE_DIRECTORY, xsltSecurityForbid) | xsltSetSecurityPrefs(secprefs, XSLT_SECPREF_READ_NETWORK, xsltSecurityForbid) | xsltSetSecurityPrefs(secprefs, XSLT_SECPREF_WRITE_NETWORK, xsltSecurityForbid); if (ret != 0) { return; } } else { xsltFreeSecurityPrefs(secprefs); secprefs = NULL; } /* cleanup only */ if (finalize) { xsltCleanupGlobals(); } } /*! * \internal * \brief Check whether a directory entry matches the upgrade XSLT pattern * * \param[in] entry Directory entry whose filename to check * * \return 1 if the entry's filename is of the form * upgrade-X.Y-ORDER.xsl with each number in the range 0 to * 255, or 0 otherwise */ static int transform_filter(const struct dirent *entry) { const char *re = NULL; unsigned int major = 0; unsigned int minor = 0; unsigned int order = 0; /* Each number is an unsigned char, which is 1 to 3 digits long. (Pacemaker * requires an 8-bit char via a configure test.) */ re = "upgrade-[[:digit:]]{1,3}\\.[[:digit:]]{1,3}-[[:digit:]]{1,3}\\.xsl"; if (!pcmk__str_eq(entry->d_name, re, pcmk__str_regex)) { return 0; } /* Performance isn't critical here and this is simpler than range-checking * within the regex */ if (sscanf(entry->d_name, "upgrade-%u.%u-%u.xsl", &major, &minor, &order) != 3) { return 0; } if ((major > UCHAR_MAX) || (minor > UCHAR_MAX) || (order > UCHAR_MAX)) { return 0; } return 1; } /*! * \internal * \brief Compare transform files based on the version strings in their names * * This is a crude version comparison that relies on the specific structure of * these filenames. * * \retval -1 if \p entry1 sorts before \p entry2 * \retval 0 if \p entry1 sorts equal to \p entry2 * \retval 1 if \p entry1 sorts after \p entry2 * * \note The GNU \c versionsort() function would be perfect here, but it's not * portable. */ static int compare_transforms(const struct dirent **entry1, const struct dirent **entry2) { unsigned char major1 = 0; unsigned char major2 = 0; unsigned char minor1 = 0; unsigned char minor2 = 0; unsigned char order1 = 0; unsigned char order2 = 0; // If these made it through the filter, they should be of the right format CRM_LOG_ASSERT(sscanf((*entry1)->d_name, "upgrade-%hhu.%hhu-%hhu.xsl", &major1, &minor1, &order1) == 3); CRM_LOG_ASSERT(sscanf((*entry2)->d_name, "upgrade-%hhu.%hhu-%hhu.xsl", &major2, &minor2, &order2) == 3); if (major1 < major2) { return -1; } else if (major1 > major2) { return 1; } if (minor1 < minor2) { return -1; } else if (minor1 > minor2) { return 1; } if (order1 < order2) { return -1; } else if (order1 > order2) { return 1; } return 0; } /*! * \internal * \brief Free a list of XSLT transform struct dirent objects * * \param[in,out] data List to free */ static void free_transform_list(void *data) { g_list_free_full((GList *) data, free); } /*! * \internal * \brief Load names of upgrade XSLT stylesheets from a directory into a table * * Stylesheets must have names of the form "upgrade-X.Y-order.xsl", where: * * X is the schema major version * * Y is the schema minor version * * ORDER is the order in which the stylesheet occurs in the transform pipeline * * \param[in] dir Directory containing XSLT stylesheets * * \return Table with schema version as key and \c GList of associated transform * files (as struct dirent) as value */ static GHashTable * load_transforms_from_dir(const char *dir) { struct dirent **namelist = NULL; GHashTable *transforms = NULL; int num_matches = scandir(dir, &namelist, transform_filter, compare_transforms); if (num_matches < 0) { int rc = errno; crm_warn("Could not load transforms from %s: %s", dir, pcmk_rc_str(rc)); goto done; } transforms = pcmk__strkey_table(free, free_transform_list); for (int i = 0; i < num_matches; i++) { pcmk__schema_version_t version = SCHEMA_ZERO; unsigned char order = 0; // Placeholder only if (sscanf(namelist[i]->d_name, "upgrade-%hhu.%hhu-%hhu.xsl", &(version.v[0]), &(version.v[1]), &order) == 3) { char *version_s = crm_strdup_printf("%hhu.%hhu", version.v[0], version.v[1]); GList *list = g_hash_table_lookup(transforms, version_s); if (list == NULL) { /* Prepend is more efficient. However, there won't be many of * these, and we want them to remain sorted by version. It's not * worth reversing all the lists at the end. * * Avoid calling g_hash_table_insert() if the list already * exists. Otherwise free_transform_list() gets called on it. */ list = g_list_append(list, namelist[i]); g_hash_table_insert(transforms, version_s, list); } else { list = g_list_append(list, namelist[i]); free(version_s); } } else { // Sanity only, should never happen thanks to transform_filter() free(namelist[i]); } } done: free(namelist); return transforms; } void pcmk__load_schemas_from_dir(const char *dir) { int lpc, max; struct dirent **namelist = NULL; GHashTable *transforms = NULL; max = scandir(dir, &namelist, schema_filter, schema_cmp_directory); if (max < 0) { int rc = errno; crm_warn("Could not load schemas from %s: %s", dir, pcmk_rc_str(rc)); goto done; } // Look for any upgrade transforms in the same directory transforms = load_transforms_from_dir(dir); for (lpc = 0; lpc < max; lpc++) { pcmk__schema_version_t version = SCHEMA_ZERO; if (version_from_filename(namelist[lpc]->d_name, &version)) { char *version_s = crm_strdup_printf("%hhu.%hhu", version.v[0], version.v[1]); char *orig_key = NULL; GList *transform_list = NULL; if (transforms != NULL) { // The schema becomes the owner of transform_list g_hash_table_lookup_extended(transforms, version_s, (gpointer *) &orig_key, (gpointer *) &transform_list); g_hash_table_steal(transforms, version_s); } add_schema(pcmk__schema_validator_rng, &version, NULL, transform_list); free(version_s); free(orig_key); } else { // Shouldn't be possible, but makes static analysis happy crm_warn("Skipping schema '%s': could not parse version", namelist[lpc]->d_name); } } for (lpc = 0; lpc < max; lpc++) { free(namelist[lpc]); } done: free(namelist); if (transforms != NULL) { g_hash_table_destroy(transforms); } } static gint schema_sort_GCompareFunc(gconstpointer a, gconstpointer b) { const pcmk__schema_t *schema_a = a; const pcmk__schema_t *schema_b = b; // @COMPAT The "none" schema is deprecated since 2.1.8 if (pcmk__str_eq(schema_a->name, PCMK_VALUE_NONE, pcmk__str_none)) { return 1; } else if (pcmk__str_eq(schema_b->name, PCMK_VALUE_NONE, pcmk__str_none)) { return -1; } else { return schema_cmp(schema_a->version, schema_b->version); } } /*! * \internal * \brief Sort the list of known schemas such that all pacemaker-X.Y are in * version order, then "none" * * This function should be called whenever additional schemas are loaded using * \c pcmk__load_schemas_from_dir(), after the initial sets in * \c pcmk__schema_init(). */ void pcmk__sort_schemas(void) { known_schemas = g_list_sort(known_schemas, schema_sort_GCompareFunc); } /*! * \internal * \brief Load pacemaker schemas into cache * * \note This currently also serves as an entry point for the * generic initialization of the libxslt library. */ void pcmk__schema_init(void) { if (!initialized) { const char *remote_schema_dir = pcmk__remote_schema_dir(); char *base = pcmk__xml_artefact_root(pcmk__xml_artefact_ns_legacy_rng); const pcmk__schema_version_t zero = SCHEMA_ZERO; int schema_index = 0; initialized = true; wrap_libxslt(false); pcmk__load_schemas_from_dir(base); pcmk__load_schemas_from_dir(remote_schema_dir); free(base); // @COMPAT Deprecated since 2.1.8 add_schema(pcmk__schema_validator_none, &zero, PCMK_VALUE_NONE, NULL); /* add_schema() prepends items to the list, so in the simple case, this * just reverses the list. However if there were any remote schemas, * sorting is necessary. */ pcmk__sort_schemas(); // Now set the schema indexes and log the final result for (GList *iter = known_schemas; iter != NULL; iter = iter->next) { pcmk__schema_t *schema = iter->data; crm_debug("Loaded schema %d: %s", schema_index, schema->name); schema->schema_index = schema_index++; } } } static bool validate_with_relaxng(xmlDocPtr doc, xmlRelaxNGValidityErrorFunc error_handler, void *error_handler_context, const char *relaxng_file, relaxng_ctx_cache_t **cached_ctx) { int rc = 0; bool valid = true; relaxng_ctx_cache_t *ctx = NULL; CRM_CHECK(doc != NULL, return false); CRM_CHECK(relaxng_file != NULL, return false); if (cached_ctx && *cached_ctx) { ctx = *cached_ctx; } else { crm_debug("Creating RNG parser context"); ctx = pcmk__assert_alloc(1, sizeof(relaxng_ctx_cache_t)); ctx->parser = xmlRelaxNGNewParserCtxt(relaxng_file); CRM_CHECK(ctx->parser != NULL, goto cleanup); if (error_handler) { xmlRelaxNGSetParserErrors(ctx->parser, (xmlRelaxNGValidityErrorFunc) error_handler, (xmlRelaxNGValidityWarningFunc) error_handler, error_handler_context); } else { xmlRelaxNGSetParserErrors(ctx->parser, (xmlRelaxNGValidityErrorFunc) fprintf, (xmlRelaxNGValidityWarningFunc) fprintf, stderr); } ctx->rng = xmlRelaxNGParse(ctx->parser); CRM_CHECK(ctx->rng != NULL, crm_err("Could not find/parse %s", relaxng_file); goto cleanup); ctx->valid = xmlRelaxNGNewValidCtxt(ctx->rng); CRM_CHECK(ctx->valid != NULL, goto cleanup); if (error_handler) { xmlRelaxNGSetValidErrors(ctx->valid, (xmlRelaxNGValidityErrorFunc) error_handler, (xmlRelaxNGValidityWarningFunc) error_handler, error_handler_context); } else { xmlRelaxNGSetValidErrors(ctx->valid, (xmlRelaxNGValidityErrorFunc) fprintf, (xmlRelaxNGValidityWarningFunc) fprintf, stderr); } } rc = xmlRelaxNGValidateDoc(ctx->valid, doc); if (rc > 0) { valid = false; } else if (rc < 0) { crm_err("Internal libxml error during validation"); } cleanup: if (cached_ctx) { *cached_ctx = ctx; } else { if (ctx->parser != NULL) { xmlRelaxNGFreeParserCtxt(ctx->parser); } if (ctx->valid != NULL) { xmlRelaxNGFreeValidCtxt(ctx->valid); } if (ctx->rng != NULL) { xmlRelaxNGFree(ctx->rng); } free(ctx); } return valid; } static void free_schema(gpointer data) { pcmk__schema_t *schema = data; relaxng_ctx_cache_t *ctx = NULL; switch (schema->validator) { case pcmk__schema_validator_none: // not cached break; case pcmk__schema_validator_rng: // cached ctx = (relaxng_ctx_cache_t *) schema->cache; if (ctx == NULL) { break; } if (ctx->parser != NULL) { xmlRelaxNGFreeParserCtxt(ctx->parser); } if (ctx->valid != NULL) { xmlRelaxNGFreeValidCtxt(ctx->valid); } if (ctx->rng != NULL) { xmlRelaxNGFree(ctx->rng); } free(ctx); schema->cache = NULL; break; } free(schema->name); g_list_free_full(schema->transforms, free); free(schema); } /*! * \internal * \brief Clean up global memory associated with XML schemas */ void pcmk__schema_cleanup(void) { if (known_schemas != NULL) { g_list_free_full(known_schemas, free_schema); known_schemas = NULL; } initialized = false; wrap_libxslt(true); } /*! * \internal * \brief Get schema list entry corresponding to a schema name * * \param[in] name Name of schema to get * * \return Schema list entry corresponding to \p name, or NULL if unknown */ GList * pcmk__get_schema(const char *name) { if (name == NULL) { return NULL; } for (GList *iter = known_schemas; iter != NULL; iter = iter->next) { pcmk__schema_t *schema = iter->data; if (pcmk__str_eq(name, schema->name, pcmk__str_none)) { return iter; } } return NULL; } /*! * \internal * \brief Compare two schema version numbers given the schema names * * \param[in] schema1 Name of first schema to compare * \param[in] schema2 Name of second schema to compare * * \return Standard comparison result (negative integer if \p schema1 has the * lower version number, positive integer if \p schema1 has the higher * version number, of 0 if the version numbers are equal) */ int pcmk__cmp_schemas_by_name(const char *schema1_name, const char *schema2_name) { GList *entry1 = pcmk__get_schema(schema1_name); GList *entry2 = pcmk__get_schema(schema2_name); if (entry1 == NULL) { return (entry2 == NULL)? 0 : -1; } else if (entry2 == NULL) { return 1; } else { pcmk__schema_t *schema1 = entry1->data; pcmk__schema_t *schema2 = entry2->data; return schema1->schema_index - schema2->schema_index; } } static bool validate_with(xmlNode *xml, pcmk__schema_t *schema, xmlRelaxNGValidityErrorFunc error_handler, void *error_handler_context) { bool valid = false; char *file = NULL; relaxng_ctx_cache_t **cache = NULL; if (schema == NULL) { return false; } if (schema->validator == pcmk__schema_validator_none) { return true; } file = pcmk__xml_artefact_path(pcmk__xml_artefact_ns_legacy_rng, schema->name); crm_trace("Validating with %s (type=%d)", pcmk__s(file, "missing schema"), schema->validator); switch (schema->validator) { case pcmk__schema_validator_rng: cache = (relaxng_ctx_cache_t **) &(schema->cache); valid = validate_with_relaxng(xml->doc, error_handler, error_handler_context, file, cache); break; default: crm_err("Unknown validator type: %d", schema->validator); break; } free(file); return valid; } static bool validate_with_silent(xmlNode *xml, pcmk__schema_t *schema) { bool rc, sl_backup = silent_logging; silent_logging = TRUE; rc = validate_with(xml, schema, (xmlRelaxNGValidityErrorFunc) xml_log, GUINT_TO_POINTER(LOG_ERR)); silent_logging = sl_backup; return rc; } bool pcmk__validate_xml(xmlNode *xml_blob, const char *validation, xmlRelaxNGValidityErrorFunc error_handler, void *error_handler_context) { GList *entry = NULL; pcmk__schema_t *schema = NULL; CRM_CHECK((xml_blob != NULL) && (xml_blob->doc != NULL), return false); if (validation == NULL) { validation = pcmk__xe_get(xml_blob, PCMK_XA_VALIDATE_WITH); } pcmk__warn_if_schema_deprecated(validation); entry = pcmk__get_schema(validation); if (entry == NULL) { pcmk__config_err("Cannot validate CIB with %s " PCMK_XA_VALIDATE_WITH " (manually edit to use a known schema)", ((validation == NULL)? "missing" : "unknown")); return false; } schema = entry->data; return validate_with(xml_blob, schema, error_handler, error_handler_context); } /*! * \internal * \brief Validate XML using its configured schema (and send errors to logs) * * \param[in] xml XML to validate * * \return true if XML validates, otherwise false */ bool pcmk__configured_schema_validates(xmlNode *xml) { return pcmk__validate_xml(xml, NULL, (xmlRelaxNGValidityErrorFunc) xml_log, GUINT_TO_POINTER(LOG_ERR)); } /* With this arrangement, an attempt to identify the message severity as explicitly signalled directly from XSLT is performed in rather a smart way (no reliance on formatting string + arguments being always specified as ["%s", purposeful_string], as it can also be ["%s: %s", some_prefix, purposeful_string] etc. so every argument pertaining %s specifier is investigated), and if such a mark found, the respective level is determined and, when the messages are to go to the native logs, the mark itself gets dropped (by the means of string shift). NOTE: whether the native logging is the right sink is decided per the ctx parameter -- NULL denotes this case, otherwise it carries a pointer to the numeric expression of the desired target logging level (messages with higher level will be suppressed) NOTE: on some architectures, this string shift may not have any effect, but that's an acceptable tradeoff The logging level for not explicitly designated messages (suspicious, likely internal errors or some runaways) is LOG_WARNING. */ static void G_GNUC_PRINTF(2, 3) cib_upgrade_err(void *ctx, const char *fmt, ...) { va_list ap, aq; char *arg_cur; bool found = false; const char *fmt_iter = fmt; uint8_t msg_log_level = LOG_WARNING; /* default for runaway messages */ const unsigned * log_level = (const unsigned *) ctx; enum { escan_seennothing, escan_seenpercent, } scan_state = escan_seennothing; va_start(ap, fmt); va_copy(aq, ap); while (!found && *fmt_iter != '\0') { /* while casing schema borrowed from libqb:qb_vsnprintf_serialize */ switch (*fmt_iter++) { case '%': if (scan_state == escan_seennothing) { scan_state = escan_seenpercent; } else if (scan_state == escan_seenpercent) { scan_state = escan_seennothing; } break; case 's': if (scan_state == escan_seenpercent) { size_t prefix_len = 0; scan_state = escan_seennothing; arg_cur = va_arg(aq, char *); - if (pcmk__starts_with(arg_cur, "WARNING: ")) { + if (arg_cur == NULL) { + break; + + } else if (g_str_has_prefix(arg_cur, "WARNING: ")) { prefix_len = sizeof("WARNING: ") - 1; msg_log_level = LOG_WARNING; - } else if (pcmk__starts_with(arg_cur, "INFO: ")) { + } else if (g_str_has_prefix(arg_cur, "INFO: ")) { prefix_len = sizeof("INFO: ") - 1; msg_log_level = LOG_INFO; - } else if (pcmk__starts_with(arg_cur, "DEBUG: ")) { + } else if (g_str_has_prefix(arg_cur, "DEBUG: ")) { prefix_len = sizeof("DEBUG: ") - 1; msg_log_level = LOG_DEBUG; } else { break; } found = true; if (ctx == NULL) { memmove(arg_cur, arg_cur + prefix_len, strlen(arg_cur + prefix_len) + 1); } } break; case '#': case '-': case ' ': case '+': case '\'': case 'I': case '.': case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': case '*': break; case 'l': case 'z': case 't': case 'j': case 'd': case 'i': case 'o': case 'u': case 'x': case 'X': case 'e': case 'E': case 'f': case 'F': case 'g': case 'G': case 'a': case 'A': case 'c': case 'p': if (scan_state == escan_seenpercent) { (void) va_arg(aq, void *); /* skip forward */ scan_state = escan_seennothing; } break; default: scan_state = escan_seennothing; break; } } if (log_level != NULL) { /* intention of the following offset is: cibadmin -V -> start showing INFO labelled messages */ if (*log_level + 4 >= msg_log_level) { vfprintf(stderr, fmt, ap); } } else { PCMK__XML_LOG_BASE(msg_log_level, TRUE, 0, "CIB upgrade: ", fmt, ap); } va_end(aq); va_end(ap); } /*! * \internal * \brief Apply a single XSL transformation to given XML * * \param[in] xml XML to transform * \param[in] transform XSL name * \param[in] to_logs If false, certain validation errors will be sent to * stderr rather than logged * * \return Transformed XML on success, otherwise NULL */ static xmlNode * apply_transformation(const xmlNode *xml, const char *transform, gboolean to_logs) { char *xform = NULL; xmlNode *out = NULL; xmlDocPtr res = NULL; xsltStylesheet *xslt = NULL; xform = pcmk__xml_artefact_path(pcmk__xml_artefact_ns_legacy_xslt, transform); /* for capturing, e.g., what's emitted via */ if (to_logs) { xsltSetGenericErrorFunc(NULL, cib_upgrade_err); } else { xsltSetGenericErrorFunc(&crm_log_level, cib_upgrade_err); } xslt = xsltParseStylesheetFile((const xmlChar *) xform); CRM_CHECK(xslt != NULL, goto cleanup); /* Caller allocates private data for final result document. Intermediate * result documents are temporary and don't need private data. */ res = xsltApplyStylesheet(xslt, xml->doc, NULL); CRM_CHECK(res != NULL, goto cleanup); xsltSetGenericErrorFunc(NULL, NULL); /* restore default one */ out = xmlDocGetRootElement(res); cleanup: if (xslt) { xsltFreeStylesheet(xslt); } free(xform); return out; } /*! * \internal * \brief Perform all transformations needed to upgrade XML to next schema * * \param[in] input_xml XML to transform * \param[in] schema_index Index of schema that successfully validates * \p original_xml * \param[in] to_logs If false, certain validation errors will be sent to * stderr rather than logged * * \return XML result of schema transforms if successful, otherwise NULL */ static xmlNode * apply_upgrade(const xmlNode *input_xml, int schema_index, gboolean to_logs) { pcmk__schema_t *schema = g_list_nth_data(known_schemas, schema_index); pcmk__schema_t *upgraded_schema = g_list_nth_data(known_schemas, schema_index + 1); xmlNode *old_xml = NULL; xmlNode *new_xml = NULL; xmlRelaxNGValidityErrorFunc error_handler = NULL; pcmk__assert((schema != NULL) && (upgraded_schema != NULL)); if (to_logs) { error_handler = (xmlRelaxNGValidityErrorFunc) xml_log; } for (GList *iter = schema->transforms; iter != NULL; iter = iter->next) { const struct dirent *entry = iter->data; const char *transform = entry->d_name; crm_debug("Upgrading schema from %s to %s: applying XSL transform %s", schema->name, upgraded_schema->name, transform); new_xml = apply_transformation(input_xml, transform, to_logs); pcmk__xml_free(old_xml); if (new_xml == NULL) { crm_err("XSL transform %s failed, aborting upgrade", transform); return NULL; } input_xml = new_xml; old_xml = new_xml; } // Final result document from upgrade pipeline needs private data pcmk__xml_new_private_data((xmlNode *) new_xml->doc); // Ensure result validates with its new schema if (!validate_with(new_xml, upgraded_schema, error_handler, GUINT_TO_POINTER(LOG_ERR))) { crm_err("Schema upgrade from %s to %s failed: " "XSL transform pipeline produced an invalid configuration", schema->name, upgraded_schema->name); crm_log_xml_debug(new_xml, "bad-transform-result"); pcmk__xml_free(new_xml); return NULL; } crm_info("Schema upgrade from %s to %s succeeded", schema->name, upgraded_schema->name); return new_xml; } /*! * \internal * \brief Get the schema list entry corresponding to XML configuration * * \param[in] xml CIB XML to check * * \return List entry of schema configured in \p xml */ static GList * get_configured_schema(const xmlNode *xml) { const char *schema_name = pcmk__xe_get(xml, PCMK_XA_VALIDATE_WITH); pcmk__warn_if_schema_deprecated(schema_name); return pcmk__get_schema(schema_name); } /*! * \brief Update CIB XML to latest schema that validates it * * \param[in,out] xml XML to update (may be freed and replaced * after being transformed) * \param[in] max_schema_name If not NULL, do not update \p xml to any * schema later than this one * \param[in] transform If false, do not update \p xml to any schema * that requires an XSL transform * \param[in] to_logs If false, certain validation errors will be * sent to stderr rather than logged * * \return Standard Pacemaker return code */ int pcmk__update_schema(xmlNode **xml, const char *max_schema_name, bool transform, bool to_logs) { int max_stable_schemas = xml_latest_schema_index(); int max_schema_index = 0; int rc = pcmk_rc_ok; GList *entry = NULL; pcmk__schema_t *best_schema = NULL; pcmk__schema_t *original_schema = NULL; xmlRelaxNGValidityErrorFunc error_handler = to_logs ? (xmlRelaxNGValidityErrorFunc) xml_log : NULL; CRM_CHECK((xml != NULL) && (*xml != NULL) && ((*xml)->doc != NULL), return EINVAL); if (max_schema_name != NULL) { GList *max_entry = pcmk__get_schema(max_schema_name); if (max_entry != NULL) { pcmk__schema_t *max_schema = max_entry->data; max_schema_index = max_schema->schema_index; } } if ((max_schema_index < 1) || (max_schema_index > max_stable_schemas)) { max_schema_index = max_stable_schemas; } entry = get_configured_schema(*xml); if (entry == NULL) { return pcmk_rc_cib_corrupt; } original_schema = entry->data; if (original_schema->schema_index >= max_schema_index) { return pcmk_rc_ok; } for (; entry != NULL; entry = entry->next) { pcmk__schema_t *current_schema = entry->data; xmlNode *upgrade = NULL; if (current_schema->schema_index > max_schema_index) { break; } if (!validate_with(*xml, current_schema, error_handler, GUINT_TO_POINTER(LOG_ERR))) { crm_debug("Schema %s does not validate", current_schema->name); if (best_schema != NULL) { /* we've satisfied the validation, no need to check further */ break; } rc = pcmk_rc_schema_validation; continue; // Try again with the next higher schema } crm_debug("Schema %s validates", current_schema->name); rc = pcmk_rc_ok; best_schema = current_schema; if (current_schema->schema_index == max_schema_index) { break; // No further transformations possible } // coverity[null_field] The index check ensures entry->next is not NULL if (!transform || (current_schema->transforms == NULL) || validate_with_silent(*xml, entry->next->data)) { /* The next schema either doesn't require a transform or validates * successfully even without the transform. Skip the transform and * try the next schema with the same XML. */ continue; } upgrade = apply_upgrade(*xml, current_schema->schema_index, to_logs); if (upgrade == NULL) { /* The transform failed, so this schema can't be used. Later * schemas are unlikely to validate, but try anyway until we * run out of options. */ rc = pcmk_rc_transform_failed; } else { best_schema = current_schema; pcmk__xml_free(*xml); *xml = upgrade; } } if ((best_schema != NULL) && (best_schema->schema_index > original_schema->schema_index)) { crm_info("%s the configuration schema to %s", (transform? "Transformed" : "Upgraded"), best_schema->name); crm_xml_add(*xml, PCMK_XA_VALIDATE_WITH, best_schema->name); } return rc; } int pcmk_update_configured_schema(xmlNode **xml) { return pcmk__update_configured_schema(xml, true); } /*! * \brief Update XML from its configured schema to the latest major series * * \param[in,out] xml XML to update * \param[in] to_logs If false, certain validation errors will be * sent to stderr rather than logged * * \return Standard Pacemaker return code */ int pcmk__update_configured_schema(xmlNode **xml, bool to_logs) { pcmk__schema_t *x_0_schema = pcmk__find_x_0_schema()->data; pcmk__schema_t *original_schema = NULL; GList *entry = NULL; if (xml == NULL) { return EINVAL; } entry = get_configured_schema(*xml); if (entry == NULL) { return pcmk_rc_cib_corrupt; } original_schema = entry->data; if (original_schema->schema_index < x_0_schema->schema_index) { // Current configuration schema is not acceptable, try to update xmlNode *converted = NULL; const char *new_schema_name = NULL; pcmk__schema_t *schema = NULL; entry = NULL; converted = pcmk__xml_copy(NULL, *xml); if (pcmk__update_schema(&converted, NULL, true, to_logs) == pcmk_rc_ok) { new_schema_name = pcmk__xe_get(converted, PCMK_XA_VALIDATE_WITH); entry = pcmk__get_schema(new_schema_name); } schema = (entry == NULL)? NULL : entry->data; if ((schema == NULL) || (schema->schema_index < x_0_schema->schema_index)) { // Updated configuration schema is still not acceptable if ((schema == NULL) || (schema->schema_index < original_schema->schema_index)) { // We couldn't validate any schema at all if (to_logs) { pcmk__config_err("Cannot upgrade configuration (claiming " "%s schema) to at least %s because it " "does not validate with any schema from " "%s to the latest", original_schema->name, x_0_schema->name, original_schema->name); } else { fprintf(stderr, "Cannot upgrade configuration (claiming " "%s schema) to at least %s because it " "does not validate with any schema from " "%s to the latest\n", original_schema->name, x_0_schema->name, original_schema->name); } } else { // We updated configuration successfully, but still too low if (to_logs) { pcmk__config_err("Cannot upgrade configuration (claiming " "%s schema) to at least %s because it " "would not upgrade past %s", original_schema->name, x_0_schema->name, pcmk__s(new_schema_name, "unspecified version")); } else { fprintf(stderr, "Cannot upgrade configuration (claiming " "%s schema) to at least %s because it " "would not upgrade past %s\n", original_schema->name, x_0_schema->name, pcmk__s(new_schema_name, "unspecified version")); } } pcmk__xml_free(converted); converted = NULL; return pcmk_rc_transform_failed; } else { // Updated configuration schema is acceptable pcmk__xml_free(*xml); *xml = converted; if (schema->schema_index < xml_latest_schema_index()) { if (to_logs) { pcmk__config_warn("Configuration with %s schema was " "internally upgraded to acceptable (but " "not most recent) %s", original_schema->name, schema->name); } } else if (to_logs) { crm_info("Configuration with %s schema was internally " "upgraded to latest version %s", original_schema->name, schema->name); } } } else if (!to_logs) { pcmk__schema_t *none_schema = NULL; entry = pcmk__get_schema(PCMK_VALUE_NONE); pcmk__assert((entry != NULL) && (entry->data != NULL)); none_schema = entry->data; if (original_schema->schema_index >= none_schema->schema_index) { // @COMPAT the none schema is deprecated since 2.1.8 fprintf(stderr, "Schema validation of configuration is " "disabled (support for " PCMK_XA_VALIDATE_WITH " set to \"" PCMK_VALUE_NONE "\" is deprecated" " and will be removed in a future release)\n"); } } return pcmk_rc_ok; } /*! * \internal * \brief Return a list of all schema files and any associated XSLT files * later than the given one * \brief Return a list of all schema versions later than the given one * * \param[in] schema The schema to compare against (for example, * "pacemaker-3.1.rng" or "pacemaker-3.1") * * \note The caller is responsible for freeing both the returned list and * the elements of the list */ GList * pcmk__schema_files_later_than(const char *name) { GList *lst = NULL; pcmk__schema_version_t ver; if (!version_from_filename(name, &ver)) { return lst; } for (GList *iter = g_list_nth(known_schemas, xml_latest_schema_index()); iter != NULL; iter = iter->prev) { pcmk__schema_t *schema = iter->data; if (schema_cmp(ver, schema->version) != -1) { continue; } for (GList *iter2 = g_list_last(schema->transforms); iter2 != NULL; iter2 = iter2->prev) { const struct dirent *entry = iter2->data; lst = g_list_prepend(lst, pcmk__str_copy(entry->d_name)); } lst = g_list_prepend(lst, crm_strdup_printf("%s.rng", schema->name)); } return lst; } static void append_href(xmlNode *xml, void *user_data) { GList **list = user_data; char *href = pcmk__xe_get_copy(xml, "href"); if (href == NULL) { return; } *list = g_list_prepend(*list, href); } static void external_refs_in_schema(GList **list, const char *contents) { /* local-name()= is needed to ignore the xmlns= setting at the top of * the XML file. Otherwise, the xpath query will always return nothing. */ const char *search = "//*[local-name()='externalRef'] | //*[local-name()='include']"; xmlNode *xml = pcmk__xml_parse(contents); pcmk__xpath_foreach_result(xml->doc, search, append_href, list); pcmk__xml_free(xml); } static int read_file_contents(const char *file, char **contents) { int rc = pcmk_rc_ok; char *path = NULL; - if (pcmk__ends_with(file, ".rng")) { + if (g_str_has_suffix(file, ".rng")) { path = pcmk__xml_artefact_path(pcmk__xml_artefact_ns_legacy_rng, file); } else { path = pcmk__xml_artefact_path(pcmk__xml_artefact_ns_legacy_xslt, file); } rc = pcmk__file_contents(path, contents); free(path); return rc; } static void add_schema_file_to_xml(xmlNode *parent, const char *file, GList **already_included) { char *contents = NULL; char *path = NULL; xmlNode *file_node = NULL; GList *includes = NULL; int rc = pcmk_rc_ok; /* If we already included this file, don't do so again. */ if (g_list_find_custom(*already_included, file, (GCompareFunc) strcmp) != NULL) { return; } /* Ensure whatever file we were given has a suffix we know about. If not, * just assume it's an RNG file. */ - if (!pcmk__ends_with(file, ".rng") && !pcmk__ends_with(file, ".xsl")) { + if (!g_str_has_suffix(file, ".rng") && !g_str_has_suffix(file, ".xsl")) { path = crm_strdup_printf("%s.rng", file); } else { path = pcmk__str_copy(file); } rc = read_file_contents(path, &contents); if (rc != pcmk_rc_ok || contents == NULL) { crm_warn("Could not read schema file %s: %s", file, pcmk_rc_str(rc)); free(path); return; } /* Create a new node with the contents of the file * as a CDATA block underneath it. */ file_node = pcmk__xe_create(parent, PCMK__XE_FILE); crm_xml_add(file_node, PCMK_XA_PATH, path); *already_included = g_list_prepend(*already_included, path); xmlAddChild(file_node, xmlNewCDataBlock(parent->doc, (const xmlChar *) contents, strlen(contents))); /* Scan the file for any or nodes and build up * a list of the files they reference. */ external_refs_in_schema(&includes, contents); /* For each referenced file, recurse to add it (and potentially anything it * references, ...) to the XML. */ for (GList *iter = includes; iter != NULL; iter = iter->next) { add_schema_file_to_xml(parent, iter->data, already_included); } free(contents); g_list_free_full(includes, free); } /*! * \internal * \brief Add an XML schema file and all the files it references as children * of a given XML node * * \param[in,out] parent The parent XML node * \param[in] name The schema version to compare against * (for example, "pacemaker-3.1" or "pacemaker-3.1.rng") * \param[in,out] already_included A list of names that have already been added * to the parent node. * * \note The caller is responsible for freeing both the returned list and * the elements of the list */ void pcmk__build_schema_xml_node(xmlNode *parent, const char *name, GList **already_included) { xmlNode *schema_node = pcmk__xe_create(parent, PCMK__XA_SCHEMA); crm_xml_add(schema_node, PCMK_XA_VERSION, name); add_schema_file_to_xml(schema_node, name, already_included); if (schema_node->children == NULL) { // Not needed if empty. May happen if name was invalid, for example. pcmk__xml_free(schema_node); } } /*! * \internal * \brief Return the directory containing any extra schema files that a * Pacemaker Remote node fetched from the cluster */ const char * pcmk__remote_schema_dir(void) { const char *dir = pcmk__env_option(PCMK__ENV_REMOTE_SCHEMA_DIRECTORY); if (pcmk__str_empty(dir)) { return PCMK__REMOTE_SCHEMA_DIR; } return dir; } /*! * \internal * \brief Warn if a given validation schema is deprecated * * \param[in] Schema name to check */ void pcmk__warn_if_schema_deprecated(const char *schema) { /* @COMPAT Disabling validation is deprecated since 2.1.8, but * resource-agents' ocf-shellfuncs (at least as of 4.15.1) uses it */ if (pcmk__str_eq(schema, PCMK_VALUE_NONE, pcmk__str_none)) { pcmk__config_warn("Support for " PCMK_XA_VALIDATE_WITH "='%s' is " "deprecated and will be removed in a future release " "without the possibility of upgrades (manually edit " "to use a supported schema)", schema); } } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START #include gboolean cli_config_update(xmlNode **xml, int *best_version, gboolean to_logs) { int rc = pcmk__update_configured_schema(xml, to_logs); if (best_version != NULL) { const char *name = pcmk__xe_get(*xml, PCMK_XA_VALIDATE_WITH); if (name == NULL) { *best_version = -1; } else { GList *entry = pcmk__get_schema(name); pcmk__schema_t *schema = (entry == NULL)? NULL : entry->data; *best_version = (schema == NULL)? -1 : schema->schema_index; } } return (rc == pcmk_rc_ok)? TRUE: FALSE; } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/common/strings.c b/lib/common/strings.c index 55eacf72be..ed85e4217d 100644 --- a/lib/common/strings.c +++ b/lib/common/strings.c @@ -1,1315 +1,1234 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include // DBL_MIN #include #include #include /*! * \internal * \brief Scan a long long integer from a string * * \param[in] text String to scan * \param[out] result If not NULL, where to store scanned value * \param[in] default_value Value to use if text is NULL or invalid * \param[out] end_text If not NULL, where to store pointer to first * non-integer character * * \return Standard Pacemaker return code (\c pcmk_rc_ok on success, * \c pcmk_rc_bad_input on failed string conversion due to invalid * input, or \c ERANGE if outside long long range) * \note Sets \c errno on error */ static int scan_ll(const char *text, long long *result, long long default_value, char **end_text) { long long local_result = default_value; char *local_end_text = NULL; int rc = pcmk_rc_ok; errno = 0; if (text != NULL) { local_result = strtoll(text, &local_end_text, 10); if (errno == ERANGE) { rc = errno; crm_debug("Integer parsed from '%s' was clipped to %lld", text, local_result); } else if (local_end_text == text) { rc = pcmk_rc_bad_input; local_result = default_value; crm_debug("Could not parse integer from '%s' (using %lld instead): " "No digits found", text, default_value); } else if (errno != 0) { rc = errno; local_result = default_value; crm_debug("Could not parse integer from '%s' (using %lld instead): " "%s", text, default_value, pcmk_rc_str(rc)); } if ((end_text == NULL) && !pcmk__str_empty(local_end_text)) { crm_debug("Characters left over after parsing '%s': '%s'", text, local_end_text); } errno = rc; } if (end_text != NULL) { *end_text = local_end_text; } if (result != NULL) { *result = local_result; } return rc; } /*! * \internal * \brief Scan a long long integer value from a string * * \param[in] text The string to scan (may be NULL) * \param[out] result Where to store result (or NULL to ignore) * \param[in] default_value Value to use if text is NULL or invalid * * \return Standard Pacemaker return code */ int pcmk__scan_ll(const char *text, long long *result, long long default_value) { long long local_result = default_value; int rc = scan_ll(text, &local_result, default_value, NULL); if (result != NULL) { *result = local_result; } return rc; } /*! * \internal * \brief Scan an integer value from a string, constrained to a minimum * * \param[in] text The string to scan (may be NULL) * \param[out] result Where to store result (or NULL to ignore) * \param[in] minimum Value to use as default and minimum * * \return Standard Pacemaker return code * \note If the value is larger than the maximum integer, EOVERFLOW will be * returned and \p result will be set to the maximum integer. */ int pcmk__scan_min_int(const char *text, int *result, int minimum) { int rc; long long result_ll; rc = pcmk__scan_ll(text, &result_ll, (long long) minimum); if (result_ll < (long long) minimum) { crm_warn("Clipped '%s' to minimum acceptable value %d", text, minimum); result_ll = (long long) minimum; } else if (result_ll > INT_MAX) { crm_warn("Clipped '%s' to maximum integer %d", text, INT_MAX); result_ll = (long long) INT_MAX; rc = EOVERFLOW; } if (result != NULL) { *result = (int) result_ll; } return rc; } /*! * \internal * \brief Scan a TCP port number from a string * * \param[in] text The string to scan * \param[out] port Where to store result (or NULL to ignore) * * \return Standard Pacemaker return code * \note \p port will be -1 if \p text is NULL or invalid */ int pcmk__scan_port(const char *text, int *port) { long long port_ll; int rc = pcmk__scan_ll(text, &port_ll, -1LL); if (rc != pcmk_rc_ok) { crm_warn("'%s' is not a valid port: %s", text, pcmk_rc_str(rc)); } else if ((text != NULL) // wasn't default or invalid && ((port_ll < 0LL) || (port_ll > 65535LL))) { crm_warn("Ignoring port specification '%s' " "not in valid range (0-65535)", text); rc = (port_ll < 0LL)? pcmk_rc_before_range : pcmk_rc_after_range; port_ll = -1LL; } if (port != NULL) { *port = (int) port_ll; } return rc; } /*! * \internal * \brief Scan a double-precision floating-point value from a string * * \param[in] text The string to parse * \param[out] result Parsed value on success, or * \c PCMK__PARSE_DBL_DEFAULT on error * \param[in] default_text Default string to parse if \p text is * \c NULL * \param[out] end_text If not \c NULL, where to store a pointer * to the position immediately after the * value * * \return Standard Pacemaker return code (\c pcmk_rc_ok on success, * \c EINVAL on failed string conversion due to invalid input, * \c EOVERFLOW on arithmetic overflow, \c pcmk_rc_underflow * on arithmetic underflow, or \c errno from \c strtod() on * other parse errors) */ int pcmk__scan_double(const char *text, double *result, const char *default_text, char **end_text) { int rc = pcmk_rc_ok; char *local_end_text = NULL; pcmk__assert(result != NULL); *result = PCMK__PARSE_DBL_DEFAULT; text = (text != NULL) ? text : default_text; if (text == NULL) { rc = EINVAL; crm_debug("No text and no default conversion value supplied"); } else { errno = 0; *result = strtod(text, &local_end_text); if (errno == ERANGE) { /* * Overflow: strtod() returns +/- HUGE_VAL and sets errno to * ERANGE * * Underflow: strtod() returns "a value whose magnitude is * no greater than the smallest normalized * positive" double. Whether ERANGE is set is * implementation-defined. */ const char *over_under; if (QB_ABS(*result) > DBL_MIN) { rc = EOVERFLOW; over_under = "over"; } else { rc = pcmk_rc_underflow; over_under = "under"; } crm_debug("Floating-point value parsed from '%s' would %sflow " "(using %g instead)", text, over_under, *result); } else if (errno != 0) { rc = errno; // strtod() set *result = 0 on parse failure *result = PCMK__PARSE_DBL_DEFAULT; crm_debug("Could not parse floating-point value from '%s' (using " "%.1f instead): %s", text, PCMK__PARSE_DBL_DEFAULT, pcmk_rc_str(rc)); } else if (local_end_text == text) { // errno == 0, but nothing was parsed rc = EINVAL; *result = PCMK__PARSE_DBL_DEFAULT; crm_debug("Could not parse floating-point value from '%s' (using " "%.1f instead): No digits found", text, PCMK__PARSE_DBL_DEFAULT); } else if (QB_ABS(*result) <= DBL_MIN) { /* * errno == 0 and text was parsed, but value might have * underflowed. * * ERANGE might not be set for underflow. Check magnitude * of *result, but also make sure the input number is not * actually zero (0 <= DBL_MIN is not underflow). * * This check must come last. A parse failure in strtod() * also sets *result == 0, so a parse failure would match * this test condition prematurely. */ for (const char *p = text; p != local_end_text; p++) { if (strchr("0.eE", *p) == NULL) { rc = pcmk_rc_underflow; crm_debug("Floating-point value parsed from '%s' would " "underflow (using %g instead)", text, *result); break; } } } else { crm_trace("Floating-point value parsed successfully from " "'%s': %g", text, *result); } if ((end_text == NULL) && !pcmk__str_empty(local_end_text)) { crm_debug("Characters left over after parsing '%s': '%s'", text, local_end_text); } } if (end_text != NULL) { *end_text = local_end_text; } return rc; } /*! * \internal * \brief Parse a guint from a string stored in a hash table * * \param[in] table Hash table to search * \param[in] key Hash table key to use to retrieve string * \param[in] default_val What to use if key has no entry in table * \param[out] result If not NULL, where to store parsed integer * * \return Standard Pacemaker return code */ int pcmk__guint_from_hash(GHashTable *table, const char *key, guint default_val, guint *result) { const char *value; long long value_ll; int rc = pcmk_rc_ok; CRM_CHECK((table != NULL) && (key != NULL), return EINVAL); if (result != NULL) { *result = default_val; } value = g_hash_table_lookup(table, key); if (value == NULL) { return pcmk_rc_ok; } rc = pcmk__scan_ll(value, &value_ll, 0LL); if (rc != pcmk_rc_ok) { crm_warn("Using default (%u) for %s because '%s' is not a " "valid integer: %s", default_val, key, value, pcmk_rc_str(rc)); return rc; } if ((value_ll < 0) || (value_ll > G_MAXUINT)) { crm_warn("Using default (%u) for %s because '%s' is not in valid range", default_val, key, value); return ERANGE; } if (result != NULL) { *result = (guint) value_ll; } return pcmk_rc_ok; } /*! * \brief Parse a time+units string and return milliseconds equivalent * * \param[in] input String with a nonnegative number and optional unit * (optionally with whitespace before and/or after the * number). If missing, the unit defaults to seconds. * * \return Milliseconds corresponding to string expression, or * \c PCMK__PARSE_INT_DEFAULT on error */ long long crm_get_msec(const char *input) { char *units = NULL; // Do not free; will point to part of input long long multiplier = 1000; long long divisor = 1; long long msec = PCMK__PARSE_INT_DEFAULT; int rc = pcmk_rc_ok; if (input == NULL) { return PCMK__PARSE_INT_DEFAULT; } // Skip initial whitespace while (isspace(*input)) { input++; } rc = scan_ll(input, &msec, PCMK__PARSE_INT_DEFAULT, &units); if ((rc == ERANGE) && (msec > 0)) { crm_warn("'%s' will be clipped to %lld", input, msec); } else if ((rc != pcmk_rc_ok) || (msec < 0)) { crm_warn("'%s' is not a valid time duration: %s", input, ((rc == pcmk_rc_ok)? "Negative" : pcmk_rc_str(rc))); return PCMK__PARSE_INT_DEFAULT; } /* If the number is a decimal, scan_ll() reads only the integer part. Skip * any remaining digits or decimal characters. * * @COMPAT Well-formed and malformed decimals are both accepted inputs. For * example, "3.14 ms" and "3.1.4 ms" are treated the same as "3ms" and * parsed successfully. At a compatibility break, decide if this is still * desired. */ while (isdigit(*units) || (*units == '.')) { units++; } // Skip any additional whitespace after the number while (isspace(*units)) { units++; } /* @COMPAT Use exact comparisons. Currently, we match too liberally, and the * second strncasecmp() in each case is redundant. */ if ((*units == '\0') || (strncasecmp(units, "s", 1) == 0) || (strncasecmp(units, "sec", 3) == 0)) { multiplier = 1000; divisor = 1; } else if ((strncasecmp(units, "ms", 2) == 0) || (strncasecmp(units, "msec", 4) == 0)) { multiplier = 1; divisor = 1; } else if ((strncasecmp(units, "us", 2) == 0) || (strncasecmp(units, "usec", 4) == 0)) { multiplier = 1; divisor = 1000; } else if ((strncasecmp(units, "m", 1) == 0) || (strncasecmp(units, "min", 3) == 0)) { multiplier = 60 * 1000; divisor = 1; } else if ((strncasecmp(units, "h", 1) == 0) || (strncasecmp(units, "hr", 2) == 0)) { multiplier = 60 * 60 * 1000; divisor = 1; } else { // Invalid units return PCMK__PARSE_INT_DEFAULT; } // Apply units, capping at LLONG_MAX if (msec > (LLONG_MAX / multiplier)) { return LLONG_MAX; } return (msec * multiplier) / divisor; } /*! * \brief Parse milliseconds from a Pacemaker interval specification * * \param[in] input Pacemaker time interval specification (a bare number * of seconds; a number with a unit, optionally with * whitespace before and/or after the number; or an ISO * 8601 duration) * \param[out] result_ms Where to store milliseconds equivalent of \p input on * success (limited to the range of an unsigned integer), * or 0 if \p input is \c NULL or invalid * * \return Standard Pacemaker return code (specifically, \c pcmk_rc_ok if * \p input is valid or \c NULL, and \c EINVAL otherwise) */ int pcmk_parse_interval_spec(const char *input, guint *result_ms) { long long msec = PCMK__PARSE_INT_DEFAULT; int rc = pcmk_rc_ok; if (input == NULL) { msec = 0; goto done; } if (input[0] == 'P') { crm_time_t *period_s = crm_time_parse_duration(input); if (period_s != NULL) { msec = crm_time_get_seconds(period_s); msec = QB_MIN(msec, G_MAXUINT / 1000) * 1000; crm_time_free(period_s); } } else { msec = crm_get_msec(input); } if (msec < 0) { crm_warn("Using 0 instead of invalid interval specification '%s'", input); msec = 0; rc = EINVAL; } done: if (result_ms != NULL) { *result_ms = (msec >= G_MAXUINT)? G_MAXUINT : (guint) msec; } return rc; } gboolean crm_is_true(const char *s) { gboolean ret = FALSE; return (crm_str_to_boolean(s, &ret) < 0)? FALSE : ret; } int crm_str_to_boolean(const char *s, int *ret) { if (s == NULL) { return -1; } if (pcmk__strcase_any_of(s, PCMK_VALUE_TRUE, "on", "yes", "y", "1", NULL)) { if (ret != NULL) { *ret = TRUE; } return 1; } if (pcmk__strcase_any_of(s, PCMK_VALUE_FALSE, PCMK_VALUE_OFF, "no", "n", "0", NULL)) { if (ret != NULL) { *ret = FALSE; } return 1; } return -1; } -/*! - * \internal - * \brief Replace any trailing newlines in a string with \0's - * - * \param[in,out] str String to trim - * - * \return \p str - */ -char * -pcmk__trim(char *str) -{ - int len; - - if (str == NULL) { - return str; - } - - for (len = strlen(str) - 1; len >= 0 && str[len] == '\n'; len--) { - str[len] = '\0'; - } - - return str; -} - -/*! - * \brief Check whether a string starts with a certain sequence - * - * \param[in] str String to check - * \param[in] prefix Sequence to match against beginning of \p str - * - * \return \c true if \p str begins with match, \c false otherwise - * \note This is equivalent to !strncmp(s, prefix, strlen(prefix)) - * but is likely less efficient when prefix is a string literal - * if the compiler optimizes away the strlen() at compile time, - * and more efficient otherwise. - */ -bool -pcmk__starts_with(const char *str, const char *prefix) -{ - const char *s = str; - const char *p = prefix; - - if (!s || !p) { - return false; - } - while (*s && *p) { - if (*s++ != *p++) { - return false; - } - } - return (*p == 0); -} - -static inline bool -ends_with(const char *s, const char *match, bool as_extension) -{ - if (pcmk__str_empty(match)) { - return true; - } else if (s == NULL) { - return false; - } else { - size_t slen, mlen; - - /* Besides as_extension, we could also check - !strchr(&match[1], match[0]) but that would be inefficient. - */ - if (as_extension) { - s = strrchr(s, match[0]); - return (s == NULL)? false : !strcmp(s, match); - } - - mlen = strlen(match); - slen = strlen(s); - return ((slen >= mlen) && !strcmp(s + slen - mlen, match)); - } -} - -/*! - * \internal - * \brief Check whether a string ends with a certain sequence - * - * \param[in] s String to check - * \param[in] match Sequence to match against end of \p s - * - * \return \c true if \p s ends case-sensitively with match, \c false otherwise - * \note pcmk__ends_with_ext() can be used if the first character of match - * does not recur in match. - */ -bool -pcmk__ends_with(const char *s, const char *match) -{ - return ends_with(s, match, false); -} - -/*! - * \internal - * \brief Check whether a string ends with a certain "extension" - * - * \param[in] s String to check - * \param[in] match Extension to match against end of \p s, that is, - * its first character must not occur anywhere - * in the rest of that very sequence (example: file - * extension where the last dot is its delimiter, - * e.g., ".html"); incorrect results may be - * returned otherwise. - * - * \return \c true if \p s ends (verbatim, i.e., case sensitively) - * with "extension" designated as \p match (including empty - * string), \c false otherwise - * - * \note Main incentive to prefer this function over \c pcmk__ends_with() - * where possible is the efficiency (at the cost of added - * restriction on \p match as stated; the complexity class - * remains the same, though: BigO(M+N) vs. BigO(M+2N)). - */ -bool -pcmk__ends_with_ext(const char *s, const char *match) -{ - return ends_with(s, match, true); -} - /*! * \internal * \brief Create a hash of a string suitable for use with GHashTable * * \param[in] v String to hash * * \return A hash of \p v compatible with g_str_hash() before glib 2.28 * \note glib changed their hash implementation: * * https://gitlab.gnome.org/GNOME/glib/commit/354d655ba8a54b754cb5a3efb42767327775696c * * Note that the new g_str_hash is presumably a *better* hash (it's actually * a correct implementation of DJB's hash), but we need to preserve existing * behaviour, because the hash key ultimately determines the "sort" order * when iterating through GHashTables, which affects allocation of scores to * clone instances when iterating through allowed nodes. It (somehow) also * appears to have some minor impact on the ordering of a few pseudo_event IDs * in the transition graph. */ static guint pcmk__str_hash(gconstpointer v) { const signed char *p; guint32 h = 0; for (p = v; *p != '\0'; p++) h = (h << 5) - h + *p; return h; } /*! * \internal * \brief Create a hash table with case-sensitive strings as keys * * \param[in] key_destroy_func Function to free a key * \param[in] value_destroy_func Function to free a value * * \return Newly allocated hash table * \note It is the caller's responsibility to free the result, using * g_hash_table_destroy(). */ GHashTable * pcmk__strkey_table(GDestroyNotify key_destroy_func, GDestroyNotify value_destroy_func) { return g_hash_table_new_full(pcmk__str_hash, g_str_equal, key_destroy_func, value_destroy_func); } /*! * \internal * \brief Insert string copies into a hash table as key and value * * \param[in,out] table Hash table to add to * \param[in] name String to add a copy of as key * \param[in] value String to add a copy of as value * * \note This asserts on invalid arguments or memory allocation failure. */ void pcmk__insert_dup(GHashTable *table, const char *name, const char *value) { pcmk__assert((table != NULL) && (name != NULL)); g_hash_table_insert(table, pcmk__str_copy(name), pcmk__str_copy(value)); } /* used with hash tables where case does not matter */ static gboolean pcmk__strcase_equal(gconstpointer a, gconstpointer b) { return pcmk__str_eq((const char *)a, (const char *)b, pcmk__str_casei); } static guint pcmk__strcase_hash(gconstpointer v) { const signed char *p; guint32 h = 0; for (p = v; *p != '\0'; p++) h = (h << 5) - h + g_ascii_tolower(*p); return h; } /*! * \internal * \brief Create a hash table with case-insensitive strings as keys * * \param[in] key_destroy_func Function to free a key * \param[in] value_destroy_func Function to free a value * * \return Newly allocated hash table * \note It is the caller's responsibility to free the result, using * g_hash_table_destroy(). */ GHashTable * pcmk__strikey_table(GDestroyNotify key_destroy_func, GDestroyNotify value_destroy_func) { return g_hash_table_new_full(pcmk__strcase_hash, pcmk__strcase_equal, key_destroy_func, value_destroy_func); } static void copy_str_table_entry(gpointer key, gpointer value, gpointer user_data) { if (key && value && user_data) { pcmk__insert_dup((GHashTable *) user_data, (const char *) key, (const char *) value); } } /*! * \internal * \brief Copy a hash table that uses dynamically allocated strings * * \param[in,out] old_table Hash table to duplicate * * \return New hash table with copies of everything in \p old_table * \note This assumes the hash table uses dynamically allocated strings -- that * is, both the key and value free functions are free(). */ GHashTable * pcmk__str_table_dup(GHashTable *old_table) { GHashTable *new_table = NULL; if (old_table) { new_table = pcmk__strkey_table(free, free); g_hash_table_foreach(old_table, copy_str_table_entry, new_table); } return new_table; } /*! * \internal * \brief Add a word to a string list of words * * \param[in,out] list Pointer to current string list (may not be \p NULL) * \param[in] init_size \p list will be initialized to at least this size, * if it needs initialization (if 0, use GLib's default * initial string size) * \param[in] word String to add to \p list (\p list will be * unchanged if this is \p NULL or the empty string) * \param[in] separator String to separate words in \p list * (a space will be used if this is NULL) * * \note \p word may contain \p separator, though that would be a bad idea if * the string needs to be parsed later. */ void pcmk__add_separated_word(GString **list, size_t init_size, const char *word, const char *separator) { pcmk__assert(list != NULL); if (pcmk__str_empty(word)) { return; } if (*list == NULL) { if (init_size > 0) { *list = g_string_sized_new(init_size); } else { *list = g_string_new(NULL); } } if ((*list)->len == 0) { // Don't add a separator before the first word in the list separator = ""; } else if (separator == NULL) { // Default to space-separated separator = " "; } g_string_append(*list, separator); g_string_append(*list, word); } /*! * \internal * \brief Compress data * * \param[in] data Data to compress * \param[in] length Number of characters of data to compress * \param[in] max Maximum size of compressed data (or 0 to estimate) * \param[out] result Where to store newly allocated compressed result * \param[out] result_len Where to store actual compressed length of result * * \return Standard Pacemaker return code */ int pcmk__compress(const char *data, unsigned int length, unsigned int max, char **result, unsigned int *result_len) { int rc; char *compressed = NULL; char *uncompressed = strdup(data); #ifdef CLOCK_MONOTONIC struct timespec after_t; struct timespec before_t; #endif if (max == 0) { max = (length * 1.01) + 601; // Size guaranteed to hold result } #ifdef CLOCK_MONOTONIC clock_gettime(CLOCK_MONOTONIC, &before_t); #endif compressed = pcmk__assert_alloc((size_t) max, sizeof(char)); *result_len = max; rc = BZ2_bzBuffToBuffCompress(compressed, result_len, uncompressed, length, CRM_BZ2_BLOCKS, 0, CRM_BZ2_WORK); rc = pcmk__bzlib2rc(rc); free(uncompressed); if (rc != pcmk_rc_ok) { crm_err("Compression of %d bytes failed: %s " QB_XS " rc=%d", length, pcmk_rc_str(rc), rc); free(compressed); return rc; } #ifdef CLOCK_MONOTONIC clock_gettime(CLOCK_MONOTONIC, &after_t); crm_trace("Compressed %d bytes into %d (ratio %d:1) in %.0fms", length, *result_len, length / (*result_len), (after_t.tv_sec - before_t.tv_sec) * 1000 + (after_t.tv_nsec - before_t.tv_nsec) / 1e6); #else crm_trace("Compressed %d bytes into %d (ratio %d:1)", length, *result_len, length / (*result_len)); #endif *result = compressed; return pcmk_rc_ok; } char * crm_strdup_printf(char const *format, ...) { va_list ap; int len = 0; char *string = NULL; va_start(ap, format); len = vasprintf(&string, format, ap); pcmk__assert(len > 0); va_end(ap); return string; } +/*! + * \internal + * \brief Parse a range specification string + * + * A valid range specification string can be in any of the following forms, + * where \c "X", \c "Y", and \c "Z" are nonnegative integers that fit into a + * long long variable: + * * "X-Y" + * * "X-" + * * "-Y" + * * "Z" + * + * In the list above, \c "X" is the start value and \c "Y" is the end value of + * the range. Either the start value or the end value, but not both, can be + * empty. \c "Z", a single integer with no \c '-' character, is both the start + * value and the end value of its range. + * + * If the start value or end value is empty, then the parsed result stored in + * \p *start or \p *end (respectively) is \c PCMK__PARSE_INT_DEFAULT after a + * successful parse. + * + * If the specification string consists of only a single number, then the same + * value is stored in both \p *start and \p *end on a successful parse. + * + * \param[in] text String to parse + * \param[out] start Where to store start value (can be \c NULL) + * \param[out] end Where to store end value (can be \c NULL) + * + * \return Standard Pacemaker return code + * + * \note The values stored in \p *start and \p *end are undefined if the return + * value is not \c pcmk_rc_ok. + */ int -pcmk__parse_ll_range(const char *srcstring, long long *start, long long *end) +pcmk__parse_ll_range(const char *text, long long *start, long long *end) { - char *remainder = NULL; int rc = pcmk_rc_ok; + long long local_start = 0; + long long local_end = 0; + gchar **split = NULL; + guint length = 0; + const gchar *start_s = NULL; + const gchar *end_s = NULL; + + // Do not free + char *remainder = NULL; - pcmk__assert((start != NULL) && (end != NULL)); - + if (start == NULL) { + start = &local_start; + } + if (end == NULL) { + end = &local_end; + } *start = PCMK__PARSE_INT_DEFAULT; - // cppcheck doesn't understand the above pcmk__assert line - // cppcheck-suppress ctunullpointer *end = PCMK__PARSE_INT_DEFAULT; - crm_trace("Attempting to decode: [%s]", srcstring); - if (pcmk__str_eq(srcstring, "", pcmk__str_null_matches)) { - return ENODATA; - } else if (pcmk__str_eq(srcstring, "-", pcmk__str_none)) { - return pcmk_rc_bad_input; + if (pcmk__str_empty(text)) { + rc = ENODATA; + goto done; } - /* String starts with a dash, so this is either a range with - * no beginning or garbage. - * */ - if (*srcstring == '-') { - int rc = scan_ll(srcstring+1, end, PCMK__PARSE_INT_DEFAULT, &remainder); + split = g_strsplit(text, "-", 2); + length = g_strv_length(split); + start_s = split[0]; + if (length == 2) { + end_s = split[1]; + } - if ((rc == pcmk_rc_ok) && (*remainder != '\0')) { + if (pcmk__str_empty(start_s) && pcmk__str_empty(end_s)) { + rc = pcmk_rc_bad_input; + goto done; + } + + if (!pcmk__str_empty(start_s)) { + rc = scan_ll(start_s, start, PCMK__PARSE_INT_DEFAULT, &remainder); + if (rc != pcmk_rc_ok) { + goto done; + } + if (!pcmk__str_empty(remainder)) { rc = pcmk_rc_bad_input; + goto done; } - return rc; } - rc = scan_ll(srcstring, start, PCMK__PARSE_INT_DEFAULT, &remainder); - if (rc != pcmk_rc_ok) { - return rc; + if (length == 1) { + // String contains only a single number, which is both start and end + *end = *start; + goto done; } - if (*remainder && *remainder == '-') { - if (*(remainder+1)) { - char *more_remainder = NULL; - int rc = scan_ll(remainder+1, end, PCMK__PARSE_INT_DEFAULT, - &more_remainder); + if (!pcmk__str_empty(end_s)) { + rc = scan_ll(end_s, end, PCMK__PARSE_INT_DEFAULT, &remainder); - if (rc != pcmk_rc_ok) { - return rc; - } else if (*more_remainder != '\0') { - return pcmk_rc_bad_input; - } + if ((rc == pcmk_rc_ok) && !pcmk__str_empty(remainder)) { + rc = pcmk_rc_bad_input; } - } else if (*remainder && *remainder != '-') { - *start = PCMK__PARSE_INT_DEFAULT; - return pcmk_rc_bad_input; - } else { - /* The input string contained only one number. Set start and end - * to the same value and return pcmk_rc_ok. This gives the caller - * a way to tell this condition apart from a range with no end. - */ - *end = *start; } - return pcmk_rc_ok; +done: + g_strfreev(split); + return rc; } /*! * \internal * \brief Find a string in a list of strings * * \note This function takes the same flags and has the same behavior as * pcmk__str_eq(). * * \note No matter what input string or flags are provided, an empty * list will always return FALSE. * * \param[in] s String to search for * \param[in] lst List to search * \param[in] flags A bitfield of pcmk__str_flags to modify operation * * \return \c TRUE if \p s is in \p lst, or \c FALSE otherwise */ gboolean pcmk__str_in_list(const gchar *s, const GList *lst, uint32_t flags) { for (const GList *ele = lst; ele != NULL; ele = ele->next) { if (pcmk__str_eq(s, ele->data, flags)) { return TRUE; } } return FALSE; } static bool str_any_of(const char *s, va_list args, uint32_t flags) { if (s == NULL) { return false; } while (1) { const char *ele = va_arg(args, const char *); if (ele == NULL) { break; } else if (pcmk__str_eq(s, ele, flags)) { return true; } } return false; } /*! * \internal * \brief Is a string a member of a list of strings? * * \param[in] s String to search for in \p ... * \param[in] ... Strings to compare \p s against. The final string * must be NULL. * * \note The comparison is done case-insensitively. The function name is * meant to be reminiscent of strcasecmp. * * \return \c true if \p s is in \p ..., or \c false otherwise */ bool pcmk__strcase_any_of(const char *s, ...) { va_list ap; bool rc; va_start(ap, s); rc = str_any_of(s, ap, pcmk__str_casei); va_end(ap); return rc; } /*! * \internal * \brief Is a string a member of a list of strings? * * \param[in] s String to search for in \p ... * \param[in] ... Strings to compare \p s against. The final string * must be NULL. * * \note The comparison is done taking case into account. * * \return \c true if \p s is in \p ..., or \c false otherwise */ bool pcmk__str_any_of(const char *s, ...) { va_list ap; bool rc; va_start(ap, s); rc = str_any_of(s, ap, pcmk__str_none); va_end(ap); return rc; } /*! * \internal * \brief Sort strings, with numeric portions sorted numerically * * Sort two strings case-insensitively like strcasecmp(), but with any numeric * portions of the string sorted numerically. This is particularly useful for * node names (for example, "node10" will sort higher than "node9" but lower * than "remotenode9"). * * \param[in] s1 First string to compare (must not be NULL) * \param[in] s2 Second string to compare (must not be NULL) * * \retval -1 \p s1 comes before \p s2 * \retval 0 \p s1 and \p s2 are equal * \retval 1 \p s1 comes after \p s2 */ int pcmk__numeric_strcasecmp(const char *s1, const char *s2) { pcmk__assert((s1 != NULL) && (s2 != NULL)); while (*s1 && *s2) { if (isdigit(*s1) && isdigit(*s2)) { // If node names contain a number, sort numerically char *end1 = NULL; char *end2 = NULL; long num1 = strtol(s1, &end1, 10); long num2 = strtol(s2, &end2, 10); // allow ordering e.g. 007 > 7 size_t len1 = end1 - s1; size_t len2 = end2 - s2; if (num1 < num2) { return -1; } else if (num1 > num2) { return 1; } else if (len1 < len2) { return -1; } else if (len1 > len2) { return 1; } s1 = end1; s2 = end2; } else { // Compare non-digits case-insensitively int lower1 = tolower(*s1); int lower2 = tolower(*s2); if (lower1 < lower2) { return -1; } else if (lower1 > lower2) { return 1; } ++s1; ++s2; } } if (!*s1 && *s2) { return -1; } else if (*s1 && !*s2) { return 1; } return 0; } /*! * \internal * \brief Sort strings. * * This is your one-stop function for string comparison. By default, this * function works like \p g_strcmp0. That is, like \p strcmp but a \p NULL * string sorts before a non-NULL string. * * The \p pcmk__str_none flag produces the default behavior. Behavior can be * changed with various flags: * * - \p pcmk__str_regex - The second string is a regular expression that the * first string will be matched against. * - \p pcmk__str_casei - By default, comparisons are done taking case into * account. This flag makes comparisons case- * insensitive. This can be combined with * \p pcmk__str_regex. * - \p pcmk__str_null_matches - If one string is \p NULL and the other is not, * still return \p 0. * - \p pcmk__str_star_matches - If one string is \p "*" and the other is not, * still return \p 0. * * \param[in] s1 First string to compare * \param[in] s2 Second string to compare, or a regular expression to * match if \p pcmk__str_regex is set * \param[in] flags A bitfield of \p pcmk__str_flags to modify operation * * \retval negative \p s1 is \p NULL or comes before \p s2 * \retval 0 \p s1 and \p s2 are equal, or \p s1 is found in \p s2 if * \c pcmk__str_regex is set * \retval positive \p s2 is \p NULL or \p s1 comes after \p s2, or \p s2 * is an invalid regular expression, or \p s1 was not found * in \p s2 if \p pcmk__str_regex is set. */ int pcmk__strcmp(const char *s1, const char *s2, uint32_t flags) { /* If this flag is set, the second string is a regex. */ if (pcmk_is_set(flags, pcmk__str_regex)) { regex_t r_patt; int reg_flags = REG_EXTENDED | REG_NOSUB; int regcomp_rc = 0; int rc = 0; if (s1 == NULL || s2 == NULL) { return 1; } if (pcmk_is_set(flags, pcmk__str_casei)) { reg_flags |= REG_ICASE; } regcomp_rc = regcomp(&r_patt, s2, reg_flags); if (regcomp_rc != 0) { rc = 1; crm_err("Bad regex '%s' for update: %s", s2, strerror(regcomp_rc)); } else { rc = regexec(&r_patt, s1, 0, NULL, 0); regfree(&r_patt); if (rc != 0) { rc = 1; } } return rc; } /* If the strings are the same pointer, return 0 immediately. */ if (s1 == s2) { return 0; } /* If this flag is set, return 0 if either (or both) of the input strings * are NULL. If neither one is NULL, we need to continue and compare * them normally. */ if (pcmk_is_set(flags, pcmk__str_null_matches)) { if (s1 == NULL || s2 == NULL) { return 0; } } /* Handle the cases where one is NULL and the str_null_matches flag is not set. * A NULL string always sorts to the beginning. */ if (s1 == NULL) { return -1; } else if (s2 == NULL) { return 1; } /* If this flag is set, return 0 if either (or both) of the input strings * are "*". If neither one is, we need to continue and compare them * normally. */ if (pcmk_is_set(flags, pcmk__str_star_matches)) { if (strcmp(s1, "*") == 0 || strcmp(s2, "*") == 0) { return 0; } } if (pcmk_is_set(flags, pcmk__str_casei)) { return strcasecmp(s1, s2); } else { return strcmp(s1, s2); } } /*! * \internal * \brief Copy a string, asserting on failure * * \param[in] file File where \p function is located * \param[in] function Calling function * \param[in] line Line within \p file * \param[in] str String to copy (can be \c NULL) * * \return Newly allocated copy of \p str, or \c NULL if \p str is \c NULL * * \note The caller is responsible for freeing the return value using \c free(). */ char * pcmk__str_copy_as(const char *file, const char *function, uint32_t line, const char *str) { if (str != NULL) { char *result = strdup(str); if (result == NULL) { crm_abort(file, function, line, "Out of memory", FALSE, TRUE); crm_exit(CRM_EX_OSERR); } return result; } return NULL; } /*! * \internal * \brief Update a dynamically allocated string with a new value * * Given a dynamically allocated string and a new value for it, if the string * is different from the new value, free the string and replace it with either a * newly allocated duplicate of the value or NULL as appropriate. * * \param[in,out] str Pointer to dynamically allocated string * \param[in] value New value to duplicate (or NULL) * * \note The caller remains responsibile for freeing \p *str. */ void pcmk__str_update(char **str, const char *value) { if ((str != NULL) && !pcmk__str_eq(*str, value, pcmk__str_none)) { free(*str); *str = pcmk__str_copy(value); } } /*! * \internal * \brief Append a list of strings to a destination \p GString * * \param[in,out] buffer Where to append the strings (must not be \p NULL) * \param[in] ... A NULL-terminated list of strings * * \note This tends to be more efficient than a single call to * \p g_string_append_printf(). */ void pcmk__g_strcat(GString *buffer, ...) { va_list ap; pcmk__assert(buffer != NULL); va_start(ap, buffer); while (true) { const char *ele = va_arg(ap, const char *); if (ele == NULL) { break; } g_string_append(buffer, ele); } va_end(ap); } diff --git a/lib/common/tests/strings/Makefile.am b/lib/common/tests/strings/Makefile.am index e61f41f51d..0f007fcd94 100644 --- a/lib/common/tests/strings/Makefile.am +++ b/lib/common/tests/strings/Makefile.am @@ -1,41 +1,38 @@ # -# Copyright 2020-2024 the Pacemaker project contributors +# Copyright 2020-2025 the Pacemaker project contributors # # The version control history for this file may have further details. # # This source code is licensed under the GNU General Public License version 2 # or later (GPLv2+) WITHOUT ANY WARRANTY. # include $(top_srcdir)/mk/common.mk include $(top_srcdir)/mk/tap.mk include $(top_srcdir)/mk/unittest.mk # Add "_test" to the end of all test program names to simplify .gitignore. check_PROGRAMS = crm_get_msec_test \ crm_is_true_test \ crm_str_to_boolean_test \ pcmk__add_word_test \ pcmk__btoa_test \ pcmk__compress_test \ - pcmk__ends_with_test \ pcmk__g_strcat_test \ pcmk__guint_from_hash_test \ pcmk__numeric_strcasecmp_test \ pcmk__parse_ll_range_test \ pcmk__s_test \ pcmk__scan_double_test \ pcmk__scan_ll_test \ pcmk__scan_min_int_test \ pcmk__scan_port_test \ - pcmk__starts_with_test \ pcmk__str_any_of_test \ pcmk__str_in_list_test \ pcmk__str_table_dup_test \ pcmk__str_update_test \ pcmk__strcmp_test \ pcmk__strkey_table_test \ - pcmk__strikey_table_test \ - pcmk__trim_test + pcmk__strikey_table_test TESTS = $(check_PROGRAMS) diff --git a/lib/common/tests/strings/pcmk__ends_with_test.c b/lib/common/tests/strings/pcmk__ends_with_test.c deleted file mode 100644 index 75035714fa..0000000000 --- a/lib/common/tests/strings/pcmk__ends_with_test.c +++ /dev/null @@ -1,57 +0,0 @@ -/* - * Copyright 2021 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU General Public License version 2 - * or later (GPLv2+) WITHOUT ANY WARRANTY. - */ - -#include - -#include - -static void -bad_input(void **state) { - assert_false(pcmk__ends_with(NULL, "xyz")); - - assert_true(pcmk__ends_with(NULL, NULL)); - assert_true(pcmk__ends_with(NULL, "")); - assert_true(pcmk__ends_with("", NULL)); - assert_true(pcmk__ends_with("", "")); - assert_true(pcmk__ends_with("abc", NULL)); - assert_true(pcmk__ends_with("abc", "")); -} - -static void -ends_with(void **state) { - assert_true(pcmk__ends_with("abc", "abc")); - assert_true(pcmk__ends_with("abc", "bc")); - assert_true(pcmk__ends_with("abc", "c")); - assert_true(pcmk__ends_with("abcbc", "bc")); - - assert_false(pcmk__ends_with("abc", "def")); - assert_false(pcmk__ends_with("abc", "defg")); - assert_false(pcmk__ends_with("abc", "bcd")); - assert_false(pcmk__ends_with("abc", "ab")); - - assert_false(pcmk__ends_with("abc", "BC")); -} - -static void -ends_with_ext(void **state) { - assert_true(pcmk__ends_with_ext("ab.c", ".c")); - assert_true(pcmk__ends_with_ext("ab.cb.c", ".c")); - - assert_false(pcmk__ends_with_ext("ab.c", ".def")); - assert_false(pcmk__ends_with_ext("ab.c", ".defg")); - assert_false(pcmk__ends_with_ext("ab.c", ".cd")); - assert_false(pcmk__ends_with_ext("ab.c", "ab")); - - assert_false(pcmk__ends_with_ext("ab.c", ".C")); -} - -PCMK__UNIT_TEST(NULL, NULL, - cmocka_unit_test(bad_input), - cmocka_unit_test(ends_with), - cmocka_unit_test(ends_with_ext)) diff --git a/lib/common/tests/strings/pcmk__parse_ll_range_test.c b/lib/common/tests/strings/pcmk__parse_ll_range_test.c index 35c21672b8..b6f20febe4 100644 --- a/lib/common/tests/strings/pcmk__parse_ll_range_test.c +++ b/lib/common/tests/strings/pcmk__parse_ll_range_test.c @@ -1,120 +1,118 @@ /* - * Copyright 2020-2024 the Pacemaker project contributors + * Copyright 2020-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include static void empty_input_string(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range(NULL, &start, &end), ENODATA); assert_int_equal(pcmk__parse_ll_range("", &start, &end), ENODATA); } static void null_input_variables(void **state) { long long start, end; - pcmk__assert_asserts(pcmk__parse_ll_range("1234", NULL, &end)); - pcmk__assert_asserts(pcmk__parse_ll_range("1234", &start, NULL)); + assert_int_equal(pcmk__parse_ll_range("1234", NULL, NULL), pcmk_rc_ok); + + assert_int_equal(pcmk__parse_ll_range("1234", &start, NULL), pcmk_rc_ok); + assert_int_equal(start, 1234); + + assert_int_equal(pcmk__parse_ll_range("1234", NULL, &end), pcmk_rc_ok); + assert_int_equal(end, 1234); } static void missing_separator(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("1234", &start, &end), pcmk_rc_ok); assert_int_equal(start, 1234); assert_int_equal(end, 1234); } static void only_separator(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("-", &start, &end), pcmk_rc_bad_input); - assert_int_equal(start, PCMK__PARSE_INT_DEFAULT); - assert_int_equal(end, PCMK__PARSE_INT_DEFAULT); } static void no_range_end(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("2000-", &start, &end), pcmk_rc_ok); assert_int_equal(start, 2000); assert_int_equal(end, PCMK__PARSE_INT_DEFAULT); } static void no_range_start(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("-2020", &start, &end), pcmk_rc_ok); assert_int_equal(start, PCMK__PARSE_INT_DEFAULT); assert_int_equal(end, 2020); } static void range_start_and_end(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("2000-2020", &start, &end), pcmk_rc_ok); assert_int_equal(start, 2000); assert_int_equal(end, 2020); assert_int_equal(pcmk__parse_ll_range("2000-2020-2030", &start, &end), pcmk_rc_bad_input); } static void garbage(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("2000x-", &start, &end), pcmk_rc_bad_input); - assert_int_equal(start, PCMK__PARSE_INT_DEFAULT); - assert_int_equal(end, PCMK__PARSE_INT_DEFAULT); - assert_int_equal(pcmk__parse_ll_range("-x2000", &start, &end), pcmk_rc_bad_input); - assert_int_equal(start, PCMK__PARSE_INT_DEFAULT); - assert_int_equal(end, PCMK__PARSE_INT_DEFAULT); } static void strtoll_errors(void **state) { long long start, end; assert_int_equal(pcmk__parse_ll_range("20000000000000000000-", &start, &end), ERANGE); assert_int_equal(pcmk__parse_ll_range("100-20000000000000000000", &start, &end), ERANGE); } PCMK__UNIT_TEST(NULL, NULL, cmocka_unit_test(empty_input_string), cmocka_unit_test(null_input_variables), cmocka_unit_test(missing_separator), cmocka_unit_test(only_separator), cmocka_unit_test(no_range_end), cmocka_unit_test(no_range_start), cmocka_unit_test(range_start_and_end), cmocka_unit_test(strtoll_errors), cmocka_unit_test(garbage)) diff --git a/lib/common/tests/strings/pcmk__starts_with_test.c b/lib/common/tests/strings/pcmk__starts_with_test.c deleted file mode 100644 index 5429417139..0000000000 --- a/lib/common/tests/strings/pcmk__starts_with_test.c +++ /dev/null @@ -1,35 +0,0 @@ -/* - * Copyright 2021 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU General Public License version 2 - * or later (GPLv2+) WITHOUT ANY WARRANTY. - */ - -#include - -#include - -static void -bad_input(void **state) { - assert_false(pcmk__starts_with(NULL, "x")); - assert_false(pcmk__starts_with("abc", NULL)); -} - -static void -starts_with(void **state) { - assert_true(pcmk__starts_with("abc", "a")); - assert_true(pcmk__starts_with("abc", "ab")); - assert_true(pcmk__starts_with("abc", "abc")); - - assert_false(pcmk__starts_with("abc", "A")); - assert_false(pcmk__starts_with("abc", "bc")); - - assert_false(pcmk__starts_with("", "x")); - assert_true(pcmk__starts_with("xyz", "")); -} - -PCMK__UNIT_TEST(NULL, NULL, - cmocka_unit_test(bad_input), - cmocka_unit_test(starts_with)) diff --git a/lib/common/tests/strings/pcmk__trim_test.c b/lib/common/tests/strings/pcmk__trim_test.c deleted file mode 100644 index 56bdd17b10..0000000000 --- a/lib/common/tests/strings/pcmk__trim_test.c +++ /dev/null @@ -1,72 +0,0 @@ -/* - * Copyright 2022 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU General Public License version 2 - * or later (GPLv2+) WITHOUT ANY WARRANTY. - */ - -#include - -#include - -#include - -static void -empty_input(void **state) -{ - char *s = strdup(""); - - assert_null(pcmk__trim(NULL)); - assert_string_equal(pcmk__trim(s), ""); - - free(s); -} - -static void -leading_newline(void **state) -{ - char *s = strdup("\nabcd"); - - assert_string_equal(pcmk__trim(s), "\nabcd"); - free(s); -} - -static void -middle_newline(void **state) -{ - char *s = strdup("ab\ncd"); - - assert_string_equal(pcmk__trim(s), "ab\ncd"); - free(s); -} - -static void -trailing_newline(void **state) -{ - char *s = strdup("abcd\n\n"); - - assert_string_equal(pcmk__trim(s), "abcd"); - free(s); - - s = strdup("abcd\n "); - assert_string_equal(pcmk__trim(s), "abcd\n "); - free(s); -} - -static void -other_whitespace(void **state) -{ - char *s = strdup(" ab\t\ncd \t"); - - assert_string_equal(pcmk__trim(s), " ab\t\ncd \t"); - free(s); -} - -PCMK__UNIT_TEST(NULL, NULL, - cmocka_unit_test(empty_input), - cmocka_unit_test(leading_newline), - cmocka_unit_test(middle_newline), - cmocka_unit_test(trailing_newline), - cmocka_unit_test(other_whitespace)) diff --git a/lib/common/watchdog.c b/lib/common/watchdog.c index 2f464eed47..7342df8d20 100644 --- a/lib/common/watchdog.c +++ b/lib/common/watchdog.c @@ -1,293 +1,297 @@ /* - * Copyright 2013-2024 the Pacemaker project contributors + * Copyright 2013-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include +#include // g_str_has_prefix() + static pid_t sbd_pid = 0; /*! * \internal * \brief Tell pacemakerd to panic the local host * * \param[in] ppid Process ID of parent process */ static void panic_local_nonroot(pid_t ppid) { if (ppid > 1) { // pacemakerd is still our parent crm_emerg("Escalating panic to " PCMK__SERVER_PACEMAKERD "[%lld]", (long long) ppid); } else { // Signal (non-parent) pacemakerd if possible ppid = pcmk__procfs_pid_of(PCMK__SERVER_PACEMAKERD); if (ppid > 0) { union sigval signal_value; crm_emerg("Signaling " PCMK__SERVER_PACEMAKERD "[%lld] to panic", (long long) ppid); memset(&signal_value, 0, sizeof(signal_value)); if (sigqueue(ppid, SIGQUIT, signal_value) < 0) { crm_emerg("Exiting after signal failure: %s", strerror(errno)); } } else { crm_emerg("Exiting with no known " PCMK__SERVER_PACEMAKERD "process"); } } crm_exit(CRM_EX_PANIC); } /*! * \internal * \brief Panic the local host (if root) or tell pacemakerd to do so */ static void panic_local(void) { const char *full_panic_action = pcmk__env_option(PCMK__ENV_PANIC_ACTION); const char *panic_action = full_panic_action; int reboot_cmd = RB_AUTOBOOT; // Default panic action is reboot if (geteuid() != 0) { // Non-root caller such as the controller panic_local_nonroot(getppid()); return; } - if (pcmk__starts_with(full_panic_action, "sync-")) { + if ((full_panic_action != NULL) + && g_str_has_prefix(full_panic_action, "sync-")) { + panic_action += sizeof("sync-") - 1; sync(); } if (pcmk__str_empty(full_panic_action) || pcmk__str_eq(panic_action, PCMK_VALUE_REBOOT, pcmk__str_none)) { pcmk__sysrq_trigger('b'); } else if (pcmk__str_eq(panic_action, PCMK_VALUE_CRASH, pcmk__str_none)) { pcmk__sysrq_trigger('c'); } else if (pcmk__str_eq(panic_action, PCMK_VALUE_OFF, pcmk__str_none)) { pcmk__sysrq_trigger('o'); #ifdef RB_POWER_OFF reboot_cmd = RB_POWER_OFF; #elif defined(RB_POWEROFF) reboot_cmd = RB_POWEROFF; #endif } else { crm_warn("Using default '" PCMK_VALUE_REBOOT "' for local option PCMK_" PCMK__ENV_PANIC_ACTION " because '%s' is not a valid value", full_panic_action); pcmk__sysrq_trigger('b'); } // sysrq failed or is not supported on this platform, so fall back to reboot reboot(reboot_cmd); // Even reboot failed, nothing left to do but exit crm_emerg("Exiting after reboot failed: %s", strerror(errno)); if (getppid() > 1) { // pacemakerd is parent process crm_exit(CRM_EX_PANIC); } else { // This is pacemakerd, or an orphaned subdaemon crm_exit(CRM_EX_FATAL); } } /*! * \internal * \brief Tell sbd to kill the local host, then exit */ static void panic_sbd(void) { union sigval signal_value; pid_t ppid = getppid(); memset(&signal_value, 0, sizeof(signal_value)); /* TODO: Arrange for a slightly less brutal option? */ if(sigqueue(sbd_pid, SIGKILL, signal_value) < 0) { crm_emerg("Panicking directly because couldn't signal sbd"); panic_local(); } if(ppid > 1) { /* child daemon */ crm_exit(CRM_EX_PANIC); } else { /* pacemakerd or orphan child */ crm_exit(CRM_EX_FATAL); } } /*! * \internal * \brief Panic the local host * * Panic the local host either by sbd (if running), directly, or by asking * pacemakerd. If trace logging this function, exit instead. * * \param[in] reason Why panic is needed (for logging only) */ void pcmk__panic(const char *reason) { if (pcmk__locate_sbd() > 1) { crm_emerg("Signaling sbd[%lld] to panic the system: %s", (long long) sbd_pid, reason); panic_sbd(); } else { crm_emerg("Panicking the system directly: %s", reason); panic_local(); } } /*! * \internal * \brief Return the process ID of sbd (or 0 if it is not running) */ pid_t pcmk__locate_sbd(void) { const char *pidfile = PCMK__RUN_DIR "/sbd.pid"; int rc; if(sbd_pid > 1) { return sbd_pid; } /* Read the pid file */ rc = pcmk__pidfile_matches(pidfile, 0, SBIN_DIR "/sbd", &sbd_pid); if (rc == pcmk_rc_ok) { crm_trace("SBD detected at pid %lld (via PID file %s)", (long long) sbd_pid, pidfile); } else { /* Fall back to /proc for systems that support it */ sbd_pid = pcmk__procfs_pid_of("sbd"); if (sbd_pid != 0) { crm_trace("SBD detected at pid %lld (via procfs)", (long long) sbd_pid); } } if(sbd_pid < 0) { sbd_pid = 0; crm_trace("SBD not detected"); } return sbd_pid; } long pcmk__get_sbd_watchdog_timeout(void) { static long sbd_timeout = -2; if (sbd_timeout == -2) { sbd_timeout = crm_get_msec(getenv("SBD_WATCHDOG_TIMEOUT")); } return sbd_timeout; } bool pcmk__get_sbd_sync_resource_startup(void) { static int sync_resource_startup = PCMK__SBD_SYNC_DEFAULT; static bool checked_sync_resource_startup = false; if (!checked_sync_resource_startup) { const char *sync_env = getenv("SBD_SYNC_RESOURCE_STARTUP"); if (sync_env == NULL) { crm_trace("Defaulting to %sstart-up synchronization with sbd", (PCMK__SBD_SYNC_DEFAULT? "" : "no ")); } else if (crm_str_to_boolean(sync_env, &sync_resource_startup) < 0) { crm_warn("Defaulting to %sstart-up synchronization with sbd " "because environment value '%s' is invalid", (PCMK__SBD_SYNC_DEFAULT? "" : "no "), sync_env); } checked_sync_resource_startup = true; } return sync_resource_startup != 0; } long pcmk__auto_stonith_watchdog_timeout(void) { long sbd_timeout = pcmk__get_sbd_watchdog_timeout(); return (sbd_timeout <= 0)? 0 : (2 * sbd_timeout); } bool pcmk__valid_stonith_watchdog_timeout(const char *value) { /* @COMPAT At a compatibility break, accept either negative values or a * specific string like "auto" (but not both) to mean "auto-calculate the * timeout." Reject other values that aren't parsable as timeouts. */ long long st_timeout = 0; if (value != NULL) { /* @COMPAT So far it has been documented that a negative value is * valid. Parse it as an integer first to avoid the warning from * crm_get_msec(). */ int rc = pcmk__scan_ll(value, &st_timeout, PCMK__PARSE_INT_DEFAULT); if (rc != pcmk_rc_ok || st_timeout >= 0) { st_timeout = crm_get_msec(value); } } if (st_timeout < 0) { st_timeout = pcmk__auto_stonith_watchdog_timeout(); crm_debug("Using calculated value %lld for " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " (%s)", st_timeout, value); } if (st_timeout == 0) { crm_debug("Watchdog may be enabled but " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " is disabled (%s)", value? value : "default"); } else if (pcmk__locate_sbd() == 0) { crm_emerg("Shutting down: " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " configured (%s) but SBD not active", pcmk__s(value, "auto")); crm_exit(CRM_EX_FATAL); return false; } else { long sbd_timeout = pcmk__get_sbd_watchdog_timeout(); if (st_timeout < sbd_timeout) { crm_emerg("Shutting down: " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " (%s) too short (must be >%ldms)", value, sbd_timeout); crm_exit(CRM_EX_FATAL); return false; } crm_info("Watchdog configured with " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " %s and SBD timeout %ldms", value, sbd_timeout); } return true; } diff --git a/lib/common/xml.c b/lib/common/xml.c index 30383f9ba9..6e5484daaa 100644 --- a/lib/common/xml.c +++ b/lib/common/xml.c @@ -1,1858 +1,1858 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include // uint32_t #include #include #include #include // stat(), S_ISREG, etc. #include #include // gboolean, GString #include // xmlNode, etc. #include // xmlChar, xmlGetUTF8Char() #include #include #include // PCMK__XML_LOG_BASE, etc. #include "crmcommon_private.h" //! libxml2 supports only XML version 1.0, at least as of libxml2-2.12.5 #define XML_VERSION ((const xmlChar *) "1.0") /*! * \internal * \brief Get a string representation of an XML element type for logging * * \param[in] type XML element type * * \return String representation of \p type */ const char * pcmk__xml_element_type_text(xmlElementType type) { static const char *const element_type_names[] = { [XML_ELEMENT_NODE] = "element", [XML_ATTRIBUTE_NODE] = "attribute", [XML_TEXT_NODE] = "text", [XML_CDATA_SECTION_NODE] = "CDATA section", [XML_ENTITY_REF_NODE] = "entity reference", [XML_ENTITY_NODE] = "entity", [XML_PI_NODE] = "PI", [XML_COMMENT_NODE] = "comment", [XML_DOCUMENT_NODE] = "document", [XML_DOCUMENT_TYPE_NODE] = "document type", [XML_DOCUMENT_FRAG_NODE] = "document fragment", [XML_NOTATION_NODE] = "notation", [XML_HTML_DOCUMENT_NODE] = "HTML document", [XML_DTD_NODE] = "DTD", [XML_ELEMENT_DECL] = "element declaration", [XML_ATTRIBUTE_DECL] = "attribute declaration", [XML_ENTITY_DECL] = "entity declaration", [XML_NAMESPACE_DECL] = "namespace declaration", [XML_XINCLUDE_START] = "XInclude start", [XML_XINCLUDE_END] = "XInclude end", }; // Assumes the numeric values of the indices are in ascending order if ((type < XML_ELEMENT_NODE) || (type > XML_XINCLUDE_END)) { return "unrecognized type"; } return element_type_names[type]; } /*! * \internal * \brief Apply a function to each XML node in a tree (pre-order, depth-first) * * \param[in,out] xml XML tree to traverse * \param[in,out] fn Function to call for each node (returns \c true to * continue traversing the tree or \c false to stop) * \param[in,out] user_data Argument to \p fn * * \return \c false if any \p fn call returned \c false, or \c true otherwise * * \note This function is recursive. */ bool pcmk__xml_tree_foreach(xmlNode *xml, bool (*fn)(xmlNode *, void *), void *user_data) { if (xml == NULL) { return true; } if (!fn(xml, user_data)) { return false; } for (xml = pcmk__xml_first_child(xml); xml != NULL; xml = pcmk__xml_next(xml)) { if (!pcmk__xml_tree_foreach(xml, fn, user_data)) { return false; } } return true; } void pcmk__xml_set_parent_flags(xmlNode *xml, uint64_t flags) { for (; xml != NULL; xml = xml->parent) { xml_node_private_t *nodepriv = xml->_private; if (nodepriv != NULL) { pcmk__set_xml_flags(nodepriv, flags); } } } /*! * \internal * \brief Set flags for an XML document * * \param[in,out] doc XML document * \param[in] flags Group of enum pcmk__xml_flags */ void pcmk__xml_doc_set_flags(xmlDoc *doc, uint32_t flags) { if (doc != NULL) { xml_doc_private_t *docpriv = doc->_private; pcmk__set_xml_flags(docpriv, flags); } } /*! * \internal * \brief Check whether the given flags are set for an XML document * * \param[in] doc XML document to check * \param[in] flags Group of enum pcmk__xml_flags * * \return \c true if all of \p flags are set for \p doc, or \c false otherwise */ bool pcmk__xml_doc_all_flags_set(const xmlDoc *doc, uint32_t flags) { if (doc != NULL) { xml_doc_private_t *docpriv = doc->_private; return (docpriv != NULL) && pcmk_all_flags_set(docpriv->flags, flags); } return false; } // Mark document, element, and all element's parents as changed void pcmk__mark_xml_node_dirty(xmlNode *xml) { if (xml == NULL) { return; } pcmk__xml_doc_set_flags(xml->doc, pcmk__xf_dirty); pcmk__xml_set_parent_flags(xml, pcmk__xf_dirty); } /*! * \internal * \brief Clear flags on an XML node * * \param[in,out] xml XML node whose flags to reset * \param[in,out] user_data Ignored * * \return \c true (to continue traversing the tree) * * \note This is compatible with \c pcmk__xml_tree_foreach(). */ bool pcmk__xml_reset_node_flags(xmlNode *xml, void *user_data) { xml_node_private_t *nodepriv = xml->_private; if (nodepriv != NULL) { nodepriv->flags = pcmk__xf_none; } return true; } /*! * \internal * \brief Set the \c pcmk__xf_dirty and \c pcmk__xf_created flags on an XML node * * \param[in,out] xml Node whose flags to set * \param[in] user_data Ignored * * \return \c true (to continue traversing the tree) * * \note This is compatible with \c pcmk__xml_tree_foreach(). */ static bool mark_xml_dirty_created(xmlNode *xml, void *user_data) { xml_node_private_t *nodepriv = xml->_private; if (nodepriv != NULL) { pcmk__set_xml_flags(nodepriv, pcmk__xf_dirty|pcmk__xf_created); } return true; } /*! * \internal * \brief Mark an XML tree as dirty and created, and mark its parents dirty * * Also mark the document dirty. * * \param[in,out] xml Tree to mark as dirty and created */ static void mark_xml_tree_dirty_created(xmlNode *xml) { pcmk__assert(xml != NULL); if (!pcmk__xml_doc_all_flags_set(xml->doc, pcmk__xf_tracking)) { // Tracking is disabled for entire document return; } // Mark all parents and document dirty pcmk__mark_xml_node_dirty(xml); pcmk__xml_tree_foreach(xml, mark_xml_dirty_created, NULL); } // Free an XML object previously marked as deleted static void free_deleted_object(void *data) { if(data) { pcmk__deleted_xml_t *deleted_obj = data; g_free(deleted_obj->path); free(deleted_obj); } } // Free and NULL user, ACLs, and deleted objects in an XML node's private data static void reset_xml_private_data(xml_doc_private_t *docpriv) { if (docpriv != NULL) { pcmk__assert(docpriv->check == PCMK__XML_DOC_PRIVATE_MAGIC); pcmk__str_update(&(docpriv->acl_user), NULL); if (docpriv->acls != NULL) { pcmk__free_acls(docpriv->acls); docpriv->acls = NULL; } if(docpriv->deleted_objs) { g_list_free_full(docpriv->deleted_objs, free_deleted_object); docpriv->deleted_objs = NULL; } } } /*! * \internal * \brief Allocate and initialize private data for an XML node * * \param[in,out] node XML node whose private data to initialize * \param[in] user_data Ignored * * \return \c true (to continue traversing the tree) * * \note This is compatible with \c pcmk__xml_tree_foreach(). */ static bool new_private_data(xmlNode *node, void *user_data) { bool tracking = false; CRM_CHECK(node != NULL, return true); if (node->_private != NULL) { return true; } tracking = pcmk__xml_doc_all_flags_set(node->doc, pcmk__xf_tracking); switch (node->type) { case XML_DOCUMENT_NODE: { xml_doc_private_t *docpriv = pcmk__assert_alloc(1, sizeof(xml_doc_private_t)); docpriv->check = PCMK__XML_DOC_PRIVATE_MAGIC; node->_private = docpriv; } break; case XML_ELEMENT_NODE: case XML_ATTRIBUTE_NODE: case XML_COMMENT_NODE: { xml_node_private_t *nodepriv = pcmk__assert_alloc(1, sizeof(xml_node_private_t)); nodepriv->check = PCMK__XML_NODE_PRIVATE_MAGIC; node->_private = nodepriv; if (tracking) { pcmk__set_xml_flags(nodepriv, pcmk__xf_dirty|pcmk__xf_created); } for (xmlAttr *iter = pcmk__xe_first_attr(node); iter != NULL; iter = iter->next) { new_private_data((xmlNode *) iter, user_data); } } break; case XML_TEXT_NODE: case XML_DTD_NODE: case XML_CDATA_SECTION_NODE: return true; default: CRM_LOG_ASSERT(node->type == XML_ELEMENT_NODE); return true; } if (tracking) { pcmk__mark_xml_node_dirty(node); } return true; } /*! * \internal * \brief Free private data for an XML node * * \param[in,out] node XML node whose private data to free * \param[in] user_data Ignored * * \return \c true (to continue traversing the tree) * * \note This is compatible with \c pcmk__xml_tree_foreach(). */ static bool free_private_data(xmlNode *node, void *user_data) { CRM_CHECK(node != NULL, return true); if (node->_private == NULL) { return true; } if (node->type == XML_DOCUMENT_NODE) { reset_xml_private_data((xml_doc_private_t *) node->_private); } else { xml_node_private_t *nodepriv = node->_private; pcmk__assert(nodepriv->check == PCMK__XML_NODE_PRIVATE_MAGIC); for (xmlAttr *iter = pcmk__xe_first_attr(node); iter != NULL; iter = iter->next) { free_private_data((xmlNode *) iter, user_data); } } free(node->_private); node->_private = NULL; return true; } /*! * \internal * \brief Allocate and initialize private data recursively for an XML tree * * \param[in,out] node XML node whose private data to initialize */ void pcmk__xml_new_private_data(xmlNode *xml) { pcmk__xml_tree_foreach(xml, new_private_data, NULL); } /*! * \internal * \brief Free private data recursively for an XML tree * * \param[in,out] node XML node whose private data to free */ void pcmk__xml_free_private_data(xmlNode *xml) { pcmk__xml_tree_foreach(xml, free_private_data, NULL); } /*! * \internal * \brief Return ordinal position of an XML node among its siblings * * \param[in] xml XML node to check * \param[in] ignore_if_set Don't count siblings with this flag set * * \return Ordinal position of \p xml (starting with 0) */ int pcmk__xml_position(const xmlNode *xml, enum pcmk__xml_flags ignore_if_set) { int position = 0; for (const xmlNode *cIter = xml; cIter->prev; cIter = cIter->prev) { xml_node_private_t *nodepriv = ((xmlNode*)cIter->prev)->_private; if (!pcmk_is_set(nodepriv->flags, ignore_if_set)) { position++; } } return position; } /*! * \internal * \brief Remove all attributes marked as deleted from an XML node * * \param[in,out] xml XML node whose deleted attributes to remove * \param[in,out] user_data Ignored * * \return \c true (to continue traversing the tree) * * \note This is compatible with \c pcmk__xml_tree_foreach(). */ static bool commit_attr_deletions(xmlNode *xml, void *user_data) { pcmk__xml_reset_node_flags(xml, NULL); pcmk__xe_remove_matching_attrs(xml, true, pcmk__marked_as_deleted, NULL); return true; } /*! * \internal * \brief Finalize all pending changes to an XML document and reset private data * * Clear the ACL user and all flags, unpacked ACLs, and deleted node records for * the document; clear all flags on each node in the tree; and delete any * attributes that are marked for deletion. * * \param[in,out] doc XML document * * \note When change tracking is enabled, "deleting" an attribute simply marks * it for deletion (using \c pcmk__xf_deleted) until changes are * committed. Freeing a node (using \c pcmk__xml_free()) adds a deleted * node record (\c pcmk__deleted_xml_t) to the node's document before * freeing it. * \note This function clears all flags, not just flags that indicate changes. * In particular, note that it clears the \c pcmk__xf_tracking flag, thus * disabling tracking. */ void pcmk__xml_commit_changes(xmlDoc *doc) { xml_doc_private_t *docpriv = NULL; if (doc == NULL) { return; } docpriv = doc->_private; if (docpriv == NULL) { return; } if (pcmk_is_set(docpriv->flags, pcmk__xf_dirty)) { pcmk__xml_tree_foreach(xmlDocGetRootElement(doc), commit_attr_deletions, NULL); } reset_xml_private_data(docpriv); docpriv->flags = pcmk__xf_none; } /*! * \internal * \brief Create a new XML document * * \return Newly allocated XML document (guaranteed not to be \c NULL) * * \note The caller is responsible for freeing the return value using * \c pcmk__xml_free_doc(). */ xmlDoc * pcmk__xml_new_doc(void) { xmlDoc *doc = xmlNewDoc(XML_VERSION); pcmk__mem_assert(doc); pcmk__xml_new_private_data((xmlNode *) doc); return doc; } /*! * \internal * \brief Free a new XML document * * \param[in,out] doc XML document to free */ void pcmk__xml_free_doc(xmlDoc *doc) { if (doc != NULL) { pcmk__xml_free_private_data((xmlNode *) doc); xmlFreeDoc(doc); } } /*! * \internal * \brief Check whether the first character of a string is an XML NameStartChar * * See https://www.w3.org/TR/xml/#NT-NameStartChar. * * This is almost identical to libxml2's \c xmlIsDocNameStartChar(), but they * don't expose it as part of the public API. * * \param[in] utf8 UTF-8 encoded string * \param[out] len If not \c NULL, where to store size in bytes of first * character in \p utf8 * * \return \c true if \p utf8 begins with a valid XML NameStartChar, or \c false * otherwise */ bool pcmk__xml_is_name_start_char(const char *utf8, int *len) { int c = 0; int local_len = 0; if (len == NULL) { len = &local_len; } /* xmlGetUTF8Char() abuses the len argument. At call time, it must be set to * "the minimum number of bytes present in the sequence... to assure the * next character is completely contained within the sequence." It's similar * to the "n" in the strn*() functions. However, this doesn't make any sense * for null-terminated strings, and there's no value that indicates "keep * going until '\0'." So we set it to 4, the max number of bytes in a UTF-8 * character. * * At return, it's set to the actual number of bytes in the char, or 0 on * error. */ *len = 4; // Note: xmlGetUTF8Char() assumes a 32-bit int c = xmlGetUTF8Char((const xmlChar *) utf8, len); if (c < 0) { GString *buf = g_string_sized_new(32); for (int i = 0; (i < 4) && (utf8[i] != '\0'); i++) { g_string_append_printf(buf, " 0x%.2X", utf8[i]); } crm_info("Invalid UTF-8 character (bytes:%s)", (pcmk__str_empty(buf->str)? " " : buf->str)); g_string_free(buf, TRUE); return false; } return (c == '_') || (c == ':') || ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z')) || ((c >= 0xC0) && (c <= 0xD6)) || ((c >= 0xD8) && (c <= 0xF6)) || ((c >= 0xF8) && (c <= 0x2FF)) || ((c >= 0x370) && (c <= 0x37D)) || ((c >= 0x37F) && (c <= 0x1FFF)) || ((c >= 0x200C) && (c <= 0x200D)) || ((c >= 0x2070) && (c <= 0x218F)) || ((c >= 0x2C00) && (c <= 0x2FEF)) || ((c >= 0x3001) && (c <= 0xD7FF)) || ((c >= 0xF900) && (c <= 0xFDCF)) || ((c >= 0xFDF0) && (c <= 0xFFFD)) || ((c >= 0x10000) && (c <= 0xEFFFF)); } /*! * \internal * \brief Check whether the first character of a string is an XML NameChar * * See https://www.w3.org/TR/xml/#NT-NameChar. * * This is almost identical to libxml2's \c xmlIsDocNameChar(), but they don't * expose it as part of the public API. * * \param[in] utf8 UTF-8 encoded string * \param[out] len If not \c NULL, where to store size in bytes of first * character in \p utf8 * * \return \c true if \p utf8 begins with a valid XML NameChar, or \c false * otherwise */ bool pcmk__xml_is_name_char(const char *utf8, int *len) { int c = 0; int local_len = 0; if (len == NULL) { len = &local_len; } // See comment regarding len in pcmk__xml_is_name_start_char() *len = 4; // Note: xmlGetUTF8Char() assumes a 32-bit int c = xmlGetUTF8Char((const xmlChar *) utf8, len); if (c < 0) { GString *buf = g_string_sized_new(32); for (int i = 0; (i < 4) && (utf8[i] != '\0'); i++) { g_string_append_printf(buf, " 0x%.2X", utf8[i]); } crm_info("Invalid UTF-8 character (bytes:%s)", (pcmk__str_empty(buf->str)? " " : buf->str)); g_string_free(buf, TRUE); return false; } return ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z')) || ((c >= '0') && (c <= '9')) || (c == '_') || (c == ':') || (c == '-') || (c == '.') || (c == 0xB7) || ((c >= 0xC0) && (c <= 0xD6)) || ((c >= 0xD8) && (c <= 0xF6)) || ((c >= 0xF8) && (c <= 0x2FF)) || ((c >= 0x300) && (c <= 0x36F)) || ((c >= 0x370) && (c <= 0x37D)) || ((c >= 0x37F) && (c <= 0x1FFF)) || ((c >= 0x200C) && (c <= 0x200D)) || ((c >= 0x203F) && (c <= 0x2040)) || ((c >= 0x2070) && (c <= 0x218F)) || ((c >= 0x2C00) && (c <= 0x2FEF)) || ((c >= 0x3001) && (c <= 0xD7FF)) || ((c >= 0xF900) && (c <= 0xFDCF)) || ((c >= 0xFDF0) && (c <= 0xFFFD)) || ((c >= 0x10000) && (c <= 0xEFFFF)); } /*! * \internal * \brief Sanitize a string so it is usable as an XML ID * * An ID must match the Name production as defined here: * https://www.w3.org/TR/xml/#NT-Name. * * Convert an invalid start character to \c '_'. Convert an invalid character * after the start character to \c '.'. * * \param[in,out] id String to sanitize */ void pcmk__xml_sanitize_id(char *id) { bool valid = true; int len = 0; // If id is empty or NULL, there's no way to make it a valid XML ID pcmk__assert(!pcmk__str_empty(id)); /* @TODO Suppose there are two strings and each has an invalid ID character * in the same position. The strings are otherwise identical. Both strings * will be sanitized to the same valid ID, which is incorrect. * * The caller is responsible for ensuring the sanitized ID does not already * exist in a given XML document before using it, if uniqueness is desired. */ valid = pcmk__xml_is_name_start_char(id, &len); CRM_CHECK(len > 0, return); // UTF-8 encoding error if (!valid) { *id = '_'; for (int i = 1; i < len; i++) { id[i] = '.'; } } for (id += len; *id != '\0'; id += len) { valid = pcmk__xml_is_name_char(id, &len); CRM_CHECK(len > 0, return); // UTF-8 encoding error if (!valid) { for (int i = 0; i < len; i++) { id[i] = '.'; } } } } /*! * \internal * \brief Free an XML tree without ACL checks or change tracking * * \param[in,out] xml XML node to free */ void pcmk__xml_free_node(xmlNode *xml) { pcmk__xml_free_private_data(xml); xmlUnlinkNode(xml); xmlFreeNode(xml); } /*! * \internal * \brief Free an XML tree if ACLs allow; track deletion if tracking is enabled * * If \p node is the root of its document, free the entire document. * * \param[in,out] node XML node to free * \param[in] position Position of \p node among its siblings for change * tracking (negative to calculate automatically if * needed) * * \return Standard Pacemaker return code */ static int free_xml_with_position(xmlNode *node, int position) { xmlDoc *doc = NULL; xml_node_private_t *nodepriv = NULL; if (node == NULL) { return pcmk_rc_ok; } doc = node->doc; nodepriv = node->_private; if ((doc != NULL) && (xmlDocGetRootElement(doc) == node)) { /* @TODO Should we check ACLs first? Otherwise it seems like we could * free the root element without write permission. */ pcmk__xml_free_doc(doc); return pcmk_rc_ok; } if (!pcmk__check_acl(node, NULL, pcmk__xf_acl_write)) { pcmk__if_tracing( { GString *xpath = pcmk__element_xpath(node); qb_log_from_external_source(__func__, __FILE__, "Cannot remove %s %x", LOG_TRACE, __LINE__, 0, xpath->str, nodepriv->flags); g_string_free(xpath, TRUE); }, {} ); return EACCES; } if (pcmk__xml_doc_all_flags_set(node->doc, pcmk__xf_tracking) && !pcmk_is_set(nodepriv->flags, pcmk__xf_created)) { xml_doc_private_t *docpriv = doc->_private; GString *xpath = pcmk__element_xpath(node); if (xpath != NULL) { pcmk__deleted_xml_t *deleted_obj = NULL; crm_trace("Deleting %s %p from %p", xpath->str, node, doc); deleted_obj = pcmk__assert_alloc(1, sizeof(pcmk__deleted_xml_t)); deleted_obj->path = g_string_free(xpath, FALSE); deleted_obj->position = -1; // Record the position only for XML comments for now if (node->type == XML_COMMENT_NODE) { if (position >= 0) { deleted_obj->position = position; } else { deleted_obj->position = pcmk__xml_position(node, pcmk__xf_skip); } } docpriv->deleted_objs = g_list_append(docpriv->deleted_objs, deleted_obj); pcmk__xml_doc_set_flags(node->doc, pcmk__xf_dirty); } } pcmk__xml_free_node(node); return pcmk_rc_ok; } /*! * \internal * \brief Free an XML tree if ACLs allow; track deletion if tracking is enabled * * If \p xml is the root of its document, free the entire document. * * \param[in,out] xml XML node to free */ void pcmk__xml_free(xmlNode *xml) { free_xml_with_position(xml, -1); } /*! * \internal * \brief Make a deep copy of an XML node under a given parent * * \param[in,out] parent XML element that will be the copy's parent (\c NULL * to create a new XML document with the copy as root) * \param[in] src XML node to copy * * \return Deep copy of \p src, or \c NULL if \p src is \c NULL */ xmlNode * pcmk__xml_copy(xmlNode *parent, xmlNode *src) { xmlNode *copy = NULL; if (src == NULL) { return NULL; } if (parent == NULL) { xmlDoc *doc = NULL; // The copy will be the root element of a new document pcmk__assert(src->type == XML_ELEMENT_NODE); doc = pcmk__xml_new_doc(); copy = xmlDocCopyNode(src, doc, 1); pcmk__mem_assert(copy); xmlDocSetRootElement(doc, copy); } else { copy = xmlDocCopyNode(src, parent->doc, 1); pcmk__mem_assert(copy); xmlAddChild(parent, copy); } pcmk__xml_new_private_data(copy); return copy; } /*! * \internal * \brief Remove XML text nodes from specified XML and all its children * * \param[in,out] xml XML to strip text from */ void pcmk__strip_xml_text(xmlNode *xml) { xmlNode *iter = xml->children; while (iter) { xmlNode *next = iter->next; switch (iter->type) { case XML_TEXT_NODE: pcmk__xml_free_node(iter); break; case XML_ELEMENT_NODE: /* Search it */ pcmk__strip_xml_text(iter); break; default: /* Leave it */ break; } iter = next; } } /*! * \internal * \brief Check whether a string has XML special characters that must be escaped * * See \c pcmk__xml_escape() and \c pcmk__xml_escape_type for more details. * * \param[in] text String to check * \param[in] type Type of escaping * * \return \c true if \p text has special characters that need to be escaped, or * \c false otherwise */ bool pcmk__xml_needs_escape(const char *text, enum pcmk__xml_escape_type type) { if (text == NULL) { return false; } while (*text != '\0') { switch (type) { case pcmk__xml_escape_text: switch (*text) { case '<': case '>': case '&': return true; case '\n': case '\t': break; default: if (g_ascii_iscntrl(*text)) { return true; } break; } break; case pcmk__xml_escape_attr: switch (*text) { case '<': case '>': case '&': case '"': return true; default: if (g_ascii_iscntrl(*text)) { return true; } break; } break; case pcmk__xml_escape_attr_pretty: switch (*text) { case '\n': case '\r': case '\t': case '"': return true; default: break; } break; default: // Invalid enum value pcmk__assert(false); break; } text = g_utf8_next_char(text); } return false; } /*! * \internal * \brief Replace special characters with their XML escape sequences * * \param[in] text Text to escape * \param[in] type Type of escaping * * \return Newly allocated string equivalent to \p text but with special * characters replaced with XML escape sequences (or \c NULL if \p text * is \c NULL). If \p text is not \c NULL, the return value is * guaranteed not to be \c NULL. * * \note There are libxml functions that purport to do this: * \c xmlEncodeEntitiesReentrant() and \c xmlEncodeSpecialChars(). * However, their escaping is incomplete. See: * https://discourse.gnome.org/t/intended-use-of-xmlencodeentitiesreentrant-vs-xmlencodespecialchars/19252 * \note The caller is responsible for freeing the return value using * \c g_free(). */ gchar * pcmk__xml_escape(const char *text, enum pcmk__xml_escape_type type) { GString *copy = NULL; if (text == NULL) { return NULL; } copy = g_string_sized_new(strlen(text)); while (*text != '\0') { // Don't escape any non-ASCII characters if ((*text & 0x80) != 0) { size_t bytes = g_utf8_next_char(text) - text; g_string_append_len(copy, text, bytes); text += bytes; continue; } switch (type) { case pcmk__xml_escape_text: switch (*text) { case '<': g_string_append(copy, PCMK__XML_ENTITY_LT); break; case '>': g_string_append(copy, PCMK__XML_ENTITY_GT); break; case '&': g_string_append(copy, PCMK__XML_ENTITY_AMP); break; case '\n': case '\t': g_string_append_c(copy, *text); break; default: if (g_ascii_iscntrl(*text)) { g_string_append_printf(copy, "&#x%.2X;", *text); } else { g_string_append_c(copy, *text); } break; } break; case pcmk__xml_escape_attr: switch (*text) { case '<': g_string_append(copy, PCMK__XML_ENTITY_LT); break; case '>': g_string_append(copy, PCMK__XML_ENTITY_GT); break; case '&': g_string_append(copy, PCMK__XML_ENTITY_AMP); break; case '"': g_string_append(copy, PCMK__XML_ENTITY_QUOT); break; default: if (g_ascii_iscntrl(*text)) { g_string_append_printf(copy, "&#x%.2X;", *text); } else { g_string_append_c(copy, *text); } break; } break; case pcmk__xml_escape_attr_pretty: switch (*text) { case '"': g_string_append(copy, "\\\""); break; case '\n': g_string_append(copy, "\\n"); break; case '\r': g_string_append(copy, "\\r"); break; case '\t': g_string_append(copy, "\\t"); break; default: g_string_append_c(copy, *text); break; } break; default: // Invalid enum value pcmk__assert(false); break; } text = g_utf8_next_char(text); } return g_string_free(copy, FALSE); } /*! * \internal * \brief Add an XML attribute to a node, marked as deleted * * When calculating XML changes, we need to know when an attribute has been * deleted. Add the attribute back to the new XML, so that we can check the * removal against ACLs, and mark it as deleted for later removal after * differences have been calculated. * * \param[in,out] new_xml XML to modify * \param[in] element Name of XML element that changed (for logging) * \param[in] attr_name Name of attribute that was deleted * \param[in] old_value Value of attribute that was deleted */ static void mark_attr_deleted(xmlNode *new_xml, const char *element, const char *attr_name, const char *old_value) { xml_doc_private_t *docpriv = new_xml->doc->_private; xmlAttr *attr = NULL; xml_node_private_t *nodepriv; /* Restore the old value (without setting dirty flag recursively upwards or * checking ACLs) */ pcmk__clear_xml_flags(docpriv, pcmk__xf_tracking); crm_xml_add(new_xml, attr_name, old_value); pcmk__set_xml_flags(docpriv, pcmk__xf_tracking); // Reset flags (so the attribute doesn't appear as newly created) attr = xmlHasProp(new_xml, (const xmlChar *) attr_name); nodepriv = attr->_private; nodepriv->flags = 0; // Check ACLs and mark restored value for later removal pcmk__xa_remove(attr, false); crm_trace("XML attribute %s=%s was removed from %s", attr_name, old_value, element); } /* * \internal * \brief Check ACLs for a changed XML attribute */ static void mark_attr_changed(xmlNode *new_xml, const char *element, const char *attr_name, const char *old_value) { xml_doc_private_t *docpriv = new_xml->doc->_private; char *vcopy = pcmk__xe_get_copy(new_xml, attr_name); crm_trace("XML attribute %s was changed from '%s' to '%s' in %s", attr_name, old_value, vcopy, element); // Restore the original value (without checking ACLs) pcmk__clear_xml_flags(docpriv, pcmk__xf_tracking); crm_xml_add(new_xml, attr_name, old_value); pcmk__set_xml_flags(docpriv, pcmk__xf_tracking); // Change it back to the new value, to check ACLs crm_xml_add(new_xml, attr_name, vcopy); free(vcopy); } /*! * \internal * \brief Mark an XML attribute as having changed position * * \param[in,out] new_xml XML to modify * \param[in] element Name of XML element that changed (for logging) * \param[in,out] old_attr Attribute that moved, in original XML * \param[in,out] new_attr Attribute that moved, in \p new_xml * \param[in] p_old Ordinal position of \p old_attr in original XML * \param[in] p_new Ordinal position of \p new_attr in \p new_xml */ static void mark_attr_moved(xmlNode *new_xml, const char *element, xmlAttr *old_attr, xmlAttr *new_attr, int p_old, int p_new) { xml_node_private_t *nodepriv = new_attr->_private; crm_trace("XML attribute %s moved from position %d to %d in %s", old_attr->name, p_old, p_new, element); // Mark document, element, and all element's parents as changed pcmk__mark_xml_node_dirty(new_xml); // Mark attribute as changed pcmk__set_xml_flags(nodepriv, pcmk__xf_dirty|pcmk__xf_moved); nodepriv = (p_old > p_new)? old_attr->_private : new_attr->_private; pcmk__set_xml_flags(nodepriv, pcmk__xf_skip); } /*! * \internal * \brief Calculate differences in all previously existing XML attributes * * \param[in,out] old_xml Original XML to compare * \param[in,out] new_xml New XML to compare */ static void xml_diff_old_attrs(xmlNode *old_xml, xmlNode *new_xml) { xmlAttr *attr_iter = pcmk__xe_first_attr(old_xml); while (attr_iter != NULL) { const char *name = (const char *) attr_iter->name; xmlAttr *old_attr = attr_iter; xmlAttr *new_attr = xmlHasProp(new_xml, attr_iter->name); const char *old_value = pcmk__xml_attr_value(attr_iter); attr_iter = attr_iter->next; if (new_attr == NULL) { mark_attr_deleted(new_xml, (const char *) old_xml->name, name, old_value); } else { xml_node_private_t *nodepriv = new_attr->_private; int new_pos = pcmk__xml_position((xmlNode*) new_attr, pcmk__xf_skip); int old_pos = pcmk__xml_position((xmlNode*) old_attr, pcmk__xf_skip); const char *new_value = pcmk__xe_get(new_xml, name); // This attribute isn't new pcmk__clear_xml_flags(nodepriv, pcmk__xf_created); if (strcmp(new_value, old_value) != 0) { mark_attr_changed(new_xml, (const char *) old_xml->name, name, old_value); } else if ((old_pos != new_pos) && !pcmk__xml_doc_all_flags_set(new_xml->doc, pcmk__xf_ignore_attr_pos |pcmk__xf_tracking)) { /* pcmk__xf_tracking is always set by pcmk__xml_mark_changes() * before this function is called, so only the * pcmk__xf_ignore_attr_pos check is truly relevant. */ mark_attr_moved(new_xml, (const char *) old_xml->name, old_attr, new_attr, old_pos, new_pos); } } } } /*! * \internal * \brief Check all attributes in new XML for creation * * For each of a given XML element's attributes marked as newly created, accept * (and mark as dirty) or reject the creation according to ACLs. * * \param[in,out] new_xml XML to check */ static void mark_created_attrs(xmlNode *new_xml) { xmlAttr *attr_iter = pcmk__xe_first_attr(new_xml); while (attr_iter != NULL) { xmlAttr *new_attr = attr_iter; xml_node_private_t *nodepriv = attr_iter->_private; attr_iter = attr_iter->next; if (pcmk_is_set(nodepriv->flags, pcmk__xf_created)) { const char *attr_name = (const char *) new_attr->name; crm_trace("Created new attribute %s=%s in %s", attr_name, pcmk__xml_attr_value(new_attr), new_xml->name); /* Check ACLs (we can't use the remove-then-create trick because it * would modify the attribute position). */ if (pcmk__check_acl(new_xml, attr_name, pcmk__xf_acl_write)) { pcmk__mark_xml_attr_dirty(new_attr); } else { // Creation was not allowed, so remove the attribute pcmk__xa_remove(new_attr, true); } } } } /*! * \internal * \brief Calculate differences in attributes between two XML nodes * * \param[in,out] old_xml Original XML to compare * \param[in,out] new_xml New XML to compare */ static void xml_diff_attrs(xmlNode *old_xml, xmlNode *new_xml) { // Cleared later if attributes are not really new for (xmlAttr *attr = pcmk__xe_first_attr(new_xml); attr != NULL; attr = attr->next) { xml_node_private_t *nodepriv = attr->_private; pcmk__set_xml_flags(nodepriv, pcmk__xf_created); } xml_diff_old_attrs(old_xml, new_xml); mark_created_attrs(new_xml); } /*! * \internal * \brief Add a deleted object record for an old XML child if ACLs allow * * This is intended to be called for a child of an old XML element that is not * present as a child of a new XML element. * * Add a temporary copy of the old child to the new XML. Then check whether ACLs * would have allowed the deletion of that element. If so, add a deleted object * record for it to the new XML's document, and set the \c pcmk__xf_skip flag on * the old child. * * The temporary copy is removed before returning. The new XML and all of its * ancestors will have the \c pcmk__xf_dirty flag set because of the creation, * however. * * \param[in,out] old_child Child of old XML * \param[in,out] new_parent New XML that does not contain \p old_child */ static void mark_child_deleted(xmlNode *old_child, xmlNode *new_parent) { int pos = pcmk__xml_position(old_child, pcmk__xf_skip); // Re-create the child element so we can check ACLs xmlNode *candidate = pcmk__xml_copy(new_parent, old_child); // Clear flags on new child and its children pcmk__xml_tree_foreach(candidate, pcmk__xml_reset_node_flags, NULL); // free_xml_with_position() will check whether ACLs allow the deletion pcmk__apply_acl(xmlDocGetRootElement(candidate->doc)); /* Try to remove the child again (which will track it in document's * deleted_objs on success) */ if (free_xml_with_position(candidate, pos) != pcmk_rc_ok) { // ACLs denied deletion in free_xml_with_position. Free candidate here. pcmk__xml_free_node(candidate); } pcmk__set_xml_flags((xml_node_private_t *) old_child->_private, pcmk__xf_skip); } /*! * \internal * \brief Mark a new child as moved and set \c pcmk__xf_skip as appropriate * * \param[in,out] old_child Child of old XML * \param[in,out] new_child Child of new XML that matches \p old_child * \param[in] old_pos Position of \p old_child among its siblings * \param[in] new_pos Position of \p new_child among its siblings */ static void mark_child_moved(xmlNode *old_child, xmlNode *new_child, int old_pos, int new_pos) { const char *id_s = pcmk__s(pcmk__xe_id(new_child), ""); xmlNode *new_parent = new_child->parent; xml_node_private_t *nodepriv = new_child->_private; crm_trace("Child element %s with " PCMK_XA_ID "='%s' moved from position " "%d to %d under %s", new_child->name, id_s, old_pos, new_pos, new_parent->name); pcmk__mark_xml_node_dirty(new_parent); pcmk__set_xml_flags(nodepriv, pcmk__xf_moved); /* @TODO Figure out and document why we skip the old child in future * position calculations if the old position is higher, and skip the new * child in future position calculations if the new position is higher. This * goes back to d028b52, and there's no explanation in the commit message. */ if (old_pos > new_pos) { nodepriv = old_child->_private; } pcmk__set_xml_flags(nodepriv, pcmk__xf_skip); } /*! * \internal * \brief Check whether a new XML child comment matches an old XML child comment * * Two comments match if they have the same position among their siblings and * the same contents. * * If \p new_comment has the \c pcmk__xf_skip flag set, then it is automatically * considered not to match. * * \param[in] old_comment Old XML child element * \param[in] new_comment New XML child element * * \retval \c true if \p new_comment matches \p old_comment * \retval \c false otherwise */ static bool new_comment_matches(const xmlNode *old_comment, const xmlNode *new_comment) { xml_node_private_t *nodepriv = new_comment->_private; if (pcmk_is_set(nodepriv->flags, pcmk__xf_skip)) { /* @TODO Should we also return false if old_comment has pcmk__xf_skip * set? This preserves existing behavior at time of writing. */ return false; } if (pcmk__xml_position(old_comment, pcmk__xf_skip) != pcmk__xml_position(new_comment, pcmk__xf_skip)) { return false; } return pcmk__xc_matches(old_comment, new_comment); } /*! * \internal * \brief Check whether a new XML child element matches an old XML child element * * Two elements match if they have the same name and the same ID. (Both IDs can * be \c NULL.) * * For XML attributes other than \c PCMK_XA_ID, we can treat a value change as * an in-place modification. However, when Pacemaker applies a patchset, it uses * the \c PCMK_XA_ID attribute to find the node to update (modify, delete, or * move). If we treat two nodes with different \c PCMK_XA_ID attributes as * matching and then mark that attribute as changed, it can cause this lookup to * fail. * * There's unlikely to ever be much practical reason to treat elements with * different IDs as a change. Unless that changes, we'll treat them as a * mismatch. * * \param[in] old_element Old XML child element * \param[in] new_element New XML child element * * \retval \c true if \p new_element matches \p old_element * \retval \c false otherwise */ static bool new_element_matches(const xmlNode *old_element, const xmlNode *new_element) { return pcmk__xe_is(new_element, (const char *) old_element->name) && pcmk__str_eq(pcmk__xe_id(old_element), pcmk__xe_id(new_element), pcmk__str_none); } /*! * \internal * \brief Check whether a new XML child node matches an old XML child node * * Node types must be the same in order to match. * * For comments, a match is a comment at the same position with the same * content. * * For elements, a match is an element with the same name and the same ID. (Both * IDs can be \c NULL.) * * For other node types, there is no match. * * \param[in] old_child Child of old XML * \param[in] new_child Child of new XML * * \retval \c true if \p new_child matches \p old_child * \retval \c false otherwise */ static bool new_child_matches(const xmlNode *old_child, const xmlNode *new_child) { if (old_child->type != new_child->type) { return false; } switch (old_child->type) { case XML_COMMENT_NODE: return new_comment_matches(old_child, new_child); case XML_ELEMENT_NODE: return new_element_matches(old_child, new_child); default: return false; } } /*! * \internal * \brief Find matching XML node pairs between old and new XML's children * * A node that is part of a matching pair gets its _private:match * member set to the matching node. * * \param[in,out] old_xml Old XML * \param[in,out] new_xml New XML */ static void find_matching_children(xmlNode *old_xml, xmlNode *new_xml) { for (xmlNode *old_child = pcmk__xml_first_child(old_xml); old_child != NULL; old_child = pcmk__xml_next(old_child)) { xml_node_private_t *old_nodepriv = old_child->_private; if ((old_nodepriv == NULL) || (old_nodepriv->match != NULL)) { // Can't process, or we already found a match for this old child continue; } for (xmlNode *new_child = pcmk__xml_first_child(new_xml); new_child != NULL; new_child = pcmk__xml_next(new_child)) { xml_node_private_t *new_nodepriv = new_child->_private; if ((new_nodepriv == NULL) || (new_nodepriv->match != NULL)) { /* Can't process, or this new child already matched some old * child */ continue; } if (new_child_matches(old_child, new_child)) { old_nodepriv->match = new_child; new_nodepriv->match = old_child; break; } } } } /*! * \internal * \brief Mark changes between two XML trees * * Set flags in a new XML tree to indicate changes relative to an old XML tree. * * \param[in,out] old_xml XML before changes * \param[in,out] new_xml XML after changes * * \note This may set \c pcmk__xf_skip on parts of \p old_xml. */ void pcmk__xml_mark_changes(xmlNode *old_xml, xmlNode *new_xml) { /* This function may set the xml_node_private_t:match member on children of * old_xml and new_xml, but it clears that member before returning. * * @TODO Ensure we handle (for example, by copying) or reject user-created * XML that is missing xml_node_private_t at top level or in any children. * Similarly, check handling of node types for which we don't create private * data. For now, we'll skip them in the loops below. */ CRM_CHECK((old_xml != NULL) && (new_xml != NULL), return); if ((old_xml->_private == NULL) || (new_xml->_private == NULL)) { return; } pcmk__xml_doc_set_flags(new_xml->doc, pcmk__xf_tracking); xml_diff_attrs(old_xml, new_xml); find_matching_children(old_xml, new_xml); // Process matches (changed children) and deletions for (xmlNode *old_child = pcmk__xml_first_child(old_xml); old_child != NULL; old_child = pcmk__xml_next(old_child)) { xml_node_private_t *nodepriv = old_child->_private; xmlNode *new_child = NULL; if (nodepriv == NULL) { continue; } if (nodepriv->match == NULL) { // No match in new XML means the old child was deleted mark_child_deleted(old_child, new_xml); continue; } /* Fetch the match and clear old_child->_private's match member. * new_child->_private's match member is handled in the new_xml loop. */ new_child = nodepriv->match; nodepriv->match = NULL; pcmk__assert(old_child->type == new_child->type); if (old_child->type == XML_COMMENT_NODE) { // Comments match only if their positions and contents match continue; } pcmk__xml_mark_changes(old_child, new_child); } /* Mark unmatched new children as created, and mark matched new children as * moved if their positions changed. Grab the next new child in advance, * since new_child may get freed in the loop body. */ for (xmlNode *new_child = pcmk__xml_first_child(new_xml), *next = pcmk__xml_next(new_child); new_child != NULL; new_child = next, next = pcmk__xml_next(new_child)) { xml_node_private_t *nodepriv = new_child->_private; if (nodepriv == NULL) { continue; } if (nodepriv->match != NULL) { /* Fetch the match and clear new_child->_private's match member. Any * changes were marked in the old_xml loop. Mark the move. * * We might be able to mark the move earlier, when we mark changes * for matches in the old_xml loop, consolidating both actions. We'd * have to think about whether the timing of setting the * pcmk__xf_skip flag makes any difference. */ xmlNode *old_child = nodepriv->match; int old_pos = pcmk__xml_position(old_child, pcmk__xf_skip); int new_pos = pcmk__xml_position(new_child, pcmk__xf_skip); if (old_pos != new_pos) { mark_child_moved(old_child, new_child, old_pos, new_pos); } nodepriv->match = NULL; continue; } // No match in old XML means the new child is newly created pcmk__set_xml_flags(nodepriv, pcmk__xf_skip); mark_xml_tree_dirty_created(new_child); // Check whether creation was allowed (may free new_child) pcmk__apply_creation_acl(new_child, true); } } char * pcmk__xml_artefact_root(enum pcmk__xml_artefact_ns ns) { static const char *base = NULL; char *ret = NULL; if (base == NULL) { base = pcmk__env_option(PCMK__ENV_SCHEMA_DIRECTORY); } if (pcmk__str_empty(base)) { base = PCMK_SCHEMA_DIR; } switch (ns) { case pcmk__xml_artefact_ns_legacy_rng: case pcmk__xml_artefact_ns_legacy_xslt: ret = strdup(base); break; case pcmk__xml_artefact_ns_base_rng: case pcmk__xml_artefact_ns_base_xslt: ret = crm_strdup_printf("%s/base", base); break; default: crm_err("XML artefact family specified as %u not recognized", ns); } return ret; } static char * find_artefact(enum pcmk__xml_artefact_ns ns, const char *path, const char *filespec) { char *ret = NULL; switch (ns) { case pcmk__xml_artefact_ns_legacy_rng: case pcmk__xml_artefact_ns_base_rng: - if (pcmk__ends_with(filespec, ".rng")) { + if (g_str_has_suffix(filespec, ".rng")) { ret = crm_strdup_printf("%s/%s", path, filespec); } else { ret = crm_strdup_printf("%s/%s.rng", path, filespec); } break; case pcmk__xml_artefact_ns_legacy_xslt: case pcmk__xml_artefact_ns_base_xslt: - if (pcmk__ends_with(filespec, ".xsl")) { + if (g_str_has_suffix(filespec, ".xsl")) { ret = crm_strdup_printf("%s/%s", path, filespec); } else { ret = crm_strdup_printf("%s/%s.xsl", path, filespec); } break; default: crm_err("XML artefact family specified as %u not recognized", ns); } return ret; } char * pcmk__xml_artefact_path(enum pcmk__xml_artefact_ns ns, const char *filespec) { struct stat sb; char *base = pcmk__xml_artefact_root(ns); char *ret = NULL; ret = find_artefact(ns, base, filespec); free(base); if (stat(ret, &sb) != 0 || !S_ISREG(sb.st_mode)) { const char *remote_schema_dir = pcmk__remote_schema_dir(); free(ret); ret = find_artefact(ns, remote_schema_dir, filespec); } return ret; } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START #include // xmlCleanupParser() #include xmlNode * copy_xml(xmlNode *src) { xmlDoc *doc = pcmk__xml_new_doc(); xmlNode *copy = NULL; copy = xmlDocCopyNode(src, doc, 1); pcmk__mem_assert(copy); xmlDocSetRootElement(doc, copy); pcmk__xml_new_private_data(copy); return copy; } void crm_xml_init(void) { pcmk__schema_init(); } void crm_xml_cleanup(void) { pcmk__schema_cleanup(); xmlCleanupParser(); } void pcmk_free_xml_subtree(xmlNode *xml) { pcmk__xml_free_node(xml); } void free_xml(xmlNode *child) { pcmk__xml_free(child); } void crm_xml_sanitize_id(char *id) { char *c; for (c = id; *c; ++c) { switch (*c) { case ':': case '#': *c = '.'; } } } bool xml_tracking_changes(xmlNode *xml) { return (xml != NULL) && pcmk__xml_doc_all_flags_set(xml->doc, pcmk__xf_tracking); } bool xml_document_dirty(xmlNode *xml) { return (xml != NULL) && pcmk__xml_doc_all_flags_set(xml->doc, pcmk__xf_dirty); } void xml_accept_changes(xmlNode *xml) { if (xml != NULL) { pcmk__xml_commit_changes(xml->doc); } } void xml_track_changes(xmlNode *xml, const char *user, xmlNode *acl_source, bool enforce_acls) { if (xml == NULL) { return; } pcmk__xml_commit_changes(xml->doc); crm_trace("Tracking changes%s to %p", (enforce_acls? " with ACLs" : ""), xml); pcmk__xml_doc_set_flags(xml->doc, pcmk__xf_tracking); if (enforce_acls) { if (acl_source == NULL) { acl_source = xml; } pcmk__xml_doc_set_flags(xml->doc, pcmk__xf_acl_enabled); pcmk__unpack_acl(acl_source, xml, user); pcmk__apply_acl(xml); } } void xml_calculate_changes(xmlNode *old_xml, xmlNode *new_xml) { CRM_CHECK((old_xml != NULL) && (new_xml != NULL) && pcmk__xe_is(old_xml, (const char *) new_xml->name) && pcmk__str_eq(pcmk__xe_id(old_xml), pcmk__xe_id(new_xml), pcmk__str_none), return); if (!pcmk__xml_doc_all_flags_set(new_xml->doc, pcmk__xf_tracking)) { // Ensure tracking has a clean start (pcmk__xml_mark_changes() enables) pcmk__xml_commit_changes(new_xml->doc); } pcmk__xml_mark_changes(old_xml, new_xml); } void xml_calculate_significant_changes(xmlNode *old_xml, xmlNode *new_xml) { CRM_CHECK((old_xml != NULL) && (new_xml != NULL) && pcmk__xe_is(old_xml, (const char *) new_xml->name) && pcmk__str_eq(pcmk__xe_id(old_xml), pcmk__xe_id(new_xml), pcmk__str_none), return); /* BUG: If pcmk__xf_tracking is not set for new_xml when this function is * called, then we unset pcmk__xf_ignore_attr_pos via * pcmk__xml_commit_changes(). Since this function is about to be * deprecated, it's not worth fixing this and changing the user-facing * behavior. */ pcmk__xml_doc_set_flags(new_xml->doc, pcmk__xf_ignore_attr_pos); if (!pcmk__xml_doc_all_flags_set(new_xml->doc, pcmk__xf_tracking)) { // Ensure tracking has a clean start (pcmk__xml_mark_changes() enables) pcmk__xml_commit_changes(new_xml->doc); } pcmk__xml_mark_changes(old_xml, new_xml); } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/common/xml_io.c b/lib/common/xml_io.c index 559bb06f22..acb0fc3eb9 100644 --- a/lib/common/xml_io.c +++ b/lib/common/xml_io.c @@ -1,619 +1,619 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include // xmlOutputBuffer* #include // xmlChar #include #include #include #include "crmcommon_private.h" /*! * \internal * \brief Decompress a bzip2-compressed file into a string buffer * * \param[in] filename Name of file to decompress * * \return Newly allocated string with the decompressed contents of \p filename, * or \c NULL on error. * * \note The caller is responsible for freeing the return value using \c free(). */ static char * decompress_file(const char *filename) { char *buffer = NULL; int rc = pcmk_rc_ok; size_t length = 0; BZFILE *bz_file = NULL; FILE *input = fopen(filename, "r"); if (input == NULL) { crm_perror(LOG_ERR, "Could not open %s for reading", filename); return NULL; } bz_file = BZ2_bzReadOpen(&rc, input, 0, 0, NULL, 0); rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { crm_err("Could not prepare to read compressed %s: %s " QB_XS " rc=%d", filename, pcmk_rc_str(rc), rc); goto done; } do { int read_len = 0; buffer = pcmk__realloc(buffer, length + PCMK__BUFFER_SIZE + 1); read_len = BZ2_bzRead(&rc, bz_file, buffer + length, PCMK__BUFFER_SIZE); if ((rc == BZ_OK) || (rc == BZ_STREAM_END)) { crm_trace("Read %ld bytes from file: %d", (long) read_len, rc); length += read_len; } } while (rc == BZ_OK); rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { rc = pcmk__bzlib2rc(rc); crm_err("Could not read compressed %s: %s " QB_XS " rc=%d", filename, pcmk_rc_str(rc), rc); free(buffer); buffer = NULL; } else { buffer[length] = '\0'; } done: BZ2_bzReadClose(&rc, bz_file); fclose(input); return buffer; } /*! * \internal * \brief Parse XML from a file * * \param[in] filename Name of file containing XML (\c NULL or \c "-" for * \c stdin); if \p filename ends in \c ".bz2", the file * will be decompressed using \c bzip2 * * \return XML tree parsed from the given file on success, otherwise \c NULL */ xmlNode * pcmk__xml_read(const char *filename) { bool use_stdin = pcmk__str_eq(filename, "-", pcmk__str_null_matches); xmlNode *xml = NULL; xmlDoc *output = NULL; xmlParserCtxt *ctxt = NULL; const xmlError *last_error = NULL; // Create a parser context ctxt = xmlNewParserCtxt(); CRM_CHECK(ctxt != NULL, return NULL); xmlCtxtResetLastError(ctxt); xmlSetGenericErrorFunc(ctxt, pcmk__log_xmllib_err); if (use_stdin) { output = xmlCtxtReadFd(ctxt, STDIN_FILENO, NULL, NULL, XML_PARSE_NOBLANKS); - } else if (pcmk__ends_with_ext(filename, ".bz2")) { + } else if (g_str_has_suffix(filename, ".bz2")) { char *input = decompress_file(filename); if (input != NULL) { output = xmlCtxtReadDoc(ctxt, (const xmlChar *) input, NULL, NULL, XML_PARSE_NOBLANKS); free(input); } } else { output = xmlCtxtReadFile(ctxt, filename, NULL, XML_PARSE_NOBLANKS); } if (output != NULL) { pcmk__xml_new_private_data((xmlNode *) output); xml = xmlDocGetRootElement(output); if (xml != NULL) { /* @TODO Should we really be stripping out text? This seems like an * overly broad way to get rid of whitespace, if that's the goal. * Text nodes may be invalid in most or all Pacemaker inputs, but * stripping them in a generic "parse XML from file" function may * not be the best way to ignore them. */ pcmk__strip_xml_text(xml); } } last_error = xmlCtxtGetLastError(ctxt); if ((last_error != NULL) && (xml != NULL)) { crm_log_xml_debug(xml, "partial"); pcmk__xml_free(xml); xml = NULL; } xmlFreeParserCtxt(ctxt); return xml; } /*! * \internal * \brief Parse XML from a string * * \param[in] input String to parse * * \return XML tree parsed from the given string on success, otherwise \c NULL */ xmlNode * pcmk__xml_parse(const char *input) { xmlNode *xml = NULL; xmlDoc *output = NULL; xmlParserCtxt *ctxt = NULL; const xmlError *last_error = NULL; if (input == NULL) { return NULL; } ctxt = xmlNewParserCtxt(); if (ctxt == NULL) { return NULL; } xmlCtxtResetLastError(ctxt); xmlSetGenericErrorFunc(ctxt, pcmk__log_xmllib_err); output = xmlCtxtReadDoc(ctxt, (const xmlChar *) input, NULL, NULL, XML_PARSE_NOBLANKS); if (output != NULL) { pcmk__xml_new_private_data((xmlNode *) output); xml = xmlDocGetRootElement(output); } last_error = xmlCtxtGetLastError(ctxt); if ((last_error != NULL) && (xml != NULL)) { crm_log_xml_debug(xml, "partial"); pcmk__xml_free(xml); xml = NULL; } xmlFreeParserCtxt(ctxt); return xml; } /*! * \internal * \brief Append a string representation of an XML element to a buffer * * \param[in] data XML whose representation to append * \param[in] options Group of \p pcmk__xml_fmt_options flags * \param[in,out] buffer Where to append the content (must not be \p NULL) * \param[in] depth Current indentation level */ static void dump_xml_element(const xmlNode *data, uint32_t options, GString *buffer, int depth) { bool pretty = pcmk_is_set(options, pcmk__xml_fmt_pretty); bool filtered = pcmk_is_set(options, pcmk__xml_fmt_filtered); int spaces = pretty? (2 * depth) : 0; for (int lpc = 0; lpc < spaces; lpc++) { g_string_append_c(buffer, ' '); } pcmk__g_strcat(buffer, "<", data->name, NULL); for (const xmlAttr *attr = pcmk__xe_first_attr(data); attr != NULL; attr = attr->next) { if (!filtered || !pcmk__xa_filterable((const char *) (attr->name))) { pcmk__dump_xml_attr(attr, buffer); } } if (data->children == NULL) { g_string_append(buffer, "/>"); } else { g_string_append_c(buffer, '>'); } if (pretty) { g_string_append_c(buffer, '\n'); } if (data->children) { for (const xmlNode *child = data->children; child != NULL; child = child->next) { pcmk__xml_string(child, options, buffer, depth + 1); } for (int lpc = 0; lpc < spaces; lpc++) { g_string_append_c(buffer, ' '); } pcmk__g_strcat(buffer, "name, ">", NULL); if (pretty) { g_string_append_c(buffer, '\n'); } } } /*! * \internal * \brief Append XML text content to a buffer * * \param[in] data XML whose content to append * \param[in] options Group of enum pcmk__xml_fmt_options * \param[in,out] buffer Where to append the content (must not be \p NULL) * \param[in] depth Current indentation level */ static void dump_xml_text(const xmlNode *data, uint32_t options, GString *buffer, int depth) { bool pretty = pcmk_is_set(options, pcmk__xml_fmt_pretty); int spaces = pretty? (2 * depth) : 0; const char *content = (const char *) data->content; gchar *content_esc = NULL; if (pcmk__xml_needs_escape(content, pcmk__xml_escape_text)) { content_esc = pcmk__xml_escape(content, pcmk__xml_escape_text); content = content_esc; } for (int lpc = 0; lpc < spaces; lpc++) { g_string_append_c(buffer, ' '); } g_string_append(buffer, content); if (pretty) { g_string_append_c(buffer, '\n'); } g_free(content_esc); } /*! * \internal * \brief Append XML CDATA content to a buffer * * \param[in] data XML whose content to append * \param[in] options Group of \p pcmk__xml_fmt_options flags * \param[in,out] buffer Where to append the content (must not be \p NULL) * \param[in] depth Current indentation level */ static void dump_xml_cdata(const xmlNode *data, uint32_t options, GString *buffer, int depth) { bool pretty = pcmk_is_set(options, pcmk__xml_fmt_pretty); int spaces = pretty? (2 * depth) : 0; for (int lpc = 0; lpc < spaces; lpc++) { g_string_append_c(buffer, ' '); } pcmk__g_strcat(buffer, "content, "]]>", NULL); if (pretty) { g_string_append_c(buffer, '\n'); } } /*! * \internal * \brief Append an XML comment to a buffer * * \param[in] data XML whose content to append * \param[in] options Group of \p pcmk__xml_fmt_options flags * \param[in,out] buffer Where to append the content (must not be \p NULL) * \param[in] depth Current indentation level */ static void dump_xml_comment(const xmlNode *data, uint32_t options, GString *buffer, int depth) { bool pretty = pcmk_is_set(options, pcmk__xml_fmt_pretty); int spaces = pretty? (2 * depth) : 0; for (int lpc = 0; lpc < spaces; lpc++) { g_string_append_c(buffer, ' '); } pcmk__g_strcat(buffer, "", NULL); if (pretty) { g_string_append_c(buffer, '\n'); } } /*! * \internal * \brief Create a string representation of an XML object * * libxml2's \c xmlNodeDumpOutput() doesn't allow filtering, doesn't escape * special characters thoroughly, and doesn't allow a const argument. * * \param[in] data XML to convert * \param[in] options Group of \p pcmk__xml_fmt_options flags * \param[in,out] buffer Where to store the text (must not be \p NULL) * \param[in] depth Current indentation level * * \todo Create a wrapper that doesn't require \p depth. Only used with * recursive calls currently. */ void pcmk__xml_string(const xmlNode *data, uint32_t options, GString *buffer, int depth) { if (data == NULL) { crm_trace("Nothing to dump"); return; } pcmk__assert(buffer != NULL); CRM_CHECK(depth >= 0, depth = 0); switch(data->type) { case XML_ELEMENT_NODE: /* Handle below */ dump_xml_element(data, options, buffer, depth); break; case XML_TEXT_NODE: if (pcmk_is_set(options, pcmk__xml_fmt_text)) { dump_xml_text(data, options, buffer, depth); } break; case XML_COMMENT_NODE: dump_xml_comment(data, options, buffer, depth); break; case XML_CDATA_SECTION_NODE: dump_xml_cdata(data, options, buffer, depth); break; default: crm_warn("Cannot convert XML %s node to text " QB_XS " type=%d", pcmk__xml_element_type_text(data->type), data->type); break; } } /*! * \internal * \brief Write a string to a file stream, compressed using \c bzip2 * * \param[in] text String to write * \param[in] filename Name of file being written (for logging only) * \param[in,out] stream Open file stream to write to * \param[out] bytes_out Number of bytes written (valid only on success) * * \return Standard Pacemaker return code */ static int write_compressed_stream(char *text, const char *filename, FILE *stream, unsigned int *bytes_out) { unsigned int bytes_in = 0; int rc = pcmk_rc_ok; // (5, 0, 0): (intermediate block size, silent, default workFactor) BZFILE *bz_file = BZ2_bzWriteOpen(&rc, stream, 5, 0, 0); rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { crm_warn("Not compressing %s: could not prepare file stream: %s " QB_XS " rc=%d", filename, pcmk_rc_str(rc), rc); goto done; } BZ2_bzWrite(&rc, bz_file, text, strlen(text)); rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { crm_warn("Not compressing %s: could not compress data: %s " QB_XS " rc=%d errno=%d", filename, pcmk_rc_str(rc), rc, errno); goto done; } BZ2_bzWriteClose(&rc, bz_file, 0, &bytes_in, bytes_out); bz_file = NULL; rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { crm_warn("Not compressing %s: could not write compressed data: %s " QB_XS " rc=%d errno=%d", filename, pcmk_rc_str(rc), rc, errno); goto done; } crm_trace("Compressed XML for %s from %u bytes to %u", filename, bytes_in, *bytes_out); done: if (bz_file != NULL) { BZ2_bzWriteClose(&rc, bz_file, 0, NULL, NULL); } return rc; } /*! * \internal * \brief Write XML to a file stream * * \param[in] xml XML to write * \param[in] filename Name of file being written (for logging only) * \param[in,out] stream Open file stream corresponding to filename (closed * when this function returns) * \param[in] compress Whether to compress XML before writing * * \return Standard Pacemaker return code */ static int write_xml_stream(const xmlNode *xml, const char *filename, FILE *stream, bool compress) { GString *buffer = g_string_sized_new(1024); unsigned int bytes_out = 0; int rc = pcmk_rc_ok; pcmk__xml_string(xml, pcmk__xml_fmt_pretty, buffer, 0); CRM_CHECK(!pcmk__str_empty(buffer->str), crm_log_xml_info(xml, "dump-failed"); rc = pcmk_rc_error; goto done); crm_log_xml_trace(xml, "writing"); if (compress && (write_compressed_stream(buffer->str, filename, stream, &bytes_out) == pcmk_rc_ok)) { goto done; } rc = fprintf(stream, "%s", buffer->str); if (rc < 0) { rc = EIO; crm_perror(LOG_ERR, "writing %s", filename); goto done; } bytes_out = (unsigned int) rc; rc = pcmk_rc_ok; done: if (fflush(stream) != 0) { rc = errno; crm_perror(LOG_ERR, "flushing %s", filename); } // Don't report error if the file does not support synchronization if ((fsync(fileno(stream)) < 0) && (errno != EROFS) && (errno != EINVAL)) { rc = errno; crm_perror(LOG_ERR, "synchronizing %s", filename); } fclose(stream); crm_trace("Saved %u bytes to %s as XML", bytes_out, filename); g_string_free(buffer, TRUE); return rc; } /*! * \internal * \brief Write XML to a file descriptor * * \param[in] xml XML to write * \param[in] filename Name of file being written (for logging only) * \param[in] fd Open file descriptor corresponding to \p filename * * \return Standard Pacemaker return code */ int pcmk__xml_write_fd(const xmlNode *xml, const char *filename, int fd) { FILE *stream = NULL; CRM_CHECK((xml != NULL) && (fd > 0), return EINVAL); stream = fdopen(fd, "w"); if (stream == NULL) { return errno; } return write_xml_stream(xml, pcmk__s(filename, "unnamed file"), stream, false); } /*! * \internal * \brief Write XML to a file * * \param[in] xml XML to write * \param[in] filename Name of file to write * \param[in] compress If \c true, compress XML before writing * * \return Standard Pacemaker return code */ int pcmk__xml_write_file(const xmlNode *xml, const char *filename, bool compress) { FILE *stream = NULL; CRM_CHECK((xml != NULL) && (filename != NULL), return EINVAL); stream = fopen(filename, "w"); if (stream == NULL) { return errno; } return write_xml_stream(xml, filename, stream, compress); } /*! * \internal * \brief Serialize XML (using libxml) into provided descriptor * * \param[in] fd File descriptor to (piece-wise) write to * \param[in] cur XML subtree to proceed * * \return a standard Pacemaker return code */ int pcmk__xml2fd(int fd, xmlNode *cur) { bool success; xmlOutputBuffer *fd_out = xmlOutputBufferCreateFd(fd, NULL); pcmk__mem_assert(fd_out); xmlNodeDumpOutput(fd_out, cur->doc, cur, 0, pcmk__xml_fmt_pretty, NULL); success = xmlOutputBufferWrite(fd_out, sizeof("\n") - 1, "\n") != -1; success = xmlOutputBufferClose(fd_out) != -1 && success; if (!success) { return EIO; } fsync(fd); return pcmk_rc_ok; } void save_xml_to_file(const xmlNode *xml, const char *desc, const char *filename) { char *f = NULL; if (filename == NULL) { char *uuid = crm_generate_uuid(); f = crm_strdup_printf("%s/%s", pcmk__get_tmpdir(), uuid); filename = f; free(uuid); } crm_info("Saving %s to %s", desc, filename); pcmk__xml_write_file(xml, filename, false); free(f); } diff --git a/lib/fencing/st_rhcs.c b/lib/fencing/st_rhcs.c index 04b1c79ab1..b728b28816 100644 --- a/lib/fencing/st_rhcs.c +++ b/lib/fencing/st_rhcs.c @@ -1,324 +1,324 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include // xmlXPathObject, etc. #include #include #include #include #include "fencing_private.h" /*! * \internal * \brief \c scandir() filter for RHCS fence agents * * \param[in] entry Directory entry * * \retval 1 if \p entry is a regular file whose name begins with \c "fence_" * \retval 0 otherwise */ static int rhcs_agent_filter(const struct dirent *entry) { char *buf = NULL; struct stat sb; int rc = 0; - if (!pcmk__starts_with(entry->d_name, "fence_")) { + if (!g_str_has_prefix(entry->d_name, "fence_")) { goto done; } // glibc doesn't enforce PATH_MAX, so don't limit buf size buf = crm_strdup_printf(PCMK__FENCE_BINDIR "/%s", entry->d_name); if ((stat(buf, &sb) != 0) || !S_ISREG(sb.st_mode)) { goto done; } rc = 1; done: free(buf); return rc; } /*! * \internal * \brief Add available RHCS-compatible agents to a list * * \param[in,out] List to add to * * \return Number of agents added */ int stonith__list_rhcs_agents(stonith_key_value_t **devices) { struct dirent **namelist = NULL; const int file_num = scandir(PCMK__FENCE_BINDIR, &namelist, rhcs_agent_filter, alphasort); if (file_num < 0) { int rc = errno; crm_err("Could not list " PCMK__FENCE_BINDIR ": %s", pcmk_rc_str(rc)); free(namelist); return 0; } for (int i = 0; i < file_num; i++) { *devices = stonith__key_value_add(*devices, NULL, namelist[i]->d_name); free(namelist[i]); } free(namelist); return file_num; } static void stonith_rhcs_parameter_not_required(xmlNode *metadata, const char *parameter) { char *xpath = NULL; xmlXPathObject *xpathObj = NULL; CRM_CHECK(metadata != NULL, return); CRM_CHECK(parameter != NULL, return); xpath = crm_strdup_printf("//" PCMK_XE_PARAMETER "[@" PCMK_XA_NAME "='%s']", parameter); /* Fudge metadata so that the parameter isn't required in config * Pacemaker handles and adds it */ xpathObj = pcmk__xpath_search(metadata->doc, xpath); if (pcmk__xpath_num_results(xpathObj) > 0) { xmlNode *tmp = pcmk__xpath_result(xpathObj, 0); if (tmp != NULL) { crm_xml_add(tmp, "required", "0"); } } xmlXPathFreeObject(xpathObj); free(xpath); } /*! * \brief Execute RHCS-compatible agent's metadata action * * \param[in] agent Agent to execute * \param[in] timeout_sec Action timeout * \param[out] metadata Where to store output xmlNode (or NULL to ignore) */ static int stonith__rhcs_get_metadata(const char *agent, int timeout_sec, xmlNode **metadata) { xmlNode *xml = NULL; xmlNode *actions = NULL; xmlXPathObject *xpathObj = NULL; stonith_action_t *action = stonith__action_create(agent, PCMK_ACTION_METADATA, NULL, timeout_sec, NULL, NULL, NULL); int rc = stonith__execute(action); pcmk__action_result_t *result = stonith__action_result(action); if (result == NULL) { if (rc < 0) { crm_warn("Could not execute metadata action for %s: %s " QB_XS " rc=%d", agent, pcmk_strerror(rc), rc); } stonith__destroy_action(action); return rc; } if (result->execution_status != PCMK_EXEC_DONE) { crm_warn("Could not execute metadata action for %s: %s", agent, pcmk_exec_status_str(result->execution_status)); rc = pcmk_rc2legacy(stonith__result2rc(result)); stonith__destroy_action(action); return rc; } if (!pcmk__result_ok(result)) { crm_warn("Metadata action for %s returned error code %d", agent, result->exit_status); rc = pcmk_rc2legacy(stonith__result2rc(result)); stonith__destroy_action(action); return rc; } if (result->action_stdout == NULL) { crm_warn("Metadata action for %s returned no data", agent); stonith__destroy_action(action); return -ENODATA; } xml = pcmk__xml_parse(result->action_stdout); stonith__destroy_action(action); if (xml == NULL) { crm_warn("Metadata for %s is invalid", agent); return -pcmk_err_schema_validation; } xpathObj = pcmk__xpath_search(xml->doc, "//" PCMK_XE_ACTIONS); if (pcmk__xpath_num_results(xpathObj) > 0) { actions = pcmk__xpath_result(xpathObj, 0); } xmlXPathFreeObject(xpathObj); // Add start and stop (implemented by pacemaker, not agent) to meta-data xpathObj = pcmk__xpath_search(xml->doc, "//" PCMK_XE_ACTION "[@" PCMK_XA_NAME "='" PCMK_ACTION_STOP "']"); if (pcmk__xpath_num_results(xpathObj) == 0) { xmlNode *tmp = NULL; const char *timeout_str = NULL; timeout_str = pcmk__readable_interval(PCMK_DEFAULT_ACTION_TIMEOUT_MS); tmp = pcmk__xe_create(actions, PCMK_XE_ACTION); crm_xml_add(tmp, PCMK_XA_NAME, PCMK_ACTION_STOP); crm_xml_add(tmp, PCMK_META_TIMEOUT, timeout_str); tmp = pcmk__xe_create(actions, PCMK_XE_ACTION); crm_xml_add(tmp, PCMK_XA_NAME, PCMK_ACTION_START); crm_xml_add(tmp, PCMK_META_TIMEOUT, timeout_str); } xmlXPathFreeObject(xpathObj); // Fudge metadata so parameters are not required in config (pacemaker adds them) stonith_rhcs_parameter_not_required(xml, STONITH_ATTR_ACTION_OP); stonith_rhcs_parameter_not_required(xml, "plug"); stonith_rhcs_parameter_not_required(xml, "port"); if (metadata) { *metadata = xml; } else { pcmk__xml_free(xml); } return pcmk_ok; } /*! * \brief Retrieve metadata for RHCS-compatible fence agent * * \param[in] agent Agent to execute * \param[in] timeout_sec Action timeout * \param[out] output Where to store action output (or NULL to ignore) */ int stonith__rhcs_metadata(const char *agent, int timeout_sec, char **output) { GString *buffer = NULL; xmlNode *xml = NULL; int rc = stonith__rhcs_get_metadata(agent, timeout_sec, &xml); if (rc != pcmk_ok) { goto done; } buffer = g_string_sized_new(1024); pcmk__xml_string(xml, pcmk__xml_fmt_pretty|pcmk__xml_fmt_text, buffer, 0); if (pcmk__str_empty(buffer->str)) { rc = -pcmk_err_schema_validation; goto done; } if (output != NULL) { pcmk__str_update(output, buffer->str); } done: if (buffer != NULL) { g_string_free(buffer, TRUE); } pcmk__xml_free(xml); return rc; } bool stonith__agent_is_rhcs(const char *agent) { struct stat prop; char *buffer = crm_strdup_printf(PCMK__FENCE_BINDIR "/%s", agent); int rc = stat(buffer, &prop); free(buffer); return (rc >= 0) && S_ISREG(prop.st_mode); } int stonith__rhcs_validate(stonith_t *st, int call_options, const char *target, const char *agent, GHashTable *params, const char * host_arg, int timeout, char **output, char **error_output) { int rc = pcmk_ok; int remaining_timeout = timeout; xmlNode *metadata = NULL; stonith_action_t *action = NULL; pcmk__action_result_t *result = NULL; if (host_arg == NULL) { time_t start_time = time(NULL); rc = stonith__rhcs_get_metadata(agent, remaining_timeout, &metadata); if (rc == pcmk_ok) { host_arg = stonith__default_host_arg(metadata); crm_trace("Using '%s' as default " PCMK_STONITH_HOST_ARGUMENT " for %s", pcmk__s(host_arg, PCMK_VALUE_NONE), agent); } pcmk__xml_free(metadata); remaining_timeout -= time(NULL) - start_time; if (rc == -ETIME || remaining_timeout <= 0 ) { return -ETIME; } } else if (pcmk__str_eq(host_arg, PCMK_VALUE_NONE, pcmk__str_casei)) { host_arg = NULL; } action = stonith__action_create(agent, PCMK_ACTION_VALIDATE_ALL, target, remaining_timeout, params, NULL, host_arg); rc = stonith__execute(action); result = stonith__action_result(action); if (result != NULL) { rc = pcmk_rc2legacy(stonith__result2rc(result)); // Take ownership of output so stonith__destroy_action() doesn't free it if (output != NULL) { *output = result->action_stdout; result->action_stdout = NULL; } if (error_output != NULL) { *error_output = result->action_stderr; result->action_stderr = NULL; } } stonith__destroy_action(action); return rc; } diff --git a/lib/pacemaker/pcmk_graph_producer.c b/lib/pacemaker/pcmk_graph_producer.c index e6cff3e26c..d293683ccf 100644 --- a/lib/pacemaker/pcmk_graph_producer.c +++ b/lib/pacemaker/pcmk_graph_producer.c @@ -1,1108 +1,1108 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include "libpacemaker_private.h" // Convenience macros for logging action properties #define action_type_str(flags) \ (pcmk_is_set((flags), pcmk__action_pseudo)? "pseudo-action" : "action") #define action_optional_str(flags) \ (pcmk_is_set((flags), pcmk__action_optional)? "optional" : "required") #define action_runnable_str(flags) \ (pcmk_is_set((flags), pcmk__action_runnable)? "runnable" : "unrunnable") #define action_node_str(a) \ (((a)->node == NULL)? "no node" : (a)->node->priv->name) /*! * \internal * \brief Add an XML node tag for a specified ID * * \param[in] id Node UUID to add * \param[in,out] xml Parent XML tag to add to */ static xmlNode* add_node_to_xml_by_id(const char *id, xmlNode *xml) { xmlNode *node_xml; node_xml = pcmk__xe_create(xml, PCMK_XE_NODE); crm_xml_add(node_xml, PCMK_XA_ID, id); return node_xml; } /*! * \internal * \brief Add an XML node tag for a specified node * * \param[in] node Node to add * \param[in,out] xml XML to add node to */ static void add_node_to_xml(const pcmk_node_t *node, void *xml) { add_node_to_xml_by_id(node->priv->id, (xmlNode *) xml); } /*! * \internal * \brief Count (optionally add to XML) nodes needing maintenance state update * * \param[in,out] xml Parent XML tag to add to, if any * \param[in] scheduler Scheduler data * * \return Count of nodes added * \note Only Pacemaker Remote nodes are considered currently */ static int add_maintenance_nodes(xmlNode *xml, const pcmk_scheduler_t *scheduler) { xmlNode *maintenance = NULL; int count = 0; if (xml != NULL) { maintenance = pcmk__xe_create(xml, PCMK__XE_MAINTENANCE); } for (const GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) { const pcmk_node_t *node = iter->data; if (!pcmk__is_pacemaker_remote_node(node)) { continue; } if ((node->details->maintenance && !pcmk_is_set(node->priv->flags, pcmk__node_remote_maint)) || (!node->details->maintenance && pcmk_is_set(node->priv->flags, pcmk__node_remote_maint))) { if (maintenance != NULL) { crm_xml_add(add_node_to_xml_by_id(node->priv->id, maintenance), PCMK__XA_NODE_IN_MAINTENANCE, (node->details->maintenance? "1" : "0")); } count++; } } crm_trace("%s %d nodes in need of maintenance mode update in state", ((maintenance == NULL)? "Counted" : "Added"), count); return count; } /*! * \internal * \brief Add pseudo action with nodes needing maintenance state update * * \param[in,out] scheduler Scheduler data */ static void add_maintenance_update(pcmk_scheduler_t *scheduler) { pcmk_action_t *action = NULL; if (add_maintenance_nodes(NULL, scheduler) != 0) { action = get_pseudo_op(PCMK_ACTION_MAINTENANCE_NODES, scheduler); pcmk__set_action_flags(action, pcmk__action_always_in_graph); } } /*! * \internal * \brief Add XML with nodes that an action is expected to bring down * * If a specified action is expected to bring any nodes down, add an XML block * with their UUIDs. When a node is lost, this allows the controller to * determine whether it was expected. * * \param[in,out] xml Parent XML tag to add to * \param[in] action Action to check for downed nodes */ static void add_downed_nodes(xmlNode *xml, const pcmk_action_t *action) { CRM_CHECK((xml != NULL) && (action != NULL) && (action->node != NULL), return); if (pcmk__str_eq(action->task, PCMK_ACTION_DO_SHUTDOWN, pcmk__str_none)) { /* Shutdown makes the action's node down */ xmlNode *downed = pcmk__xe_create(xml, PCMK__XE_DOWNED); add_node_to_xml_by_id(action->node->priv->id, downed); } else if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH, pcmk__str_none)) { /* Fencing makes the action's node and any hosted guest nodes down */ const char *fence = g_hash_table_lookup(action->meta, PCMK__META_STONITH_ACTION); if (pcmk__is_fencing_action(fence)) { xmlNode *downed = pcmk__xe_create(xml, PCMK__XE_DOWNED); add_node_to_xml_by_id(action->node->priv->id, downed); pe_foreach_guest_node(action->node->priv->scheduler, action->node, add_node_to_xml, downed); } } else if ((action->rsc != NULL) && pcmk_is_set(action->rsc->flags, pcmk__rsc_is_remote_connection) && pcmk__str_eq(action->task, PCMK_ACTION_STOP, pcmk__str_none)) { /* Stopping a remote connection resource makes connected node down, * unless it's part of a migration */ GList *iter; pcmk_action_t *input; bool migrating = false; for (iter = action->actions_before; iter != NULL; iter = iter->next) { input = ((pcmk__related_action_t *) iter->data)->action; if ((input->rsc != NULL) && pcmk__str_eq(action->rsc->id, input->rsc->id, pcmk__str_none) && pcmk__str_eq(input->task, PCMK_ACTION_MIGRATE_FROM, pcmk__str_none)) { migrating = true; break; } } if (!migrating) { xmlNode *downed = pcmk__xe_create(xml, PCMK__XE_DOWNED); add_node_to_xml_by_id(action->rsc->id, downed); } } } /*! * \internal * \brief Create a transition graph operation key for a clone action * * \param[in] action Clone action * \param[in] interval_ms Action interval in milliseconds * * \return Newly allocated string with transition graph operation key */ static char * clone_op_key(const pcmk_action_t *action, guint interval_ms) { if (pcmk__str_eq(action->task, PCMK_ACTION_NOTIFY, pcmk__str_none)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_operation"); return pcmk__notify_key(action->rsc->priv->history_id, n_type, n_task); } return pcmk__op_key(action->rsc->priv->history_id, pcmk__s(action->cancel_task, action->task), interval_ms); } /*! * \internal * \brief Add node details to transition graph action XML * * \param[in] action Scheduled action * \param[in,out] xml Transition graph action XML for \p action */ static void add_node_details(const pcmk_action_t *action, xmlNode *xml) { pcmk_node_t *router_node = pcmk__connection_host_for_action(action); crm_xml_add(xml, PCMK__META_ON_NODE, action->node->priv->name); crm_xml_add(xml, PCMK__META_ON_NODE_UUID, action->node->priv->id); if (router_node != NULL) { crm_xml_add(xml, PCMK__XA_ROUTER_NODE, router_node->priv->name); } } /*! * \internal * \brief Add resource details to transition graph action XML * * \param[in] action Scheduled action * \param[in,out] action_xml Transition graph action XML for \p action */ static void add_resource_details(const pcmk_action_t *action, xmlNode *action_xml) { xmlNode *rsc_xml = NULL; const char *attr_list[] = { PCMK_XA_CLASS, PCMK_XA_PROVIDER, PCMK_XA_TYPE, }; /* If a resource is locked to a node via PCMK_OPT_SHUTDOWN_LOCK, mark its * actions so the controller can preserve the lock when the action * completes. */ if (pcmk__action_locks_rsc_to_node(action)) { crm_xml_add_ll(action_xml, PCMK_OPT_SHUTDOWN_LOCK, (long long) action->rsc->priv->lock_time); } // List affected resource rsc_xml = pcmk__xe_create(action_xml, (const char *) action->rsc->priv->xml->name); if (pcmk_is_set(action->rsc->flags, pcmk__rsc_removed) && (action->rsc->priv->history_id != NULL)) { /* Use the numbered instance name here, because if there is more * than one instance on a node, we need to make sure the command * goes to the right one. * * This is important even for anonymous clones, because the clone's * unique meta-attribute might have just been toggled from on to * off. */ crm_debug("Using orphan clone name %s instead of history ID %s", action->rsc->id, action->rsc->priv->history_id); crm_xml_add(rsc_xml, PCMK_XA_ID, action->rsc->priv->history_id); crm_xml_add(rsc_xml, PCMK__XA_LONG_ID, action->rsc->id); } else if (!pcmk_is_set(action->rsc->flags, pcmk__rsc_unique)) { const char *xml_id = pcmk__xe_id(action->rsc->priv->xml); crm_debug("Using anonymous clone name %s for %s (aka %s)", xml_id, action->rsc->id, action->rsc->priv->history_id); /* ID is what we'd like client to use * LONG_ID is what they might know it as instead * * LONG_ID is only strictly needed /here/ during the * transition period until all nodes in the cluster * are running the new software /and/ have rebooted * once (meaning that they've only ever spoken to a DC * supporting this feature). (@TODO The effect of removing this on * regression tests suggests that it is still needed for unique clones) * * If anyone toggles the unique flag to 'on', the * 'instance free' name will correspond to an orphan * and fall into the clause above instead */ crm_xml_add(rsc_xml, PCMK_XA_ID, xml_id); if ((action->rsc->priv->history_id != NULL) && !pcmk__str_eq(xml_id, action->rsc->priv->history_id, pcmk__str_none)) { crm_xml_add(rsc_xml, PCMK__XA_LONG_ID, action->rsc->priv->history_id); } else { crm_xml_add(rsc_xml, PCMK__XA_LONG_ID, action->rsc->id); } } else { pcmk__assert(action->rsc->priv->history_id == NULL); crm_xml_add(rsc_xml, PCMK_XA_ID, action->rsc->id); } for (int lpc = 0; lpc < PCMK__NELEM(attr_list); lpc++) { crm_xml_add(rsc_xml, attr_list[lpc], g_hash_table_lookup(action->rsc->priv->meta, attr_list[lpc])); } } /*! * \internal * \brief Add action attributes to transition graph action XML * * \param[in,out] action Scheduled action * \param[in,out] action_xml Transition graph action XML for \p action */ static void add_action_attributes(pcmk_action_t *action, xmlNode *action_xml) { xmlNode *args_xml = NULL; pcmk_resource_t *rsc = action->rsc; /* We create free-standing XML to start, so we can sort the attributes * before adding it to action_xml, which keeps the scheduler regression * test graphs comparable. */ args_xml = pcmk__xe_create(action_xml, PCMK__XE_ATTRIBUTES); crm_xml_add(args_xml, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET); g_hash_table_foreach(action->extra, hash2field, args_xml); if ((rsc != NULL) && (action->node != NULL)) { // Get the resource instance attributes, evaluated properly for node GHashTable *params = pe_rsc_params(rsc, action->node, rsc->priv->scheduler); pcmk__substitute_remote_addr(rsc, params); g_hash_table_foreach(params, hash2smartfield, args_xml); } else if ((rsc != NULL) && (rsc->priv->variant <= pcmk__rsc_variant_primitive)) { GHashTable *params = pe_rsc_params(rsc, NULL, rsc->priv->scheduler); g_hash_table_foreach(params, hash2smartfield, args_xml); } g_hash_table_foreach(action->meta, hash2metafield, args_xml); if (rsc != NULL) { pcmk_resource_t *parent = rsc; while (parent != NULL) { parent->priv->cmds->add_graph_meta(parent, args_xml); parent = parent->priv->parent; } pcmk__add_guest_meta_to_xml(args_xml, action); } pcmk__xe_sort_attrs(args_xml); } /*! * \internal * \brief Create the transition graph XML for a scheduled action * * \param[in,out] parent Parent XML element to add action to * \param[in,out] action Scheduled action * \param[in] skip_details If false, add action details as sub-elements * \param[in] scheduler Scheduler data */ static void create_graph_action(xmlNode *parent, pcmk_action_t *action, bool skip_details, const pcmk_scheduler_t *scheduler) { bool needs_node_info = true; bool needs_maintenance_info = false; xmlNode *action_xml = NULL; if ((action == NULL) || (scheduler == NULL)) { return; } // Create the top-level element based on task if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH, pcmk__str_none)) { /* All fences need node info; guest node fences are pseudo-events */ if (pcmk_is_set(action->flags, pcmk__action_pseudo)) { action_xml = pcmk__xe_create(parent, PCMK__XE_PSEUDO_EVENT); } else { action_xml = pcmk__xe_create(parent, PCMK__XE_CRM_EVENT); } } else if (pcmk__str_any_of(action->task, PCMK_ACTION_DO_SHUTDOWN, PCMK_ACTION_CLEAR_FAILCOUNT, NULL)) { action_xml = pcmk__xe_create(parent, PCMK__XE_CRM_EVENT); } else if (pcmk__str_eq(action->task, PCMK_ACTION_LRM_DELETE, pcmk__str_none)) { // CIB-only clean-up for shutdown locks action_xml = pcmk__xe_create(parent, PCMK__XE_CRM_EVENT); crm_xml_add(action_xml, PCMK__XA_MODE, PCMK__VALUE_CIB); } else if (pcmk_is_set(action->flags, pcmk__action_pseudo)) { if (pcmk__str_eq(action->task, PCMK_ACTION_MAINTENANCE_NODES, pcmk__str_none)) { needs_maintenance_info = true; } action_xml = pcmk__xe_create(parent, PCMK__XE_PSEUDO_EVENT); needs_node_info = false; } else { action_xml = pcmk__xe_create(parent, PCMK__XE_RSC_OP); } crm_xml_add_int(action_xml, PCMK_XA_ID, action->id); crm_xml_add(action_xml, PCMK_XA_OPERATION, action->task); if ((action->rsc != NULL) && (action->rsc->priv->history_id != NULL)) { char *clone_key = NULL; guint interval_ms; if (pcmk__guint_from_hash(action->meta, PCMK_META_INTERVAL, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } clone_key = clone_op_key(action, interval_ms); crm_xml_add(action_xml, PCMK__XA_OPERATION_KEY, clone_key); crm_xml_add(action_xml, "internal_" PCMK__XA_OPERATION_KEY, action->uuid); free(clone_key); } else { crm_xml_add(action_xml, PCMK__XA_OPERATION_KEY, action->uuid); } if (needs_node_info && (action->node != NULL)) { add_node_details(action, action_xml); pcmk__insert_dup(action->meta, PCMK__META_ON_NODE, action->node->priv->name); pcmk__insert_dup(action->meta, PCMK__META_ON_NODE_UUID, action->node->priv->id); } if (skip_details) { return; } if ((action->rsc != NULL) && !pcmk_is_set(action->flags, pcmk__action_pseudo)) { // This is a real resource action, so add resource details add_resource_details(action, action_xml); } /* List any attributes in effect */ add_action_attributes(action, action_xml); /* List any nodes this action is expected to make down */ if (needs_node_info && (action->node != NULL)) { add_downed_nodes(action_xml, action); } if (needs_maintenance_info) { add_maintenance_nodes(action_xml, scheduler); } } /*! * \internal * \brief Check whether an action should be added to the transition graph * * \param[in,out] action Action to check * * \return true if action should be added to graph, otherwise false */ static bool should_add_action_to_graph(pcmk_action_t *action) { if (!pcmk_is_set(action->flags, pcmk__action_runnable)) { crm_trace("Ignoring action %s (%d): unrunnable", action->uuid, action->id); return false; } if (pcmk_is_set(action->flags, pcmk__action_optional) && !pcmk_is_set(action->flags, pcmk__action_always_in_graph)) { crm_trace("Ignoring action %s (%d): optional", action->uuid, action->id); return false; } /* Actions for unmanaged resources should be excluded from the graph, * with the exception of monitors and cancellation of recurring monitors. */ if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pcmk__rsc_managed) && !pcmk__str_eq(action->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { const char *interval_ms_s; /* A cancellation of a recurring monitor will get here because the task * is cancel rather than monitor, but the interval can still be used to * recognize it. The interval has been normalized to milliseconds by * this point, so a string comparison is sufficient. */ interval_ms_s = g_hash_table_lookup(action->meta, PCMK_META_INTERVAL); if (pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches)) { crm_trace("Ignoring action %s (%d): for unmanaged resource (%s)", action->uuid, action->id, action->rsc->id); return false; } } /* Always add pseudo-actions, fence actions, and shutdown actions (already * determined to be required and runnable by this point) */ if (pcmk_is_set(action->flags, pcmk__action_pseudo) || pcmk__strcase_any_of(action->task, PCMK_ACTION_STONITH, PCMK_ACTION_DO_SHUTDOWN, NULL)) { return true; } if (action->node == NULL) { pcmk__sched_err(action->scheduler, "Skipping action %s (%d) " "because it was not assigned to a node (bug?)", action->uuid, action->id); pcmk__log_action("Unassigned", action, false); return false; } if (pcmk_is_set(action->flags, pcmk__action_on_dc)) { crm_trace("Action %s (%d) should be dumped: " "can run on DC instead of %s", action->uuid, action->id, pcmk__node_name(action->node)); } else if (pcmk__is_guest_or_bundle_node(action->node) && !pcmk_is_set(action->node->priv->flags, pcmk__node_remote_reset)) { crm_trace("Action %s (%d) should be dumped: " "assuming will be runnable on guest %s", action->uuid, action->id, pcmk__node_name(action->node)); } else if (!action->node->details->online) { pcmk__sched_err(action->scheduler, "Skipping action %s (%d) " "because it was scheduled for offline node (bug?)", action->uuid, action->id); pcmk__log_action("Offline node", action, false); return false; } else if (action->node->details->unclean) { pcmk__sched_err(action->scheduler, "Skipping action %s (%d) " "because it was scheduled for unclean node (bug?)", action->uuid, action->id); pcmk__log_action("Unclean node", action, false); return false; } return true; } /*! * \internal * \brief Check whether an ordering's flags can change an action * * \param[in] ordering Ordering to check * * \return true if ordering has flags that can change an action, false otherwise */ static bool ordering_can_change_actions(const pcmk__related_action_t *ordering) { return pcmk_any_flags_set(ordering->flags, ~(pcmk__ar_then_implies_first_graphed |pcmk__ar_first_implies_then_graphed |pcmk__ar_ordered)); } /*! * \internal * \brief Check whether an action input should be in the transition graph * * \param[in] action Action to check * \param[in,out] input Action input to check * * \return true if input should be in graph, false otherwise * \note This function may not only check an input, but disable it under certian * circumstances (load or anti-colocation orderings that are not needed). */ static bool should_add_input_to_graph(const pcmk_action_t *action, pcmk__related_action_t *input) { if (input->graphed) { return true; } if (input->flags == pcmk__ar_none) { crm_trace("Ignoring %s (%d) input %s (%d): " "ordering disabled", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pcmk__action_runnable) && !ordering_can_change_actions(input)) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pcmk__action_runnable) && pcmk_is_set(input->flags, pcmk__ar_min_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "minimum number of instances required but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->flags, pcmk__ar_unmigratable_then_blocks) && !pcmk_is_set(input->action->flags, pcmk__action_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "input blocked if 'then' unmigratable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->flags, pcmk__ar_if_first_unmigratable) && pcmk_is_set(input->action->flags, pcmk__action_migratable)) { crm_trace("Ignoring %s (%d) input %s (%d): ordering applies " "only if input is unmigratable, but it is migratable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if ((input->flags == pcmk__ar_ordered) && pcmk_is_set(input->action->flags, pcmk__action_migratable) - && pcmk__ends_with(input->action->uuid, "_stop_0")) { + && g_str_has_suffix(input->action->uuid, "_stop_0")) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional but stop in migration", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (input->flags == pcmk__ar_if_on_same_node_or_target) { pcmk_node_t *input_node = input->action->node; if ((action->rsc != NULL) && pcmk__str_eq(action->task, PCMK_ACTION_MIGRATE_TO, pcmk__str_none)) { pcmk_node_t *assigned = action->rsc->priv->assigned_node; /* For load_stopped -> migrate_to orderings, we care about where * the resource has been assigned, not where migrate_to will be * executed. */ if (!pcmk__same_node(input_node, assigned)) { crm_trace("Ignoring %s (%d) input %s (%d): " "migration target %s is not same as input node %s", action->uuid, action->id, input->action->uuid, input->action->id, (assigned? assigned->priv->name : ""), (input_node? input_node->priv->name : "")); input->flags = pcmk__ar_none; return false; } } else if (!pcmk__same_node(input_node, action->node)) { crm_trace("Ignoring %s (%d) input %s (%d): " "not on same node (%s vs %s)", action->uuid, action->id, input->action->uuid, input->action->id, (action->node? action->node->priv->name : ""), (input_node? input_node->priv->name : "")); input->flags = pcmk__ar_none; return false; } else if (pcmk_is_set(input->action->flags, pcmk__action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "ordering optional", action->uuid, action->id, input->action->uuid, input->action->id); input->flags = pcmk__ar_none; return false; } } else if (input->flags == pcmk__ar_if_required_on_same_node) { if (input->action->node && action->node && !pcmk__same_node(input->action->node, action->node)) { crm_trace("Ignoring %s (%d) input %s (%d): " "not on same node (%s vs %s)", action->uuid, action->id, input->action->uuid, input->action->id, pcmk__node_name(action->node), pcmk__node_name(input->action->node)); input->flags = pcmk__ar_none; return false; } else if (pcmk_is_set(input->action->flags, pcmk__action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): optional", action->uuid, action->id, input->action->uuid, input->action->id); input->flags = pcmk__ar_none; return false; } } else if (input->action->rsc && input->action->rsc != action->rsc && pcmk_is_set(input->action->rsc->flags, pcmk__rsc_failed) && !pcmk_is_set(input->action->rsc->flags, pcmk__rsc_managed) - && pcmk__ends_with(input->action->uuid, "_stop_0") + && g_str_has_suffix(input->action->uuid, "_stop_0") && pcmk__is_clone(action->rsc)) { crm_warn("Ignoring requirement that %s complete before %s:" " unmanaged failed resources cannot prevent clone shutdown", input->action->uuid, action->uuid); return false; } else if (pcmk_is_set(input->action->flags, pcmk__action_optional) && !pcmk_any_flags_set(input->action->flags, pcmk__action_always_in_graph |pcmk__action_added_to_graph) && !should_add_action_to_graph(input->action)) { crm_trace("Ignoring %s (%d) input %s (%d): " "input optional", action->uuid, action->id, input->action->uuid, input->action->id); return false; } crm_trace("%s (%d) input %s %s (%d) on %s should be dumped: %s %s %#.6x", action->uuid, action->id, action_type_str(input->action->flags), input->action->uuid, input->action->id, action_node_str(input->action), action_runnable_str(input->action->flags), action_optional_str(input->action->flags), input->flags); return true; } /*! * \internal * \brief Check whether an ordering creates an ordering loop * * \param[in] init_action "First" action in ordering * \param[in] action Callers should always set this the same as * \p init_action (this function may use a different * value for recursive calls) * \param[in,out] input Action wrapper for "then" action in ordering * * \return true if the ordering creates a loop, otherwise false */ bool pcmk__graph_has_loop(const pcmk_action_t *init_action, const pcmk_action_t *action, pcmk__related_action_t *input) { bool has_loop = false; if (pcmk_is_set(input->action->flags, pcmk__action_detect_loop)) { crm_trace("Breaking tracking loop: %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->priv->name : "", action->uuid, action->node? action->node->priv->name : "", input->flags); return false; } // Don't need to check inputs that won't be used if (!should_add_input_to_graph(action, input)) { return false; } if (input->action == init_action) { crm_debug("Input loop found in %s@%s ->...-> %s@%s", action->uuid, action->node? action->node->priv->name : "", init_action->uuid, init_action->node? init_action->node->priv->name : ""); return true; } pcmk__set_action_flags(input->action, pcmk__action_detect_loop); crm_trace("Checking inputs of action %s@%s input %s@%s (%#.6x)" "for graph loop with %s@%s ", action->uuid, action->node? action->node->priv->name : "", input->action->uuid, input->action->node? input->action->node->priv->name : "", input->flags, init_action->uuid, init_action->node? init_action->node->priv->name : ""); // Recursively check input itself for loops for (GList *iter = input->action->actions_before; iter != NULL; iter = iter->next) { if (pcmk__graph_has_loop(init_action, input->action, (pcmk__related_action_t *) iter->data)) { // Recursive call already logged a debug message has_loop = true; break; } } pcmk__clear_action_flags(input->action, pcmk__action_detect_loop); if (!has_loop) { crm_trace("No input loop found in %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->priv->name : "", action->uuid, action->node? action->node->priv->name : "", input->flags); } return has_loop; } /*! * \internal * \brief Create a synapse XML element for a transition graph * * \param[in] action Action that synapse is for * \param[in,out] scheduler Scheduler data containing graph * * \return Newly added XML element for new graph synapse */ static xmlNode * create_graph_synapse(const pcmk_action_t *action, pcmk_scheduler_t *scheduler) { int synapse_priority = 0; xmlNode *syn = pcmk__xe_create(scheduler->priv->graph, PCMK__XE_SYNAPSE); crm_xml_add_int(syn, PCMK_XA_ID, scheduler->priv->synapse_count++); if (action->rsc != NULL) { synapse_priority = action->rsc->priv->priority; } if (action->priority > synapse_priority) { synapse_priority = action->priority; } if (synapse_priority > 0) { crm_xml_add_int(syn, PCMK__XA_PRIORITY, synapse_priority); } return syn; } /*! * \internal * \brief Add an action to the transition graph XML if appropriate * * \param[in,out] data Action to possibly add * \param[in,out] user_data Scheduler data * * \note This will de-duplicate the action inputs, meaning that the * pcmk__related_action_t:type flags can no longer be relied on to retain * their original settings. That means this MUST be called after * pcmk__apply_orderings() is complete, and nothing after this should rely * on those type flags. (For example, some code looks for type equal to * some flag rather than whether the flag is set, and some code looks for * particular combinations of flags -- such code must be done before * pcmk__create_graph().) */ static void add_action_to_graph(gpointer data, gpointer user_data) { pcmk_action_t *action = (pcmk_action_t *) data; pcmk_scheduler_t *scheduler = (pcmk_scheduler_t *) user_data; xmlNode *syn = NULL; xmlNode *set = NULL; xmlNode *in = NULL; /* If we haven't already, de-duplicate inputs (even if we won't be adding * the action to the graph, so that crm_simulate's dot graphs don't have * duplicates). */ if (!pcmk_is_set(action->flags, pcmk__action_inputs_deduplicated)) { pcmk__deduplicate_action_inputs(action); pcmk__set_action_flags(action, pcmk__action_inputs_deduplicated); } if (pcmk_is_set(action->flags, pcmk__action_added_to_graph) || !should_add_action_to_graph(action)) { return; // Already added, or shouldn't be } pcmk__set_action_flags(action, pcmk__action_added_to_graph); crm_trace("Adding action %d (%s%s%s) to graph", action->id, action->uuid, ((action->node == NULL)? "" : " on "), ((action->node == NULL)? "" : action->node->priv->name)); syn = create_graph_synapse(action, scheduler); set = pcmk__xe_create(syn, PCMK__XE_ACTION_SET); in = pcmk__xe_create(syn, PCMK__XE_INPUTS); create_graph_action(set, action, false, scheduler); for (GList *lpc = action->actions_before; lpc != NULL; lpc = lpc->next) { pcmk__related_action_t *input = lpc->data; if (should_add_input_to_graph(action, input)) { xmlNode *input_xml = pcmk__xe_create(in, PCMK__XE_TRIGGER); input->graphed = true; create_graph_action(input_xml, input->action, true, scheduler); } } } static int transition_id = 0; /*! * \internal * \brief Log a message after calculating a transition * * \param[in] scheduler Scheduler data * \param[in] filename Where transition input is stored */ void pcmk__log_transition_summary(const pcmk_scheduler_t *scheduler, const char *filename) { if (pcmk_is_set(scheduler->flags, pcmk__sched_processing_error) || pcmk__config_has_error) { crm_err("Calculated transition %d (with errors)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else if (pcmk_is_set(scheduler->flags, pcmk__sched_processing_warning) || pcmk__config_has_warning) { crm_warn("Calculated transition %d (with warnings)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else { crm_notice("Calculated transition %d%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } if (pcmk__config_has_error) { crm_notice("Configuration errors found during scheduler processing," " please run \"crm_verify -L\" to identify issues"); } } /*! * \internal * \brief Add a resource's actions to the transition graph * * \param[in,out] rsc Resource whose actions should be added */ void pcmk__add_rsc_actions_to_graph(pcmk_resource_t *rsc) { GList *iter = NULL; pcmk__assert(rsc != NULL); pcmk__rsc_trace(rsc, "Adding actions for %s to graph", rsc->id); // First add the resource's own actions g_list_foreach(rsc->priv->actions, add_action_to_graph, rsc->priv->scheduler); // Then recursively add its children's actions (appropriate to variant) for (iter = rsc->priv->children; iter != NULL; iter = iter->next) { pcmk_resource_t *child_rsc = (pcmk_resource_t *) iter->data; child_rsc->priv->cmds->add_actions_to_graph(child_rsc); } } /*! * \internal * \brief Create a transition graph with all cluster actions needed * * \param[in,out] scheduler Scheduler data */ void pcmk__create_graph(pcmk_scheduler_t *scheduler) { GList *iter = NULL; const char *value = NULL; long long limit = 0LL; GHashTable *config_hash = scheduler->priv->options; int rc = pcmk_rc_ok; transition_id++; crm_trace("Creating transition graph %d", transition_id); scheduler->priv->graph = pcmk__xe_create(NULL, PCMK__XE_TRANSITION_GRAPH); value = pcmk__cluster_option(config_hash, PCMK_OPT_CLUSTER_DELAY); crm_xml_add(scheduler->priv->graph, PCMK_OPT_CLUSTER_DELAY, value); value = pcmk__cluster_option(config_hash, PCMK_OPT_STONITH_TIMEOUT); crm_xml_add(scheduler->priv->graph, PCMK_OPT_STONITH_TIMEOUT, value); crm_xml_add(scheduler->priv->graph, PCMK__XA_FAILED_STOP_OFFSET, PCMK_VALUE_INFINITY); if (pcmk_is_set(scheduler->flags, pcmk__sched_start_failure_fatal)) { crm_xml_add(scheduler->priv->graph, PCMK__XA_FAILED_START_OFFSET, PCMK_VALUE_INFINITY); } else { crm_xml_add(scheduler->priv->graph, PCMK__XA_FAILED_START_OFFSET, "1"); } value = pcmk__cluster_option(config_hash, PCMK_OPT_BATCH_LIMIT); crm_xml_add(scheduler->priv->graph, PCMK_OPT_BATCH_LIMIT, value); crm_xml_add_int(scheduler->priv->graph, "transition_id", transition_id); value = pcmk__cluster_option(config_hash, PCMK_OPT_MIGRATION_LIMIT); rc = pcmk__scan_ll(value, &limit, 0LL); if (rc != pcmk_rc_ok) { crm_warn("Ignoring invalid value '%s' for " PCMK_OPT_MIGRATION_LIMIT ": %s", value, pcmk_rc_str(rc)); } else if (limit > 0) { crm_xml_add(scheduler->priv->graph, PCMK_OPT_MIGRATION_LIMIT, value); } if (scheduler->priv->recheck_by > 0) { char *recheck_epoch = NULL; recheck_epoch = crm_strdup_printf("%llu", (unsigned long long) scheduler->priv->recheck_by); crm_xml_add(scheduler->priv->graph, "recheck-by", recheck_epoch); free(recheck_epoch); } /* The following code will de-duplicate action inputs, so nothing past this * should rely on the action input type flags retaining their original * values. */ // Add resource actions to graph for (iter = scheduler->priv->resources; iter != NULL; iter = iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data; pcmk__rsc_trace(rsc, "Processing actions for %s", rsc->id); rsc->priv->cmds->add_actions_to_graph(rsc); } // Add pseudo-action for list of nodes with maintenance state update add_maintenance_update(scheduler); // Add non-resource (node) actions for (iter = scheduler->priv->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = (pcmk_action_t *) iter->data; if ((action->rsc != NULL) && (action->node != NULL) && action->node->details->shutdown && !pcmk_is_set(action->rsc->flags, pcmk__rsc_maintenance) && !pcmk_any_flags_set(action->flags, pcmk__action_optional|pcmk__action_runnable) && pcmk__str_eq(action->task, PCMK_ACTION_STOP, pcmk__str_none)) { /* Eventually we should just ignore the 'fence' case, but for now * it's the best way to detect (in CTS) when CIB resource updates * are being lost. */ if (pcmk_is_set(scheduler->flags, pcmk__sched_quorate) || (scheduler->no_quorum_policy == pcmk_no_quorum_ignore)) { const bool managed = pcmk_is_set(action->rsc->flags, pcmk__rsc_managed); const bool failed = pcmk_is_set(action->rsc->flags, pcmk__rsc_failed); crm_crit("Cannot %s %s because of %s:%s%s (%s)", action->node->details->unclean? "fence" : "shut down", pcmk__node_name(action->node), action->rsc->id, (managed? " blocked" : " unmanaged"), (failed? " failed" : ""), action->uuid); } } add_action_to_graph((gpointer) action, (gpointer) scheduler); } crm_log_xml_trace(scheduler->priv->graph, "graph"); } diff --git a/lib/pacemaker/pcmk_resource.c b/lib/pacemaker/pcmk_resource.c index 6a3e50ae57..c783ac499d 100644 --- a/lib/pacemaker/pcmk_resource.c +++ b/lib/pacemaker/pcmk_resource.c @@ -1,253 +1,253 @@ /* * Copyright 2021-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include // xmlNode #include #include #include #include #include #include #include // Search path for resource operation history (takes node name and resource ID) #define XPATH_OP_HISTORY "//" PCMK_XE_STATUS \ "/" PCMK__XE_NODE_STATE \ "[@" PCMK_XA_UNAME "='%s']" \ "/" PCMK__XE_LRM "/" PCMK__XE_LRM_RESOURCES \ "/" PCMK__XE_LRM_RESOURCE "[@" PCMK_XA_ID "='%s']" static xmlNode * best_op(const pcmk_resource_t *rsc, const pcmk_node_t *node) { char *xpath = NULL; xmlNode *history = NULL; xmlNode *best = NULL; bool best_effective_op = false; guint best_interval = 0; bool best_failure = false; const char *best_digest = NULL; // Find node's resource history xpath = crm_strdup_printf(XPATH_OP_HISTORY, node->priv->name, rsc->id); history = pcmk__xpath_find_one(rsc->priv->scheduler->input->doc, xpath, LOG_NEVER); free(xpath); // Examine each history entry for (xmlNode *lrm_rsc_op = pcmk__xe_first_child(history, PCMK__XE_LRM_RSC_OP, NULL, NULL); lrm_rsc_op != NULL; lrm_rsc_op = pcmk__xe_next(lrm_rsc_op, PCMK__XE_LRM_RSC_OP)) { const char *digest = pcmk__xe_get(lrm_rsc_op, PCMK__XA_OP_RESTART_DIGEST); guint interval_ms = 0; const char *task = pcmk__xe_get(lrm_rsc_op, PCMK_XA_OPERATION); bool effective_op = false; - bool failure = pcmk__ends_with(pcmk__xe_id(lrm_rsc_op), - "_last_failure_0"); + const char *id = pcmk__xe_id(lrm_rsc_op); + bool failure = (id != NULL) && g_str_has_suffix(id, "_last_failure_0"); pcmk__xe_get_guint(lrm_rsc_op, PCMK_META_INTERVAL, &interval_ms); effective_op = interval_ms == 0 && pcmk__strcase_any_of(task, PCMK_ACTION_MONITOR, PCMK_ACTION_START, PCMK_ACTION_PROMOTE, PCMK_ACTION_MIGRATE_FROM, NULL); if (best == NULL) { goto is_best; } if (best_effective_op) { // Do not use an ineffective op if there's an effective one. if (!effective_op) { continue; } // Do not use an ineffective non-recurring op if there's a recurring one } else if (best_interval != 0 && !effective_op && interval_ms == 0) { continue; } // Do not use last failure if there's a successful one. if (!best_failure && failure) { continue; } // Do not use an op without a restart digest if there's one with. if (best_digest != NULL && digest == NULL) { continue; } // Do not use an older op if there's a newer one. if (pe__is_newer_op(best, lrm_rsc_op) > 0) { continue; } is_best: best = lrm_rsc_op; best_effective_op = effective_op; best_interval = interval_ms; best_failure = failure; best_digest = digest; } return best; } /*! * \internal * \brief Remove a resource * * \param[in,out] cib An open connection to the CIB * \param[in] cib_opts Options to use in the CIB operation call * \param[in] rsc_id Resource to remove * \param[in] rsc_type Type of the resource ("primitive", "group", etc.) * * \return Standard Pacemaker return code */ int pcmk__resource_delete(cib_t *cib, uint32_t cib_opts, const char *rsc_id, const char *rsc_type) { int rc = pcmk_rc_ok; xmlNode *msg_data = NULL; if (cib == NULL) { return ENOTCONN; } if (rsc_id == NULL || rsc_type == NULL) { return EINVAL; } msg_data = pcmk__xe_create(NULL, rsc_type); crm_xml_add(msg_data, PCMK_XA_ID, rsc_id); rc = cib->cmds->remove(cib, PCMK_XE_RESOURCES, msg_data, cib_opts); rc = pcmk_legacy2rc(rc); pcmk__xml_free(msg_data); return rc; } int pcmk_resource_delete(xmlNodePtr *xml, const char *rsc_id, const char *rsc_type) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; uint32_t cib_opts = cib_sync_call; cib_t *cib = NULL; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } cib = cib_new(); if (cib == NULL) { rc = pcmk_rc_cib_corrupt; goto done; } rc = cib->cmds->signon(cib, crm_system_name, cib_command); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { goto done; } rc = pcmk__resource_delete(cib, cib_opts, rsc_id, rsc_type); done: if (cib != NULL) { cib__clean_up_connection(&cib); } pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /*! * \internal * \brief Calculate and output resource operation digests * * \param[in,out] out Output object * \param[in,out] rsc Resource to calculate digests for * \param[in] node Node whose operation history should be used * \param[in] overrides Hash table of configuration parameters to override * * \return Standard Pacemaker return code */ int pcmk__resource_digests(pcmk__output_t *out, pcmk_resource_t *rsc, const pcmk_node_t *node, GHashTable *overrides) { const char *task = NULL; xmlNode *xml_op = NULL; pcmk__op_digest_t *digests = NULL; guint interval_ms = 0; int rc = pcmk_rc_ok; if ((out == NULL) || (rsc == NULL) || (node == NULL)) { return EINVAL; } if (!pcmk__is_primitive(rsc)) { // Only primitives get operation digests return EOPNOTSUPP; } // Find XML of operation history to use xml_op = best_op(rsc, node); // Generate an operation key if (xml_op != NULL) { task = pcmk__xe_get(xml_op, PCMK_XA_OPERATION); pcmk__xe_get_guint(xml_op, PCMK_META_INTERVAL, &interval_ms); } if (task == NULL) { // Assume start if no history is available task = PCMK_ACTION_START; interval_ms = 0; } // Calculate and show digests digests = pe__calculate_digests(rsc, task, &interval_ms, node, xml_op, overrides, true, rsc->priv->scheduler); rc = out->message(out, "digests", rsc, node, task, interval_ms, digests); pe__free_digests(digests); return rc; } int pcmk_resource_digests(xmlNodePtr *xml, pcmk_resource_t *rsc, const pcmk_node_t *node, GHashTable *overrides) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__resource_digests(out, rsc, node, overrides); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } diff --git a/lib/pacemaker/pcmk_sched_instances.c b/lib/pacemaker/pcmk_sched_instances.c index f2bc1a432c..b5b5af5c13 100644 --- a/lib/pacemaker/pcmk_sched_instances.c +++ b/lib/pacemaker/pcmk_sched_instances.c @@ -1,1713 +1,1717 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ /* This file is intended for code usable with both clone instances and bundle * replica containers. */ #include + +#include // g_str_has_suffix() #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Check whether a node is allowed to run an instance * * \param[in] instance Clone instance or bundle container to check * \param[in] node Node to check * \param[in] max_per_node Maximum number of instances allowed to run on a node * * \return true if \p node is allowed to run \p instance, otherwise false */ static bool can_run_instance(const pcmk_resource_t *instance, const pcmk_node_t *node, int max_per_node) { pcmk_node_t *allowed_node = NULL; if (pcmk_is_set(instance->flags, pcmk__rsc_removed)) { pcmk__rsc_trace(instance, "%s cannot run on %s: orphaned", instance->id, pcmk__node_name(node)); return false; } if (!pcmk__node_available(node, false, false)) { pcmk__rsc_trace(instance, "%s cannot run on %s: node cannot run resources", instance->id, pcmk__node_name(node)); return false; } allowed_node = pcmk__top_allowed_node(instance, node); if (allowed_node == NULL) { crm_warn("%s cannot run on %s: node not allowed", instance->id, pcmk__node_name(node)); return false; } if (allowed_node->assign->score < 0) { pcmk__rsc_trace(instance, "%s cannot run on %s: parent score is %s there", instance->id, pcmk__node_name(node), pcmk_readable_score(allowed_node->assign->score)); return false; } if (allowed_node->assign->count >= max_per_node) { pcmk__rsc_trace(instance, "%s cannot run on %s: node already has %d instance%s", instance->id, pcmk__node_name(node), max_per_node, pcmk__plural_s(max_per_node)); return false; } pcmk__rsc_trace(instance, "%s can run on %s (%d already running)", instance->id, pcmk__node_name(node), allowed_node->assign->count); return true; } /*! * \internal * \brief Ban a clone instance or bundle replica from unavailable allowed nodes * * \param[in,out] instance Clone instance or bundle replica to ban * \param[in] max_per_node Maximum instances allowed to run on a node */ static void ban_unavailable_allowed_nodes(pcmk_resource_t *instance, int max_per_node) { if (instance->priv->allowed_nodes != NULL) { GHashTableIter iter; pcmk_node_t *node = NULL; g_hash_table_iter_init(&iter, instance->priv->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (!can_run_instance(instance, node, max_per_node)) { pcmk__rsc_trace(instance, "Banning %s from unavailable node %s", instance->id, pcmk__node_name(node)); node->assign->score = -PCMK_SCORE_INFINITY; for (GList *child_iter = instance->priv->children; child_iter != NULL; child_iter = child_iter->next) { pcmk_resource_t *child = child_iter->data; pcmk_node_t *child_node = NULL; child_node = g_hash_table_lookup(child->priv->allowed_nodes, node->priv->id); if (child_node != NULL) { pcmk__rsc_trace(instance, "Banning %s child %s " "from unavailable node %s", instance->id, child->id, pcmk__node_name(node)); child_node->assign->score = -PCMK_SCORE_INFINITY; } } } } } } /*! * \internal * \brief Create a hash table with a single node in it * * \param[in] node Node to copy into new table * * \return Newly created hash table containing a copy of \p node * \note The caller is responsible for freeing the result with * g_hash_table_destroy(). */ static GHashTable * new_node_table(pcmk_node_t *node) { GHashTable *table = pcmk__strkey_table(NULL, pcmk__free_node_copy); node = pe__copy_node(node); g_hash_table_insert(table, (gpointer) node->priv->id, node); return table; } /*! * \internal * \brief Apply a resource's parent's colocation scores to a node table * * \param[in] rsc Resource whose colocations should be applied * \param[in,out] nodes Node table to apply colocations to */ static void apply_parent_colocations(const pcmk_resource_t *rsc, GHashTable **nodes) { GList *colocations = pcmk__this_with_colocations(rsc); for (const GList *iter = colocations; iter != NULL; iter = iter->next) { const pcmk__colocation_t *colocation = iter->data; pcmk_resource_t *other = colocation->primary; float factor = colocation->score / (float) PCMK_SCORE_INFINITY; other->priv->cmds->add_colocated_node_scores(other, rsc, rsc->id, nodes, colocation, factor, pcmk__coloc_select_default); } g_list_free(colocations); colocations = pcmk__with_this_colocations(rsc); for (const GList *iter = colocations; iter != NULL; iter = iter->next) { const pcmk__colocation_t *colocation = iter->data; pcmk_resource_t *other = colocation->dependent; float factor = colocation->score / (float) PCMK_SCORE_INFINITY; if (!pcmk__colocation_has_influence(colocation, rsc)) { continue; } other->priv->cmds->add_colocated_node_scores(other, rsc, rsc->id, nodes, colocation, factor, pcmk__coloc_select_nonnegative); } g_list_free(colocations); } /*! * \internal * \brief Compare clone or bundle instances based on colocation scores * * Determine the relative order in which two clone or bundle instances should be * assigned to nodes, considering the scores of colocation constraints directly * or indirectly involving them. * * \param[in] instance1 First instance to compare * \param[in] instance2 Second instance to compare * * \return A negative number if \p instance1 should be assigned first, * a positive number if \p instance2 should be assigned first, * or 0 if assignment order doesn't matter */ static int cmp_instance_by_colocation(const pcmk_resource_t *instance1, const pcmk_resource_t *instance2) { int rc = 0; pcmk_node_t *node1 = NULL; pcmk_node_t *node2 = NULL; pcmk_node_t *current_node1 = pcmk__current_node(instance1); pcmk_node_t *current_node2 = pcmk__current_node(instance2); GHashTable *colocated_scores1 = NULL; GHashTable *colocated_scores2 = NULL; pcmk__assert((instance1 != NULL) && (instance1->priv->parent != NULL) && (instance2 != NULL) && (instance2->priv->parent != NULL) && (current_node1 != NULL) && (current_node2 != NULL)); // Create node tables initialized with each node colocated_scores1 = new_node_table(current_node1); colocated_scores2 = new_node_table(current_node2); // Apply parental colocations apply_parent_colocations(instance1, &colocated_scores1); apply_parent_colocations(instance2, &colocated_scores2); // Find original nodes again, with scores updated for colocations node1 = g_hash_table_lookup(colocated_scores1, current_node1->priv->id); node2 = g_hash_table_lookup(colocated_scores2, current_node2->priv->id); // Compare nodes by updated scores if (node1->assign->score < node2->assign->score) { crm_trace("Assign %s (%d on %s) after %s (%d on %s)", instance1->id, node1->assign->score, pcmk__node_name(node1), instance2->id, node2->assign->score, pcmk__node_name(node2)); rc = 1; } else if (node1->assign->score > node2->assign->score) { crm_trace("Assign %s (%d on %s) before %s (%d on %s)", instance1->id, node1->assign->score, pcmk__node_name(node1), instance2->id, node2->assign->score, pcmk__node_name(node2)); rc = -1; } g_hash_table_destroy(colocated_scores1); g_hash_table_destroy(colocated_scores2); return rc; } /*! * \internal * \brief Check whether a resource or any of its children are failed * * \param[in] rsc Resource to check * * \return true if \p rsc or any of its children are failed, otherwise false */ static bool did_fail(const pcmk_resource_t *rsc) { if (pcmk_is_set(rsc->flags, pcmk__rsc_failed)) { return true; } for (GList *iter = rsc->priv->children; iter != NULL; iter = iter->next) { if (did_fail((const pcmk_resource_t *) iter->data)) { return true; } } return false; } /*! * \internal * \brief Check whether a node is allowed to run a resource * * \param[in] rsc Resource to check * \param[in,out] node Node to check (will be set NULL if not allowed) * * \return true if *node is either NULL or allowed for \p rsc, otherwise false */ static bool node_is_allowed(const pcmk_resource_t *rsc, pcmk_node_t **node) { if (*node != NULL) { pcmk_node_t *allowed = g_hash_table_lookup(rsc->priv->allowed_nodes, (*node)->priv->id); if ((allowed == NULL) || (allowed->assign->score < 0)) { pcmk__rsc_trace(rsc, "%s: current location (%s) is unavailable", rsc->id, pcmk__node_name(*node)); *node = NULL; return false; } } return true; } /*! * \internal * \brief Compare two clone or bundle instances' instance numbers * * \param[in] a First instance to compare * \param[in] b Second instance to compare * * \return A negative number if \p a's instance number is lower, * a positive number if \p b's instance number is lower, * or 0 if their instance numbers are the same */ gint pcmk__cmp_instance_number(gconstpointer a, gconstpointer b) { const pcmk_resource_t *instance1 = (const pcmk_resource_t *) a; const pcmk_resource_t *instance2 = (const pcmk_resource_t *) b; char *div1 = NULL; char *div2 = NULL; pcmk__assert((instance1 != NULL) && (instance2 != NULL)); // Clone numbers are after a colon, bundle numbers after a dash div1 = strrchr(instance1->id, ':'); if (div1 == NULL) { div1 = strrchr(instance1->id, '-'); } div2 = strrchr(instance2->id, ':'); if (div2 == NULL) { div2 = strrchr(instance2->id, '-'); } pcmk__assert((div1 != NULL) && (div2 != NULL)); return (gint) (strtol(div1 + 1, NULL, 10) - strtol(div2 + 1, NULL, 10)); } /*! * \internal * \brief Compare clone or bundle instances according to assignment order * * Compare two clone or bundle instances according to the order they should be * assigned to nodes, preferring (in order): * * - Active instance that is less multiply active * - Instance that is not active on a disallowed node * - Instance with higher configured priority * - Active instance whose current node can run resources * - Active instance whose parent is allowed on current node * - Active instance whose current node has fewer other instances * - Active instance * - Instance that isn't failed * - Instance whose colocations result in higher score on current node * - Instance with lower ID in lexicographic order * * \param[in] a First instance to compare * \param[in] b Second instance to compare * * \return A negative number if \p a should be assigned first, * a positive number if \p b should be assigned first, * or 0 if assignment order doesn't matter */ gint pcmk__cmp_instance(gconstpointer a, gconstpointer b) { int rc = 0; pcmk_node_t *node1 = NULL; pcmk_node_t *node2 = NULL; unsigned int nnodes1 = 0; unsigned int nnodes2 = 0; bool can1 = true; bool can2 = true; const pcmk_resource_t *instance1 = (const pcmk_resource_t *) a; const pcmk_resource_t *instance2 = (const pcmk_resource_t *) b; pcmk__assert((instance1 != NULL) && (instance2 != NULL)); node1 = instance1->priv->fns->active_node(instance1, &nnodes1, NULL); node2 = instance2->priv->fns->active_node(instance2, &nnodes2, NULL); /* If both instances are running and at least one is multiply * active, prefer instance that's running on fewer nodes. */ if ((nnodes1 > 0) && (nnodes2 > 0)) { if (nnodes1 < nnodes2) { crm_trace("Assign %s (active on %d) before %s (active on %d): " "less multiply active", instance1->id, nnodes1, instance2->id, nnodes2); return -1; } else if (nnodes1 > nnodes2) { crm_trace("Assign %s (active on %d) after %s (active on %d): " "more multiply active", instance1->id, nnodes1, instance2->id, nnodes2); return 1; } } /* An instance that is either inactive or active on an allowed node is * preferred over an instance that is active on a no-longer-allowed node. */ can1 = node_is_allowed(instance1, &node1); can2 = node_is_allowed(instance2, &node2); if (can1 && !can2) { crm_trace("Assign %s before %s: not active on a disallowed node", instance1->id, instance2->id); return -1; } else if (!can1 && can2) { crm_trace("Assign %s after %s: active on a disallowed node", instance1->id, instance2->id); return 1; } // Prefer instance with higher configured priority if (instance1->priv->priority > instance2->priv->priority) { crm_trace("Assign %s before %s: priority (%d > %d)", instance1->id, instance2->id, instance1->priv->priority, instance2->priv->priority); return -1; } else if (instance1->priv->priority < instance2->priv->priority) { crm_trace("Assign %s after %s: priority (%d < %d)", instance1->id, instance2->id, instance1->priv->priority, instance2->priv->priority); return 1; } // Prefer active instance if ((node1 == NULL) && (node2 == NULL)) { crm_trace("No assignment preference for %s vs. %s: inactive", instance1->id, instance2->id); return 0; } else if (node1 == NULL) { crm_trace("Assign %s after %s: active", instance1->id, instance2->id); return 1; } else if (node2 == NULL) { crm_trace("Assign %s before %s: active", instance1->id, instance2->id); return -1; } // Prefer instance whose current node can run resources can1 = pcmk__node_available(node1, false, false); can2 = pcmk__node_available(node2, false, false); if (can1 && !can2) { crm_trace("Assign %s before %s: current node can run resources", instance1->id, instance2->id); return -1; } else if (!can1 && can2) { crm_trace("Assign %s after %s: current node can't run resources", instance1->id, instance2->id); return 1; } // Prefer instance whose parent is allowed to run on instance's current node node1 = pcmk__top_allowed_node(instance1, node1); node2 = pcmk__top_allowed_node(instance2, node2); if ((node1 == NULL) && (node2 == NULL)) { crm_trace("No assignment preference for %s vs. %s: " "parent not allowed on either instance's current node", instance1->id, instance2->id); return 0; } else if (node1 == NULL) { crm_trace("Assign %s after %s: parent not allowed on current node", instance1->id, instance2->id); return 1; } else if (node2 == NULL) { crm_trace("Assign %s before %s: parent allowed on current node", instance1->id, instance2->id); return -1; } // Prefer instance whose current node is running fewer other instances if (node1->assign->count < node2->assign->count) { crm_trace("Assign %s before %s: fewer active instances on current node", instance1->id, instance2->id); return -1; } else if (node1->assign->count > node2->assign->count) { crm_trace("Assign %s after %s: more active instances on current node", instance1->id, instance2->id); return 1; } // Prefer instance that isn't failed can1 = did_fail(instance1); can2 = did_fail(instance2); if (!can1 && can2) { crm_trace("Assign %s before %s: not failed", instance1->id, instance2->id); return -1; } else if (can1 && !can2) { crm_trace("Assign %s after %s: failed", instance1->id, instance2->id); return 1; } // Prefer instance with higher cumulative colocation score on current node rc = cmp_instance_by_colocation(instance1, instance2); if (rc != 0) { return rc; } // Prefer instance with lower instance number rc = pcmk__cmp_instance_number(instance1, instance2); if (rc < 0) { crm_trace("Assign %s before %s: instance number", instance1->id, instance2->id); } else if (rc > 0) { crm_trace("Assign %s after %s: instance number", instance1->id, instance2->id); } else { crm_trace("No assignment preference for %s vs. %s", instance1->id, instance2->id); } return rc; } /*! * \internal * \brief Increment the parent's instance count after assigning an instance * * An instance's parent tracks how many instances have been assigned to each * node via its pcmk_node_t:count member. After assigning an instance to a node, * find the corresponding node in the parent's allowed table and increment it. * * \param[in,out] instance Instance whose parent to update * \param[in] assigned_to Node to which the instance was assigned */ static void increment_parent_count(pcmk_resource_t *instance, const pcmk_node_t *assigned_to) { pcmk_node_t *allowed = NULL; if (assigned_to == NULL) { return; } allowed = pcmk__top_allowed_node(instance, assigned_to); if (allowed == NULL) { /* The instance is allowed on the node, but its parent isn't. This * shouldn't be possible if the resource is managed, and we won't be * able to limit the number of instances assigned to the node. */ CRM_LOG_ASSERT(!pcmk_is_set(instance->flags, pcmk__rsc_managed)); } else { allowed->assign->count++; } } /*! * \internal * \brief Assign an instance to a node * * \param[in,out] instance Clone instance or bundle replica container * \param[in] prefer If not NULL, attempt early assignment to this * node, if still the best choice; otherwise, * perform final assignment * \param[in] max_per_node Assign at most this many instances to one node * * \return Node to which \p instance is assigned */ static const pcmk_node_t * assign_instance(pcmk_resource_t *instance, const pcmk_node_t *prefer, int max_per_node) { pcmk_node_t *chosen = NULL; pcmk__rsc_trace(instance, "Assigning %s (preferring %s)", instance->id, ((prefer == NULL)? "no node" : prefer->priv->name)); if (pcmk_is_set(instance->flags, pcmk__rsc_assigning)) { pcmk__rsc_debug(instance, "Assignment loop detected involving %s colocations", instance->id); return NULL; } ban_unavailable_allowed_nodes(instance, max_per_node); // Failed early assignments are reversible (stop_if_fail=false) chosen = instance->priv->cmds->assign(instance, prefer, (prefer == NULL)); increment_parent_count(instance, chosen); return chosen; } /*! * \internal * \brief Try to assign an instance to its current node early * * \param[in] rsc Clone or bundle being assigned (for logs only) * \param[in] instance Clone instance or bundle replica container * \param[in] current Instance's current node * \param[in] max_per_node Maximum number of instances per node * \param[in] available Number of instances still available for assignment * * \return \c true if \p instance was successfully assigned to its current node, * or \c false otherwise */ static bool assign_instance_early(const pcmk_resource_t *rsc, pcmk_resource_t *instance, const pcmk_node_t *current, int max_per_node, int available) { const pcmk_node_t *chosen = NULL; int reserved = 0; pcmk_resource_t *parent = instance->priv->parent; GHashTable *allowed_orig = NULL; GHashTable *allowed_orig_parent = parent->priv->allowed_nodes; const pcmk_node_t *allowed_node = NULL; pcmk__rsc_trace(instance, "Trying to assign %s to its current node %s", instance->id, pcmk__node_name(current)); allowed_node = g_hash_table_lookup(instance->priv->allowed_nodes, current->priv->id); if (!pcmk__node_available(allowed_node, true, false)) { pcmk__rsc_info(instance, "Not assigning %s to current node %s: unavailable", instance->id, pcmk__node_name(current)); return false; } /* On each iteration, if instance gets assigned to a node other than its * current one, we reserve one instance for the chosen node, unassign * instance, restore instance's original node tables, and try again. This * way, instances are proportionally assigned to nodes based on preferences, * but shuffling of specific instances is minimized. If a node will be * assigned instances at all, it preferentially receives instances that are * currently active there. * * parent->private->allowed_nodes tracks the number of instances assigned to * each node. If a node already has max_per_node instances assigned, * ban_unavailable_allowed_nodes() marks it as unavailable. * * In the end, we restore the original parent->private->allowed_nodes to * undo the changes to counts during tentative assignments. If we * successfully assigned an instance to its current node, we increment that * node's counter. */ // Back up the allowed node tables of instance and its children recursively pcmk__copy_node_tables(instance, &allowed_orig); // Update instances-per-node counts in a scratch table parent->priv->allowed_nodes = pcmk__copy_node_table(allowed_orig_parent); while (reserved < available) { chosen = assign_instance(instance, current, max_per_node); if (pcmk__same_node(chosen, current)) { // Successfully assigned to current node break; } // Assignment updates scores, so restore to original state pcmk__rsc_debug(instance, "Rolling back node scores for %s", instance->id); pcmk__restore_node_tables(instance, allowed_orig); if (chosen == NULL) { // Assignment failed, so give up pcmk__rsc_info(instance, "Not assigning %s to current node %s: unavailable", instance->id, pcmk__node_name(current)); pcmk__set_rsc_flags(instance, pcmk__rsc_unassigned); break; } // We prefer more strongly to assign an instance to the chosen node pcmk__rsc_debug(instance, "Not assigning %s to current node %s: %s is better", instance->id, pcmk__node_name(current), pcmk__node_name(chosen)); // Reserve one instance for the chosen node and try again if (++reserved >= available) { pcmk__rsc_info(instance, "Not assigning %s to current node %s: " "other assignments are more important", instance->id, pcmk__node_name(current)); } else { pcmk__rsc_debug(instance, "Reserved an instance of %s for %s. Retrying " "assignment of %s to %s", rsc->id, pcmk__node_name(chosen), instance->id, pcmk__node_name(current)); } // Clear this assignment (frees chosen); leave instance counts in parent pcmk__unassign_resource(instance); chosen = NULL; } g_hash_table_destroy(allowed_orig); // Restore original instances-per-node counts g_hash_table_destroy(parent->priv->allowed_nodes); parent->priv->allowed_nodes = allowed_orig_parent; if (chosen == NULL) { // Couldn't assign instance to current node return false; } pcmk__rsc_trace(instance, "Assigned %s to current node %s", instance->id, pcmk__node_name(current)); increment_parent_count(instance, chosen); return true; } /*! * \internal * \brief Reset the node counts of a resource's allowed nodes to zero * * \param[in,out] rsc Resource to reset * * \return Number of nodes that are available to run resources */ static unsigned int reset_allowed_node_counts(pcmk_resource_t *rsc) { unsigned int available_nodes = 0; pcmk_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->priv->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { node->assign->count = 0; if (pcmk__node_available(node, false, false)) { available_nodes++; } } return available_nodes; } /*! * \internal * \brief Check whether an instance has a preferred node * * \param[in] instance Clone instance or bundle replica container * \param[in] optimal_per_node Optimal number of instances per node * * \return Instance's current node if still available, otherwise NULL */ static const pcmk_node_t * preferred_node(const pcmk_resource_t *instance, int optimal_per_node) { const pcmk_node_t *node = NULL; const pcmk_node_t *parent_node = NULL; // Check whether instance is active, healthy, and not yet assigned if ((instance->priv->active_nodes == NULL) || !pcmk_is_set(instance->flags, pcmk__rsc_unassigned) || pcmk_is_set(instance->flags, pcmk__rsc_failed)) { return NULL; } // Check whether instance's current node can run resources node = pcmk__current_node(instance); if (!pcmk__node_available(node, true, false)) { pcmk__rsc_trace(instance, "Not assigning %s to %s early (unavailable)", instance->id, pcmk__node_name(node)); return NULL; } // Check whether node already has optimal number of instances assigned parent_node = pcmk__top_allowed_node(instance, node); if ((parent_node != NULL) && (parent_node->assign->count >= optimal_per_node)) { pcmk__rsc_trace(instance, "Not assigning %s to %s early " "(optimal instances already assigned)", instance->id, pcmk__node_name(node)); return NULL; } return node; } /*! * \internal * \brief Assign collective instances to nodes * * \param[in,out] collective Clone or bundle resource being assigned * \param[in,out] instances List of clone instances or bundle containers * \param[in] max_total Maximum instances to assign in total * \param[in] max_per_node Maximum instances to assign to any one node */ void pcmk__assign_instances(pcmk_resource_t *collective, GList *instances, int max_total, int max_per_node) { // Reuse node count to track number of assigned instances unsigned int available_nodes = reset_allowed_node_counts(collective); int optimal_per_node = 0; int assigned = 0; GList *iter = NULL; pcmk_resource_t *instance = NULL; const pcmk_node_t *current = NULL; if (available_nodes > 0) { optimal_per_node = max_total / available_nodes; } if (optimal_per_node < 1) { optimal_per_node = 1; } pcmk__rsc_debug(collective, "Assigning up to %d %s instance%s to up to %u node%s " "(at most %d per host, %d optimal)", max_total, collective->id, pcmk__plural_s(max_total), available_nodes, pcmk__plural_s(available_nodes), max_per_node, optimal_per_node); // Assign as many instances as possible to their current location for (iter = instances; (iter != NULL) && (assigned < max_total); iter = iter->next) { int available = max_total - assigned; instance = iter->data; if (!pcmk_is_set(instance->flags, pcmk__rsc_unassigned)) { continue; // Already assigned } current = preferred_node(instance, optimal_per_node); if ((current != NULL) && assign_instance_early(collective, instance, current, max_per_node, available)) { assigned++; } } pcmk__rsc_trace(collective, "Assigned %d of %d instance%s to current node", assigned, max_total, pcmk__plural_s(max_total)); for (iter = instances; iter != NULL; iter = iter->next) { instance = (pcmk_resource_t *) iter->data; if (!pcmk_is_set(instance->flags, pcmk__rsc_unassigned)) { continue; // Already assigned } if (instance->priv->active_nodes != NULL) { current = pcmk__current_node(instance); if (pcmk__top_allowed_node(instance, current) == NULL) { const char *unmanaged = ""; if (!pcmk_is_set(instance->flags, pcmk__rsc_managed)) { unmanaged = "Unmanaged resource "; } crm_notice("%s%s is running on %s which is no longer allowed", unmanaged, instance->id, pcmk__node_name(current)); } } if (assigned >= max_total) { pcmk__rsc_debug(collective, "Not assigning %s because maximum %d instances " "already assigned", instance->id, max_total); resource_location(instance, NULL, -PCMK_SCORE_INFINITY, "collective_limit_reached", collective->priv->scheduler); } else if (assign_instance(instance, NULL, max_per_node) != NULL) { assigned++; } } pcmk__rsc_debug(collective, "Assigned %d of %d possible instance%s of %s", assigned, max_total, pcmk__plural_s(max_total), collective->id); } enum instance_state { instance_starting = (1 << 0), instance_stopping = (1 << 1), /* This indicates that some instance is restarting. It's not the same as * instance_starting|instance_stopping, which would indicate that some * instance is starting, and some instance (not necessarily the same one) is * stopping. */ instance_restarting = (1 << 2), instance_active = (1 << 3), instance_all = instance_starting|instance_stopping |instance_restarting|instance_active, }; /*! * \internal * \brief Check whether an instance is active, starting, and/or stopping * * \param[in] instance Clone instance or bundle replica container * \param[in,out] state Whether any instance is starting, stopping, etc. */ static void check_instance_state(const pcmk_resource_t *instance, uint32_t *state) { const GList *iter = NULL; uint32_t instance_state = 0; // State of just this instance // No need to check further if all conditions have already been detected if (pcmk_all_flags_set(*state, instance_all)) { return; } // If instance is a collective (a cloned group), check its children instead if (instance->priv->variant > pcmk__rsc_variant_primitive) { for (iter = instance->priv->children; (iter != NULL) && !pcmk_all_flags_set(*state, instance_all); iter = iter->next) { check_instance_state((const pcmk_resource_t *) iter->data, state); } return; } // If we get here, instance is a primitive if (instance->priv->active_nodes != NULL) { instance_state |= instance_active; } // Check each of the instance's actions for runnable start or stop for (iter = instance->priv->actions; (iter != NULL) && !pcmk_all_flags_set(instance_state, instance_starting |instance_stopping); iter = iter->next) { const pcmk_action_t *action = (const pcmk_action_t *) iter->data; const bool optional = pcmk_is_set(action->flags, pcmk__action_optional); if (pcmk__str_eq(PCMK_ACTION_START, action->task, pcmk__str_none)) { if (!optional && pcmk_is_set(action->flags, pcmk__action_runnable)) { pcmk__rsc_trace(instance, "Instance is starting due to %s", action->uuid); instance_state |= instance_starting; } else { pcmk__rsc_trace(instance, "%s doesn't affect %s state (%s)", action->uuid, instance->id, (optional? "optional" : "unrunnable")); } } else if (pcmk__str_eq(PCMK_ACTION_STOP, action->task, pcmk__str_none)) { /* Only stop actions can be pseudo-actions for primitives. That * indicates that the node they are on is being fenced, so the stop * is implied rather than actually executed. */ if (!optional && pcmk_any_flags_set(action->flags, pcmk__action_pseudo |pcmk__action_runnable)) { pcmk__rsc_trace(instance, "Instance is stopping due to %s", action->uuid); instance_state |= instance_stopping; } else { pcmk__rsc_trace(instance, "%s doesn't affect %s state (%s)", action->uuid, instance->id, (optional? "optional" : "unrunnable")); } } } if (pcmk_all_flags_set(instance_state, instance_starting|instance_stopping)) { instance_state |= instance_restarting; } *state |= instance_state; } /*! * \internal * \brief Create actions for collective resource instances * * \param[in,out] collective Clone or bundle resource to create actions for * \param[in,out] instances List of clone instances or bundle containers */ void pcmk__create_instance_actions(pcmk_resource_t *collective, GList *instances) { uint32_t state = 0; pcmk_action_t *stop = NULL; pcmk_action_t *stopped = NULL; pcmk_action_t *start = NULL; pcmk_action_t *started = NULL; pcmk__rsc_trace(collective, "Creating collective instance actions for %s", collective->id); // Create actions for each instance appropriate to its variant for (GList *iter = instances; iter != NULL; iter = iter->next) { pcmk_resource_t *instance = (pcmk_resource_t *) iter->data; instance->priv->cmds->create_actions(instance); check_instance_state(instance, &state); } // Create pseudo-actions for rsc start and started start = pe__new_rsc_pseudo_action(collective, PCMK_ACTION_START, !pcmk_is_set(state, instance_starting), true); started = pe__new_rsc_pseudo_action(collective, PCMK_ACTION_RUNNING, !pcmk_is_set(state, instance_starting), false); started->priority = PCMK_SCORE_INFINITY; if (pcmk_any_flags_set(state, instance_active|instance_starting)) { pcmk__set_action_flags(started, pcmk__action_runnable); } // Create pseudo-actions for rsc stop and stopped stop = pe__new_rsc_pseudo_action(collective, PCMK_ACTION_STOP, !pcmk_is_set(state, instance_stopping), true); stopped = pe__new_rsc_pseudo_action(collective, PCMK_ACTION_STOPPED, !pcmk_is_set(state, instance_stopping), true); stopped->priority = PCMK_SCORE_INFINITY; if (!pcmk_is_set(state, instance_restarting)) { pcmk__set_action_flags(stop, pcmk__action_migratable); } if (pcmk__is_clone(collective)) { pe__create_clone_notif_pseudo_ops(collective, start, started, stop, stopped); } } /*! * \internal * \brief Get a list of clone instances or bundle replica containers * * \param[in] rsc Clone or bundle resource * * \return Clone instances if \p rsc is a clone, or a newly created list of * \p rsc's replica containers if \p rsc is a bundle * \note The caller must call free_instance_list() on the result when the list * is no longer needed. */ static inline GList * get_instance_list(const pcmk_resource_t *rsc) { if (pcmk__is_bundle(rsc)) { return pe__bundle_containers(rsc); } else { return rsc->priv->children; } } /*! * \internal * \brief Free any memory created by get_instance_list() * * \param[in] rsc Clone or bundle resource passed to get_instance_list() * \param[in,out] list Return value of get_instance_list() for \p rsc */ static inline void free_instance_list(const pcmk_resource_t *rsc, GList *list) { if (list != rsc->priv->children) { g_list_free(list); } } /*! * \internal * \brief Check whether an instance is compatible with a role and node * * \param[in] instance Clone instance or bundle replica container * \param[in] node Instance must match this node * \param[in] role If not pcmk_role_unknown, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return true if \p instance is compatible with \p node and \p role, * otherwise false */ bool pcmk__instance_matches(const pcmk_resource_t *instance, const pcmk_node_t *node, enum rsc_role_e role, bool current) { pcmk_node_t *instance_node = NULL; CRM_CHECK((instance != NULL) && (node != NULL), return false); if ((role != pcmk_role_unknown) && (role != instance->priv->fns->state(instance, current))) { pcmk__rsc_trace(instance, "%s is not a compatible instance (role is not %s)", instance->id, pcmk_role_text(role)); return false; } if (!is_set_recursive(instance, pcmk__rsc_blocked, true)) { uint32_t target = pcmk__rsc_node_assigned; if (current) { target = pcmk__rsc_node_current; } // We only want instances that haven't failed instance_node = instance->priv->fns->location(instance, NULL, target); } if (instance_node == NULL) { pcmk__rsc_trace(instance, "%s is not a compatible instance " "(not assigned to a node)", instance->id); return false; } if (!pcmk__same_node(instance_node, node)) { pcmk__rsc_trace(instance, "%s is not a compatible instance " "(assigned to %s not %s)", instance->id, pcmk__node_name(instance_node), pcmk__node_name(node)); return false; } return true; } #define display_role(r) \ (((r) == pcmk_role_unknown)? "matching" : pcmk_role_text(r)) /*! * \internal * \brief Find an instance that matches a given resource by node and role * * \param[in] match_rsc Resource that instance must match (for logging only) * \param[in] rsc Clone or bundle resource to check for matching instance * \param[in] node Instance must match this node * \param[in] role If not pcmk_role_unknown, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return \p rsc instance matching \p node and \p role if any, otherwise NULL */ static pcmk_resource_t * find_compatible_instance_on_node(const pcmk_resource_t *match_rsc, const pcmk_resource_t *rsc, const pcmk_node_t *node, enum rsc_role_e role, bool current) { GList *instances = NULL; instances = get_instance_list(rsc); for (GList *iter = instances; iter != NULL; iter = iter->next) { pcmk_resource_t *instance = (pcmk_resource_t *) iter->data; if (pcmk__instance_matches(instance, node, role, current)) { pcmk__rsc_trace(match_rsc, "Found %s %s instance %s compatible with %s on %s", display_role(role), rsc->id, instance->id, match_rsc->id, pcmk__node_name(node)); free_instance_list(rsc, instances); // Only frees list, not contents return instance; } } free_instance_list(rsc, instances); pcmk__rsc_trace(match_rsc, "No %s %s instance found compatible with %s on %s", display_role(role), rsc->id, match_rsc->id, pcmk__node_name(node)); return NULL; } /*! * \internal * \brief Find a clone instance or bundle container compatible with a resource * * \param[in] match_rsc Resource that instance must match * \param[in] rsc Clone or bundle resource to check for matching instance * \param[in] role If not pcmk_role_unknown, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return Compatible (by \p role and \p match_rsc location) instance of \p rsc * if any, otherwise NULL */ pcmk_resource_t * pcmk__find_compatible_instance(const pcmk_resource_t *match_rsc, const pcmk_resource_t *rsc, enum rsc_role_e role, bool current) { pcmk_resource_t *instance = NULL; GList *nodes = NULL; const pcmk_node_t *node = NULL; GHashTable *allowed_nodes = match_rsc->priv->allowed_nodes; uint32_t target = pcmk__rsc_node_assigned; if (current) { target = pcmk__rsc_node_current; } // If match_rsc has a node, check only that node node = match_rsc->priv->fns->location(match_rsc, NULL, target); if (node != NULL) { return find_compatible_instance_on_node(match_rsc, rsc, node, role, current); } // Otherwise check for an instance matching any of match_rsc's allowed nodes nodes = pcmk__sort_nodes(g_hash_table_get_values(allowed_nodes), NULL); for (GList *iter = nodes; (iter != NULL) && (instance == NULL); iter = iter->next) { instance = find_compatible_instance_on_node(match_rsc, rsc, (pcmk_node_t *) iter->data, role, current); } if (instance == NULL) { pcmk__rsc_debug(rsc, "No %s instance found compatible with %s", rsc->id, match_rsc->id); } g_list_free(nodes); return instance; } /*! * \internal * \brief Unassign an instance if mandatory ordering has no interleave match * * \param[in] first 'First' action in an ordering * \param[in] then 'Then' action in an ordering * \param[in,out] then_instance 'Then' instance that has no interleave match * \param[in] type Group of enum pcmk__action_relation_flags * \param[in] current If true, "then" action is stopped or demoted * * \return true if \p then_instance was unassigned, otherwise false */ static bool unassign_if_mandatory(const pcmk_action_t *first, const pcmk_action_t *then, pcmk_resource_t *then_instance, uint32_t type, bool current) { // Allow "then" instance to go down even without an interleave match if (current) { pcmk__rsc_trace(then->rsc, "%s has no instance to order before stopping " "or demoting %s", first->rsc->id, then_instance->id); /* If the "first" action must be runnable, but there is no "first" * instance, the "then" instance must not be allowed to come up. */ } else if (pcmk_any_flags_set(type, pcmk__ar_unrunnable_first_blocks |pcmk__ar_first_implies_then)) { pcmk__rsc_info(then->rsc, "Inhibiting %s from being active " "because there is no %s instance to interleave", then_instance->id, first->rsc->id); return pcmk__assign_resource(then_instance, NULL, true, true); } return false; } /*! * \internal * \brief Find first matching action for a clone instance or bundle container * * \param[in] action Action in an interleaved ordering * \param[in] instance Clone instance or bundle container being interleaved * \param[in] action_name Action to look for * \param[in] node If not NULL, require action to be on this node * \param[in] for_first If true, \p instance is the 'first' resource in the * ordering, otherwise it is the 'then' resource * * \return First action for \p instance (or in some cases if \p instance is a * bundle container, its containerized resource) that matches * \p action_name and \p node if any, otherwise NULL */ static pcmk_action_t * find_instance_action(const pcmk_action_t *action, const pcmk_resource_t *instance, const char *action_name, const pcmk_node_t *node, bool for_first) { const pcmk_resource_t *rsc = NULL; pcmk_action_t *matching_action = NULL; /* If instance is a bundle container, sometimes we should interleave the * action for the container itself, and sometimes for the containerized * resource. * * For example, given "start bundle A then bundle B", B likely requires the * service inside A's container to be active, rather than just the * container, so we should interleave the action for A's containerized * resource. On the other hand, it's possible B's container itself requires * something from A, so we should interleave the action for B's container. * * Essentially, for 'first', we should use the containerized resource for * everything except stop, and for 'then', we should use the container for * everything except promote and demote (which can only be performed on the * containerized resource). */ if ((for_first && !pcmk__str_any_of(action->task, PCMK_ACTION_STOP, PCMK_ACTION_STOPPED, NULL)) || (!for_first && pcmk__str_any_of(action->task, PCMK_ACTION_PROMOTE, PCMK_ACTION_PROMOTED, PCMK_ACTION_DEMOTE, PCMK_ACTION_DEMOTED, NULL))) { rsc = pe__get_rsc_in_container(instance); } if (rsc == NULL) { rsc = instance; // No containerized resource, use instance itself } else { node = NULL; // Containerized actions are on bundle-created guest } matching_action = find_first_action(rsc->priv->actions, NULL, action_name, node); if (matching_action != NULL) { return matching_action; } if (pcmk_is_set(instance->flags, pcmk__rsc_removed) || pcmk__is_down_action(action_name)) { crm_trace("No %s action found for %s%s", action_name, pcmk_is_set(instance->flags, pcmk__rsc_removed)? "orphan " : "", instance->id); } else { crm_err("No %s action found for %s to interleave (bug?)", action_name, instance->id); } return NULL; } /*! * \internal * \brief Get the original action name of a bundle or clone action * * Given an action for a bundle or clone, get the original action name, * mapping notify to the action being notified, and if the instances are * primitives, mapping completion actions to the action that was completed * (for example, stopped to stop). * * \param[in] action Clone or bundle action to check * * \return Original action name for \p action */ static const char * orig_action_name(const pcmk_action_t *action) { // Any instance will do const pcmk_resource_t *instance = action->rsc->priv->children->data; char *action_type = NULL; const char *action_name = action->task; enum pcmk__action_type orig_task = pcmk__action_unspecified; if (pcmk__strcase_any_of(action->task, PCMK_ACTION_NOTIFY, PCMK_ACTION_NOTIFIED, NULL)) { // action->uuid is RSC_(confirmed-){pre,post}_notify_ACTION_INTERVAL CRM_CHECK(parse_op_key(action->uuid, NULL, &action_type, NULL), return pcmk__action_text(pcmk__action_unspecified)); action_name = strstr(action_type, "_notify_"); CRM_CHECK(action_name != NULL, return pcmk__action_text(pcmk__action_unspecified)); action_name += strlen("_notify_"); } orig_task = get_complex_task(instance, action_name); free(action_type); return pcmk__action_text(orig_task); } /*! * \internal * \brief Update two interleaved actions according to an ordering between them * * Given information about an ordering of two interleaved actions, update the * actions' flags (and runnable_before members if appropriate) as appropriate * for the ordering. Effects may cascade to other orderings involving the * actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * \param[in] filter Action flags to limit scope of certain updates (may * include pcmk__action_optional to affect only * mandatory actions, and pcmk__action_runnable to * affect only runnable actions) * \param[in] type Group of enum pcmk__action_relation_flags to apply * * \return Group of enum pcmk__updated flags indicating what was updated */ static uint32_t update_interleaved_actions(pcmk_action_t *first, pcmk_action_t *then, const pcmk_node_t *node, uint32_t filter, uint32_t type) { GList *instances = NULL; uint32_t changed = pcmk__updated_none; const char *orig_first_task = orig_action_name(first); // Stops and demotes must be interleaved with instance on current node - bool current = pcmk__ends_with(first->uuid, "_" PCMK_ACTION_STOPPED "_0") - || pcmk__ends_with(first->uuid, - "_" PCMK_ACTION_DEMOTED "_0"); + bool current = (first->uuid != NULL) + && (g_str_has_suffix(first->uuid, + "_" PCMK_ACTION_STOPPED "_0") + || g_str_has_suffix(first->uuid, + "_" PCMK_ACTION_DEMOTED "_0")); // Update the specified actions for each "then" instance individually instances = get_instance_list(then->rsc); for (GList *iter = instances; iter != NULL; iter = iter->next) { pcmk_resource_t *first_instance = NULL; pcmk_resource_t *then_instance = iter->data; pcmk_action_t *first_action = NULL; pcmk_action_t *then_action = NULL; // Find a "first" instance to interleave with this "then" instance first_instance = pcmk__find_compatible_instance(then_instance, first->rsc, pcmk_role_unknown, current); if (first_instance == NULL) { // No instance can be interleaved if (unassign_if_mandatory(first, then, then_instance, type, current)) { pcmk__set_updated_flags(changed, first, pcmk__updated_then); } continue; } first_action = find_instance_action(first, first_instance, orig_first_task, node, true); if (first_action == NULL) { continue; } then_action = find_instance_action(then, then_instance, then->task, node, false); if (then_action == NULL) { continue; } if (order_actions(first_action, then_action, type)) { pcmk__set_updated_flags(changed, first, pcmk__updated_first|pcmk__updated_then); } changed |= then_instance->priv->cmds->update_ordered_actions( first_action, then_action, node, first_instance->priv->cmds->action_flags(first_action, node), filter, type, then->rsc->priv->scheduler); } free_instance_list(then->rsc, instances); return changed; } /*! * \internal * \brief Check whether two actions in an ordering can be interleaved * * \param[in] first 'First' action in the ordering * \param[in] then 'Then' action in the ordering * * \return true if \p first and \p then can be interleaved, otherwise false */ static bool can_interleave_actions(const pcmk_action_t *first, const pcmk_action_t *then) { bool interleave = false; pcmk_resource_t *rsc = NULL; if ((first->rsc == NULL) || (then->rsc == NULL)) { crm_trace("Not interleaving %s with %s: not resource actions", first->uuid, then->uuid); return false; } if (first->rsc == then->rsc) { crm_trace("Not interleaving %s with %s: same resource", first->uuid, then->uuid); return false; } if ((first->rsc->priv->variant < pcmk__rsc_variant_clone) || (then->rsc->priv->variant < pcmk__rsc_variant_clone)) { crm_trace("Not interleaving %s with %s: not clones or bundles", first->uuid, then->uuid); return false; } - if (pcmk__ends_with(then->uuid, "_stop_0") - || pcmk__ends_with(then->uuid, "_demote_0")) { + if (g_str_has_suffix(then->uuid, "_stop_0") + || g_str_has_suffix(then->uuid, "_demote_0")) { rsc = first->rsc; } else { rsc = then->rsc; } interleave = crm_is_true(g_hash_table_lookup(rsc->priv->meta, PCMK_META_INTERLEAVE)); pcmk__rsc_trace(rsc, "'%s then %s' will %sbe interleaved (based on %s)", first->uuid, then->uuid, (interleave? "" : "not "), rsc->id); return interleave; } /*! * \internal * \brief Update non-interleaved instance actions according to an ordering * * Given information about an ordering of two non-interleaved actions, update * the actions' flags (and runnable_before members if appropriate) as * appropriate for the ordering. Effects may cascade to other orderings * involving the actions as well. * * \param[in,out] instance Clone instance or bundle container * \param[in,out] first "First" action in ordering * \param[in] then "Then" action in ordering (for \p instance's parent) * \param[in] node If not NULL, limit scope of ordering to this node * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pcmk__action_optional to affect only * mandatory actions, and pcmk__action_runnable to * affect only runnable actions) * \param[in] type Group of enum pcmk__action_relation_flags to apply * * \return Group of enum pcmk__updated flags indicating what was updated */ static uint32_t update_noninterleaved_actions(pcmk_resource_t *instance, pcmk_action_t *first, const pcmk_action_t *then, const pcmk_node_t *node, uint32_t flags, uint32_t filter, uint32_t type) { pcmk_action_t *instance_action = NULL; pcmk_scheduler_t *scheduler = instance->priv->scheduler; uint32_t instance_flags = 0; uint32_t changed = pcmk__updated_none; // Check whether instance has an equivalent of "then" action instance_action = find_first_action(instance->priv->actions, NULL, then->task, node); if (instance_action == NULL) { return changed; } // Check whether action is runnable instance_flags = instance->priv->cmds->action_flags(instance_action, node); if (!pcmk_is_set(instance_flags, pcmk__action_runnable)) { return changed; } // If so, update actions for the instance changed = instance->priv->cmds->update_ordered_actions(first, instance_action, node, flags, filter, type, scheduler); // Propagate any changes to later actions if (pcmk_is_set(changed, pcmk__updated_then)) { for (GList *after_iter = instance_action->actions_after; after_iter != NULL; after_iter = after_iter->next) { pcmk__related_action_t *after = after_iter->data; pcmk__update_action_for_orderings(after->action, scheduler); } } return changed; } /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two clone or bundle actions, update * the actions' flags (and runnable_before members if appropriate) as * appropriate for the ordering. Effects may cascade to other orderings * involving the actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * (only used when interleaving instances) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pcmk__action_optional to affect only * mandatory actions, and pcmk__action_runnable to * affect only runnable actions) * \param[in] type Group of enum pcmk__action_relation_flags to apply * \param[in,out] scheduler Scheduler data * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t pcmk__instance_update_ordered_actions(pcmk_action_t *first, pcmk_action_t *then, const pcmk_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pcmk_scheduler_t *scheduler) { pcmk__assert((first != NULL) && (then != NULL) && (scheduler != NULL)); if (then->rsc == NULL) { return pcmk__updated_none; } else if (can_interleave_actions(first, then)) { return update_interleaved_actions(first, then, node, filter, type); } else { uint32_t changed = pcmk__updated_none; GList *instances = get_instance_list(then->rsc); // Update actions for the clone or bundle resource itself changed |= pcmk__update_ordered_actions(first, then, node, flags, filter, type, scheduler); // Update the 'then' clone instances or bundle containers individually for (GList *iter = instances; iter != NULL; iter = iter->next) { pcmk_resource_t *instance = iter->data; changed |= update_noninterleaved_actions(instance, first, then, node, flags, filter, type); } free_instance_list(then->rsc, instances); return changed; } } #define pe__clear_action_summary_flags(flags, action, flag) do { \ flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Action summary", action->rsc->id, \ flags, flag, #flag); \ } while (0) /*! * \internal * \brief Return action flags for a given clone or bundle action * * \param[in,out] action Action for a clone or bundle * \param[in] instances Clone instances or bundle containers * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__collective_action_flags(pcmk_action_t *action, const GList *instances, const pcmk_node_t *node) { bool any_runnable = false; const char *action_name = orig_action_name(action); // Set original assumptions (optional and runnable may be cleared below) uint32_t flags = pcmk__action_optional |pcmk__action_runnable |pcmk__action_pseudo; for (const GList *iter = instances; iter != NULL; iter = iter->next) { const pcmk_resource_t *instance = iter->data; const pcmk_node_t *instance_node = NULL; pcmk_action_t *instance_action = NULL; uint32_t instance_flags; // Node is relevant only to primitive instances if (pcmk__is_primitive(instance)) { instance_node = node; } instance_action = find_first_action(instance->priv->actions, NULL, action_name, instance_node); if (instance_action == NULL) { pcmk__rsc_trace(action->rsc, "%s has no %s action on %s", instance->id, action_name, pcmk__node_name(node)); continue; } pcmk__rsc_trace(action->rsc, "%s has %s for %s on %s", instance->id, instance_action->uuid, action_name, pcmk__node_name(node)); instance_flags = instance->priv->cmds->action_flags(instance_action, node); // If any instance action is mandatory, so is the collective action if (pcmk_is_set(flags, pcmk__action_optional) && !pcmk_is_set(instance_flags, pcmk__action_optional)) { pcmk__rsc_trace(instance, "%s is mandatory because %s is", action->uuid, instance_action->uuid); pe__clear_action_summary_flags(flags, action, pcmk__action_optional); pcmk__clear_action_flags(action, pcmk__action_optional); } // If any instance action is runnable, so is the collective action if (pcmk_is_set(instance_flags, pcmk__action_runnable)) { any_runnable = true; } } if (!any_runnable) { pcmk__rsc_trace(action->rsc, "%s is not runnable because no instance can run %s", action->uuid, action_name); pe__clear_action_summary_flags(flags, action, pcmk__action_runnable); if (node == NULL) { pcmk__clear_action_flags(action, pcmk__action_runnable); } } return flags; } diff --git a/lib/pacemaker/pcmk_sched_probes.c b/lib/pacemaker/pcmk_sched_probes.c index bda90cee19..a84e382572 100644 --- a/lib/pacemaker/pcmk_sched_probes.c +++ b/lib/pacemaker/pcmk_sched_probes.c @@ -1,905 +1,905 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Add the expected result to a newly created probe * * \param[in,out] probe Probe action to add expected result to * \param[in] rsc Resource that probe is for * \param[in] node Node that probe will run on */ static void add_expected_result(pcmk_action_t *probe, const pcmk_resource_t *rsc, const pcmk_node_t *node) { // Check whether resource is currently active on node pcmk_node_t *running = pe_find_node_id(rsc->priv->active_nodes, node->priv->id); // The expected result is what we think the resource's current state is if (running == NULL) { pe__add_action_expected_result(probe, CRM_EX_NOT_RUNNING); } else if (rsc->priv->orig_role == pcmk_role_promoted) { pe__add_action_expected_result(probe, CRM_EX_PROMOTED); } } /*! * \internal * \brief Create any needed robes on a node for a list of resources * * \param[in,out] rscs List of resources to create probes for * \param[in,out] node Node to create probes on * * \return true if any probe was created, otherwise false */ bool pcmk__probe_resource_list(GList *rscs, pcmk_node_t *node) { bool any_created = false; for (GList *iter = rscs; iter != NULL; iter = iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data; if (rsc->priv->cmds->create_probe(rsc, node)) { any_created = true; } } return any_created; } /*! * \internal * \brief Order one resource's start after another's start-up probe * * \param[in,out] rsc1 Resource that might get start-up probe * \param[in] rsc2 Resource that might be started */ static void probe_then_start(pcmk_resource_t *rsc1, pcmk_resource_t *rsc2) { const pcmk_node_t *rsc1_node = rsc1->priv->assigned_node; if ((rsc1_node != NULL) && (g_hash_table_lookup(rsc1->priv->probed_nodes, rsc1_node->priv->id) == NULL)) { pcmk__new_ordering(rsc1, pcmk__op_key(rsc1->id, PCMK_ACTION_MONITOR, 0), NULL, rsc2, pcmk__op_key(rsc2->id, PCMK_ACTION_START, 0), NULL, pcmk__ar_ordered, rsc1->priv->scheduler); } } /*! * \internal * \brief Check whether a guest resource will stop * * \param[in] node Guest node to check * * \return true if guest resource will likely stop, otherwise false */ static bool guest_resource_will_stop(const pcmk_node_t *node) { const pcmk_resource_t *guest_rsc = NULL; const pcmk_node_t *guest_node = NULL; guest_rsc = node->priv->remote->priv->launcher; guest_node = guest_rsc->priv->assigned_node; /* Ideally, we'd check whether the guest has a required stop, but that * information doesn't exist yet, so approximate it ... */ return pcmk_is_set(node->priv->flags, pcmk__node_remote_reset) || node->details->unclean || pcmk_is_set(guest_rsc->flags, pcmk__rsc_failed) || (guest_rsc->priv->next_role == pcmk_role_stopped) // Guest is moving || ((guest_rsc->priv->orig_role > pcmk_role_stopped) && (guest_node != NULL) && pcmk__find_node_in_list(guest_rsc->priv->active_nodes, guest_node->priv->name) == NULL); } /*! * \internal * \brief Create a probe action for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return Newly created probe action */ static pcmk_action_t * probe_action(pcmk_resource_t *rsc, pcmk_node_t *node) { pcmk_action_t *probe = NULL; char *key = pcmk__op_key(rsc->id, PCMK_ACTION_MONITOR, 0); crm_debug("Scheduling probe of %s %s on %s", pcmk_role_text(rsc->priv->orig_role), rsc->id, pcmk__node_name(node)); probe = custom_action(rsc, key, PCMK_ACTION_MONITOR, node, FALSE, rsc->priv->scheduler); pcmk__clear_action_flags(probe, pcmk__action_optional); pcmk__order_vs_unfence(rsc, node, probe, pcmk__ar_ordered); add_expected_result(probe, rsc, node); return probe; } /*! * \internal * \brief Create probes for a resource on a node, if needed * * \brief Schedule any probes needed for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return true if any probe was created, otherwise false */ bool pcmk__probe_rsc_on_node(pcmk_resource_t *rsc, pcmk_node_t *node) { uint32_t flags = pcmk__ar_ordered; pcmk_action_t *probe = NULL; pcmk_node_t *allowed = NULL; pcmk_resource_t *top = uber_parent(rsc); const char *reason = NULL; pcmk__assert((rsc != NULL) && (node != NULL)); if (!pcmk_is_set(rsc->priv->scheduler->flags, pcmk__sched_probe_resources)) { reason = "start-up probes are disabled"; goto no_probe; } if (pcmk__is_pacemaker_remote_node(node)) { if (pcmk_is_set(rsc->flags, pcmk__rsc_fence_device)) { reason = "Pacemaker Remote nodes cannot run stonith agents"; goto no_probe; } else if (pcmk__is_guest_or_bundle_node(node) && pe__resource_contains_guest_node(rsc->priv->scheduler, rsc)) { reason = "guest nodes cannot run resources containing guest nodes"; goto no_probe; } else if (pcmk_is_set(rsc->flags, pcmk__rsc_is_remote_connection)) { reason = "Pacemaker Remote nodes cannot host remote connections"; goto no_probe; } } // If this is a collective resource, probes are created for its children if (rsc->priv->children != NULL) { return pcmk__probe_resource_list(rsc->priv->children, node); } if ((rsc->priv->launcher != NULL) && !pcmk_is_set(rsc->flags, pcmk__rsc_is_remote_connection)) { reason = "resource is inside a container"; goto no_probe; } else if (pcmk_is_set(rsc->flags, pcmk__rsc_removed)) { reason = "resource is orphaned"; goto no_probe; } else if (g_hash_table_lookup(rsc->priv->probed_nodes, node->priv->id) != NULL) { reason = "resource state is already known"; goto no_probe; } allowed = g_hash_table_lookup(rsc->priv->allowed_nodes, node->priv->id); if (pcmk_is_set(rsc->flags, pcmk__rsc_exclusive_probes) || pcmk_is_set(top->flags, pcmk__rsc_exclusive_probes)) { // Exclusive discovery is enabled ... if (allowed == NULL) { // ... but this node is not allowed to run the resource reason = "resource has exclusive discovery but is not allowed " "on node"; goto no_probe; } else if (allowed->assign->probe_mode != pcmk__probe_exclusive) { // ... but no constraint marks this node for discovery of resource reason = "resource has exclusive discovery but is not enabled " "on node"; goto no_probe; } } if (allowed == NULL) { allowed = node; } if (allowed->assign->probe_mode == pcmk__probe_never) { reason = "node has discovery disabled"; goto no_probe; } if (pcmk__is_guest_or_bundle_node(node)) { pcmk_resource_t *guest = node->priv->remote->priv->launcher; if (guest->priv->orig_role == pcmk_role_stopped) { // The guest is stopped, so we know no resource is active there reason = "node's guest is stopped"; probe_then_start(guest, top); goto no_probe; } else if (guest_resource_will_stop(node)) { reason = "node's guest will stop"; // Order resource start after guest stop (in case it's restarting) pcmk__new_ordering(guest, pcmk__op_key(guest->id, PCMK_ACTION_STOP, 0), NULL, top, pcmk__op_key(top->id, PCMK_ACTION_START, 0), NULL, pcmk__ar_ordered, rsc->priv->scheduler); goto no_probe; } } // We've eliminated all cases where a probe is not needed, so now it is probe = probe_action(rsc, node); /* Below, we will order the probe relative to start or reload. If this is a * clone instance, the start or reload is for the entire clone rather than * just the instance. Otherwise, the start or reload is for the resource * itself. */ if (!pcmk__is_clone(top)) { top = rsc; } /* Prevent a start if the resource can't be probed, but don't cause the * resource or entire clone to stop if already active. */ if (!pcmk_is_set(probe->flags, pcmk__action_runnable) && (top->priv->active_nodes == NULL)) { pcmk__set_relation_flags(flags, pcmk__ar_unrunnable_first_blocks); } // Start or reload after probing the resource pcmk__new_ordering(rsc, NULL, probe, top, pcmk__op_key(top->id, PCMK_ACTION_START, 0), NULL, flags, rsc->priv->scheduler); pcmk__new_ordering(rsc, NULL, probe, top, reload_key(rsc), NULL, pcmk__ar_ordered, rsc->priv->scheduler); return true; no_probe: pcmk__rsc_trace(rsc, "Skipping probe for %s on %s because %s", rsc->id, node->priv->id, reason); return false; } /*! * \internal * \brief Check whether a probe should be ordered before another action * * \param[in] probe Probe action to check * \param[in] then Other action to check * * \return true if \p probe should be ordered before \p then, otherwise false */ static bool probe_needed_before_action(const pcmk_action_t *probe, const pcmk_action_t *then) { // Probes on a node are performed after unfencing it, not before if (pcmk__str_eq(then->task, PCMK_ACTION_STONITH, pcmk__str_none) && pcmk__same_node(probe->node, then->node)) { const char *op = g_hash_table_lookup(then->meta, PCMK__META_STONITH_ACTION); if (pcmk__str_eq(op, PCMK_ACTION_ON, pcmk__str_casei)) { return false; } } // Probes should be done on a node before shutting it down if (pcmk__str_eq(then->task, PCMK_ACTION_DO_SHUTDOWN, pcmk__str_none) && (probe->node != NULL) && (then->node != NULL) && !pcmk__same_node(probe->node, then->node)) { return false; } // Otherwise probes should always be done before any other action return true; } /*! * \internal * \brief Add implicit "probe then X" orderings for "stop then X" orderings * * If the state of a resource is not known yet, a probe will be scheduled, * expecting a "not running" result. If the probe fails, a stop will not be * scheduled until the next transition. Thus, if there are ordering constraints * like "stop this resource then do something else that's not for the same * resource", add implicit "probe this resource then do something" equivalents * so the relation is upheld until we know whether a stop is needed. * * \param[in,out] scheduler Scheduler data */ static void add_probe_orderings_for_stops(pcmk_scheduler_t *scheduler) { for (GList *iter = scheduler->priv->ordering_constraints; iter != NULL; iter = iter->next) { pcmk__action_relation_t *order = iter->data; uint32_t order_flags = pcmk__ar_ordered; GList *probes = NULL; GList *then_actions = NULL; pcmk_action_t *first = NULL; pcmk_action_t *then = NULL; // Skip disabled orderings if (order->flags == pcmk__ar_none) { continue; } // Skip non-resource orderings, and orderings for the same resource if ((order->rsc1 == NULL) || (order->rsc1 == order->rsc2)) { continue; } // Skip invalid orderings (shouldn't be possible) first = order->action1; then = order->action2; if (((first == NULL) && (order->task1 == NULL)) || ((then == NULL) && (order->task2 == NULL))) { continue; } // Skip orderings for first actions other than stop if ((first != NULL) && !pcmk__str_eq(first->task, PCMK_ACTION_STOP, pcmk__str_none)) { continue; } else if ((first == NULL) - && !pcmk__ends_with(order->task1, - "_" PCMK_ACTION_STOP "_0")) { + && !g_str_has_suffix(order->task1, + "_" PCMK_ACTION_STOP "_0")) { continue; } /* Do not imply a probe ordering for a resource inside of a stopping * launcher. Otherwise, it might introduce a transition loop, since a * probe could be scheduled after the launcher starts again. */ if ((order->rsc2 != NULL) && (order->rsc1->priv->launcher == order->rsc2)) { if ((then != NULL) && pcmk__str_eq(then->task, PCMK_ACTION_STOP, pcmk__str_none)) { continue; } else if ((then == NULL) - && pcmk__ends_with(order->task2, - "_" PCMK_ACTION_STOP "_0")) { + && g_str_has_suffix(order->task2, + "_" PCMK_ACTION_STOP "_0")) { continue; } } // Preserve certain order options for future filtering if (pcmk_is_set(order->flags, pcmk__ar_if_first_unmigratable)) { pcmk__set_relation_flags(order_flags, pcmk__ar_if_first_unmigratable); } if (pcmk_is_set(order->flags, pcmk__ar_if_on_same_node)) { pcmk__set_relation_flags(order_flags, pcmk__ar_if_on_same_node); } // Preserve certain order types for future filtering if ((order->flags == pcmk__ar_if_required_on_same_node) || (order->flags == pcmk__ar_if_on_same_node_or_target)) { order_flags = order->flags; } // List all scheduled probes for the first resource probes = pe__resource_actions(order->rsc1, NULL, PCMK_ACTION_MONITOR, FALSE); if (probes == NULL) { // There aren't any continue; } // List all relevant "then" actions if (then != NULL) { then_actions = g_list_prepend(NULL, then); } else if (order->rsc2 != NULL) { then_actions = find_actions(order->rsc2->priv->actions, order->task2, NULL); if (then_actions == NULL) { // There aren't any g_list_free(probes); continue; } } crm_trace("Implying 'probe then' orderings for '%s then %s' " "(id=%d, type=%.6x)", ((first == NULL)? order->task1 : first->uuid), ((then == NULL)? order->task2 : then->uuid), order->id, order->flags); for (GList *probe_iter = probes; probe_iter != NULL; probe_iter = probe_iter->next) { pcmk_action_t *probe = (pcmk_action_t *) probe_iter->data; for (GList *then_iter = then_actions; then_iter != NULL; then_iter = then_iter->next) { pcmk_action_t *then = (pcmk_action_t *) then_iter->data; if (probe_needed_before_action(probe, then)) { order_actions(probe, then, order_flags); } } } g_list_free(then_actions); g_list_free(probes); } } /*! * \internal * \brief Add necessary orderings between probe and starts of clone instances * * , in additon to the ordering with the parent resource added upon creating * the probe. * * \param[in,out] probe Probe as 'first' action in an ordering * \param[in,out] after 'then' action wrapper in the ordering */ static void add_start_orderings_for_probe(pcmk_action_t *probe, pcmk__related_action_t *after) { uint32_t flags = pcmk__ar_ordered|pcmk__ar_unrunnable_first_blocks; /* Although the ordering between the probe of the clone instance and the * start of its parent has been added in pcmk__probe_rsc_on_node(), we * avoided enforcing `pcmk__ar_unrunnable_first_blocks` order type for that * as long as any of the clone instances are running to prevent them from * being unexpectedly stopped. * * On the other hand, we still need to prevent any inactive instances from * starting unless the probe is runnable so that we don't risk starting too * many instances before we know the state on all nodes. */ if ((after->action->rsc->priv->variant <= pcmk__rsc_variant_group) || pcmk_is_set(probe->flags, pcmk__action_runnable) // The order type is already enforced for its parent. || pcmk_is_set(after->flags, pcmk__ar_unrunnable_first_blocks) || (pe__const_top_resource(probe->rsc, false) != after->action->rsc) || !pcmk__str_eq(after->action->task, PCMK_ACTION_START, pcmk__str_none)) { return; } crm_trace("Adding probe start orderings for 'unrunnable %s@%s " "then instances of %s@%s'", probe->uuid, pcmk__node_name(probe->node), after->action->uuid, pcmk__node_name(after->action->node)); for (GList *then_iter = after->action->actions_after; then_iter != NULL; then_iter = then_iter->next) { pcmk__related_action_t *then = then_iter->data; if ((then->action->rsc->priv->active_nodes != NULL) || (pe__const_top_resource(then->action->rsc, false) != after->action->rsc) || !pcmk__str_eq(then->action->task, PCMK_ACTION_START, pcmk__str_none)) { continue; } crm_trace("Adding probe start ordering for 'unrunnable %s@%s " "then %s@%s' (type=%#.6x)", probe->uuid, pcmk__node_name(probe->node), then->action->uuid, pcmk__node_name(then->action->node), flags); /* Prevent the instance from starting if the instance can't, but don't * cause any other intances to stop if already active. */ order_actions(probe, then->action, flags); } return; } /*! * \internal * \brief Order probes before restarts and re-promotes * * If a given ordering is a "probe then start" or "probe then promote" ordering, * add an implicit "probe then stop/demote" ordering in case the action is part * of a restart/re-promote, and do the same recursively for all actions ordered * after the "then" action. * * \param[in,out] probe Probe as 'first' action in an ordering * \param[in,out] after 'then' action in the ordering */ static void add_restart_orderings_for_probe(pcmk_action_t *probe, pcmk_action_t *after) { GList *iter = NULL; bool interleave = false; pcmk_resource_t *compatible_rsc = NULL; // Validate that this is a resource probe followed by some action if ((after == NULL) || (probe == NULL) || !pcmk__is_primitive(probe->rsc) || !pcmk__str_eq(probe->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { return; } // Avoid running into any possible loop if (pcmk_is_set(after->flags, pcmk__action_detect_loop)) { return; } pcmk__set_action_flags(after, pcmk__action_detect_loop); crm_trace("Adding probe restart orderings for '%s@%s then %s@%s'", probe->uuid, pcmk__node_name(probe->node), after->uuid, pcmk__node_name(after->node)); /* Add restart orderings if "then" is for a different primitive. * Orderings for collective resources will be added later. */ if (pcmk__is_primitive(after->rsc) && (probe->rsc != after->rsc)) { GList *then_actions = NULL; if (pcmk__str_eq(after->task, PCMK_ACTION_START, pcmk__str_none)) { then_actions = pe__resource_actions(after->rsc, NULL, PCMK_ACTION_STOP, FALSE); } else if (pcmk__str_eq(after->task, PCMK_ACTION_PROMOTE, pcmk__str_none)) { then_actions = pe__resource_actions(after->rsc, NULL, PCMK_ACTION_DEMOTE, FALSE); } for (iter = then_actions; iter != NULL; iter = iter->next) { pcmk_action_t *then = (pcmk_action_t *) iter->data; // Skip pseudo-actions (for example, those implied by fencing) if (!pcmk_is_set(then->flags, pcmk__action_pseudo)) { order_actions(probe, then, pcmk__ar_ordered); } } g_list_free(then_actions); } /* Detect whether "then" is an interleaved clone action. For these, we want * to add orderings only for the relevant instance. */ if ((after->rsc != NULL) && (after->rsc->priv->variant > pcmk__rsc_variant_group)) { interleave = crm_is_true(g_hash_table_lookup(after->rsc->priv->meta, PCMK_META_INTERLEAVE)); if (interleave) { compatible_rsc = pcmk__find_compatible_instance(probe->rsc, after->rsc, pcmk_role_unknown, false); } } /* Now recursively do the same for all actions ordered after "then". This * also handles collective resources since the collective action will be * ordered before its individual instances' actions. */ for (iter = after->actions_after; iter != NULL; iter = iter->next) { pcmk__related_action_t *after_wrapper = iter->data; const pcmk_resource_t *chained_rsc = NULL; /* pcmk__ar_first_implies_then is the reason why a required A.start * implies/enforces B.start to be required too, which is the cause of * B.restart/re-promote. * * Not sure about pcmk__ar_first_implies_same_node_then though. It's now * only used for unfencing case, which tends to introduce transition * loops... */ if (!pcmk_is_set(after_wrapper->flags, pcmk__ar_first_implies_then)) { /* The order type between a group/clone and its child such as * B.start-> B_child.start is: * pcmk__ar_then_implies_first_graphed * |pcmk__ar_unrunnable_first_blocks * * Proceed through the ordering chain and build dependencies with * its children. */ if ((after->rsc == NULL) || (after->rsc->priv->variant < pcmk__rsc_variant_group) || (probe->rsc->priv->parent == after->rsc) || (after_wrapper->action->rsc == NULL)) { continue; } chained_rsc = after_wrapper->action->rsc; if ((chained_rsc->priv->variant > pcmk__rsc_variant_group) || (after->rsc != chained_rsc->priv->parent)) { continue; } /* Proceed to the children of a group or a non-interleaved clone. * For an interleaved clone, proceed only to the relevant child. */ if ((after->rsc->priv->variant > pcmk__rsc_variant_group) && interleave && ((compatible_rsc == NULL) || (compatible_rsc != chained_rsc))) { continue; } } crm_trace("Recursively adding probe restart orderings for " "'%s@%s then %s@%s' (type=%#.6x)", after->uuid, pcmk__node_name(after->node), after_wrapper->action->uuid, pcmk__node_name(after_wrapper->action->node), after_wrapper->flags); add_restart_orderings_for_probe(probe, after_wrapper->action); } } /*! * \internal * \brief Clear the tracking flag on all scheduled actions * * \param[in,out] scheduler Scheduler data */ static void clear_actions_tracking_flag(pcmk_scheduler_t *scheduler) { for (GList *iter = scheduler->priv->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = iter->data; pcmk__clear_action_flags(action, pcmk__action_detect_loop); } } /*! * \internal * \brief Add start and restart orderings for probes scheduled for a resource * * \param[in,out] data Resource whose probes should be ordered * \param[in] user_data Unused */ static void add_start_restart_orderings_for_rsc(gpointer data, gpointer user_data) { pcmk_resource_t *rsc = data; GList *probes = NULL; // For collective resources, order each instance recursively if (!pcmk__is_primitive(rsc)) { g_list_foreach(rsc->priv->children, add_start_restart_orderings_for_rsc, NULL); return; } // Find all probes for given resource probes = pe__resource_actions(rsc, NULL, PCMK_ACTION_MONITOR, FALSE); // Add probe restart orderings for each probe found for (GList *iter = probes; iter != NULL; iter = iter->next) { pcmk_action_t *probe = (pcmk_action_t *) iter->data; for (GList *then_iter = probe->actions_after; then_iter != NULL; then_iter = then_iter->next) { pcmk__related_action_t *then = then_iter->data; add_start_orderings_for_probe(probe, then); add_restart_orderings_for_probe(probe, then->action); clear_actions_tracking_flag(rsc->priv->scheduler); } } g_list_free(probes); } /*! * \internal * \brief Add "A then probe B" orderings for "A then B" orderings * * \param[in,out] scheduler Scheduler data * * \note This function is currently disabled (see next comment). */ static void order_then_probes(pcmk_scheduler_t *scheduler) { #if 0 /* Given an ordering "A then B", we would prefer to wait for A to be started * before probing B. * * For example, if A is a filesystem which B can't even run without, it * would be helpful if the author of B's agent could assume that A is * running before B.monitor will be called. * * However, we can't _only_ probe after A is running, otherwise we wouldn't * detect the state of B if A could not be started. We can't even do an * opportunistic version of this, because B may be moving: * * A.stop -> A.start -> B.probe -> B.stop -> B.start * * and if we add B.stop -> A.stop here, we get a loop: * * A.stop -> A.start -> B.probe -> B.stop -> A.stop * * We could kill the "B.probe -> B.stop" dependency, but that could mean * stopping B "too" soon, because B.start must wait for the probe, and * we don't want to stop B if we can't start it. * * We could add the ordering only if A is an anonymous clone with * clone-max == node-max (since we'll never be moving it). However, we could * still be stopping one instance at the same time as starting another. * * The complexity of checking for allowed conditions combined with the ever * narrowing use case suggests that this code should remain disabled until * someone gets smarter. */ for (GList *iter = scheduler->priv->resources; iter != NULL; iter = iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data; pcmk_action_t *start = NULL; GList *actions = NULL; GList *probes = NULL; actions = pe__resource_actions(rsc, NULL, PCMK_ACTION_START, FALSE); if (actions) { start = actions->data; g_list_free(actions); } if (start == NULL) { crm_debug("No start action for %s", rsc->id); continue; } probes = pe__resource_actions(rsc, NULL, PCMK_ACTION_MONITOR, FALSE); for (actions = start->actions_before; actions != NULL; actions = actions->next) { pcmk__related_action_t *before = actions->data; pcmk_action_t *first = before->action; pcmk_resource_t *first_rsc = first->rsc; if (first->required_runnable_before) { for (GList *clone_actions = first->actions_before; clone_actions != NULL; clone_actions = clone_actions->next) { before = clone_actions->data; crm_trace("Testing '%s then %s' for %s", first->uuid, before->action->uuid, start->uuid); pcmk__assert(before->action->rsc != NULL); first_rsc = before->action->rsc; break; } } else if (!pcmk__str_eq(first->task, PCMK_ACTION_START, pcmk__str_none)) { crm_trace("Not a start op %s for %s", first->uuid, start->uuid); } if (first_rsc == NULL) { continue; } else if (pe__const_top_resource(first_rsc, false) == pe__const_top_resource(start->rsc, false)) { crm_trace("Same parent %s for %s", first_rsc->id, start->uuid); continue; } else if (!pcmk__is_clone(pe__const_top_resource(first_rsc, false))) { crm_trace("Not a clone %s for %s", first_rsc->id, start->uuid); continue; } crm_debug("Applying %s before %s", first->uuid, start->uuid); for (GList *probe_iter = probes; probe_iter != NULL; probe_iter = probe_iter->next) { pcmk_action_t *probe = (pcmk_action_t *) probe_iter->data; crm_debug("Ordering %s before %s", first->uuid, probe->uuid); order_actions(first, probe, pcmk__ar_ordered); } } } #endif } void pcmk__order_probes(pcmk_scheduler_t *scheduler) { // Add orderings for "probe then X" g_list_foreach(scheduler->priv->resources, add_start_restart_orderings_for_rsc, NULL); add_probe_orderings_for_stops(scheduler); order_then_probes(scheduler); } /*! * \internal * \brief Schedule any probes needed * * \param[in,out] scheduler Scheduler data * * \note This may also schedule fencing of failed remote nodes. */ void pcmk__schedule_probes(pcmk_scheduler_t *scheduler) { // Schedule probes on each node in the cluster as needed for (GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) { pcmk_node_t *node = (pcmk_node_t *) iter->data; if (!node->details->online) { // Don't probe offline nodes if (pcmk__is_failed_remote_node(node)) { pe_fence_node(scheduler, node, "the connection is unrecoverable", FALSE); } continue; } if (node->details->unclean) { // Don't probe nodes that need fencing continue; } if (!pcmk_is_set(node->priv->flags, pcmk__node_probes_allowed)) { // The user requested that probes not be done on this node continue; } // Probe each resource in the cluster on this node, as needed pcmk__probe_resource_list(scheduler->priv->resources, node); } } diff --git a/lib/pacemaker/pcmk_simulate.c b/lib/pacemaker/pcmk_simulate.c index e24269f80f..2805e543db 100644 --- a/lib/pacemaker/pcmk_simulate.c +++ b/lib/pacemaker/pcmk_simulate.c @@ -1,1065 +1,1067 @@ /* * Copyright 2021-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include // uint32_t, uint64_t #include #include #include +#include // g_str_has_suffix() + #include "libpacemaker_private.h" static const char *profiling_dir = NULL; static pcmk__output_t *out = NULL; static cib_t *fake_cib = NULL; static GList *fake_resource_list = NULL; static const GList *fake_op_fail_list = NULL; static void set_effective_date(pcmk_scheduler_t *scheduler, bool print_original, const char *use_date); /*! * \internal * \brief Create an action name for use in a dot graph * * \param[in] action Action to create name for * \param[in] verbose If true, add action ID to name * * \return Newly allocated string with action name * \note It is the caller's responsibility to free the result. */ static char * create_action_name(const pcmk_action_t *action, bool verbose) { char *action_name = NULL; const char *prefix = ""; const char *action_host = NULL; const char *history_id = NULL; const char *task = action->task; if (action->node != NULL) { action_host = action->node->priv->name; } else if (!pcmk_is_set(action->flags, pcmk__action_pseudo)) { action_host = ""; } if (pcmk__str_eq(action->task, PCMK_ACTION_CANCEL, pcmk__str_none)) { prefix = "Cancel "; task = action->cancel_task; } if (action->rsc != NULL) { history_id = action->rsc->priv->history_id; } if (history_id != NULL) { char *key = NULL; guint interval_ms = 0; if (pcmk__guint_from_hash(action->meta, PCMK_META_INTERVAL, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } if (pcmk__strcase_any_of(action->task, PCMK_ACTION_NOTIFY, PCMK_ACTION_NOTIFIED, NULL)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_key_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_key_operation"); pcmk__assert((n_type != NULL) && (n_task != NULL)); key = pcmk__notify_key(history_id, n_type, n_task); } else { key = pcmk__op_key(history_id, task, interval_ms); } if (action_host != NULL) { action_name = crm_strdup_printf("%s%s %s", prefix, key, action_host); } else { action_name = crm_strdup_printf("%s%s", prefix, key); } free(key); } else if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH, pcmk__str_none)) { const char *op = g_hash_table_lookup(action->meta, PCMK__META_STONITH_ACTION); action_name = crm_strdup_printf("%s%s '%s' %s", prefix, action->task, op, action_host); } else if (action->rsc && action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->uuid, action_host); } else if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->task, action_host); } else { action_name = crm_strdup_printf("%s", action->uuid); } if (verbose) { char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id); free(action_name); action_name = with_id; } return action_name; } /*! * \internal * \brief Display the status of a cluster * * \param[in,out] scheduler Scheduler data * \param[in] show_opts How to modify display (as pcmk_show_opt_e flags) * \param[in] section_opts Sections to display (as pcmk_section_e flags) * \param[in] title What to use as list title * \param[in] print_spacer Whether to display a spacer first */ static void print_cluster_status(pcmk_scheduler_t *scheduler, uint32_t show_opts, uint32_t section_opts, const char *title, bool print_spacer) { pcmk__output_t *out = scheduler->priv->out; GList *all = NULL; crm_exit_t stonith_rc = 0; enum pcmk_pacemakerd_state state = pcmk_pacemakerd_state_invalid; section_opts |= pcmk_section_nodes | pcmk_section_resources; show_opts |= pcmk_show_inactive_rscs | pcmk_show_failed_detail; all = g_list_prepend(all, (gpointer) "*"); PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "%s", title); out->message(out, "cluster-status", scheduler, state, stonith_rc, NULL, pcmk__fence_history_none, section_opts, show_opts, NULL, all, all); out->end_list(out); g_list_free(all); } /*! * \internal * \brief Display a summary of all actions scheduled in a transition * * \param[in,out] scheduler Scheduler data (fully scheduled) * \param[in] print_spacer Whether to display a spacer first */ static void print_transition_summary(pcmk_scheduler_t *scheduler, bool print_spacer) { pcmk__output_t *out = scheduler->priv->out; PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "Transition Summary"); pcmk__output_actions(scheduler); out->end_list(out); } /*! * \internal * \brief Reset scheduler, set some members, and unpack status * * \param[in,out] scheduler Scheduler data * \param[in] input What to set as cluster input * \param[in] out What to set as cluster output object * \param[in] use_date What to set as cluster's current timestamp * \param[in] flags Group of enum pcmk__scheduler_flags to set */ static void reset(pcmk_scheduler_t *scheduler, xmlNodePtr input, pcmk__output_t *out, const char *use_date, unsigned int flags) { pcmk_reset_scheduler(scheduler); scheduler->input = input; scheduler->priv->out = out; set_effective_date(scheduler, true, use_date); if (pcmk_is_set(flags, pcmk_sim_sanitized)) { pcmk__set_scheduler_flags(scheduler, pcmk__sched_sanitized); } if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pcmk__set_scheduler_flags(scheduler, pcmk__sched_output_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pcmk__set_scheduler_flags(scheduler, pcmk__sched_show_utilization); } cluster_status(scheduler); } /*! * \brief Write out a file in dot(1) format describing the actions that will * be taken by the scheduler in response to an input CIB file. * * \param[in,out] scheduler Scheduler data * \param[in] dot_file The filename to write * \param[in] all_actions Write all actions, even those that are optional * or are on unmanaged resources * \param[in] verbose Add extra information, such as action IDs, to the * output * * \return Standard Pacemaker return code */ static int write_sim_dotfile(pcmk_scheduler_t *scheduler, const char *dot_file, bool all_actions, bool verbose) { GList *iter = NULL; FILE *dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { return errno; } fprintf(dot_strm, " digraph \"g\" {\n"); for (iter = scheduler->priv->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = (pcmk_action_t *) iter->data; const char *style = "dashed"; const char *font = "black"; const char *color = NULL; char *action_name = create_action_name(action, verbose); if (pcmk_is_set(action->flags, pcmk__action_pseudo)) { font = "orange"; } if (pcmk_is_set(action->flags, pcmk__action_added_to_graph)) { style = PCMK__VALUE_BOLD; color = "green"; } else if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pcmk__rsc_managed)) { color = "red"; font = "purple"; if (!all_actions) { goto do_not_write; } } else if (pcmk_is_set(action->flags, pcmk__action_optional)) { color = "blue"; if (!all_actions) { goto do_not_write; } } else { color = "red"; CRM_LOG_ASSERT(!pcmk_is_set(action->flags, pcmk__action_runnable)); } pcmk__set_action_flags(action, pcmk__action_added_to_graph); fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n", action_name, style, color, font); do_not_write: free(action_name); } for (iter = scheduler->priv->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = (pcmk_action_t *) iter->data; for (GList *before_iter = action->actions_before; before_iter != NULL; before_iter = before_iter->next) { pcmk__related_action_t *before = before_iter->data; char *before_name = NULL; char *after_name = NULL; const char *style = "dashed"; bool optional = true; if (before->graphed) { optional = false; style = PCMK__VALUE_BOLD; } else if (before->flags == pcmk__ar_none) { continue; } else if (pcmk_is_set(before->action->flags, pcmk__action_added_to_graph) && pcmk_is_set(action->flags, pcmk__action_added_to_graph) && before->flags != pcmk__ar_if_on_same_node_or_target) { optional = false; } if (all_actions || !optional) { before_name = create_action_name(before->action, verbose); after_name = create_action_name(action, verbose); fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n", before_name, after_name, style); free(before_name); free(after_name); } } } fprintf(dot_strm, "}\n"); fflush(dot_strm); fclose(dot_strm); return pcmk_rc_ok; } /*! * \internal * \brief \c scandir() filter for scheduler input CIB files to profile * * \param[in] entry Directory entry * * \retval 1 if the filename ends with ".xml", does not begin with ".", and * refers to a regular file * \retval 0 otherwise */ static int profile_filter(const struct dirent *entry) { const char *filename = entry->d_name; char *buf = NULL; struct stat sb; int rc = 0; if (pcmk__str_any_of(filename, ".", "..", NULL)) { // Skip current (".") and parent ("..") directory links goto done; } if (filename[0] == '.') { crm_trace("Not profiling hidden file '%s'", filename); goto done; } - if (!pcmk__ends_with_ext(filename, ".xml")) { + if (!g_str_has_suffix(filename, ".xml")) { crm_trace("Not profiling file '%s' without '.xml' extension", filename); goto done; } buf = crm_strdup_printf("%s/%s", profiling_dir, filename); if ((stat(buf, &sb) != 0) || !S_ISREG(sb.st_mode)) { crm_trace("Not profiling file '%s': not a regular file", filename); goto done; } rc = 1; done: free(buf); return rc; } /*! * \internal * \brief Profile the configuration updates and scheduler actions in a single * CIB file, printing the profiling timings. * * \note \p scheduler->priv->out must have been set to a valid \c pcmk__output_t * object before this function is called. * * \param[in] xml_file The CIB file to profile * \param[in] repeat Number of times to run * \param[in,out] scheduler Scheduler data * \param[in,out] flags Group of enum pcmk__scheduler_flags to set * in addition to defaults * \param[in] use_date The date to set the cluster's time to (may be NULL) */ static void profile_file(const char *xml_file, unsigned int repeat, pcmk_scheduler_t *scheduler, uint64_t flags, const char *use_date) { pcmk__output_t *out = scheduler->priv->out; xmlNode *cib_object = NULL; clock_t start = 0; clock_t end; pcmk__assert(out != NULL); cib_object = pcmk__xml_read(xml_file); start = clock(); if (pcmk_find_cib_element(cib_object, PCMK_XE_STATUS) == NULL) { pcmk__xe_create(cib_object, PCMK_XE_STATUS); } if (pcmk__update_configured_schema(&cib_object, false) != pcmk_rc_ok) { goto done; } if (!pcmk__validate_xml(cib_object, NULL, NULL, NULL)) { goto done; } for (int i = 0; i < repeat; ++i) { pcmk_reset_scheduler(scheduler); scheduler->input = pcmk__xml_copy(NULL, cib_object); pcmk__set_scheduler_flags(scheduler, flags); set_effective_date(scheduler, false, use_date); pcmk__schedule_actions(scheduler); } pcmk_reset_scheduler(scheduler); end = clock(); out->message(out, "profile", xml_file, start, end); done: pcmk__xml_free(cib_object); } int pcmk__profile_dir(pcmk__output_t *out, uint32_t flags, const char *dir, unsigned int repeat, const char *use_date) { pcmk_scheduler_t *scheduler = NULL; uint64_t scheduler_flags = pcmk__sched_none; struct dirent **namelist = NULL; int num_files = 0; int rc = pcmk_rc_ok; pcmk__assert(out != NULL); scheduler = pcmk_new_scheduler(); if (scheduler == NULL) { return ENOMEM; } scheduler->priv->out = out; if (pcmk_is_set(flags, pcmk_sim_show_scores)) { scheduler_flags |= pcmk__sched_output_scores; } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { scheduler_flags |= pcmk__sched_show_utilization; } // Hack to pass user data to profile_filter profiling_dir = dir; num_files = scandir(dir, &namelist, profile_filter, alphasort); profiling_dir = NULL; if (num_files < 0) { rc = errno; goto done; } if (num_files == 0) { goto done; } out->begin_list(out, NULL, NULL, "Timings"); for (int i = 0; i < num_files; i++) { // glibc doesn't enforce PATH_MAX, so don't limit the buffer size char *path = crm_strdup_printf("%s/%s", dir, namelist[i]->d_name); profile_file(path, repeat, scheduler, scheduler_flags, use_date); free(path); free(namelist[i]); } out->end_list(out); done: pcmk_free_scheduler(scheduler); free(namelist); return rc; } /*! * \brief Set the date of the cluster, either to the value given by * \p use_date, or to the \c PCMK_XA_EXECUTION_DATE value in the CIB. * * \note \p scheduler->priv->out must have been set to a valid \p pcmk__output_t * object before this function is called. * * \param[in,out] scheduler Scheduler data * \param[in] print_original If \p true, the \c PCMK_XA_EXECUTION_DATE * should also be printed * \param[in] use_date The date to set the cluster's time to * (may be NULL) */ static void set_effective_date(pcmk_scheduler_t *scheduler, bool print_original, const char *use_date) { pcmk__output_t *out = scheduler->priv->out; time_t original_date = 0; pcmk__assert(out != NULL); pcmk__xe_get_time(scheduler->input, PCMK_XA_EXECUTION_DATE, &original_date); if (use_date) { scheduler->priv->now = crm_time_new(use_date); out->info(out, "Setting effective cluster time: %s", use_date); crm_time_log(LOG_NOTICE, "Pretending 'now' is", scheduler->priv->now, crm_time_log_date | crm_time_log_timeofday); } else if (original_date != 0) { scheduler->priv->now = pcmk__copy_timet(original_date); if (print_original) { char *when = crm_time_as_string(scheduler->priv->now, crm_time_log_date |crm_time_log_timeofday); out->info(out, "Using the original execution date of: %s", when); free(when); } } } /*! * \internal * \brief Simulate successfully executing a pseudo-action in a graph * * \param[in,out] graph Graph to update with pseudo-action result * \param[in,out] action Pseudo-action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_pseudo_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = pcmk__xe_get(action->xml, PCMK__META_ON_NODE); const char *task = pcmk__xe_get(action->xml, PCMK__XA_OPERATION_KEY); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-pseudo-action", node, task); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate executing a resource action in a graph * * \param[in,out] graph Graph to update with resource action result * \param[in,out] action Resource action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_resource_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { int rc; lrmd_event_data_t *op = NULL; int target_outcome = PCMK_OCF_OK; const char *rtype = NULL; const char *rclass = NULL; const char *resource = NULL; const char *rprovider = NULL; const char *resource_config_name = NULL; const char *operation = pcmk__xe_get(action->xml, PCMK_XA_OPERATION); const char *target_rc_s = crm_meta_value(action->params, PCMK__META_OP_TARGET_RC); xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; xmlNode *action_rsc = pcmk__xe_first_child(action->xml, PCMK_XE_PRIMITIVE, NULL, NULL); char *node = pcmk__xe_get_copy(action->xml, PCMK__META_ON_NODE); char *uuid = NULL; const char *router_node = pcmk__xe_get(action->xml, PCMK__XA_ROUTER_NODE); // Certain actions don't need to be displayed or history entries if (pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) { crm_debug("No history injection for %s op on %s", operation, node); goto done; // Confirm action and update graph } if (action_rsc == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "Bad"); free(node); return EPROTO; } /* A resource might be known by different names in the configuration and in * the action (for example, a clone instance). Grab the configuration name * (which is preferred when writing history), and if necessary, the instance * name. */ resource_config_name = pcmk__xe_get(action_rsc, PCMK_XA_ID); if (resource_config_name == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "No ID"); free(node); return EPROTO; } resource = resource_config_name; if (pe_find_resource(fake_resource_list, resource) == NULL) { const char *longname = pcmk__xe_get(action_rsc, PCMK__XA_LONG_ID); if ((longname != NULL) && (pe_find_resource(fake_resource_list, longname) != NULL)) { resource = longname; } } // Certain actions need to be displayed but don't need history entries if (pcmk__strcase_any_of(operation, PCMK_ACTION_DELETE, PCMK_ACTION_META_DATA, NULL)) { out->message(out, "inject-rsc-action", resource, operation, node, (guint) 0); goto done; // Confirm action and update graph } rclass = pcmk__xe_get(action_rsc, PCMK_XA_CLASS); rtype = pcmk__xe_get(action_rsc, PCMK_XA_TYPE); rprovider = pcmk__xe_get(action_rsc, PCMK_XA_PROVIDER); pcmk__scan_min_int(target_rc_s, &target_outcome, 0); pcmk__assert(fake_cib->cmds->query(fake_cib, NULL, NULL, cib_sync_call) == pcmk_ok); // Ensure the action node is in the CIB uuid = pcmk__xe_get_copy(action->xml, PCMK__META_ON_NODE_UUID); cib_node = pcmk__inject_node(fake_cib, node, ((router_node == NULL)? uuid: node)); free(uuid); pcmk__assert(cib_node != NULL); // Add a history entry for the action cib_resource = pcmk__inject_resource_history(out, cib_node, resource, resource_config_name, rclass, rtype, rprovider); if (cib_resource == NULL) { crm_err("Could not simulate action %d history for resource %s", action->id, resource); free(node); pcmk__xml_free(cib_node); return EINVAL; } // Simulate and display an executor event for the action result op = pcmk__event_from_graph_action(cib_resource, action, PCMK_EXEC_DONE, target_outcome, "User-injected result"); out->message(out, "inject-rsc-action", resource, op->op_type, node, op->interval_ms); // Check whether action is in a list of desired simulated failures for (const GList *iter = fake_op_fail_list; iter != NULL; iter = iter->next) { const char *spec = (const char *) iter->data; char *key = NULL; const char *match_name = NULL; const char *offset = NULL; // Allow user to specify anonymous clone with or without instance number key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource; } free(key); // If not found, try the resource's name in the configuration if ((match_name == NULL) && (strcmp(resource, resource_config_name) != 0)) { key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource_config_name, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource_config_name; } free(key); } if (match_name == NULL) { continue; // This failed action entry doesn't match } // ${match_name}_${task}_${interval_in_ms}@${node}=${rc} rc = sscanf(spec, "%*[^=]=%d", (int *) &op->rc); if (rc != 1) { out->err(out, "Invalid failed operation '%s' " "(result code must be integer)", spec); continue; // Keep checking other list entries } out->info(out, "Pretending action %d failed with rc=%d", action->id, op->rc); pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); graph->abort_priority = PCMK_SCORE_INFINITY; if (pcmk__str_eq(op->op_type, PCMK_ACTION_START, pcmk__str_none)) { offset = pcmk__s(graph->failed_start_offset, PCMK_VALUE_INFINITY); } else if (pcmk__str_eq(op->op_type, PCMK_ACTION_STOP, pcmk__str_none)) { offset = pcmk__s(graph->failed_stop_offset, PCMK_VALUE_INFINITY); } pcmk__inject_failcount(out, fake_cib, cib_node, match_name, op->op_type, op->interval_ms, op->rc, pcmk_str_is_infinity(offset)); break; } pcmk__inject_action_result(cib_resource, op, node, target_outcome); lrmd_free_event(op); rc = fake_cib->cmds->modify(fake_cib, PCMK_XE_STATUS, cib_node, cib_sync_call); pcmk__assert(rc == pcmk_ok); done: free(node); pcmk__xml_free(cib_node); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a cluster action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Cluster action to simulate * * \return Standard Pacemaker return code */ static int simulate_cluster_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = pcmk__xe_get(action->xml, PCMK__META_ON_NODE); const char *task = pcmk__xe_get(action->xml, PCMK_XA_OPERATION); xmlNode *rsc = pcmk__xe_first_child(action->xml, PCMK_XE_PRIMITIVE, NULL, NULL); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-cluster-action", node, task, rsc); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a fencing action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Fencing action to simulate * * \return Standard Pacemaker return code */ static int simulate_fencing_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *op = crm_meta_value(action->params, PCMK__META_STONITH_ACTION); char *target = pcmk__xe_get_copy(action->xml, PCMK__META_ON_NODE); out->message(out, "inject-fencing-action", target, op); if (!pcmk__str_eq(op, PCMK_ACTION_ON, pcmk__str_casei)) { int rc = pcmk_ok; GString *xpath = g_string_sized_new(512); // Set node state to offline xmlNode *cib_node = pcmk__inject_node_state_change(fake_cib, target, false); pcmk__assert(cib_node != NULL); crm_xml_add(cib_node, PCMK_XA_CRM_DEBUG_ORIGIN, __func__); rc = fake_cib->cmds->replace(fake_cib, PCMK_XE_STATUS, cib_node, cib_sync_call); pcmk__assert(rc == pcmk_ok); // Simulate controller clearing node's resource history and attributes pcmk__g_strcat(xpath, "//" PCMK__XE_NODE_STATE "[@" PCMK_XA_UNAME "='", target, "']/" PCMK__XE_LRM, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call); g_string_truncate(xpath, 0); pcmk__g_strcat(xpath, "//" PCMK__XE_NODE_STATE "[@" PCMK_XA_UNAME "='", target, "']" "/" PCMK__XE_TRANSIENT_ATTRIBUTES, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call); pcmk__xml_free(cib_node); g_string_free(xpath, TRUE); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); free(target); return pcmk_rc_ok; } enum pcmk__graph_status pcmk__simulate_transition(pcmk_scheduler_t *scheduler, cib_t *cib, const GList *op_fail_list) { pcmk__graph_t *transition = NULL; enum pcmk__graph_status graph_rc; pcmk__graph_functions_t simulation_fns = { simulate_pseudo_action, simulate_resource_action, simulate_cluster_action, simulate_fencing_action, }; out = scheduler->priv->out; fake_cib = cib; fake_op_fail_list = op_fail_list; if (!out->is_quiet(out)) { out->begin_list(out, NULL, NULL, "Executing Cluster Transition"); } pcmk__set_graph_functions(&simulation_fns); transition = pcmk__unpack_graph(scheduler->priv->graph, crm_system_name); pcmk__log_graph(LOG_DEBUG, transition); fake_resource_list = scheduler->priv->resources; do { graph_rc = pcmk__execute_graph(transition); } while (graph_rc == pcmk__graph_active); fake_resource_list = NULL; if (graph_rc != pcmk__graph_complete) { out->err(out, "Transition failed: %s", pcmk__graph_status2text(graph_rc)); pcmk__log_graph(LOG_ERR, transition); out->err(out, "An invalid transition was produced"); } pcmk__free_graph(transition); if (!out->is_quiet(out)) { // If not quiet, we'll need the resulting CIB for later display xmlNode *cib_object = NULL; int rc = fake_cib->cmds->query(fake_cib, NULL, &cib_object, cib_sync_call); pcmk__assert(rc == pcmk_ok); pcmk_reset_scheduler(scheduler); scheduler->input = cib_object; out->end_list(out); } return graph_rc; } int pcmk__simulate(pcmk_scheduler_t *scheduler, pcmk__output_t *out, const pcmk_injections_t *injections, uint32_t flags, uint32_t section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { int printed = pcmk_rc_no_output; int rc = pcmk_rc_ok; xmlNodePtr input = NULL; cib_t *cib = NULL; rc = cib__signon_query(out, &cib, &input); if (rc != pcmk_rc_ok) { goto simulate_done; } reset(scheduler, input, out, use_date, flags); if (!out->is_quiet(out)) { const bool show_pending = pcmk_is_set(flags, pcmk_sim_show_pending); if (pcmk_is_set(scheduler->flags, pcmk__sched_in_maintenance)) { printed = out->message(out, "maint-mode", scheduler->flags); } if ((scheduler->priv->disabled_resources > 0) || (scheduler->priv->blocked_resources > 0)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); printed = out->info(out, "%d of %d resource instances DISABLED and " "%d BLOCKED from further action due to failure", scheduler->priv->disabled_resources, scheduler->priv->ninstances, scheduler->priv->blocked_resources); } /* Most formatted output headers use caps for each word, but this one * only has the first word capitalized for compatibility with pcs. */ print_cluster_status(scheduler, (show_pending? pcmk_show_pending : 0), section_opts, "Current cluster status", (printed == pcmk_rc_ok)); printed = pcmk_rc_ok; } // If the user requested any injections, handle them if ((injections->node_down != NULL) || (injections->node_fail != NULL) || (injections->node_up != NULL) || (injections->op_inject != NULL) || (injections->ticket_activate != NULL) || (injections->ticket_grant != NULL) || (injections->ticket_revoke != NULL) || (injections->ticket_standby != NULL) || (injections->watchdog != NULL)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); pcmk__inject_scheduler_input(scheduler, cib, injections); printed = pcmk_rc_ok; rc = cib->cmds->query(cib, NULL, &input, cib_sync_call); if (rc != pcmk_rc_ok) { rc = pcmk_legacy2rc(rc); goto simulate_done; } reset(scheduler, input, out, use_date, flags); } if (input_file != NULL) { rc = pcmk__xml_write_file(input, input_file, false); if (rc != pcmk_rc_ok) { goto simulate_done; } } if (pcmk_any_flags_set(flags, pcmk_sim_process | pcmk_sim_simulate)) { pcmk__output_t *logger_out = NULL; if (pcmk_all_flags_set(scheduler->flags, pcmk__sched_output_scores |pcmk__sched_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Assignment Scores and Utilization Information"); printed = pcmk_rc_ok; } else if (pcmk_is_set(scheduler->flags, pcmk__sched_output_scores)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Assignment Scores"); printed = pcmk_rc_ok; } else if (pcmk_is_set(scheduler->flags, pcmk__sched_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Utilization Information"); printed = pcmk_rc_ok; } else { rc = pcmk__log_output_new(&logger_out); if (rc != pcmk_rc_ok) { goto simulate_done; } pe__register_messages(logger_out); pcmk__register_lib_messages(logger_out); scheduler->priv->out = logger_out; } pcmk__schedule_actions(scheduler); if (logger_out == NULL) { out->end_list(out); } else { logger_out->finish(logger_out, CRM_EX_OK, true, NULL); pcmk__output_free(logger_out); scheduler->priv->out = out; } input = NULL; /* Don't try and free it twice */ if (graph_file != NULL) { rc = pcmk__xml_write_file(scheduler->priv->graph, graph_file, false); if (rc != pcmk_rc_ok) { rc = pcmk_rc_graph_error; goto simulate_done; } } if (dot_file != NULL) { rc = write_sim_dotfile(scheduler, dot_file, pcmk_is_set(flags, pcmk_sim_all_actions), pcmk_is_set(flags, pcmk_sim_verbose)); if (rc != pcmk_rc_ok) { rc = pcmk_rc_dot_error; goto simulate_done; } } if (!out->is_quiet(out)) { print_transition_summary(scheduler, printed == pcmk_rc_ok); } } rc = pcmk_rc_ok; if (!pcmk_is_set(flags, pcmk_sim_simulate)) { goto simulate_done; } PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); if (pcmk__simulate_transition(scheduler, cib, injections->op_fail) != pcmk__graph_complete) { rc = pcmk_rc_invalid_transition; } if (out->is_quiet(out)) { goto simulate_done; } set_effective_date(scheduler, true, use_date); if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pcmk__set_scheduler_flags(scheduler, pcmk__sched_output_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pcmk__set_scheduler_flags(scheduler, pcmk__sched_show_utilization); } cluster_status(scheduler); print_cluster_status(scheduler, 0, section_opts, "Revised Cluster Status", true); simulate_done: cib__clean_up_connection(&cib); return rc; } // @COMPAT Use uint32_t for flags int pcmk_simulate(xmlNodePtr *xml, pcmk_scheduler_t *scheduler, const pcmk_injections_t *injections, unsigned int flags, unsigned int section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pe__register_messages(out); pcmk__register_lib_messages(out); rc = pcmk__simulate(scheduler, out, injections, (uint32_t) flags, (uint32_t) section_opts, use_date, input_file, graph_file, dot_file); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } diff --git a/lib/pengine/pe_health.c b/lib/pengine/pe_health.c index f90a21d26f..c5320f4992 100644 --- a/lib/pengine/pe_health.c +++ b/lib/pengine/pe_health.c @@ -1,183 +1,189 @@ /* - * Copyright 2004-2024 the Pacemaker project contributors + * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include +#include // g_str_has_prefix() + #include #include #include "pe_status_private.h" /*! * \internal * \brief Set the node health values to use for \c PCMK_VALUE_RED, * \c PCMK_VALUE_YELLOW, and \c PCMK_VALUE_GREEN * * \param[in,out] scheduler Scheduler data */ void pe__unpack_node_health_scores(pcmk_scheduler_t *scheduler) { switch (pe__health_strategy(scheduler)) { case pcmk__health_strategy_none: pcmk__score_red = 0; pcmk__score_yellow = 0; pcmk__score_green = 0; break; case pcmk__health_strategy_no_red: pcmk__score_red = -PCMK_SCORE_INFINITY; pcmk__score_yellow = 0; pcmk__score_green = 0; break; case pcmk__health_strategy_only_green: pcmk__score_red = -PCMK_SCORE_INFINITY; pcmk__score_yellow = -PCMK_SCORE_INFINITY; pcmk__score_green = 0; break; default: // progressive or custom pcmk__score_red = pcmk__health_score(PCMK_OPT_NODE_HEALTH_RED, scheduler); pcmk__score_green = pcmk__health_score(PCMK_OPT_NODE_HEALTH_GREEN, scheduler); pcmk__score_yellow = pcmk__health_score(PCMK_OPT_NODE_HEALTH_YELLOW, scheduler); break; } if ((pcmk__score_red != 0) || (pcmk__score_yellow != 0) || (pcmk__score_green != 0)) { crm_debug("Values of node health scores: " PCMK_VALUE_RED "=%d " PCMK_VALUE_YELLOW "=%d " PCMK_VALUE_GREEN "=%d", pcmk__score_red, pcmk__score_yellow, pcmk__score_green); } } struct health_sum { const pcmk_node_t *node; // Node that health score is being summed for int sum; // Sum of health scores checked so far }; /*! * \internal * \brief Add node attribute value to an integer, if it is a health attribute * * \param[in] key Name of node attribute * \param[in] value String value of node attribute * \param[in,out] user_data Address of integer to which \p value should be * added if \p key is a node health attribute */ static void add_node_health_value(gpointer key, gpointer value, gpointer user_data) { - if (pcmk__starts_with((const char *) key, "#health")) { + if (key == NULL) { + return; + } + + if (g_str_has_prefix((const gchar *) key, "#health")) { struct health_sum *health_sum = user_data; int score = 0; int rc = pcmk_parse_score((const char *) value, &score, 0); if (rc != pcmk_rc_ok) { crm_warn("Ignoring %s for %s because '%s' is not a valid value: %s", (const char *) key, pcmk__node_name(health_sum->node), (const char *) value, pcmk_rc_str(rc)); return; } health_sum->sum = pcmk__add_scores(score, health_sum->sum); crm_trace("Combined '%s' into node health score (now %s)", (const char *) value, pcmk_readable_score(health_sum->sum)); } } /*! * \internal * \brief Sum a node's health attribute scores * * \param[in] node Node whose health attributes should be added * \param[in] base_health Add this number to the total * * \return Sum of all health attribute scores of \p node plus \p base_health */ int pe__sum_node_health_scores(const pcmk_node_t *node, int base_health) { struct health_sum health_sum = { node, base_health, }; pcmk__assert(node != NULL); g_hash_table_foreach(node->priv->attrs, add_node_health_value, &health_sum); return health_sum.sum; } /*! * \internal * \brief Check the general health status for a node * * \param[in,out] node Node to check * * \return A negative value if any health attribute for \p node is red, * otherwise 0 if any attribute is yellow, otherwise a positive value. */ int pe__node_health(pcmk_node_t *node) { GHashTableIter iter; const char *name = NULL; const char *value = NULL; enum pcmk__health_strategy strategy; int score = 0; int rc = 1; pcmk__assert(node != NULL); strategy = pe__health_strategy(node->priv->scheduler); if (strategy == pcmk__health_strategy_none) { return rc; } g_hash_table_iter_init(&iter, node->priv->attrs); while (g_hash_table_iter_next(&iter, (gpointer *) &name, (gpointer *) &value)) { - if (pcmk__starts_with(name, "#health")) { + if ((name != NULL) && g_str_has_prefix(name, "#health")) { int parse_rc = pcmk_rc_ok; /* It's possible that pcmk__score_red equals pcmk__score_yellow, * or pcmk__score_yellow equals pcmk__score_green, so check the * textual value first to be able to distinguish those. */ if (pcmk__str_eq(value, PCMK_VALUE_RED, pcmk__str_casei)) { return -1; } else if (pcmk__str_eq(value, PCMK_VALUE_YELLOW, pcmk__str_casei)) { rc = 0; continue; } parse_rc = pcmk_parse_score(value, &score, 0); if (parse_rc != pcmk_rc_ok) { crm_warn("Ignoring %s for %s " "because '%s' is not a valid value: %s", name, pcmk__node_name(node), value, pcmk_rc_str(parse_rc)); continue; } // The value is an integer, so compare numerically if (score <= pcmk__score_red) { return -1; } else if ((score <= pcmk__score_yellow) && (pcmk__score_yellow != pcmk__score_green)) { rc = 0; } } } return rc; } diff --git a/lib/pengine/pe_output.c b/lib/pengine/pe_output.c index 83d0205361..a9af3aca62 100644 --- a/lib/pengine/pe_output.c +++ b/lib/pengine/pe_output.c @@ -1,3473 +1,3474 @@ /* * Copyright 2019-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include +#include // g_strchomp() #include // xmlNode #include #include #include #include #include #include const char * pe__resource_description(const pcmk_resource_t *rsc, uint32_t show_opts) { const char * desc = NULL; // User-supplied description if (pcmk_any_flags_set(show_opts, pcmk_show_rsc_only|pcmk_show_description)) { desc = pcmk__xe_get(rsc->priv->xml, PCMK_XA_DESCRIPTION); } return desc; } /* Never display node attributes whose name starts with one of these prefixes */ #define FILTER_STR { PCMK__FAIL_COUNT_PREFIX, PCMK__LAST_FAILURE_PREFIX, \ PCMK__NODE_ATTR_SHUTDOWN, PCMK_NODE_ATTR_TERMINATE, \ PCMK_NODE_ATTR_STANDBY, "#", NULL } static int compare_attribute(gconstpointer a, gconstpointer b) { int rc; rc = strcmp((const char *)a, (const char *)b); return rc; } /*! * \internal * \brief Determine whether extended information about an attribute should be added. * * \param[in] node Node that ran this resource * \param[in,out] rsc_list List of resources for this node * \param[in,out] scheduler Scheduler data * \param[in] attrname Attribute to find * \param[out] expected_score Expected value for this attribute * * \return true if extended information should be printed, false otherwise * \note Currently, extended information is only supported for ping/pingd * resources, for which a message will be printed if connectivity is lost * or degraded. */ static bool add_extra_info(const pcmk_node_t *node, GList *rsc_list, pcmk_scheduler_t *scheduler, const char *attrname, int *expected_score) { GList *gIter = NULL; for (gIter = rsc_list; gIter != NULL; gIter = gIter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter->data; const char *type = g_hash_table_lookup(rsc->priv->meta, PCMK_XA_TYPE); const char *name = NULL; GHashTable *params = NULL; if (rsc->priv->children != NULL) { if (add_extra_info(node, rsc->priv->children, scheduler, attrname, expected_score)) { return true; } } if (!pcmk__strcase_any_of(type, "ping", "pingd", NULL)) { continue; } params = pe_rsc_params(rsc, node, scheduler); name = g_hash_table_lookup(params, PCMK_XA_NAME); if (name == NULL) { name = "pingd"; } /* To identify the resource with the attribute name. */ if (pcmk__str_eq(name, attrname, pcmk__str_casei)) { int host_list_num = 0; const char *hosts = g_hash_table_lookup(params, "host_list"); const char *multiplier = g_hash_table_lookup(params, "multiplier"); int multiplier_i; if (hosts) { char **host_list = g_strsplit(hosts, " ", 0); host_list_num = g_strv_length(host_list); g_strfreev(host_list); } if ((multiplier == NULL) || (pcmk__scan_min_int(multiplier, &multiplier_i, INT_MIN) != pcmk_rc_ok)) { /* The ocf:pacemaker:ping resource agent defaults multiplier to * 1. The agent currently does not handle invalid text, but it * should, and this would be a reasonable choice ... */ multiplier_i = 1; } *expected_score = host_list_num * multiplier_i; return true; } } return false; } static GList * filter_attr_list(GList *attr_list, char *name) { int i; const char *filt_str[] = FILTER_STR; CRM_CHECK(name != NULL, return attr_list); /* filtering automatic attributes */ for (i = 0; filt_str[i] != NULL; i++) { if (g_str_has_prefix(name, filt_str[i])) { return attr_list; } } return g_list_insert_sorted(attr_list, name, compare_attribute); } static GList * get_operation_list(xmlNode *rsc_entry) { GList *op_list = NULL; xmlNode *rsc_op = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry, PCMK__XE_LRM_RSC_OP, NULL, NULL); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op, PCMK__XE_LRM_RSC_OP)) { const char *task = pcmk__xe_get(rsc_op, PCMK_XA_OPERATION); if (pcmk__str_eq(task, PCMK_ACTION_NOTIFY, pcmk__str_none)) { continue; // Ignore notify actions } else { int exit_status; pcmk__scan_min_int(pcmk__xe_get(rsc_op, PCMK__XA_RC_CODE), &exit_status, 0); if ((exit_status == CRM_EX_NOT_RUNNING) && pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_none) && pcmk__str_eq(pcmk__xe_get(rsc_op, PCMK_META_INTERVAL), "0", pcmk__str_null_matches)) { continue; // Ignore probes that found the resource not running } } op_list = g_list_append(op_list, rsc_op); } op_list = g_list_sort(op_list, sort_op_by_callid); return op_list; } static void add_dump_node(gpointer key, gpointer value, gpointer user_data) { xmlNodePtr node = user_data; node = pcmk__xe_create(node, (const char *) key); pcmk__xe_set_content(node, "%s", (const char *) value); } static void append_dump_text(gpointer key, gpointer value, gpointer user_data) { char **dump_text = user_data; char *new_text = crm_strdup_printf("%s %s=%s", *dump_text, (char *)key, (char *)value); free(*dump_text); *dump_text = new_text; } #define XPATH_STACK "//" PCMK_XE_NVPAIR \ "[@" PCMK_XA_NAME "='" \ PCMK_OPT_CLUSTER_INFRASTRUCTURE "']" static const char * get_cluster_stack(pcmk_scheduler_t *scheduler) { xmlNode *stack = pcmk__xpath_find_one(scheduler->input->doc, XPATH_STACK, LOG_DEBUG); if (stack != NULL) { return pcmk__xe_get(stack, PCMK_XA_VALUE); } return PCMK_VALUE_UNKNOWN; } static char * last_changed_string(const char *last_written, const char *user, const char *client, const char *origin) { if (last_written != NULL || user != NULL || client != NULL || origin != NULL) { return crm_strdup_printf("%s%s%s%s%s%s%s", last_written ? last_written : "", user ? " by " : "", user ? user : "", client ? " via " : "", client ? client : "", origin ? " on " : "", origin ? origin : ""); } else { return strdup(""); } } static char * op_history_string(xmlNode *xml_op, const char *task, const char *interval_ms_s, int rc, bool print_timing) { const char *call = pcmk__xe_get(xml_op, PCMK__XA_CALL_ID); char *interval_str = NULL; char *buf = NULL; if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *pair = pcmk__format_nvpair(PCMK_XA_INTERVAL, interval_ms_s, "ms"); interval_str = crm_strdup_printf(" %s", pair); free(pair); } if (print_timing) { char *last_change_str = NULL; char *exec_str = NULL; char *queue_str = NULL; const char *value = NULL; time_t epoch = 0; pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &epoch); if (epoch > 0) { char *epoch_str = pcmk__epoch2str(&epoch, 0); last_change_str = crm_strdup_printf(" %s=\"%s\"", PCMK_XA_LAST_RC_CHANGE, pcmk__s(epoch_str, "")); free(epoch_str); } value = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); if (value) { char *pair = pcmk__format_nvpair(PCMK_XA_EXEC_TIME, value, "ms"); exec_str = crm_strdup_printf(" %s", pair); free(pair); } value = pcmk__xe_get(xml_op, PCMK_XA_QUEUE_TIME); if (value) { char *pair = pcmk__format_nvpair(PCMK_XA_QUEUE_TIME, value, "ms"); queue_str = crm_strdup_printf(" %s", pair); free(pair); } buf = crm_strdup_printf("(%s) %s:%s%s%s%s rc=%d (%s)", call, task, interval_str ? interval_str : "", last_change_str ? last_change_str : "", exec_str ? exec_str : "", queue_str ? queue_str : "", rc, crm_exit_str(rc)); if (last_change_str) { free(last_change_str); } if (exec_str) { free(exec_str); } if (queue_str) { free(queue_str); } } else { buf = crm_strdup_printf("(%s) %s%s%s", call, task, interval_str ? ":" : "", interval_str ? interval_str : ""); } if (interval_str) { free(interval_str); } return buf; } static char * resource_history_string(pcmk_resource_t *rsc, const char *rsc_id, bool all, int failcount, time_t last_failure) { char *buf = NULL; if (rsc == NULL) { buf = crm_strdup_printf("%s: orphan", rsc_id); } else if (all || failcount || last_failure > 0) { char *failcount_s = NULL; char *lastfail_s = NULL; if (failcount > 0) { failcount_s = crm_strdup_printf(" %s=%d", PCMK_XA_FAIL_COUNT, failcount); } else { failcount_s = strdup(""); } if (last_failure > 0) { buf = pcmk__epoch2str(&last_failure, 0); lastfail_s = crm_strdup_printf(" %s='%s'", PCMK_XA_LAST_FAILURE, buf); free(buf); } buf = crm_strdup_printf("%s: " PCMK_META_MIGRATION_THRESHOLD "=%d%s%s", rsc_id, rsc->priv->ban_after_failures, failcount_s, pcmk__s(lastfail_s, "")); free(failcount_s); free(lastfail_s); } else { buf = crm_strdup_printf("%s:", rsc_id); } return buf; } /*! * \internal * \brief Get a node's feature set for status display purposes * * \param[in] node Node to check * * \return String representation of feature set if the node is fully up (using * "<3.15.1" for older nodes that don't set the #feature-set attribute), * otherwise NULL */ static const char * get_node_feature_set(const pcmk_node_t *node) { if (node->details->online && pcmk_is_set(node->priv->flags, pcmk__node_expected_up) && !pcmk__is_pacemaker_remote_node(node)) { const char *feature_set = g_hash_table_lookup(node->priv->attrs, CRM_ATTR_FEATURE_SET); /* The feature set attribute is present since 3.15.1. If it is missing, * then the node must be running an earlier version. */ return pcmk__s(feature_set, "<3.15.1"); } return NULL; } static bool is_mixed_version(pcmk_scheduler_t *scheduler) { const char *feature_set = NULL; for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = gIter->data; const char *node_feature_set = get_node_feature_set(node); if (node_feature_set != NULL) { if (feature_set == NULL) { feature_set = node_feature_set; } else if (strcmp(feature_set, node_feature_set) != 0) { return true; } } } return false; } static void formatted_xml_buf(const pcmk_resource_t *rsc, GString *xml_buf, bool raw) { if (raw && (rsc->priv->orig_xml != NULL)) { pcmk__xml_string(rsc->priv->orig_xml, pcmk__xml_fmt_pretty, xml_buf, 0); } else { pcmk__xml_string(rsc->priv->xml, pcmk__xml_fmt_pretty, xml_buf, 0); } } #define XPATH_DC_VERSION "//" PCMK_XE_NVPAIR \ "[@" PCMK_XA_NAME "='" PCMK_OPT_DC_VERSION "']" PCMK__OUTPUT_ARGS("cluster-summary", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "uint32_t", "uint32_t") static int cluster_summary(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(scheduler); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s, pcmkd_state); } if (pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = pcmk__xpath_find_one(scheduler->input->doc, XPATH_DC_VERSION, LOG_DEBUG); const char *dc_version_s = dc_version? pcmk__xe_get(dc_version, PCMK_XA_VALUE) : NULL; const char *quorum = pcmk__xe_get(scheduler->input, PCMK_XA_HAVE_QUORUM); char *dc_name = scheduler->dc_node? pe__node_display_name(scheduler->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; bool mixed_version = is_mixed_version(scheduler); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-dc", scheduler->dc_node, quorum, dc_version_s, dc_name, mixed_version); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = pcmk__xe_get(scheduler->input, PCMK_XA_CIB_LAST_WRITTEN); const char *user = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_USER); const char *client = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_CLIENT); const char *origin = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_ORIGIN); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-times", scheduler->priv->local_node_name, last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(scheduler->nodes), scheduler->priv->ninstances, scheduler->priv->disabled_resources, scheduler->priv->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-options", scheduler); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", scheduler->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } PCMK__OUTPUT_ARGS("cluster-summary", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "uint32_t", "uint32_t") static int cluster_summary_html(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(scheduler); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s, pcmkd_state); } /* Always print DC if none, even if not requested */ if ((scheduler->dc_node == NULL) || pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = pcmk__xpath_find_one(scheduler->input->doc, XPATH_DC_VERSION, LOG_DEBUG); const char *dc_version_s = dc_version? pcmk__xe_get(dc_version, PCMK_XA_VALUE) : NULL; const char *quorum = pcmk__xe_get(scheduler->input, PCMK_XA_HAVE_QUORUM); char *dc_name = scheduler->dc_node? pe__node_display_name(scheduler->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; bool mixed_version = is_mixed_version(scheduler); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-dc", scheduler->dc_node, quorum, dc_version_s, dc_name, mixed_version); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = pcmk__xe_get(scheduler->input, PCMK_XA_CIB_LAST_WRITTEN); const char *user = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_USER); const char *client = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_CLIENT); const char *origin = pcmk__xe_get(scheduler->input, PCMK_XA_UPDATE_ORIGIN); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-times", scheduler->priv->local_node_name, last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(scheduler->nodes), scheduler->priv->ninstances, scheduler->priv->disabled_resources, scheduler->priv->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { /* Kind of a hack - close the list we may have opened earlier in this * function so we can put all the options into their own list. We * only want to do this on HTML output, though. */ PCMK__OUTPUT_LIST_FOOTER(out, rc); out->begin_list(out, NULL, NULL, "Config Options"); out->message(out, "cluster-options", scheduler); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", scheduler->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } char * pe__node_display_name(pcmk_node_t *node, bool print_detail) { char *node_name; const char *node_host = NULL; const char *node_id = NULL; int name_len; pcmk__assert((node != NULL) && (node->priv->name != NULL)); /* Host is displayed only if this is a guest node and detail is requested */ if (print_detail && pcmk__is_guest_or_bundle_node(node)) { const pcmk_resource_t *launcher = NULL; const pcmk_node_t *host_node = NULL; launcher = node->priv->remote->priv->launcher; host_node = pcmk__current_node(launcher); if (host_node && host_node->details) { node_host = host_node->priv->name; } if (node_host == NULL) { node_host = ""; /* so we at least get "uname@" to indicate guest */ } } /* Node ID is displayed if different from uname and detail is requested */ if (print_detail && !pcmk__str_eq(node->priv->name, node->priv->id, pcmk__str_casei)) { node_id = node->priv->id; } /* Determine name length */ name_len = strlen(node->priv->name) + 1; if (node_host) { name_len += strlen(node_host) + 1; /* "@node_host" */ } if (node_id) { name_len += strlen(node_id) + 3; /* + " (node_id)" */ } /* Allocate and populate display name */ node_name = pcmk__assert_alloc(name_len, sizeof(char)); strcpy(node_name, node->priv->name); if (node_host) { strcat(node_name, "@"); strcat(node_name, node_host); } if (node_id) { strcat(node_name, " ("); strcat(node_name, node_id); strcat(node_name, ")"); } return node_name; } int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name, ...) { xmlNodePtr xml_node = NULL; va_list pairs; pcmk__assert(tag_name != NULL); xml_node = pcmk__output_xml_peek_parent(out); pcmk__assert(xml_node != NULL); xml_node = pcmk__xe_create(xml_node, tag_name); va_start(pairs, tag_name); pcmk__xe_set_propv(xml_node, pairs); va_end(pairs); if (is_list) { pcmk__output_xml_push_parent(out, xml_node); } return pcmk_rc_ok; } static const char * role_desc(enum rsc_role_e role) { if (role == pcmk_role_promoted) { return "in " PCMK_ROLE_PROMOTED " role "; } return ""; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_html(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts = va_arg(args, uint32_t); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); char *buf = crm_strdup_printf("%s\tprevents %s from running %son %s", location->id, location->rsc->id, role_desc(location->role_filter), node_name); pcmk__output_create_html_node(out, "li", NULL, NULL, buf); free(node_name); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_text(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts = va_arg(args, uint32_t); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->list_item(out, NULL, "%s\tprevents %s from running %son %s", location->id, location->rsc->id, role_desc(location->role_filter), node_name); free(node_name); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); const char *promoted_only = pcmk__btoa(location->role_filter == pcmk_role_promoted); char *weight_s = pcmk__itoa(pe_node->assign->score); pcmk__output_create_xml_node(out, PCMK_XE_BAN, PCMK_XA_ID, location->id, PCMK_XA_RESOURCE, location->rsc->id, PCMK_XA_NODE, pe_node->priv->name, PCMK_XA_WEIGHT, weight_s, PCMK_XA_PROMOTED_ONLY, promoted_only, /* This is a deprecated alias for * promoted_only. Removing it will break * backward compatibility of the API schema, * which will require an API schema major * version bump. */ PCMK__XA_PROMOTED_ONLY_LEGACY, promoted_only, NULL); free(weight_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban-list", "pcmk_scheduler_t *", "const char *", "GList *", "uint32_t", "bool") static int ban_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); const char *prefix = va_arg(args, const char *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); GList *gIter, *gIter2; int rc = pcmk_rc_no_output; /* Print each ban */ for (gIter = scheduler->priv->location_constraints; gIter != NULL; gIter = gIter->next) { pcmk__location_t *location = gIter->data; const pcmk_resource_t *rsc = location->rsc; if (prefix != NULL && !g_str_has_prefix(location->id, prefix)) { continue; } if (!pcmk__str_in_list(rsc_printable_id(rsc), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(pe__const_top_resource(rsc, false)), only_rsc, pcmk__str_star_matches)) { continue; } for (gIter2 = location->nodes; gIter2 != NULL; gIter2 = gIter2->next) { pcmk_node_t *node = (pcmk_node_t *) gIter2->data; if (node->assign->score < 0) { PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Negative Location Constraints"); out->message(out, "ban", node, location, show_opts); } } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_html(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; child = pcmk__html_create(nodes_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%d node%s configured", nnodes, pcmk__plural_s(nnodes)); if (ndisabled && nblocked) { child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "DISABLED"); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, ", %d ", nblocked); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "BLOCKED"); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " from further action due to failure)"); } else if (ndisabled && !nblocked) { child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "DISABLED"); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, ")"); } else if (!ndisabled && nblocked) { child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), nblocked); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "BLOCKED"); child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " from further action due to failure)"); } else { child = pcmk__html_create(resources_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_text(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); out->list_item(out, NULL, "%d node%s configured", nnodes, pcmk__plural_s(nnodes)); if (ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED, %d BLOCKED from " "further action due to failure)", nresources, pcmk__plural_s(nresources), ndisabled, nblocked); } else if (ndisabled && !nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED)", nresources, pcmk__plural_s(nresources), ndisabled); } else if (!ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d BLOCKED from further action " "due to failure)", nresources, pcmk__plural_s(nresources), nblocked); } else { out->list_item(out, NULL, "%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_xml(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = NULL; xmlNodePtr resources_node = NULL; char *s = NULL; nodes_node = pcmk__output_create_xml_node(out, PCMK_XE_NODES_CONFIGURED, NULL); resources_node = pcmk__output_create_xml_node(out, PCMK_XE_RESOURCES_CONFIGURED, NULL); s = pcmk__itoa(nnodes); crm_xml_add(nodes_node, PCMK_XA_NUMBER, s); free(s); s = pcmk__itoa(nresources); crm_xml_add(resources_node, PCMK_XA_NUMBER, s); free(s); s = pcmk__itoa(ndisabled); crm_xml_add(resources_node, PCMK_XA_DISABLED, s); free(s); s = pcmk__itoa(nblocked); crm_xml_add(resources_node, PCMK_XA_BLOCKED, s); free(s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_html(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); bool mixed_version = va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "Current DC: "); if (dc) { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%s (version %s) -", dc_name, pcmk__s(dc_version_s, "unknown")); if (mixed_version) { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_WARNING); pcmk__xe_set_content(child, " MIXED-VERSION"); } child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " partition"); if (crm_is_true(quorum)) { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " with"); } else { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_WARNING); pcmk__xe_set_content(child, " WITHOUT"); } child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " quorum"); } else { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_WARNING); pcmk__xe_set_content(child, "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_text(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); bool mixed_version = va_arg(args, int); if (dc) { out->list_item(out, "Current DC", "%s (version %s) - %spartition %s quorum", dc_name, dc_version_s ? dc_version_s : "unknown", mixed_version ? "MIXED-VERSION " : "", crm_is_true(quorum) ? "with" : "WITHOUT"); } else { out->list_item(out, "Current DC", "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name G_GNUC_UNUSED = va_arg(args, char *); bool mixed_version = va_arg(args, int); if (dc) { const char *with_quorum = pcmk__btoa(crm_is_true(quorum)); const char *mixed_version_s = pcmk__btoa(mixed_version); pcmk__output_create_xml_node(out, PCMK_XE_CURRENT_DC, PCMK_XA_PRESENT, PCMK_VALUE_TRUE, PCMK_XA_VERSION, pcmk__s(dc_version_s, ""), PCMK_XA_NAME, dc->priv->name, PCMK_XA_ID, dc->priv->id, PCMK_XA_WITH_QUORUM, with_quorum, PCMK_XA_MIXED_VERSION, mixed_version_s, NULL); } else { pcmk__output_create_xml_node(out, PCMK_XE_CURRENT_DC, PCMK_XA_PRESENT, PCMK_VALUE_FALSE, NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("maint-mode", "uint64_t") static int cluster_maint_mode_text(pcmk__output_t *out, va_list args) { uint64_t flags = va_arg(args, uint64_t); if (pcmk_is_set(flags, pcmk__sched_in_maintenance)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will not attempt to start, stop or recover services\n"); return pcmk_rc_ok; } else if (pcmk_is_set(flags, pcmk__sched_stop_all)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will keep all resources stopped\n"); return pcmk_rc_ok; } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_html(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { out->list_item(out, NULL, "STONITH of failed nodes enabled"); } else { out->list_item(out, NULL, "STONITH of failed nodes disabled"); } if (pcmk_is_set(scheduler->flags, pcmk__sched_symmetric_cluster)) { out->list_item(out, NULL, "Cluster is symmetric"); } else { out->list_item(out, NULL, "Cluster is asymmetric"); } switch (scheduler->no_quorum_policy) { /* @COMPAT These should say something like "resources that require * quorum" since resources with requires="nothing" are unaffected, but * it would be a good idea to investigate whether any major projects * search for this text first */ case pcmk_no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case pcmk_no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case pcmk_no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case pcmk_no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case pcmk_no_quorum_fence: out->list_item(out, NULL, "No quorum policy: Fence nodes in partition"); break; } if (pcmk_is_set(scheduler->flags, pcmk__sched_in_maintenance)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "Resource management: "); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "DISABLED"); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " (the cluster will not attempt to start, stop," " or recover services)"); } else if (pcmk_is_set(scheduler->flags, pcmk__sched_stop_all)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "Resource management: "); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "STOPPED"); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " (the cluster will keep all resources stopped)"); } else { out->list_item(out, NULL, "Resource management: enabled"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_log(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk__sched_in_maintenance)) { return out->info(out, "Resource management is DISABLED. The cluster will not attempt to start, stop or recover services."); } else if (pcmk_is_set(scheduler->flags, pcmk__sched_stop_all)) { return out->info(out, "Resource management is DISABLED. The cluster has stopped all resources."); } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_text(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { out->list_item(out, NULL, "STONITH of failed nodes enabled"); } else { out->list_item(out, NULL, "STONITH of failed nodes disabled"); } if (pcmk_is_set(scheduler->flags, pcmk__sched_symmetric_cluster)) { out->list_item(out, NULL, "Cluster is symmetric"); } else { out->list_item(out, NULL, "Cluster is asymmetric"); } switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case pcmk_no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case pcmk_no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case pcmk_no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case pcmk_no_quorum_fence: out->list_item(out, NULL, "No quorum policy: Fence nodes in partition"); break; } return pcmk_rc_ok; } /*! * \internal * \brief Get readable string representation of a no-quorum policy * * \param[in] policy No-quorum policy * * \return String representation of \p policy */ static const char * no_quorum_policy_text(enum pe_quorum_policy policy) { switch (policy) { case pcmk_no_quorum_freeze: return PCMK_VALUE_FREEZE; case pcmk_no_quorum_stop: return PCMK_VALUE_STOP; case pcmk_no_quorum_demote: return PCMK_VALUE_DEMOTE; case pcmk_no_quorum_ignore: return PCMK_VALUE_IGNORE; case pcmk_no_quorum_fence: return PCMK_VALUE_FENCE; default: return PCMK_VALUE_UNKNOWN; } } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_xml(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); const char *stonith_enabled = pcmk__flag_text(scheduler->flags, pcmk__sched_fencing_enabled); const char *symmetric_cluster = pcmk__flag_text(scheduler->flags, pcmk__sched_symmetric_cluster); const char *no_quorum_policy = no_quorum_policy_text(scheduler->no_quorum_policy); const char *maintenance_mode = pcmk__flag_text(scheduler->flags, pcmk__sched_in_maintenance); const char *stop_all_resources = pcmk__flag_text(scheduler->flags, pcmk__sched_stop_all); char *stonith_timeout_ms_s = crm_strdup_printf("%u", scheduler->priv->fence_timeout_ms); char *priority_fencing_delay_ms_s = crm_strdup_printf("%u", scheduler->priv->priority_fencing_ms); pcmk__output_create_xml_node(out, PCMK_XE_CLUSTER_OPTIONS, PCMK_XA_STONITH_ENABLED, stonith_enabled, PCMK_XA_SYMMETRIC_CLUSTER, symmetric_cluster, PCMK_XA_NO_QUORUM_POLICY, no_quorum_policy, PCMK_XA_MAINTENANCE_MODE, maintenance_mode, PCMK_XA_STOP_ALL_RESOURCES, stop_all_resources, PCMK_XA_STONITH_TIMEOUT_MS, stonith_timeout_ms_s, PCMK_XA_PRIORITY_FENCING_DELAY_MS, priority_fencing_delay_ms_s, NULL); free(stonith_timeout_ms_s); free(priority_fencing_delay_ms_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_html(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "Stack: "); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%s", stack_s); if (pcmkd_state != pcmk_pacemakerd_state_invalid) { child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " ("); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%s", pcmk__pcmkd_state_enum2friendly(pcmkd_state)); child = pcmk__html_create(node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, ")"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_text(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); if (pcmkd_state != pcmk_pacemakerd_state_invalid) { out->list_item(out, "Stack", "%s (%s)", stack_s, pcmk__pcmkd_state_enum2friendly(pcmkd_state)); } else { out->list_item(out, "Stack", "%s", stack_s); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_xml(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = NULL; if (pcmkd_state != pcmk_pacemakerd_state_invalid) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(pcmkd_state); } pcmk__output_create_xml_node(out, PCMK_XE_STACK, PCMK_XA_TYPE, stack_s, PCMK_XA_PACEMAKERD_STATE, state_s, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_html(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); xmlNodePtr updated_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr changed_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; char *time_s = NULL; child = pcmk__html_create(updated_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "Last updated: "); child = pcmk__html_create(updated_node, PCMK__XE_SPAN, NULL, NULL); time_s = pcmk__epoch2str(NULL, 0); pcmk__xe_set_content(child, "%s", time_s); free(time_s); if (our_nodename != NULL) { child = pcmk__html_create(updated_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, " on "); child = pcmk__html_create(updated_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%s", our_nodename); } child = pcmk__html_create(changed_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "Last change: "); child = pcmk__html_create(changed_node, PCMK__XE_SPAN, NULL, NULL); time_s = last_changed_string(last_written, user, client, origin); pcmk__xe_set_content(child, "%s", time_s); free(time_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_xml(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); char *time_s = pcmk__epoch2str(NULL, 0); pcmk__output_create_xml_node(out, PCMK_XE_LAST_UPDATE, PCMK_XA_TIME, time_s, PCMK_XA_ORIGIN, our_nodename, NULL); pcmk__output_create_xml_node(out, PCMK_XE_LAST_CHANGE, PCMK_XA_TIME, pcmk__s(last_written, ""), PCMK_XA_USER, pcmk__s(user, ""), PCMK_XA_CLIENT, pcmk__s(client, ""), PCMK_XA_ORIGIN, pcmk__s(origin, ""), NULL); free(time_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_text(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); char *time_s = pcmk__epoch2str(NULL, 0); out->list_item(out, "Last updated", "%s%s%s", time_s, (our_nodename != NULL)? " on " : "", pcmk__s(our_nodename, "")); free(time_s); time_s = last_changed_string(last_written, user, client, origin); out->list_item(out, "Last change", " %s", time_s); free(time_s); return pcmk_rc_ok; } /*! * \internal * \brief Display a failed action in less-technical natural language * * \param[in,out] out Output object to use for display * \param[in] xml_op XML containing failed action * \param[in] op_key Operation key of failed action * \param[in] node_name Where failed action occurred * \param[in] rc OCF exit code of failed action * \param[in] status Execution status of failed action * \param[in] exit_reason Exit reason given for failed action * \param[in] exec_time String containing execution time in milliseconds */ static void failed_action_friendly(pcmk__output_t *out, const xmlNode *xml_op, const char *op_key, const char *node_name, int rc, int status, const char *exit_reason, const char *exec_time) { char *rsc_id = NULL; char *task = NULL; guint interval_ms = 0; time_t last_change_epoch = 0; GString *str = NULL; if (pcmk__str_empty(op_key) || !parse_op_key(op_key, &rsc_id, &task, &interval_ms)) { pcmk__str_update(&rsc_id, "unknown resource"); pcmk__str_update(&task, "unknown action"); interval_ms = 0; } pcmk__assert((rsc_id != NULL) && (task != NULL)); str = g_string_sized_new(256); // Should be sufficient for most messages pcmk__g_strcat(str, rsc_id, " ", NULL); if (interval_ms != 0) { pcmk__g_strcat(str, pcmk__readable_interval(interval_ms), "-interval ", NULL); } pcmk__g_strcat(str, pcmk__readable_action(task, interval_ms), " on ", node_name, NULL); if (status == PCMK_EXEC_DONE) { pcmk__g_strcat(str, " returned '", crm_exit_str(rc), "'", NULL); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, " (", exit_reason, ")", NULL); } } else { pcmk__g_strcat(str, " could not be executed (", pcmk_exec_status_str(status), NULL); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, ": ", exit_reason, NULL); } g_string_append_c(str, ')'); } if (pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &last_change_epoch) == pcmk_rc_ok) { char *s = pcmk__epoch2str(&last_change_epoch, 0); pcmk__g_strcat(str, " at ", s, NULL); free(s); } if (!pcmk__str_empty(exec_time)) { int exec_time_ms = 0; if ((pcmk__scan_min_int(exec_time, &exec_time_ms, 0) == pcmk_rc_ok) && (exec_time_ms > 0)) { pcmk__g_strcat(str, " after ", pcmk__readable_interval(exec_time_ms), NULL); } } out->list_item(out, NULL, "%s", str->str); g_string_free(str, TRUE); free(rsc_id); free(task); } /*! * \internal * \brief Display a failed action with technical details * * \param[in,out] out Output object to use for display * \param[in] xml_op XML containing failed action * \param[in] op_key Operation key of failed action * \param[in] node_name Where failed action occurred * \param[in] rc OCF exit code of failed action * \param[in] status Execution status of failed action * \param[in] exit_reason Exit reason given for failed action * \param[in] exec_time String containing execution time in milliseconds */ static void failed_action_technical(pcmk__output_t *out, const xmlNode *xml_op, const char *op_key, const char *node_name, int rc, int status, const char *exit_reason, const char *exec_time) { const char *call_id = pcmk__xe_get(xml_op, PCMK__XA_CALL_ID); const char *queue_time = pcmk__xe_get(xml_op, PCMK_XA_QUEUE_TIME); const char *exit_status = crm_exit_str(rc); const char *lrm_status = pcmk_exec_status_str(status); time_t last_change_epoch = 0; GString *str = NULL; if (pcmk__str_empty(op_key)) { op_key = "unknown operation"; } if (pcmk__str_empty(exit_status)) { exit_status = "unknown exit status"; } if (pcmk__str_empty(call_id)) { call_id = "unknown"; } str = g_string_sized_new(256); g_string_append_printf(str, "%s on %s '%s' (%d): call=%s, status='%s'", op_key, node_name, exit_status, rc, call_id, lrm_status); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, ", exitreason='", exit_reason, "'", NULL); } if (pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &last_change_epoch) == pcmk_rc_ok) { char *last_change_str = pcmk__epoch2str(&last_change_epoch, 0); pcmk__g_strcat(str, ", " PCMK_XA_LAST_RC_CHANGE "=" "'", last_change_str, "'", NULL); free(last_change_str); } if (!pcmk__str_empty(queue_time)) { pcmk__g_strcat(str, ", queued=", queue_time, "ms", NULL); } if (!pcmk__str_empty(exec_time)) { pcmk__g_strcat(str, ", exec=", exec_time, "ms", NULL); } out->list_item(out, NULL, "%s", str->str); g_string_free(str, TRUE); } PCMK__OUTPUT_ARGS("failed-action", "xmlNode *", "uint32_t") static int failed_action_default(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); uint32_t show_opts = va_arg(args, uint32_t); const char *op_key = pcmk__xe_history_key(xml_op); const char *node_name = pcmk__xe_get(xml_op, PCMK_XA_UNAME); const char *exit_reason = pcmk__xe_get(xml_op, PCMK_XA_EXIT_REASON); const char *exec_time = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); int rc; int status; pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_RC_CODE), &rc, 0); pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_OP_STATUS), &status, 0); if (pcmk__str_empty(node_name)) { node_name = "unknown node"; } if (pcmk_is_set(show_opts, pcmk_show_failed_detail)) { failed_action_technical(out, xml_op, op_key, node_name, rc, status, exit_reason, exec_time); } else { failed_action_friendly(out, xml_op, op_key, node_name, rc, status, exit_reason, exec_time); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action", "xmlNode *", "uint32_t") static int failed_action_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); const char *op_key = pcmk__xe_history_key(xml_op); const char *op_key_name = PCMK_XA_OP_KEY; int rc; int status; const char *uname = pcmk__xe_get(xml_op, PCMK_XA_UNAME); const char *call_id = pcmk__xe_get(xml_op, PCMK__XA_CALL_ID); const char *exitstatus = NULL; const char *exit_reason = pcmk__s(pcmk__xe_get(xml_op, PCMK_XA_EXIT_REASON), "none"); const char *status_s = NULL; time_t epoch = 0; gchar *exit_reason_esc = NULL; char *rc_s = NULL; xmlNodePtr node = NULL; if (pcmk__xml_needs_escape(exit_reason, pcmk__xml_escape_attr)) { exit_reason_esc = pcmk__xml_escape(exit_reason, pcmk__xml_escape_attr); exit_reason = exit_reason_esc; } pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_RC_CODE), &rc, 0); pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_OP_STATUS), &status, 0); if (pcmk__xe_get(xml_op, PCMK__XA_OPERATION_KEY) == NULL) { op_key_name = PCMK_XA_ID; } exitstatus = crm_exit_str(rc); rc_s = pcmk__itoa(rc); status_s = pcmk_exec_status_str(status); node = pcmk__output_create_xml_node(out, PCMK_XE_FAILURE, op_key_name, op_key, PCMK_XA_NODE, uname, PCMK_XA_EXITSTATUS, exitstatus, PCMK_XA_EXITREASON, exit_reason, PCMK_XA_EXITCODE, rc_s, PCMK_XA_CALL, call_id, PCMK_XA_STATUS, status_s, NULL); free(rc_s); pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &epoch); if (epoch > 0) { const char *queue_time = pcmk__xe_get(xml_op, PCMK_XA_QUEUE_TIME); const char *exec = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); const char *task = pcmk__xe_get(xml_op, PCMK_XA_OPERATION); guint interval_ms = 0; char *interval_ms_s = NULL; char *rc_change = pcmk__epoch2str(&epoch, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); pcmk__xe_get_guint(xml_op, PCMK_META_INTERVAL, &interval_ms); interval_ms_s = crm_strdup_printf("%u", interval_ms); pcmk__xe_set_props(node, PCMK_XA_LAST_RC_CHANGE, rc_change, PCMK_XA_QUEUED, queue_time, PCMK_XA_EXEC, exec, PCMK_XA_INTERVAL, interval_ms_s, PCMK_XA_TASK, task, NULL); free(interval_ms_s); free(rc_change); } g_free(exit_reason_esc); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action-list", "pcmk_scheduler_t *", "GList *", "GList *", "uint32_t", "bool") static int failed_action_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); xmlNode *xml_op = NULL; int rc = pcmk_rc_no_output; if (xmlChildElementCount(scheduler->priv->failed) == 0) { return rc; } for (xml_op = pcmk__xe_first_child(scheduler->priv->failed, NULL, NULL, NULL); xml_op != NULL; xml_op = pcmk__xe_next(xml_op, NULL)) { char *rsc = NULL; if (!pcmk__str_in_list(pcmk__xe_get(xml_op, PCMK_XA_UNAME), only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } if (pcmk_xe_mask_probe_failure(xml_op)) { continue; } if (!parse_op_key(pcmk__xe_history_key(xml_op), &rsc, NULL, NULL)) { continue; } if (!pcmk__str_in_list(rsc, only_rsc, pcmk__str_star_matches)) { free(rsc); continue; } free(rsc); PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Failed Resource Actions"); out->message(out, "failed-action", xml_op, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } static void status_node(pcmk_node_t *node, xmlNodePtr parent, uint32_t show_opts) { int health = pe__node_health(node); xmlNode *child = NULL; // Cluster membership if (node->details->online) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK_VALUE_ONLINE); pcmk__xe_set_content(child, " online"); } else { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK_VALUE_OFFLINE); pcmk__xe_set_content(child, " OFFLINE"); } // Standby mode if (pcmk_is_set(node->priv->flags, pcmk__node_fail_standby)) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK_VALUE_STANDBY); if (node->details->running_rsc == NULL) { pcmk__xe_set_content(child, " (in standby due to " PCMK_META_ON_FAIL ")"); } else { pcmk__xe_set_content(child, " (in standby due to " PCMK_META_ON_FAIL "," " with active resources)"); } } else if (pcmk_is_set(node->priv->flags, pcmk__node_standby)) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK_VALUE_STANDBY); if (node->details->running_rsc == NULL) { pcmk__xe_set_content(child, " (in standby)"); } else { pcmk__xe_set_content(child, " (in standby, with active resources)"); } } // Maintenance mode if (node->details->maintenance) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK__VALUE_MAINT); pcmk__xe_set_content(child, " (in maintenance mode)"); } // Node health if (health < 0) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK__VALUE_HEALTH_RED); pcmk__xe_set_content(child, " (health is RED)"); } else if (health == 0) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, PCMK__VALUE_HEALTH_YELLOW); pcmk__xe_set_content(child, " (health is YELLOW)"); } // Feature set if (pcmk_is_set(show_opts, pcmk_show_feature_set)) { const char *feature_set = get_node_feature_set(node); if (feature_set != NULL) { child = pcmk__html_create(parent, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, ", feature set %s", feature_set); } } } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_html(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (full) { xmlNode *item_node = NULL; xmlNode *child = NULL; if (pcmk_all_flags_set(show_opts, pcmk_show_brief | pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); out->begin_list(out, NULL, NULL, "%s:", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "Status:"); status_node(node, item_node, show_opts); if (rscs != NULL) { uint32_t new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); } pcmk__output_xml_pop_parent(out); out->end_list(out); } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc2 = NULL; int rc = pcmk_rc_no_output; out->begin_list(out, NULL, NULL, "%s:", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "Status:"); status_node(node, item_node, show_opts); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources"); show_opts |= pcmk_show_rsc_only; out->message(out, (const char *) rsc->priv->xml->name, show_opts, rsc, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); pcmk__output_xml_pop_parent(out); out->end_list(out); } else { item_node = pcmk__output_create_xml_node(out, "li", NULL); child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "%s:", node_name); status_node(node, item_node, show_opts); } } else { out->begin_list(out, NULL, NULL, "%s:", node_name); } free(node_name); return pcmk_rc_ok; } /*! * \internal * \brief Get a human-friendly textual description of a node's status * * \param[in] node Node to check * * \return String representation of node's status */ static const char * node_text_status(const pcmk_node_t *node) { if (node->details->unclean) { if (node->details->online) { return "UNCLEAN (online)"; } else if (node->details->pending) { return "UNCLEAN (pending)"; } else { return "UNCLEAN (offline)"; } } else if (node->details->pending) { return "pending"; } else if (pcmk_is_set(node->priv->flags, pcmk__node_fail_standby) && node->details->online) { return "standby (" PCMK_META_ON_FAIL ")"; } else if (pcmk_is_set(node->priv->flags, pcmk__node_standby)) { if (!node->details->online) { return "OFFLINE (standby)"; } else if (node->details->running_rsc == NULL) { return "standby"; } else { return "standby (with active resources)"; } } else if (node->details->maintenance) { if (node->details->online) { return "maintenance"; } else { return "OFFLINE (maintenance)"; } } else if (node->details->online) { return "online"; } return "OFFLINE"; } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_text(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); GString *str = g_string_sized_new(64); int health = pe__node_health(node); // Create a summary line with node type, name, and status if (pcmk__is_guest_or_bundle_node(node)) { g_string_append(str, "GuestNode"); } else if (pcmk__is_remote_node(node)) { g_string_append(str, "RemoteNode"); } else { g_string_append(str, "Node"); } pcmk__g_strcat(str, " ", node_name, ": ", node_text_status(node), NULL); if (health < 0) { g_string_append(str, " (health is RED)"); } else if (health == 0) { g_string_append(str, " (health is YELLOW)"); } if (pcmk_is_set(show_opts, pcmk_show_feature_set)) { const char *feature_set = get_node_feature_set(node); if (feature_set != NULL) { pcmk__g_strcat(str, ", feature set ", feature_set, NULL); } } /* If we're grouping by node, print its resources */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (pcmk_is_set(show_opts, pcmk_show_brief)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); if (rscs != NULL) { uint32_t new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "%s", str->str); out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); out->end_list(out); g_list_free(rscs); } } else { GList *gIter2 = NULL; out->begin_list(out, NULL, NULL, "%s", str->str); out->begin_list(out, NULL, NULL, "Resources"); for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter2->data; show_opts |= pcmk_show_rsc_only; out->message(out, (const char *) rsc->priv->xml->name, show_opts, rsc, only_node, only_rsc); } out->end_list(out); out->end_list(out); } } else { out->list_item(out, NULL, "%s", str->str); } g_string_free(str, TRUE); free(node_name); } else { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->begin_list(out, NULL, NULL, "Node: %s", node_name); free(node_name); } return pcmk_rc_ok; } /*! * \internal * \brief Convert an integer health value to a string representation * * \param[in] health Integer health value * * \retval \c PCMK_VALUE_RED if \p health is less than 0 * \retval \c PCMK_VALUE_YELLOW if \p health is equal to 0 * \retval \c PCMK_VALUE_GREEN if \p health is greater than 0 */ static const char * health_text(int health) { if (health < 0) { return PCMK_VALUE_RED; } else if (health == 0) { return PCMK_VALUE_YELLOW; } else { return PCMK_VALUE_GREEN; } } /*! * \internal * \brief Convert a node variant to a string representation * * \param[in] variant Node variant * * \retval \c PCMK_VALUE_MEMBER if \p node_type is \c pcmk__node_variant_cluster * \retval \c PCMK_VALUE_REMOTE if \p node_type is \c pcmk__node_variant_remote * \retval \c PCMK_VALUE_UNKNOWN otherwise */ static const char * node_variant_text(enum pcmk__node_variant variant) { switch (variant) { case pcmk__node_variant_cluster: return PCMK_VALUE_MEMBER; case pcmk__node_variant_remote: return PCMK_VALUE_REMOTE; default: return PCMK_VALUE_UNKNOWN; } } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { const char *online = pcmk__btoa(node->details->online); const char *standby = pcmk__flag_text(node->priv->flags, pcmk__node_standby); const char *standby_onfail = pcmk__flag_text(node->priv->flags, pcmk__node_fail_standby); const char *maintenance = pcmk__btoa(node->details->maintenance); const char *pending = pcmk__btoa(node->details->pending); const char *unclean = pcmk__btoa(node->details->unclean); const char *health = health_text(pe__node_health(node)); const char *feature_set = get_node_feature_set(node); const char *shutdown = pcmk__btoa(node->details->shutdown); const char *expected_up = pcmk__flag_text(node->priv->flags, pcmk__node_expected_up); const bool is_dc = pcmk__same_node(node, node->priv->scheduler->dc_node); int length = g_list_length(node->details->running_rsc); char *resources_running = pcmk__itoa(length); const char *node_type = node_variant_text(node->priv->variant); int rc = pcmk_rc_ok; rc = pe__name_and_nvpairs_xml(out, true, PCMK_XE_NODE, PCMK_XA_NAME, node->priv->name, PCMK_XA_ID, node->priv->id, PCMK_XA_ONLINE, online, PCMK_XA_STANDBY, standby, PCMK_XA_STANDBY_ONFAIL, standby_onfail, PCMK_XA_MAINTENANCE, maintenance, PCMK_XA_PENDING, pending, PCMK_XA_UNCLEAN, unclean, PCMK_XA_HEALTH, health, PCMK_XA_FEATURE_SET, feature_set, PCMK_XA_SHUTDOWN, shutdown, PCMK_XA_EXPECTED_UP, expected_up, PCMK_XA_IS_DC, pcmk__btoa(is_dc), PCMK_XA_RESOURCES_RUNNING, resources_running, PCMK_XA_TYPE, node_type, NULL); free(resources_running); pcmk__assert(rc == pcmk_rc_ok); if (pcmk__is_guest_or_bundle_node(node)) { xmlNodePtr xml_node = pcmk__output_xml_peek_parent(out); crm_xml_add(xml_node, PCMK_XA_ID_AS_RESOURCE, node->priv->remote->priv->launcher->id); } if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc = NULL; for (lpc = node->details->running_rsc; lpc != NULL; lpc = lpc->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) lpc->data; show_opts |= pcmk_show_rsc_only; out->message(out, (const char *) rsc->priv->xml->name, show_opts, rsc, only_node, only_rsc); } } out->end_list(out); } else { pcmk__output_xml_create_parent(out, PCMK_XE_NODE, PCMK_XA_NAME, node->priv->name, NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_text(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); if (add_extra) { int v; if (value == NULL) { v = 0; } else { pcmk__scan_min_int(value, &v, INT_MIN); } if (v <= 0) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is lost", name, value); } else if (v < expected_score) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is degraded (Expected=%d)", name, value, expected_score); } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_html(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); if (add_extra) { int v = 0; xmlNodePtr item_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNode *child = NULL; if (value != NULL) { pcmk__scan_min_int(value, &v, INT_MIN); } child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, NULL); pcmk__xe_set_content(child, "%s: %s", name, value); if (v <= 0) { child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "(connectivity is lost)"); } else if (v < expected_score) { child = pcmk__html_create(item_node, PCMK__XE_SPAN, NULL, PCMK__VALUE_BOLD); pcmk__xe_set_content(child, "(connectivity is degraded -- expected %d)", expected_score); } } else { out->list_item(out, NULL, "%s: %s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pcmk_scheduler_t *", "xmlNode *") static int node_and_op(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pcmk_resource_t *rsc = NULL; gchar *node_str = NULL; char *last_change_str = NULL; const char *op_rsc = pcmk__xe_get(xml_op, PCMK_XA_RESOURCE); int status; time_t last_change = 0; pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_OP_STATUS), &status, PCMK_EXEC_UNKNOWN); rsc = pe_find_resource(scheduler->priv->resources, op_rsc); if (rsc) { const pcmk_node_t *node = pcmk__current_node(rsc); const char *target_role = g_hash_table_lookup(rsc->priv->meta, PCMK_META_TARGET_ROLE); uint32_t show_opts = pcmk_show_rsc_only | pcmk_show_pending; if (node == NULL) { node = rsc->priv->pending_node; } node_str = pcmk__native_output_string(rsc, rsc_printable_id(rsc), node, show_opts, target_role, false); } else { node_str = crm_strdup_printf("Unknown resource %s", op_rsc); } if (pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &last_change) == pcmk_rc_ok) { const char *exec_time = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); last_change_str = crm_strdup_printf(", %s='%s', exec=%sms", PCMK_XA_LAST_RC_CHANGE, - pcmk__trim(ctime(&last_change)), + g_strchomp(ctime(&last_change)), exec_time); } out->list_item(out, NULL, "%s: %s (node=%s, call=%s, rc=%s%s): %s", node_str, pcmk__xe_history_key(xml_op), pcmk__xe_get(xml_op, PCMK_XA_UNAME), pcmk__xe_get(xml_op, PCMK__XA_CALL_ID), pcmk__xe_get(xml_op, PCMK__XA_RC_CODE), last_change_str ? last_change_str : "", pcmk_exec_status_str(status)); g_free(node_str); free(last_change_str); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pcmk_scheduler_t *", "xmlNode *") static int node_and_op_xml(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pcmk_resource_t *rsc = NULL; const char *uname = pcmk__xe_get(xml_op, PCMK_XA_UNAME); const char *call_id = pcmk__xe_get(xml_op, PCMK__XA_CALL_ID); const char *rc_s = pcmk__xe_get(xml_op, PCMK__XA_RC_CODE); const char *status_s = NULL; const char *op_rsc = pcmk__xe_get(xml_op, PCMK_XA_RESOURCE); int status; time_t last_change = 0; xmlNode *node = NULL; pcmk__scan_min_int(pcmk__xe_get(xml_op, PCMK__XA_OP_STATUS), &status, PCMK_EXEC_UNKNOWN); status_s = pcmk_exec_status_str(status); node = pcmk__output_create_xml_node(out, PCMK_XE_OPERATION, PCMK_XA_OP, pcmk__xe_history_key(xml_op), PCMK_XA_NODE, uname, PCMK_XA_CALL, call_id, PCMK_XA_RC, rc_s, PCMK_XA_STATUS, status_s, NULL); rsc = pe_find_resource(scheduler->priv->resources, op_rsc); if (rsc) { const char *class = pcmk__xe_get(rsc->priv->xml, PCMK_XA_CLASS); const char *provider = pcmk__xe_get(rsc->priv->xml, PCMK_XA_PROVIDER); const char *kind = pcmk__xe_get(rsc->priv->xml, PCMK_XA_TYPE); bool has_provider = pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider); char *agent_tuple = crm_strdup_printf("%s:%s:%s", class, (has_provider? provider : ""), kind); pcmk__xe_set_props(node, PCMK_XA_RSC, rsc_printable_id(rsc), PCMK_XA_AGENT, agent_tuple, NULL); free(agent_tuple); } if (pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &last_change) == pcmk_rc_ok) { - const char *last_rc_change = pcmk__trim(ctime(&last_change)); + const char *last_rc_change = g_strchomp(ctime(&last_change)); const char *exec_time = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); pcmk__xe_set_props(node, PCMK_XA_LAST_RC_CHANGE, last_rc_change, PCMK_XA_EXEC_TIME, exec_time, NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, PCMK_XE_ATTRIBUTE, PCMK_XA_NAME, name, PCMK_XA_VALUE, value, NULL); if (add_extra) { char *buf = pcmk__itoa(expected_score); crm_xml_add(node, PCMK_XA_EXPECTED, buf); free(buf); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute-list", "pcmk_scheduler_t *", "uint32_t", "bool", "GList *", "GList *") static int node_attribute_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); int rc = pcmk_rc_no_output; /* Display each node's attributes */ for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = gIter->data; GList *attr_list = NULL; GHashTableIter iter; gpointer key; if (!node || !node->details || !node->details->online) { continue; } // @TODO Maybe skip filtering for XML output g_hash_table_iter_init(&iter, node->priv->attrs); while (g_hash_table_iter_next (&iter, &key, NULL)) { attr_list = filter_attr_list(attr_list, key); } if (attr_list == NULL) { continue; } if (!pcmk__str_in_list(node->priv->name, only_node, pcmk__str_star_matches|pcmk__str_casei)) { g_list_free(attr_list); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Node Attributes"); out->message(out, "node", node, show_opts, false, only_node, only_rsc); for (GList *aIter = attr_list; aIter != NULL; aIter = aIter->next) { const char *name = aIter->data; const char *value = NULL; int expected_score = 0; bool add_extra = false; value = pcmk__node_attr(node, name, NULL, pcmk__rsc_node_current); add_extra = add_extra_info(node, node->details->running_rsc, scheduler, name, &expected_score); /* Print attribute name and value */ out->message(out, "node-attribute", name, value, add_extra, expected_score); } g_list_free(attr_list); out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-capacity", "const pcmk_node_t *", "const char *") static int node_capacity(pcmk__output_t *out, va_list args) { const pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *comment = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s capacity:", comment, pcmk__node_name(node)); g_hash_table_foreach(node->priv->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-capacity", "const pcmk_node_t *", "const char *") static int node_capacity_xml(pcmk__output_t *out, va_list args) { const pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *uname = node->priv->name; const char *comment = va_arg(args, const char *); xmlNodePtr xml_node = pcmk__output_create_xml_node(out, PCMK_XE_CAPACITY, PCMK_XA_NODE, uname, PCMK_XA_COMMENT, comment, NULL); g_hash_table_foreach(node->priv->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-history-list", "pcmk_scheduler_t *", "pcmk_node_t *", "xmlNode *", "GList *", "GList *", "uint32_t", "uint32_t") static int node_history_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); xmlNode *node_state = va_arg(args, xmlNode *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); xmlNode *lrm_rsc = NULL; xmlNode *rsc_entry = NULL; int rc = pcmk_rc_no_output; lrm_rsc = pcmk__xe_first_child(node_state, PCMK__XE_LRM, NULL, NULL); lrm_rsc = pcmk__xe_first_child(lrm_rsc, PCMK__XE_LRM_RESOURCES, NULL, NULL); /* Print history of each of the node's resources */ for (rsc_entry = pcmk__xe_first_child(lrm_rsc, PCMK__XE_LRM_RESOURCE, NULL, NULL); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry, PCMK__XE_LRM_RESOURCE)) { const char *rsc_id = pcmk__xe_get(rsc_entry, PCMK_XA_ID); pcmk_resource_t *rsc = NULL; const pcmk_resource_t *parent = NULL; if (rsc_id == NULL) { continue; // Malformed entry } rsc = pe_find_resource(scheduler->priv->resources, rsc_id); if (rsc == NULL) { continue; // Resource was removed from configuration } /* We can't use is_filtered here to filter group resources. For is_filtered, * we have to decide whether to check the parent or not. If we check the * parent, all elements of a group will always be printed because that's how * is_filtered works for groups. If we do not check the parent, sometimes * this will filter everything out. * * For other resource types, is_filtered is okay. */ parent = pe__const_top_resource(rsc, false); if (pcmk__is_group(parent)) { if (!pcmk__str_in_list(rsc_printable_id(rsc), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(parent), only_rsc, pcmk__str_star_matches)) { continue; } } else if (rsc->priv->fns->is_filtered(rsc, only_rsc, true)) { continue; } if (!pcmk_is_set(section_opts, pcmk_section_operations)) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pcmk__fc_default, NULL); if (failcount <= 0) { continue; } if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, false, only_node, only_rsc); } out->message(out, "resource-history", rsc, rsc_id, false, failcount, last_failure, false); } else { GList *op_list = get_operation_list(rsc_entry); pcmk_resource_t *rsc = NULL; if (op_list == NULL) { continue; } rsc = pe_find_resource(scheduler->priv->resources, pcmk__xe_get(rsc_entry, PCMK_XA_ID)); if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, false, only_node, only_rsc); } out->message(out, "resource-operation-list", scheduler, rsc, node, op_list, show_opts); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_html(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer G_GNUC_UNUSED = va_arg(args, int); int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; if (!pcmk__str_in_list(node->priv->name, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Node List"); out->message(out, "node", node, show_opts, true, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_text(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); /* space-separated lists of node names */ GString *online_nodes = NULL; GString *online_remote_nodes = NULL; GString *online_guest_nodes = NULL; GString *offline_nodes = NULL; GString *offline_remote_nodes = NULL; int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (!pcmk__str_in_list(node->priv->name, only_node, pcmk__str_star_matches|pcmk__str_casei)) { free(node_name); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Node List"); // Determine whether to display node individually or in a list if (node->details->unclean || node->details->pending || (pcmk_is_set(node->priv->flags, pcmk__node_fail_standby) && node->details->online) || pcmk_is_set(node->priv->flags, pcmk__node_standby) || node->details->maintenance || pcmk_is_set(show_opts, pcmk_show_rscs_by_node) || pcmk_is_set(show_opts, pcmk_show_feature_set) || (pe__node_health(node) <= 0)) { // Display node individually } else if (node->details->online) { // Display online node in a list if (pcmk__is_guest_or_bundle_node(node)) { pcmk__add_word(&online_guest_nodes, 1024, node_name); } else if (pcmk__is_remote_node(node)) { pcmk__add_word(&online_remote_nodes, 1024, node_name); } else { pcmk__add_word(&online_nodes, 1024, node_name); } free(node_name); continue; } else { // Display offline node in a list if (pcmk__is_remote_node(node)) { pcmk__add_word(&offline_remote_nodes, 1024, node_name); } else if (pcmk__is_guest_or_bundle_node(node)) { /* ignore offline guest nodes */ } else { pcmk__add_word(&offline_nodes, 1024, node_name); } free(node_name); continue; } /* If we get here, node is in bad state, or we're grouping by node */ out->message(out, "node", node, show_opts, true, only_node, only_rsc); free(node_name); } /* If we're not grouping by node, summarize nodes by status */ if (online_nodes != NULL) { out->list_item(out, "Online", "[ %s ]", (const char *) online_nodes->str); g_string_free(online_nodes, TRUE); } if (offline_nodes != NULL) { out->list_item(out, "OFFLINE", "[ %s ]", (const char *) offline_nodes->str); g_string_free(offline_nodes, TRUE); } if (online_remote_nodes) { out->list_item(out, "RemoteOnline", "[ %s ]", (const char *) online_remote_nodes->str); g_string_free(online_remote_nodes, TRUE); } if (offline_remote_nodes) { out->list_item(out, "RemoteOFFLINE", "[ %s ]", (const char *) offline_remote_nodes->str); g_string_free(offline_remote_nodes, TRUE); } if (online_guest_nodes != NULL) { out->list_item(out, "GuestOnline", "[ %s ]", (const char *) online_guest_nodes->str); g_string_free(online_guest_nodes, TRUE); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_xml(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer G_GNUC_UNUSED = va_arg(args, int); /* PCMK_XE_NODES acts as the list's element name for CLI tools that use * pcmk__output_enable_list_element. Otherwise PCMK_XE_NODES is the * value of the list's PCMK_XA_NAME attribute. */ out->begin_list(out, NULL, NULL, PCMK_XE_NODES); for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; if (!pcmk__str_in_list(node->priv->name, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } out->message(out, "node", node, show_opts, true, only_node, only_rsc); } out->end_list(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-summary", "pcmk_scheduler_t *", "GList *", "GList *", "uint32_t", "uint32_t", "bool") static int node_summary(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); xmlNode *node_state = NULL; xmlNode *cib_status = pcmk_find_cib_element(scheduler->input, PCMK_XE_STATUS); int rc = pcmk_rc_no_output; if (xmlChildElementCount(cib_status) == 0) { return rc; } for (node_state = pcmk__xe_first_child(cib_status, PCMK__XE_NODE_STATE, NULL, NULL); node_state != NULL; node_state = pcmk__xe_next(node_state, PCMK__XE_NODE_STATE)) { pcmk_node_t *node = pe_find_node_id(scheduler->nodes, pcmk__xe_id(node_state)); if (!node || !node->details || !node->details->online) { continue; } if (!pcmk__str_in_list(node->priv->name, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, pcmk_is_set(section_opts, pcmk_section_operations) ? "Operations" : "Migration Summary"); out->message(out, "node-history-list", scheduler, node, node_state, only_node, only_rsc, section_opts, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-weight", "const pcmk_resource_t *", "const char *", "const char *", "const char *") static int node_weight(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); const char *score = va_arg(args, const char *); if (rsc) { out->list_item(out, NULL, "%s: %s allocation score on %s: %s", prefix, rsc->id, uname, score); } else { out->list_item(out, NULL, "%s: %s = %s", prefix, uname, score); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-weight", "const pcmk_resource_t *", "const char *", "const char *", "const char *") static int node_weight_xml(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); const char *score = va_arg(args, const char *); xmlNodePtr node = pcmk__output_create_xml_node(out, PCMK_XE_NODE_WEIGHT, PCMK_XA_FUNCTION, prefix, PCMK_XA_NODE, uname, PCMK_XA_SCORE, score, NULL); if (rsc) { crm_xml_add(node, PCMK_XA_ID, rsc->id); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNode *", "const char *", "const char *", "int", "uint32_t") static int op_history_text(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); uint32_t show_opts = va_arg(args, uint32_t); char *buf = op_history_string(xml_op, task, interval_ms_s, rc, pcmk_is_set(show_opts, pcmk_show_timing)); out->list_item(out, NULL, "%s", buf); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNode *", "const char *", "const char *", "int", "uint32_t") static int op_history_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); uint32_t show_opts = va_arg(args, uint32_t); const char *call_id = pcmk__xe_get(xml_op, PCMK__XA_CALL_ID); char *rc_s = pcmk__itoa(rc); const char *rc_text = crm_exit_str(rc); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, PCMK_XE_OPERATION_HISTORY, PCMK_XA_CALL, call_id, PCMK_XA_TASK, task, PCMK_XA_RC, rc_s, PCMK_XA_RC_TEXT, rc_text, NULL); free(rc_s); if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *s = crm_strdup_printf("%sms", interval_ms_s); crm_xml_add(node, PCMK_XA_INTERVAL, s); free(s); } if (pcmk_is_set(show_opts, pcmk_show_timing)) { const char *value = NULL; time_t epoch = 0; pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &epoch); if (epoch > 0) { char *s = pcmk__epoch2str(&epoch, 0); crm_xml_add(node, PCMK_XA_LAST_RC_CHANGE, s); free(s); } value = pcmk__xe_get(xml_op, PCMK_XA_EXEC_TIME); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, PCMK_XA_EXEC_TIME, s); free(s); } value = pcmk__xe_get(xml_op, PCMK_XA_QUEUE_TIME); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, PCMK_XA_QUEUE_TIME, s); free(s); } } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int promotion_score(pcmk__output_t *out, va_list args) { pcmk_resource_t *child_rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *chosen = va_arg(args, pcmk_node_t *); const char *score = va_arg(args, const char *); if (chosen == NULL) { out->list_item(out, NULL, "%s promotion score (inactive): %s", child_rsc->id, score); } else { out->list_item(out, NULL, "%s promotion score on %s: %s", child_rsc->id, pcmk__node_name(chosen), score); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int promotion_score_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *child_rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *chosen = va_arg(args, pcmk_node_t *); const char *score = va_arg(args, const char *); xmlNodePtr node = pcmk__output_create_xml_node(out, PCMK_XE_PROMOTION_SCORE, PCMK_XA_ID, child_rsc->id, PCMK_XA_SCORE, score, NULL); if (chosen) { crm_xml_add(node, PCMK_XA_NODE, chosen->priv->name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-config", "const pcmk_resource_t *", "bool") static int resource_config(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); GString *xml_buf = g_string_sized_new(1024); bool raw = va_arg(args, int); formatted_xml_buf(rsc, xml_buf, raw); out->output_xml(out, PCMK_XE_XML, xml_buf->str); g_string_free(xml_buf, TRUE); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-config", "const pcmk_resource_t *", "bool") static int resource_config_text(pcmk__output_t *out, va_list args) { pcmk__formatted_printf(out, "Resource XML:\n"); return resource_config(out, args); } PCMK__OUTPUT_ARGS("resource-history", "pcmk_resource_t *", "const char *", "bool", "int", "time_t", "bool") static int resource_history_text(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); const char *rsc_id = va_arg(args, const char *); bool all = va_arg(args, int); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, time_t); bool as_header = va_arg(args, int); char *buf = resource_history_string(rsc, rsc_id, all, failcount, last_failure); if (as_header) { out->begin_list(out, NULL, NULL, "%s", buf); } else { out->list_item(out, NULL, "%s", buf); } free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-history", "pcmk_resource_t *", "const char *", "bool", "int", "time_t", "bool") static int resource_history_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); const char *rsc_id = va_arg(args, const char *); bool all = va_arg(args, int); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, time_t); bool as_header = va_arg(args, int); xmlNodePtr node = pcmk__output_xml_create_parent(out, PCMK_XE_RESOURCE_HISTORY, PCMK_XA_ID, rsc_id, NULL); if (rsc == NULL) { pcmk__xe_set_bool_attr(node, PCMK_XA_ORPHAN, true); } else if (all || failcount || last_failure > 0) { char *migration_s = pcmk__itoa(rsc->priv->ban_after_failures); pcmk__xe_set_props(node, PCMK_XA_ORPHAN, PCMK_VALUE_FALSE, PCMK_META_MIGRATION_THRESHOLD, migration_s, NULL); free(migration_s); if (failcount > 0) { char *s = pcmk__itoa(failcount); crm_xml_add(node, PCMK_XA_FAIL_COUNT, s); free(s); } if (last_failure > 0) { char *s = pcmk__epoch2str(&last_failure, 0); crm_xml_add(node, PCMK_XA_LAST_FAILURE, s); free(s); } } if (!as_header) { pcmk__output_xml_pop_parent(out); } return pcmk_rc_ok; } static void print_resource_header(pcmk__output_t *out, uint32_t show_opts) { if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { /* Active resources have already been printed by node */ out->begin_list(out, NULL, NULL, "Inactive Resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->begin_list(out, NULL, NULL, "Full List of Resources"); } else { out->begin_list(out, NULL, NULL, "Active Resources"); } } PCMK__OUTPUT_ARGS("resource-list", "pcmk_scheduler_t *", "uint32_t", "bool", "GList *", "GList *", "bool") static int resource_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); uint32_t show_opts = va_arg(args, uint32_t); bool print_summary = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); bool print_spacer = va_arg(args, int); GList *rsc_iter; int rc = pcmk_rc_no_output; bool printed_header = false; /* If we already showed active resources by node, and * we're not showing inactive resources, we have nothing to do */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node) && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { return rc; } /* If we haven't already printed resources grouped by node, * and brief output was requested, print resource summary */ if (pcmk_is_set(show_opts, pcmk_show_brief) && !pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(scheduler->priv->resources, only_rsc); PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; rc = pe__rscs_brief_output(out, rscs, show_opts); g_list_free(rscs); } /* For each resource, display it if appropriate */ for (rsc_iter = scheduler->priv->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) rsc_iter->data; int x; /* Complex resources may have some sub-resources active and some inactive */ bool is_active = rsc->priv->fns->active(rsc, true); bool partially_active = rsc->priv->fns->active(rsc, false); /* Skip inactive orphans (deleted but still in CIB) */ if (pcmk_is_set(rsc->flags, pcmk__rsc_removed) && !is_active) { continue; /* Skip active resources if we already displayed them by node */ } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (is_active) { continue; } /* Skip primitives already counted in a brief summary */ } else if (pcmk_is_set(show_opts, pcmk_show_brief) && pcmk__is_primitive(rsc)) { continue; /* Skip resources that aren't at least partially active, * unless we're displaying inactive resources */ } else if (!partially_active && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { continue; } else if (partially_active && !pe__rsc_running_on_any(rsc, only_node)) { continue; } if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } /* Print this resource */ x = out->message(out, (const char *) rsc->priv->xml->name, show_opts, rsc, only_node, only_rsc); if (x == pcmk_rc_ok) { rc = pcmk_rc_ok; } } if (print_summary && rc != pcmk_rc_ok) { if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } /* @FIXME It looks as if we can return pcmk_rc_no_output even after * writing output here. */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { out->list_item(out, NULL, "No inactive resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->list_item(out, NULL, "No resources"); } else { out->list_item(out, NULL, "No active resources"); } } if (printed_header) { out->end_list(out); } return rc; } PCMK__OUTPUT_ARGS("resource-operation-list", "pcmk_scheduler_t *", "pcmk_resource_t *", "pcmk_node_t *", "GList *", "uint32_t") static int resource_operation_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler G_GNUC_UNUSED = va_arg(args, pcmk_scheduler_t *); pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); GList *op_list = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); GList *gIter = NULL; int rc = pcmk_rc_no_output; /* Print each operation */ for (gIter = op_list; gIter != NULL; gIter = gIter->next) { xmlNode *xml_op = (xmlNode *) gIter->data; const char *task = pcmk__xe_get(xml_op, PCMK_XA_OPERATION); const char *interval_ms_s = pcmk__xe_get(xml_op, PCMK_META_INTERVAL); const char *op_rc = pcmk__xe_get(xml_op, PCMK__XA_RC_CODE); int op_rc_i; pcmk__scan_min_int(op_rc, &op_rc_i, 0); /* Display 0-interval monitors as "probe" */ if (pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei) && pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches | pcmk__str_casei)) { task = "probe"; } /* If this is the first printed operation, print heading for resource */ if (rc == pcmk_rc_no_output) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pcmk__fc_default, NULL); out->message(out, "resource-history", rsc, rsc_printable_id(rsc), true, failcount, last_failure, true); rc = pcmk_rc_ok; } /* Print the operation */ out->message(out, "op-history", xml_op, task, interval_ms_s, op_rc_i, show_opts); } /* Free the list we created (no need to free the individual items) */ g_list_free(op_list); PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("resource-util", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int resource_util(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *fn = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s utilization on %s:", fn, rsc->id, pcmk__node_name(node)); g_hash_table_foreach(rsc->priv->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-util", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int resource_util_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *uname = node->priv->name; const char *fn = va_arg(args, const char *); xmlNodePtr xml_node = NULL; xml_node = pcmk__output_create_xml_node(out, PCMK_XE_UTILIZATION, PCMK_XA_RESOURCE, rsc->id, PCMK_XA_NODE, uname, PCMK_XA_FUNCTION, fn, NULL); g_hash_table_foreach(rsc->priv->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } static inline const char * ticket_status(pcmk__ticket_t *ticket) { if (pcmk_is_set(ticket->flags, pcmk__ticket_granted)) { return PCMK_VALUE_GRANTED; } return PCMK_VALUE_REVOKED; } static inline const char * ticket_standby_text(pcmk__ticket_t *ticket) { return pcmk_is_set(ticket->flags, pcmk__ticket_standby)? " [standby]" : ""; } PCMK__OUTPUT_ARGS("ticket", "pcmk__ticket_t *", "bool", "bool") static int ticket_default(pcmk__output_t *out, va_list args) { pcmk__ticket_t *ticket = va_arg(args, pcmk__ticket_t *); bool raw = va_arg(args, int); bool details = va_arg(args, int); GString *detail_str = NULL; if (raw) { out->list_item(out, ticket->id, "%s", ticket->id); return pcmk_rc_ok; } if (details && g_hash_table_size(ticket->state) > 0) { GHashTableIter iter; const char *name = NULL; const char *value = NULL; bool already_added = false; detail_str = g_string_sized_new(100); pcmk__g_strcat(detail_str, "\t(", NULL); g_hash_table_iter_init(&iter, ticket->state); while (g_hash_table_iter_next(&iter, (void **) &name, (void **) &value)) { if (already_added) { g_string_append_printf(detail_str, ", %s=", name); } else { g_string_append_printf(detail_str, "%s=", name); already_added = true; } if (pcmk__str_any_of(name, PCMK_XA_LAST_GRANTED, "expires", NULL)) { char *epoch_str = NULL; long long time_ll; (void) pcmk__scan_ll(value, &time_ll, 0); epoch_str = pcmk__epoch2str((const time_t *) &time_ll, 0); pcmk__g_strcat(detail_str, epoch_str, NULL); free(epoch_str); } else { pcmk__g_strcat(detail_str, value, NULL); } } pcmk__g_strcat(detail_str, ")", NULL); } if (ticket->last_granted > -1) { /* Prior to the introduction of the details & raw arguments to this * function, last-granted would always be added in this block. We need * to preserve that behavior. At the same time, we also need to preserve * the existing behavior from crm_ticket, which would include last-granted * as part of the (...) detail string. * * Luckily we can check detail_str - if it's NULL, either there were no * details, or we are preserving the previous behavior of this function. * If it's not NULL, we are either preserving the previous behavior of * crm_ticket or we were given details=true as an argument. */ if (detail_str == NULL) { char *epoch_str = pcmk__epoch2str(&(ticket->last_granted), 0); out->list_item(out, NULL, "%s\t%s%s last-granted=\"%s\"", ticket->id, ticket_status(ticket), ticket_standby_text(ticket), pcmk__s(epoch_str, "")); free(epoch_str); } else { out->list_item(out, NULL, "%s\t%s%s %s", ticket->id, ticket_status(ticket), ticket_standby_text(ticket), detail_str->str); } } else { out->list_item(out, NULL, "%s\t%s%s%s", ticket->id, ticket_status(ticket), ticket_standby_text(ticket), detail_str != NULL ? detail_str->str : ""); } if (detail_str != NULL) { g_string_free(detail_str, TRUE); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pcmk__ticket_t *", "bool", "bool") static int ticket_xml(pcmk__output_t *out, va_list args) { pcmk__ticket_t *ticket = va_arg(args, pcmk__ticket_t *); bool raw G_GNUC_UNUSED = va_arg(args, int); bool details G_GNUC_UNUSED = va_arg(args, int); const char *standby = pcmk__flag_text(ticket->flags, pcmk__ticket_standby); xmlNodePtr node = NULL; GHashTableIter iter; const char *name = NULL; const char *value = NULL; node = pcmk__output_create_xml_node(out, PCMK_XE_TICKET, PCMK_XA_ID, ticket->id, PCMK_XA_STATUS, ticket_status(ticket), PCMK_XA_STANDBY, standby, NULL); if (ticket->last_granted > -1) { char *buf = pcmk__epoch2str(&ticket->last_granted, 0); crm_xml_add(node, PCMK_XA_LAST_GRANTED, buf); free(buf); } g_hash_table_iter_init(&iter, ticket->state); while (g_hash_table_iter_next(&iter, (void **) &name, (void **) &value)) { /* PCMK_XA_LAST_GRANTED and "expires" are already added by the check * for ticket->last_granted above. */ if (pcmk__str_any_of(name, PCMK_XA_LAST_GRANTED, PCMK_XA_EXPIRES, NULL)) { continue; } crm_xml_add(node, name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket-list", "GHashTable *", "bool", "bool", "bool") static int ticket_list(pcmk__output_t *out, va_list args) { GHashTable *tickets = va_arg(args, GHashTable *); bool print_spacer = va_arg(args, int); bool raw = va_arg(args, int); bool details = va_arg(args, int); GHashTableIter iter; gpointer value; if (g_hash_table_size(tickets) == 0) { return pcmk_rc_no_output; } PCMK__OUTPUT_SPACER_IF(out, print_spacer); /* Print section heading */ out->begin_list(out, NULL, NULL, "Tickets"); /* Print each ticket */ g_hash_table_iter_init(&iter, tickets); while (g_hash_table_iter_next(&iter, NULL, &value)) { pcmk__ticket_t *ticket = (pcmk__ticket_t *) value; out->message(out, "ticket", ticket, raw, details); } /* Close section */ out->end_list(out); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "ban", "default", ban_text }, { "ban", "html", ban_html }, { "ban", "xml", ban_xml }, { "ban-list", "default", ban_list }, { "bundle", "default", pe__bundle_text }, { "bundle", "xml", pe__bundle_xml }, { "bundle", "html", pe__bundle_html }, { "clone", "default", pe__clone_default }, { "clone", "xml", pe__clone_xml }, { "cluster-counts", "default", cluster_counts_text }, { "cluster-counts", "html", cluster_counts_html }, { "cluster-counts", "xml", cluster_counts_xml }, { "cluster-dc", "default", cluster_dc_text }, { "cluster-dc", "html", cluster_dc_html }, { "cluster-dc", "xml", cluster_dc_xml }, { "cluster-options", "default", cluster_options_text }, { "cluster-options", "html", cluster_options_html }, { "cluster-options", "log", cluster_options_log }, { "cluster-options", "xml", cluster_options_xml }, { "cluster-summary", "default", cluster_summary }, { "cluster-summary", "html", cluster_summary_html }, { "cluster-stack", "default", cluster_stack_text }, { "cluster-stack", "html", cluster_stack_html }, { "cluster-stack", "xml", cluster_stack_xml }, { "cluster-times", "default", cluster_times_text }, { "cluster-times", "html", cluster_times_html }, { "cluster-times", "xml", cluster_times_xml }, { "failed-action", "default", failed_action_default }, { "failed-action", "xml", failed_action_xml }, { "failed-action-list", "default", failed_action_list }, { "group", "default", pe__group_default}, { "group", "xml", pe__group_xml }, { "maint-mode", "text", cluster_maint_mode_text }, { "node", "default", node_text }, { "node", "html", node_html }, { "node", "xml", node_xml }, { "node-and-op", "default", node_and_op }, { "node-and-op", "xml", node_and_op_xml }, { "node-capacity", "default", node_capacity }, { "node-capacity", "xml", node_capacity_xml }, { "node-history-list", "default", node_history_list }, { "node-list", "default", node_list_text }, { "node-list", "html", node_list_html }, { "node-list", "xml", node_list_xml }, { "node-weight", "default", node_weight }, { "node-weight", "xml", node_weight_xml }, { "node-attribute", "default", node_attribute_text }, { "node-attribute", "html", node_attribute_html }, { "node-attribute", "xml", node_attribute_xml }, { "node-attribute-list", "default", node_attribute_list }, { "node-summary", "default", node_summary }, { "op-history", "default", op_history_text }, { "op-history", "xml", op_history_xml }, { "primitive", "default", pe__resource_text }, { "primitive", "xml", pe__resource_xml }, { "primitive", "html", pe__resource_html }, { "promotion-score", "default", promotion_score }, { "promotion-score", "xml", promotion_score_xml }, { "resource-config", "default", resource_config }, { "resource-config", "text", resource_config_text }, { "resource-history", "default", resource_history_text }, { "resource-history", "xml", resource_history_xml }, { "resource-list", "default", resource_list }, { "resource-operation-list", "default", resource_operation_list }, { "resource-util", "default", resource_util }, { "resource-util", "xml", resource_util_xml }, { "ticket", "default", ticket_default }, { "ticket", "xml", ticket_xml }, { "ticket-list", "default", ticket_list }, { NULL, NULL, NULL } }; void pe__register_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c index 26a8c97ec4..5b0a7360e2 100644 --- a/lib/pengine/unpack.c +++ b/lib/pengine/unpack.c @@ -1,5119 +1,5120 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include // xmlNode #include // xmlXPathObject, etc. #include #include #include #include #include #include #include CRM_TRACE_INIT_DATA(pe_status); // A (parsed) resource action history entry struct action_history { pcmk_resource_t *rsc; // Resource that history is for pcmk_node_t *node; // Node that history is for xmlNode *xml; // History entry XML // Parsed from entry XML const char *id; // XML ID of history entry const char *key; // Operation key of action const char *task; // Action name const char *exit_reason; // Exit reason given for result guint interval_ms; // Action interval int call_id; // Call ID of action int expected_exit_status; // Expected exit status of action int exit_status; // Actual exit status of action int execution_status; // Execution status of action }; /* This uses pcmk__set_flags_as()/pcmk__clear_flags_as() directly rather than * use pcmk__set_scheduler_flags()/pcmk__clear_scheduler_flags() so that the * flag is stringified more readably in log messages. */ #define set_config_flag(scheduler, option, flag) do { \ GHashTable *config_hash = (scheduler)->priv->options; \ const char *scf_value = pcmk__cluster_option(config_hash, (option)); \ \ if (scf_value != NULL) { \ if (crm_is_true(scf_value)) { \ (scheduler)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Scheduler", \ crm_system_name, (scheduler)->flags, \ (flag), #flag); \ } else { \ (scheduler)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Scheduler", \ crm_system_name, (scheduler)->flags, \ (flag), #flag); \ } \ } \ } while(0) static void unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum pcmk__on_fail *failed); static void determine_remote_online_status(pcmk_scheduler_t *scheduler, pcmk_node_t *this_node); static void add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node, bool overwrite, pcmk_scheduler_t *scheduler); static void determine_online_status(const xmlNode *node_state, pcmk_node_t *this_node, pcmk_scheduler_t *scheduler); static void unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml, pcmk_scheduler_t *scheduler); /*! * \internal * \brief Check whether a node is a dangling guest node * * \param[in] node Node to check * * \return true if \p node had a Pacemaker Remote connection resource with a * launcher that was removed from the CIB, otherwise false. */ static bool is_dangling_guest_node(pcmk_node_t *node) { return pcmk__is_pacemaker_remote_node(node) && (node->priv->remote != NULL) && (node->priv->remote->priv->launcher == NULL) && pcmk_is_set(node->priv->remote->flags, pcmk__rsc_removed_launched); } /*! * \brief Schedule a fence action for a node * * \param[in,out] scheduler Scheduler data * \param[in,out] node Node to fence * \param[in] reason Text description of why fencing is needed * \param[in] priority_delay Whether to consider * \c PCMK_OPT_PRIORITY_FENCING_DELAY */ void pe_fence_node(pcmk_scheduler_t *scheduler, pcmk_node_t *node, const char *reason, bool priority_delay) { CRM_CHECK(node, return); if (pcmk__is_guest_or_bundle_node(node)) { // Fence a guest or bundle node by marking its launcher as failed pcmk_resource_t *rsc = node->priv->remote->priv->launcher; if (!pcmk_is_set(rsc->flags, pcmk__rsc_failed)) { if (!pcmk_is_set(rsc->flags, pcmk__rsc_managed)) { crm_notice("Not fencing guest node %s " "(otherwise would because %s): " "its guest resource %s is unmanaged", pcmk__node_name(node), reason, rsc->id); } else { pcmk__sched_warn(scheduler, "Guest node %s will be fenced " "(by recovering its guest resource %s): %s", pcmk__node_name(node), rsc->id, reason); /* We don't mark the node as unclean because that would prevent the * node from running resources. We want to allow it to run resources * in this transition if the recovery succeeds. */ pcmk__set_node_flags(node, pcmk__node_remote_reset); pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); } } } else if (is_dangling_guest_node(node)) { crm_info("Cleaning up dangling connection for guest node %s: " "fencing was already done because %s, " "and guest resource no longer exists", pcmk__node_name(node), reason); pcmk__set_rsc_flags(node->priv->remote, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); } else if (pcmk__is_remote_node(node)) { pcmk_resource_t *rsc = node->priv->remote; if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk__rsc_managed)) { crm_notice("Not fencing remote node %s " "(otherwise would because %s): connection is unmanaged", pcmk__node_name(node), reason); } else if (!pcmk_is_set(node->priv->flags, pcmk__node_remote_reset)) { pcmk__set_node_flags(node, pcmk__node_remote_reset); pcmk__sched_warn(scheduler, "Remote node %s %s: %s", pcmk__node_name(node), pe_can_fence(scheduler, node)? "will be fenced" : "is unclean", reason); } node->details->unclean = TRUE; // No need to apply PCMK_OPT_PRIORITY_FENCING_DELAY for remote nodes pe_fence_op(node, NULL, TRUE, reason, FALSE, scheduler); } else if (node->details->unclean) { crm_trace("Cluster node %s %s because %s", pcmk__node_name(node), pe_can_fence(scheduler, node)? "would also be fenced" : "also is unclean", reason); } else { pcmk__sched_warn(scheduler, "Cluster node %s %s: %s", pcmk__node_name(node), pe_can_fence(scheduler, node)? "will be fenced" : "is unclean", reason); node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, priority_delay, scheduler); } } // @TODO xpaths can't handle templates, rules, or id-refs // nvpair with provides or requires set to unfencing #define XPATH_UNFENCING_NVPAIR PCMK_XE_NVPAIR \ "[(@" PCMK_XA_NAME "='" PCMK_STONITH_PROVIDES "'" \ "or @" PCMK_XA_NAME "='" PCMK_META_REQUIRES "') " \ "and @" PCMK_XA_VALUE "='" PCMK_VALUE_UNFENCING "']" // unfencing in rsc_defaults or any resource #define XPATH_ENABLE_UNFENCING \ "/" PCMK_XE_CIB "/" PCMK_XE_CONFIGURATION "/" PCMK_XE_RESOURCES \ "//" PCMK_XE_META_ATTRIBUTES "/" XPATH_UNFENCING_NVPAIR \ "|/" PCMK_XE_CIB "/" PCMK_XE_CONFIGURATION "/" PCMK_XE_RSC_DEFAULTS \ "/" PCMK_XE_META_ATTRIBUTES "/" XPATH_UNFENCING_NVPAIR static void set_if_xpath(uint64_t flag, const char *xpath, pcmk_scheduler_t *scheduler) { xmlXPathObject *result = NULL; if (!pcmk_is_set(scheduler->flags, flag)) { result = pcmk__xpath_search(scheduler->input->doc, xpath); if (pcmk__xpath_num_results(result) > 0) { pcmk__set_scheduler_flags(scheduler, flag); } xmlXPathFreeObject(result); } } gboolean unpack_config(xmlNode *config, pcmk_scheduler_t *scheduler) { const char *value = NULL; GHashTable *config_hash = pcmk__strkey_table(free, free); const pcmk_rule_input_t rule_input = { .now = scheduler->priv->now, }; scheduler->priv->options = config_hash; pe__unpack_dataset_nvpairs(config, PCMK_XE_CLUSTER_PROPERTY_SET, &rule_input, config_hash, PCMK_VALUE_CIB_BOOTSTRAP_OPTIONS, scheduler); pcmk__validate_cluster_options(config_hash); set_config_flag(scheduler, PCMK_OPT_ENABLE_STARTUP_PROBES, pcmk__sched_probe_resources); if (!pcmk_is_set(scheduler->flags, pcmk__sched_probe_resources)) { crm_info("Startup probes: disabled (dangerous)"); } value = pcmk__cluster_option(config_hash, PCMK_OPT_HAVE_WATCHDOG); if (value && crm_is_true(value)) { crm_info("Watchdog-based self-fencing will be performed via SBD if " "fencing is required and " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " is nonzero"); pcmk__set_scheduler_flags(scheduler, pcmk__sched_have_fencing); } /* Set certain flags via xpath here, so they can be used before the relevant * configuration sections are unpacked. */ set_if_xpath(pcmk__sched_enable_unfencing, XPATH_ENABLE_UNFENCING, scheduler); value = pcmk__cluster_option(config_hash, PCMK_OPT_STONITH_TIMEOUT); pcmk_parse_interval_spec(value, &(scheduler->priv->fence_timeout_ms)); crm_debug("Default fencing action timeout: %s", pcmk__readable_interval(scheduler->priv->fence_timeout_ms)); set_config_flag(scheduler, PCMK_OPT_STONITH_ENABLED, pcmk__sched_fencing_enabled); if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { crm_debug("STONITH of failed nodes is enabled"); } else { crm_debug("STONITH of failed nodes is disabled"); } scheduler->priv->fence_action = pcmk__cluster_option(config_hash, PCMK_OPT_STONITH_ACTION); crm_trace("STONITH will %s nodes", scheduler->priv->fence_action); set_config_flag(scheduler, PCMK_OPT_CONCURRENT_FENCING, pcmk__sched_concurrent_fencing); if (pcmk_is_set(scheduler->flags, pcmk__sched_concurrent_fencing)) { crm_debug("Concurrent fencing is enabled"); } else { crm_debug("Concurrent fencing is disabled"); } value = pcmk__cluster_option(config_hash, PCMK_OPT_PRIORITY_FENCING_DELAY); if (value) { pcmk_parse_interval_spec(value, &(scheduler->priv->priority_fencing_ms)); crm_trace("Priority fencing delay is %s", pcmk__readable_interval(scheduler->priv->priority_fencing_ms)); } set_config_flag(scheduler, PCMK_OPT_STOP_ALL_RESOURCES, pcmk__sched_stop_all); crm_debug("Stop all active resources: %s", pcmk__flag_text(scheduler->flags, pcmk__sched_stop_all)); set_config_flag(scheduler, PCMK_OPT_SYMMETRIC_CLUSTER, pcmk__sched_symmetric_cluster); if (pcmk_is_set(scheduler->flags, pcmk__sched_symmetric_cluster)) { crm_debug("Cluster is symmetric" " - resources can run anywhere by default"); } value = pcmk__cluster_option(config_hash, PCMK_OPT_NO_QUORUM_POLICY); if (pcmk__str_eq(value, PCMK_VALUE_IGNORE, pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_ignore; } else if (pcmk__str_eq(value, PCMK_VALUE_FREEZE, pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_freeze; } else if (pcmk__str_eq(value, PCMK_VALUE_DEMOTE, pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_demote; } else if (pcmk__strcase_any_of(value, PCMK_VALUE_FENCE, PCMK_VALUE_FENCE_LEGACY, NULL)) { if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { int do_panic = 0; pcmk__xe_get_int(scheduler->input, PCMK_XA_NO_QUORUM_PANIC, &do_panic); if (do_panic || pcmk_is_set(scheduler->flags, pcmk__sched_quorate)) { scheduler->no_quorum_policy = pcmk_no_quorum_fence; } else { crm_notice("Resetting " PCMK_OPT_NO_QUORUM_POLICY " to 'stop': cluster has never had quorum"); scheduler->no_quorum_policy = pcmk_no_quorum_stop; } } else { pcmk__config_err("Resetting " PCMK_OPT_NO_QUORUM_POLICY " to 'stop' because fencing is disabled"); scheduler->no_quorum_policy = pcmk_no_quorum_stop; } } else { scheduler->no_quorum_policy = pcmk_no_quorum_stop; } switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: crm_debug("On loss of quorum: " "Freeze resources that require quorum"); break; case pcmk_no_quorum_stop: crm_debug("On loss of quorum: " "Stop resources that require quorum"); break; case pcmk_no_quorum_demote: crm_debug("On loss of quorum: " "Demote promotable resources and stop other resources"); break; case pcmk_no_quorum_fence: crm_notice("On loss of quorum: Fence all remaining nodes"); break; case pcmk_no_quorum_ignore: crm_notice("On loss of quorum: Ignore"); break; } set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_RESOURCES, pcmk__sched_stop_removed_resources); if (pcmk_is_set(scheduler->flags, pcmk__sched_stop_removed_resources)) { crm_trace("Orphan resources are stopped"); } else { crm_trace("Orphan resources are ignored"); } set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_ACTIONS, pcmk__sched_cancel_removed_actions); if (pcmk_is_set(scheduler->flags, pcmk__sched_cancel_removed_actions)) { crm_trace("Orphan resource actions are stopped"); } else { crm_trace("Orphan resource actions are ignored"); } set_config_flag(scheduler, PCMK_OPT_MAINTENANCE_MODE, pcmk__sched_in_maintenance); crm_trace("Maintenance mode: %s", pcmk__flag_text(scheduler->flags, pcmk__sched_in_maintenance)); set_config_flag(scheduler, PCMK_OPT_START_FAILURE_IS_FATAL, pcmk__sched_start_failure_fatal); if (pcmk_is_set(scheduler->flags, pcmk__sched_start_failure_fatal)) { crm_trace("Start failures are always fatal"); } else { crm_trace("Start failures are handled by failcount"); } if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { set_config_flag(scheduler, PCMK_OPT_STARTUP_FENCING, pcmk__sched_startup_fencing); } if (pcmk_is_set(scheduler->flags, pcmk__sched_startup_fencing)) { crm_trace("Unseen nodes will be fenced"); } else { pcmk__warn_once(pcmk__wo_blind, "Blind faith: not fencing unseen nodes"); } pe__unpack_node_health_scores(scheduler); scheduler->priv->placement_strategy = pcmk__cluster_option(config_hash, PCMK_OPT_PLACEMENT_STRATEGY); crm_trace("Placement strategy: %s", scheduler->priv->placement_strategy); set_config_flag(scheduler, PCMK_OPT_SHUTDOWN_LOCK, pcmk__sched_shutdown_lock); if (pcmk_is_set(scheduler->flags, pcmk__sched_shutdown_lock)) { value = pcmk__cluster_option(config_hash, PCMK_OPT_SHUTDOWN_LOCK_LIMIT); pcmk_parse_interval_spec(value, &(scheduler->priv->shutdown_lock_ms)); crm_trace("Resources will be locked to nodes that were cleanly " "shut down (locks expire after %s)", pcmk__readable_interval(scheduler->priv->shutdown_lock_ms)); } else { crm_trace("Resources will not be locked to nodes that were cleanly " "shut down"); } value = pcmk__cluster_option(config_hash, PCMK_OPT_NODE_PENDING_TIMEOUT); pcmk_parse_interval_spec(value, &(scheduler->priv->node_pending_ms)); if (scheduler->priv->node_pending_ms == 0U) { crm_trace("Do not fence pending nodes"); } else { crm_trace("Fence pending nodes after %s", pcmk__readable_interval(scheduler->priv->node_pending_ms)); } set_config_flag(scheduler, PCMK_OPT_FENCE_REMOTE_WITHOUT_QUORUM, pcmk__sched_fence_remote_no_quorum); if (pcmk_is_set(scheduler->flags, pcmk__sched_fence_remote_no_quorum)) { crm_trace("Pacemaker Remote nodes may be fenced without quorum"); } else { crm_trace("Pacemaker Remote nodes require quorum to be fenced"); } return TRUE; } /*! * \internal * \brief Create a new node object in scheduler data * * \param[in] id ID of new node * \param[in] uname Name of new node * \param[in] type Type of new node * \param[in] score Score of new node * \param[in,out] scheduler Scheduler data * * \return Newly created node object * \note The returned object is part of the scheduler data and should not be * freed separately. */ pcmk_node_t * pe_create_node(const char *id, const char *uname, const char *type, int score, pcmk_scheduler_t *scheduler) { enum pcmk__node_variant variant = pcmk__node_variant_cluster; pcmk_node_t *new_node = NULL; if (pcmk_find_node(scheduler, uname) != NULL) { pcmk__config_warn("More than one node entry has name '%s'", uname); } if (pcmk__str_eq(type, PCMK_VALUE_MEMBER, pcmk__str_null_matches|pcmk__str_casei)) { variant = pcmk__node_variant_cluster; } else if (pcmk__str_eq(type, PCMK_VALUE_REMOTE, pcmk__str_casei)) { variant = pcmk__node_variant_remote; } else { pcmk__config_err("Ignoring node %s with unrecognized type '%s'", pcmk__s(uname, "without name"), type); return NULL; } new_node = calloc(1, sizeof(pcmk_node_t)); if (new_node == NULL) { pcmk__sched_err(scheduler, "Could not allocate memory for node %s", uname); return NULL; } new_node->assign = calloc(1, sizeof(struct pcmk__node_assignment)); new_node->details = calloc(1, sizeof(struct pcmk__node_details)); new_node->priv = calloc(1, sizeof(pcmk__node_private_t)); if ((new_node->assign == NULL) || (new_node->details == NULL) || (new_node->priv == NULL)) { free(new_node->assign); free(new_node->details); free(new_node->priv); free(new_node); pcmk__sched_err(scheduler, "Could not allocate memory for node %s", uname); return NULL; } crm_trace("Creating node for entry %s/%s", uname, id); new_node->assign->score = score; new_node->priv->id = id; new_node->priv->name = uname; new_node->priv->flags = pcmk__node_probes_allowed; new_node->details->online = FALSE; new_node->details->shutdown = FALSE; new_node->details->running_rsc = NULL; new_node->priv->scheduler = scheduler; new_node->priv->variant = variant; new_node->priv->attrs = pcmk__strkey_table(free, free); new_node->priv->utilization = pcmk__strkey_table(free, free); new_node->priv->digest_cache = pcmk__strkey_table(free, pe__free_digests); if (pcmk__is_pacemaker_remote_node(new_node)) { pcmk__insert_dup(new_node->priv->attrs, CRM_ATTR_KIND, "remote"); pcmk__set_scheduler_flags(scheduler, pcmk__sched_have_remote_nodes); } else { pcmk__insert_dup(new_node->priv->attrs, CRM_ATTR_KIND, "cluster"); } scheduler->nodes = g_list_insert_sorted(scheduler->nodes, new_node, pe__cmp_node_name); return new_node; } static const char * expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pcmk_scheduler_t *data) { xmlNode *attr_set = NULL; xmlNode *attr = NULL; const char *container_id = pcmk__xe_id(xml_obj); const char *remote_name = NULL; const char *remote_server = NULL; const char *remote_port = NULL; const char *connect_timeout = "60s"; const char *remote_allow_migrate=NULL; const char *is_managed = NULL; // @TODO This doesn't handle rules or id-ref for (attr_set = pcmk__xe_first_child(xml_obj, PCMK_XE_META_ATTRIBUTES, NULL, NULL); attr_set != NULL; attr_set = pcmk__xe_next(attr_set, PCMK_XE_META_ATTRIBUTES)) { for (attr = pcmk__xe_first_child(attr_set, NULL, NULL, NULL); attr != NULL; attr = pcmk__xe_next(attr, NULL)) { const char *value = pcmk__xe_get(attr, PCMK_XA_VALUE); const char *name = pcmk__xe_get(attr, PCMK_XA_NAME); if (name == NULL) { // Sanity continue; } if (strcmp(name, PCMK_META_REMOTE_NODE) == 0) { remote_name = value; } else if (strcmp(name, PCMK_META_REMOTE_ADDR) == 0) { remote_server = value; } else if (strcmp(name, PCMK_META_REMOTE_PORT) == 0) { remote_port = value; } else if (strcmp(name, PCMK_META_REMOTE_CONNECT_TIMEOUT) == 0) { connect_timeout = value; } else if (strcmp(name, PCMK_META_REMOTE_ALLOW_MIGRATE) == 0) { remote_allow_migrate = value; } else if (strcmp(name, PCMK_META_IS_MANAGED) == 0) { is_managed = value; } } } if (remote_name == NULL) { return NULL; } if (pe_find_resource(data->priv->resources, remote_name) != NULL) { return NULL; } pe_create_remote_xml(parent, remote_name, container_id, remote_allow_migrate, is_managed, connect_timeout, remote_server, remote_port); return remote_name; } static void handle_startup_fencing(pcmk_scheduler_t *scheduler, pcmk_node_t *new_node) { if ((new_node->priv->variant == pcmk__node_variant_remote) && (new_node->priv->remote == NULL)) { /* Ignore fencing for remote nodes that don't have a connection resource * associated with them. This happens when remote node entries get left * in the nodes section after the connection resource is removed. */ return; } if (pcmk_is_set(scheduler->flags, pcmk__sched_startup_fencing)) { // All nodes are unclean until we've seen their status entry new_node->details->unclean = TRUE; } else { // Blind faith ... new_node->details->unclean = FALSE; } } gboolean unpack_nodes(xmlNode *xml_nodes, pcmk_scheduler_t *scheduler) { xmlNode *xml_obj = NULL; pcmk_node_t *new_node = NULL; const char *id = NULL; const char *uname = NULL; const char *type = NULL; for (xml_obj = pcmk__xe_first_child(xml_nodes, PCMK_XE_NODE, NULL, NULL); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj, PCMK_XE_NODE)) { int score = 0; int rc = pcmk__xe_get_score(xml_obj, PCMK_XA_SCORE, &score, 0); new_node = NULL; id = pcmk__xe_get(xml_obj, PCMK_XA_ID); uname = pcmk__xe_get(xml_obj, PCMK_XA_UNAME); type = pcmk__xe_get(xml_obj, PCMK_XA_TYPE); crm_trace("Processing node %s/%s", uname, id); if (id == NULL) { pcmk__config_err("Ignoring <" PCMK_XE_NODE "> entry in configuration without id"); continue; } if (rc != pcmk_rc_ok) { // Not possible with schema validation enabled pcmk__config_warn("Using 0 as score for node %s " "because '%s' is not a valid score: %s", pcmk__s(uname, "without name"), pcmk__xe_get(xml_obj, PCMK_XA_SCORE), pcmk_rc_str(rc)); } new_node = pe_create_node(id, uname, type, score, scheduler); if (new_node == NULL) { return FALSE; } handle_startup_fencing(scheduler, new_node); add_node_attrs(xml_obj, new_node, FALSE, scheduler); crm_trace("Done with node %s", pcmk__xe_get(xml_obj, PCMK_XA_UNAME)); } return TRUE; } static void unpack_launcher(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler) { const char *launcher_id = NULL; if (rsc->priv->children != NULL) { g_list_foreach(rsc->priv->children, (GFunc) unpack_launcher, scheduler); return; } launcher_id = g_hash_table_lookup(rsc->priv->meta, PCMK__META_CONTAINER); if ((launcher_id != NULL) && !pcmk__str_eq(launcher_id, rsc->id, pcmk__str_none)) { pcmk_resource_t *launcher = pe_find_resource(scheduler->priv->resources, launcher_id); if (launcher != NULL) { rsc->priv->launcher = launcher; launcher->priv->launched = g_list_append(launcher->priv->launched, rsc); pcmk__rsc_trace(rsc, "Resource %s's launcher is %s", rsc->id, launcher_id); } else { pcmk__config_err("Resource %s: Unknown " PCMK__META_CONTAINER " %s", rsc->id, launcher_id); } } } gboolean unpack_remote_nodes(xmlNode *xml_resources, pcmk_scheduler_t *scheduler) { xmlNode *xml_obj = NULL; /* Create remote nodes and guest nodes from the resource configuration * before unpacking resources. */ for (xml_obj = pcmk__xe_first_child(xml_resources, NULL, NULL, NULL); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj, NULL)) { const char *new_node_id = NULL; /* Check for remote nodes, which are defined by ocf:pacemaker:remote * primitives. */ if (xml_contains_remote_node(xml_obj)) { new_node_id = pcmk__xe_id(xml_obj); /* The pcmk_find_node() check ensures we don't iterate over an * expanded node that has already been added to the node list */ if (new_node_id && (pcmk_find_node(scheduler, new_node_id) == NULL)) { crm_trace("Found remote node %s defined by resource %s", new_node_id, pcmk__xe_id(xml_obj)); pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE, 0, scheduler); } continue; } /* Check for guest nodes, which are defined by special meta-attributes * of a primitive of any type (for example, VirtualDomain or Xen). */ if (pcmk__xe_is(xml_obj, PCMK_XE_PRIMITIVE)) { /* This will add an ocf:pacemaker:remote primitive to the * configuration for the guest node's connection, to be unpacked * later. */ new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, scheduler); if (new_node_id && (pcmk_find_node(scheduler, new_node_id) == NULL)) { crm_trace("Found guest node %s in resource %s", new_node_id, pcmk__xe_id(xml_obj)); pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE, 0, scheduler); } continue; } /* Check for guest nodes inside a group. Clones are currently not * supported as guest nodes. */ if (pcmk__xe_is(xml_obj, PCMK_XE_GROUP)) { xmlNode *xml_obj2 = NULL; for (xml_obj2 = pcmk__xe_first_child(xml_obj, NULL, NULL, NULL); xml_obj2 != NULL; xml_obj2 = pcmk__xe_next(xml_obj2, NULL)) { new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, scheduler); if (new_node_id && (pcmk_find_node(scheduler, new_node_id) == NULL)) { crm_trace("Found guest node %s in resource %s inside group %s", new_node_id, pcmk__xe_id(xml_obj2), pcmk__xe_id(xml_obj)); pe_create_node(new_node_id, new_node_id, PCMK_VALUE_REMOTE, 0, scheduler); } } } } return TRUE; } /* Call this after all the nodes and resources have been * unpacked, but before the status section is read. * * A remote node's online status is reflected by the state * of the remote node's connection resource. We need to link * the remote node to this connection resource so we can have * easy access to the connection resource during the scheduler calculations. */ static void link_rsc2remotenode(pcmk_scheduler_t *scheduler, pcmk_resource_t *new_rsc) { pcmk_node_t *remote_node = NULL; if (!pcmk_is_set(new_rsc->flags, pcmk__rsc_is_remote_connection)) { return; } if (pcmk_is_set(scheduler->flags, pcmk__sched_location_only)) { /* remote_nodes and remote_resources are not linked in quick location calculations */ return; } remote_node = pcmk_find_node(scheduler, new_rsc->id); CRM_CHECK(remote_node != NULL, return); pcmk__rsc_trace(new_rsc, "Linking remote connection resource %s to %s", new_rsc->id, pcmk__node_name(remote_node)); remote_node->priv->remote = new_rsc; if (new_rsc->priv->launcher == NULL) { /* Handle start-up fencing for remote nodes (as opposed to guest nodes) * the same as is done for cluster nodes. */ handle_startup_fencing(scheduler, remote_node); } else { /* pe_create_node() marks the new node as "remote" or "cluster"; now * that we know the node is a guest node, update it correctly. */ pcmk__insert_dup(remote_node->priv->attrs, CRM_ATTR_KIND, "container"); } } /*! * \internal * \brief Parse configuration XML for resource information * * \param[in] xml_resources Top of resource configuration XML * \param[in,out] scheduler Scheduler data * * \return TRUE * * \note unpack_remote_nodes() MUST be called before this, so that the nodes can * be used when pe__unpack_resource() calls resource_location() */ gboolean unpack_resources(const xmlNode *xml_resources, pcmk_scheduler_t *scheduler) { xmlNode *xml_obj = NULL; GList *gIter = NULL; scheduler->priv->templates = pcmk__strkey_table(free, pcmk__free_idref); for (xml_obj = pcmk__xe_first_child(xml_resources, NULL, NULL, NULL); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj, NULL)) { pcmk_resource_t *new_rsc = NULL; const char *id = pcmk__xe_id(xml_obj); if (pcmk__str_empty(id)) { pcmk__config_err("Ignoring <%s> resource without ID", xml_obj->name); continue; } if (pcmk__xe_is(xml_obj, PCMK_XE_TEMPLATE)) { if (g_hash_table_lookup_extended(scheduler->priv->templates, id, NULL, NULL) == FALSE) { /* Record the template's ID for the knowledge of its existence anyway. */ pcmk__insert_dup(scheduler->priv->templates, id, NULL); } continue; } crm_trace("Unpacking <%s " PCMK_XA_ID "='%s'>", xml_obj->name, id); if (pe__unpack_resource(xml_obj, &new_rsc, NULL, scheduler) == pcmk_rc_ok) { scheduler->priv->resources = g_list_append(scheduler->priv->resources, new_rsc); pcmk__rsc_trace(new_rsc, "Added resource %s", new_rsc->id); } else { pcmk__config_err("Ignoring <%s> resource '%s' " "because configuration is invalid", xml_obj->name, id); } } for (gIter = scheduler->priv->resources; gIter != NULL; gIter = gIter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter->data; unpack_launcher(rsc, scheduler); link_rsc2remotenode(scheduler, rsc); } scheduler->priv->resources = g_list_sort(scheduler->priv->resources, pe__cmp_rsc_priority); if (pcmk_is_set(scheduler->flags, pcmk__sched_location_only)) { /* Ignore */ } else if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled) && !pcmk_is_set(scheduler->flags, pcmk__sched_have_fencing)) { pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined"); pcmk__config_err("Either configure some or disable STONITH with the " PCMK_OPT_STONITH_ENABLED " option"); pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity"); } return TRUE; } /*! * \internal * \brief Validate the levels in a fencing topology * * \param[in] xml \c PCMK_XE_FENCING_TOPOLOGY element */ void pcmk__validate_fencing_topology(const xmlNode *xml) { if (xml == NULL) { return; } CRM_CHECK(pcmk__xe_is(xml, PCMK_XE_FENCING_TOPOLOGY), return); for (const xmlNode *level = pcmk__xe_first_child(xml, PCMK_XE_FENCING_LEVEL, NULL, NULL); level != NULL; level = pcmk__xe_next(level, PCMK_XE_FENCING_LEVEL)) { const char *id = pcmk__xe_id(level); int index = 0; if (pcmk__str_empty(id)) { pcmk__config_err("Ignoring fencing level without ID"); continue; } if (pcmk__xe_get_int(level, PCMK_XA_INDEX, &index) != pcmk_rc_ok) { pcmk__config_err("Ignoring fencing level %s with invalid index", id); continue; } if ((index < ST__LEVEL_MIN) || (index > ST__LEVEL_MAX)) { pcmk__config_err("Ignoring fencing level %s with out-of-range " "index %d", id, index); } } } gboolean unpack_tags(xmlNode *xml_tags, pcmk_scheduler_t *scheduler) { xmlNode *xml_tag = NULL; scheduler->priv->tags = pcmk__strkey_table(free, pcmk__free_idref); for (xml_tag = pcmk__xe_first_child(xml_tags, PCMK_XE_TAG, NULL, NULL); xml_tag != NULL; xml_tag = pcmk__xe_next(xml_tag, PCMK_XE_TAG)) { xmlNode *xml_obj_ref = NULL; const char *tag_id = pcmk__xe_id(xml_tag); if (tag_id == NULL) { pcmk__config_err("Ignoring <%s> without " PCMK_XA_ID, (const char *) xml_tag->name); continue; } for (xml_obj_ref = pcmk__xe_first_child(xml_tag, PCMK_XE_OBJ_REF, NULL, NULL); xml_obj_ref != NULL; xml_obj_ref = pcmk__xe_next(xml_obj_ref, PCMK_XE_OBJ_REF)) { const char *obj_ref = pcmk__xe_id(xml_obj_ref); if (obj_ref == NULL) { pcmk__config_err("Ignoring <%s> for tag '%s' without " PCMK_XA_ID, xml_obj_ref->name, tag_id); continue; } pcmk__add_idref(scheduler->priv->tags, tag_id, obj_ref); } } return TRUE; } /*! * \internal * \brief Unpack a ticket state entry * * \param[in] xml_ticket XML ticket state to unpack * \param[in,out] userdata Scheduler data * * \return pcmk_rc_ok (to always continue unpacking further entries) */ static int unpack_ticket_state(xmlNode *xml_ticket, void *userdata) { pcmk_scheduler_t *scheduler = userdata; const char *ticket_id = NULL; const char *granted = NULL; const char *last_granted = NULL; const char *standby = NULL; xmlAttrPtr xIter = NULL; pcmk__ticket_t *ticket = NULL; ticket_id = pcmk__xe_id(xml_ticket); if (pcmk__str_empty(ticket_id)) { return pcmk_rc_ok; } crm_trace("Processing ticket state for %s", ticket_id); ticket = g_hash_table_lookup(scheduler->priv->ticket_constraints, ticket_id); if (ticket == NULL) { ticket = ticket_new(ticket_id, scheduler); if (ticket == NULL) { return pcmk_rc_ok; } } for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = pcmk__xml_attr_value(xIter); if (pcmk__str_eq(prop_name, PCMK_XA_ID, pcmk__str_none)) { continue; } pcmk__insert_dup(ticket->state, prop_name, prop_value); } granted = g_hash_table_lookup(ticket->state, PCMK__XA_GRANTED); if (granted && crm_is_true(granted)) { pcmk__set_ticket_flags(ticket, pcmk__ticket_granted); crm_info("We have ticket '%s'", ticket->id); } else { pcmk__clear_ticket_flags(ticket, pcmk__ticket_granted); crm_info("We do not have ticket '%s'", ticket->id); } last_granted = g_hash_table_lookup(ticket->state, PCMK_XA_LAST_GRANTED); if (last_granted) { long long last_granted_ll = 0LL; int rc = pcmk__scan_ll(last_granted, &last_granted_ll, 0LL); if (rc != pcmk_rc_ok) { crm_warn("Using %lld instead of invalid " PCMK_XA_LAST_GRANTED " value '%s' in state for ticket %s: %s", last_granted_ll, last_granted, ticket->id, pcmk_rc_str(rc)); } ticket->last_granted = (time_t) last_granted_ll; } standby = g_hash_table_lookup(ticket->state, PCMK_XA_STANDBY); if (standby && crm_is_true(standby)) { pcmk__set_ticket_flags(ticket, pcmk__ticket_standby); if (pcmk_is_set(ticket->flags, pcmk__ticket_granted)) { crm_info("Granted ticket '%s' is in standby-mode", ticket->id); } } else { pcmk__clear_ticket_flags(ticket, pcmk__ticket_standby); } crm_trace("Done with ticket state for %s", ticket_id); return pcmk_rc_ok; } static void unpack_handle_remote_attrs(pcmk_node_t *this_node, const xmlNode *state, pcmk_scheduler_t *scheduler) { const char *discovery = NULL; const xmlNode *attrs = NULL; pcmk_resource_t *rsc = NULL; int maint = 0; if (!pcmk__xe_is(state, PCMK__XE_NODE_STATE)) { return; } if ((this_node == NULL) || !pcmk__is_pacemaker_remote_node(this_node)) { return; } crm_trace("Processing Pacemaker Remote node %s", pcmk__node_name(this_node)); pcmk__scan_min_int(pcmk__xe_get(state, PCMK__XA_NODE_IN_MAINTENANCE), &maint, 0); if (maint) { pcmk__set_node_flags(this_node, pcmk__node_remote_maint); } else { pcmk__clear_node_flags(this_node, pcmk__node_remote_maint); } rsc = this_node->priv->remote; if (!pcmk_is_set(this_node->priv->flags, pcmk__node_remote_reset)) { this_node->details->unclean = FALSE; pcmk__set_node_flags(this_node, pcmk__node_seen); } attrs = pcmk__xe_first_child(state, PCMK__XE_TRANSIENT_ATTRIBUTES, NULL, NULL); add_node_attrs(attrs, this_node, TRUE, scheduler); if (pe__shutdown_requested(this_node)) { crm_info("%s is shutting down", pcmk__node_name(this_node)); this_node->details->shutdown = TRUE; } if (crm_is_true(pcmk__node_attr(this_node, PCMK_NODE_ATTR_STANDBY, NULL, pcmk__rsc_node_current))) { crm_info("%s is in standby mode", pcmk__node_name(this_node)); pcmk__set_node_flags(this_node, pcmk__node_standby); } if (crm_is_true(pcmk__node_attr(this_node, PCMK_NODE_ATTR_MAINTENANCE, NULL, pcmk__rsc_node_current)) || ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk__rsc_managed))) { crm_info("%s is in maintenance mode", pcmk__node_name(this_node)); this_node->details->maintenance = TRUE; } discovery = pcmk__node_attr(this_node, PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED, NULL, pcmk__rsc_node_current); if ((discovery != NULL) && !crm_is_true(discovery)) { pcmk__warn_once(pcmk__wo_rdisc_enabled, "Support for the " PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED " node attribute is deprecated and will be removed" " (and behave as 'true') in a future release."); if (pcmk__is_remote_node(this_node) && !pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { pcmk__config_warn("Ignoring " PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED " attribute on Pacemaker Remote node %s" " because fencing is disabled", pcmk__node_name(this_node)); } else { /* This is either a remote node with fencing enabled, or a guest * node. We don't care whether fencing is enabled when fencing guest * nodes, because they are "fenced" by recovering their containing * resource. */ crm_info("%s has resource discovery disabled", pcmk__node_name(this_node)); pcmk__clear_node_flags(this_node, pcmk__node_probes_allowed); } } } /*! * \internal * \brief Unpack a cluster node's transient attributes * * \param[in] state CIB node state XML * \param[in,out] node Cluster node whose attributes are being unpacked * \param[in,out] scheduler Scheduler data */ static void unpack_transient_attributes(const xmlNode *state, pcmk_node_t *node, pcmk_scheduler_t *scheduler) { const char *discovery = NULL; const xmlNode *attrs = pcmk__xe_first_child(state, PCMK__XE_TRANSIENT_ATTRIBUTES, NULL, NULL); add_node_attrs(attrs, node, TRUE, scheduler); if (crm_is_true(pcmk__node_attr(node, PCMK_NODE_ATTR_STANDBY, NULL, pcmk__rsc_node_current))) { crm_info("%s is in standby mode", pcmk__node_name(node)); pcmk__set_node_flags(node, pcmk__node_standby); } if (crm_is_true(pcmk__node_attr(node, PCMK_NODE_ATTR_MAINTENANCE, NULL, pcmk__rsc_node_current))) { crm_info("%s is in maintenance mode", pcmk__node_name(node)); node->details->maintenance = TRUE; } discovery = pcmk__node_attr(node, PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED, NULL, pcmk__rsc_node_current); if ((discovery != NULL) && !crm_is_true(discovery)) { pcmk__config_warn("Ignoring " PCMK__NODE_ATTR_RESOURCE_DISCOVERY_ENABLED " attribute for %s because disabling resource" " discovery is not allowed for cluster nodes", pcmk__node_name(node)); } } /*! * \internal * \brief Unpack a node state entry (first pass) * * Unpack one node state entry from status. This unpacks information from the * \C PCMK__XE_NODE_STATE element itself and node attributes inside it, but not * the resource history inside it. Multiple passes through the status are needed * to fully unpack everything. * * \param[in] state CIB node state XML * \param[in,out] scheduler Scheduler data */ static void unpack_node_state(const xmlNode *state, pcmk_scheduler_t *scheduler) { const char *id = NULL; const char *uname = NULL; pcmk_node_t *this_node = NULL; id = pcmk__xe_get(state, PCMK_XA_ID); if (id == NULL) { pcmk__config_err("Ignoring invalid " PCMK__XE_NODE_STATE " entry without " PCMK_XA_ID); crm_log_xml_info(state, "missing-id"); return; } uname = pcmk__xe_get(state, PCMK_XA_UNAME); if (uname == NULL) { /* If a joining peer makes the cluster acquire the quorum from Corosync * but has not joined the controller CPG membership yet, it's possible * that the created PCMK__XE_NODE_STATE entry doesn't have a * PCMK_XA_UNAME yet. Recognize the node as pending and wait for it to * join CPG. */ crm_trace("Handling " PCMK__XE_NODE_STATE " entry with id=\"%s\" " "without " PCMK_XA_UNAME, id); } this_node = pe_find_node_any(scheduler->nodes, id, uname); if (this_node == NULL) { crm_notice("Ignoring recorded state for removed node with name %s and " PCMK_XA_ID " %s", pcmk__s(uname, "unknown"), id); return; } if (pcmk__is_pacemaker_remote_node(this_node)) { int remote_fenced = 0; /* We can't determine the online status of Pacemaker Remote nodes until * after all resource history has been unpacked. In this first pass, we * do need to mark whether the node has been fenced, as this plays a * role during unpacking cluster node resource state. */ pcmk__scan_min_int(pcmk__xe_get(state, PCMK__XA_NODE_FENCED), &remote_fenced, 0); if (remote_fenced) { pcmk__set_node_flags(this_node, pcmk__node_remote_fenced); } else { pcmk__clear_node_flags(this_node, pcmk__node_remote_fenced); } return; } unpack_transient_attributes(state, this_node, scheduler); /* Provisionally mark this cluster node as clean. We have at least seen it * in the current cluster's lifetime. */ this_node->details->unclean = FALSE; pcmk__set_node_flags(this_node, pcmk__node_seen); crm_trace("Determining online status of cluster node %s (id %s)", pcmk__node_name(this_node), id); determine_online_status(state, this_node, scheduler); if (!pcmk_is_set(scheduler->flags, pcmk__sched_quorate) && this_node->details->online && (scheduler->no_quorum_policy == pcmk_no_quorum_fence)) { /* Everything else should flow from this automatically * (at least until the scheduler becomes able to migrate off * healthy resources) */ pe_fence_node(scheduler, this_node, "cluster does not have quorum", FALSE); } } /*! * \internal * \brief Unpack nodes' resource history as much as possible * * Unpack as many nodes' resource history as possible in one pass through the * status. We need to process Pacemaker Remote nodes' connections/containers * before unpacking their history; the connection/container history will be * in another node's history, so it might take multiple passes to unpack * everything. * * \param[in] status CIB XML status section * \param[in] fence If true, treat any not-yet-unpacked nodes as unseen * \param[in,out] scheduler Scheduler data * * \return Standard Pacemaker return code (specifically pcmk_rc_ok if done, * or EAGAIN if more unpacking remains to be done) */ static int unpack_node_history(const xmlNode *status, bool fence, pcmk_scheduler_t *scheduler) { int rc = pcmk_rc_ok; // Loop through all PCMK__XE_NODE_STATE entries in CIB status for (const xmlNode *state = pcmk__xe_first_child(status, PCMK__XE_NODE_STATE, NULL, NULL); state != NULL; state = pcmk__xe_next(state, PCMK__XE_NODE_STATE)) { const char *id = pcmk__xe_id(state); const char *uname = pcmk__xe_get(state, PCMK_XA_UNAME); pcmk_node_t *this_node = NULL; if ((id == NULL) || (uname == NULL)) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history from malformed " PCMK__XE_NODE_STATE " without id and/or uname"); continue; } this_node = pe_find_node_any(scheduler->nodes, id, uname); if (this_node == NULL) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history for node %s because " "no longer in configuration", id); continue; } if (pcmk_is_set(this_node->priv->flags, pcmk__node_unpacked)) { crm_trace("Not unpacking resource history for node %s because " "already unpacked", id); continue; } if (fence) { // We're processing all remaining nodes } else if (pcmk__is_guest_or_bundle_node(this_node)) { /* We can unpack a guest node's history only after we've unpacked * other resource history to the point that we know that the node's * connection and containing resource are both up. */ const pcmk_resource_t *remote = this_node->priv->remote; const pcmk_resource_t *launcher = remote->priv->launcher; if ((remote->priv->orig_role != pcmk_role_started) || (launcher->priv->orig_role != pcmk_role_started)) { crm_trace("Not unpacking resource history for guest node %s " "because launcher and connection are not known to " "be up", id); continue; } } else if (pcmk__is_remote_node(this_node)) { /* We can unpack a remote node's history only after we've unpacked * other resource history to the point that we know that the node's * connection is up, with the exception of when shutdown locks are * in use. */ pcmk_resource_t *rsc = this_node->priv->remote; if ((rsc == NULL) || (!pcmk_is_set(scheduler->flags, pcmk__sched_shutdown_lock) && (rsc->priv->orig_role != pcmk_role_started))) { crm_trace("Not unpacking resource history for remote node %s " "because connection is not known to be up", id); continue; } /* If fencing and shutdown locks are disabled and we're not processing * unseen nodes, then we don't want to unpack offline nodes until online * nodes have been unpacked. This allows us to number active clone * instances first. */ } else if (!pcmk_any_flags_set(scheduler->flags, pcmk__sched_fencing_enabled |pcmk__sched_shutdown_lock) && !this_node->details->online) { crm_trace("Not unpacking resource history for offline " "cluster node %s", id); continue; } if (pcmk__is_pacemaker_remote_node(this_node)) { determine_remote_online_status(scheduler, this_node); unpack_handle_remote_attrs(this_node, state, scheduler); } crm_trace("Unpacking resource history for %snode %s", (fence? "unseen " : ""), id); pcmk__set_node_flags(this_node, pcmk__node_unpacked); unpack_node_lrm(this_node, state, scheduler); rc = EAGAIN; // Other node histories might depend on this one } return rc; } /* remove nodes that are down, stopping */ /* create positive rsc_to_node constraints between resources and the nodes they are running on */ /* anything else? */ gboolean unpack_status(xmlNode *status, pcmk_scheduler_t *scheduler) { xmlNode *state = NULL; crm_trace("Beginning unpack"); if (scheduler->priv->ticket_constraints == NULL) { scheduler->priv->ticket_constraints = pcmk__strkey_table(free, destroy_ticket); } for (state = pcmk__xe_first_child(status, NULL, NULL, NULL); state != NULL; state = pcmk__xe_next(state, NULL)) { if (pcmk__xe_is(state, PCMK_XE_TICKETS)) { pcmk__xe_foreach_child(state, PCMK__XE_TICKET_STATE, unpack_ticket_state, scheduler); } else if (pcmk__xe_is(state, PCMK__XE_NODE_STATE)) { unpack_node_state(state, scheduler); } } while (unpack_node_history(status, FALSE, scheduler) == EAGAIN) { crm_trace("Another pass through node resource histories is needed"); } // Now catch any nodes we didn't see unpack_node_history(status, pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled), scheduler); /* Now that we know where resources are, we can schedule stops of containers * with failed bundle connections */ if (scheduler->priv->stop_needed != NULL) { for (GList *item = scheduler->priv->stop_needed; item != NULL; item = item->next) { pcmk_resource_t *container = item->data; pcmk_node_t *node = pcmk__current_node(container); if (node) { stop_action(container, node, FALSE); } } g_list_free(scheduler->priv->stop_needed); scheduler->priv->stop_needed = NULL; } /* Now that we know status of all Pacemaker Remote connections and nodes, * we can stop connections for node shutdowns, and check the online status * of remote/guest nodes that didn't have any node history to unpack. */ for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *this_node = gIter->data; if (!pcmk__is_pacemaker_remote_node(this_node)) { continue; } if (this_node->details->shutdown && (this_node->priv->remote != NULL)) { pe__set_next_role(this_node->priv->remote, pcmk_role_stopped, "remote shutdown"); } if (!pcmk_is_set(this_node->priv->flags, pcmk__node_unpacked)) { determine_remote_online_status(scheduler, this_node); } } return TRUE; } /*! * \internal * \brief Unpack node's time when it became a member at the cluster layer * * \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry * \param[in,out] scheduler Scheduler data * * \return Epoch time when node became a cluster member * (or scheduler effective time for legacy entries) if a member, * 0 if not a member, or -1 if no valid information available */ static long long unpack_node_member(const xmlNode *node_state, pcmk_scheduler_t *scheduler) { const char *member_time = pcmk__xe_get(node_state, PCMK__XA_IN_CCM); int member = 0; if (member_time == NULL) { return -1LL; } else if (crm_str_to_boolean(member_time, &member) == 1) { /* If in_ccm=0, we'll return 0 here. If in_ccm=1, either the entry was * recorded as a boolean for a DC < 2.1.7, or the node is pending * shutdown and has left the CPG, in which case it was set to 1 to avoid * fencing for PCMK_OPT_NODE_PENDING_TIMEOUT. * * We return the effective time for in_ccm=1 because what's important to * avoid fencing is that effective time minus this value is less than * the pending node timeout. */ return member? (long long) pcmk__scheduler_epoch_time(scheduler) : 0LL; } else { long long when_member = 0LL; if ((pcmk__scan_ll(member_time, &when_member, 0LL) != pcmk_rc_ok) || (when_member < 0LL)) { crm_warn("Unrecognized value '%s' for " PCMK__XA_IN_CCM " in " PCMK__XE_NODE_STATE " entry", member_time); return -1LL; } return when_member; } } /*! * \internal * \brief Unpack node's time when it became online in process group * * \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry * * \return Epoch time when node became online in process group (or 0 if not * online, or 1 for legacy online entries) */ static long long unpack_node_online(const xmlNode *node_state) { const char *peer_time = pcmk__xe_get(node_state, PCMK_XA_CRMD); // @COMPAT Entries recorded for DCs < 2.1.7 have "online" or "offline" if (pcmk__str_eq(peer_time, PCMK_VALUE_OFFLINE, pcmk__str_casei|pcmk__str_null_matches)) { return 0LL; } else if (pcmk__str_eq(peer_time, PCMK_VALUE_ONLINE, pcmk__str_casei)) { return 1LL; } else { long long when_online = 0LL; if ((pcmk__scan_ll(peer_time, &when_online, 0LL) != pcmk_rc_ok) || (when_online < 0)) { crm_warn("Unrecognized value '%s' for " PCMK_XA_CRMD " in " PCMK__XE_NODE_STATE " entry, assuming offline", peer_time); return 0LL; } return when_online; } } /*! * \internal * \brief Unpack node attribute for user-requested fencing * * \param[in] node Node to check * \param[in] node_state Node's \c PCMK__XE_NODE_STATE entry in CIB status * * \return \c true if fencing has been requested for \p node, otherwise \c false */ static bool unpack_node_terminate(const pcmk_node_t *node, const xmlNode *node_state) { long long value = 0LL; int value_i = 0; int rc = pcmk_rc_ok; const char *value_s = pcmk__node_attr(node, PCMK_NODE_ATTR_TERMINATE, NULL, pcmk__rsc_node_current); // Value may be boolean or an epoch time if (crm_str_to_boolean(value_s, &value_i) == 1) { return (value_i != 0); } rc = pcmk__scan_ll(value_s, &value, 0LL); if (rc == pcmk_rc_ok) { return (value > 0); } crm_warn("Ignoring unrecognized value '%s' for " PCMK_NODE_ATTR_TERMINATE "node attribute for %s: %s", value_s, pcmk__node_name(node), pcmk_rc_str(rc)); return false; } static gboolean determine_online_status_no_fencing(pcmk_scheduler_t *scheduler, const xmlNode *node_state, pcmk_node_t *this_node) { gboolean online = FALSE; const char *join = pcmk__xe_get(node_state, PCMK__XA_JOIN); const char *exp_state = pcmk__xe_get(node_state, PCMK_XA_EXPECTED); long long when_member = unpack_node_member(node_state, scheduler); long long when_online = unpack_node_online(node_state); if (when_member <= 0) { crm_trace("Node %s is %sdown", pcmk__node_name(this_node), ((when_member < 0)? "presumed " : "")); } else if (when_online > 0) { if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { online = TRUE; } else { crm_debug("Node %s is not ready to run resources: %s", pcmk__node_name(this_node), join); } } else if (!pcmk_is_set(this_node->priv->flags, pcmk__node_expected_up)) { crm_trace("Node %s controller is down: " "member@%lld online@%lld join=%s expected=%s", pcmk__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, "")); } else { /* mark it unclean */ pe_fence_node(scheduler, this_node, "peer is unexpectedly down", FALSE); crm_info("Node %s member@%lld online@%lld join=%s expected=%s", pcmk__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, "")); } return online; } /*! * \internal * \brief Check whether a node has taken too long to join controller group * * \param[in,out] scheduler Scheduler data * \param[in] node Node to check * \param[in] when_member Epoch time when node became a cluster member * \param[in] when_online Epoch time when node joined controller group * * \return true if node has been pending (on the way up) longer than * \c PCMK_OPT_NODE_PENDING_TIMEOUT, otherwise false * \note This will also update the cluster's recheck time if appropriate. */ static inline bool pending_too_long(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, long long when_member, long long when_online) { if ((scheduler->priv->node_pending_ms > 0U) && (when_member > 0) && (when_online <= 0)) { // There is a timeout on pending nodes, and node is pending time_t timeout = when_member + pcmk__timeout_ms2s(scheduler->priv->node_pending_ms); if (pcmk__scheduler_epoch_time(node->priv->scheduler) >= timeout) { return true; // Node has timed out } // Node is pending, but still has time pcmk__update_recheck_time(timeout, scheduler, "pending node timeout"); } return false; } static bool determine_online_status_fencing(pcmk_scheduler_t *scheduler, const xmlNode *node_state, pcmk_node_t *this_node) { bool termination_requested = unpack_node_terminate(this_node, node_state); const char *join = pcmk__xe_get(node_state, PCMK__XA_JOIN); const char *exp_state = pcmk__xe_get(node_state, PCMK_XA_EXPECTED); long long when_member = unpack_node_member(node_state, scheduler); long long when_online = unpack_node_online(node_state); /* - PCMK__XA_JOIN ::= member|down|pending|banned - PCMK_XA_EXPECTED ::= member|down @COMPAT with entries recorded for DCs < 2.1.7 - PCMK__XA_IN_CCM ::= true|false - PCMK_XA_CRMD ::= online|offline Since crm_feature_set 3.18.0 (pacemaker-2.1.7): - PCMK__XA_IN_CCM ::= |0 Since when node has been a cluster member. A value 0 of means the node is not a cluster member. - PCMK_XA_CRMD ::= |0 Since when peer has been online in CPG. A value 0 means the peer is offline in CPG. */ crm_trace("Node %s member@%lld online@%lld join=%s expected=%s%s", pcmk__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, ""), (termination_requested? " (termination requested)" : "")); if (this_node->details->shutdown) { crm_debug("%s is shutting down", pcmk__node_name(this_node)); /* Slightly different criteria since we can't shut down a dead peer */ return (when_online > 0); } if (when_member < 0) { pe_fence_node(scheduler, this_node, "peer has not been seen by the cluster", FALSE); return false; } if (pcmk__str_eq(join, CRMD_JOINSTATE_NACK, pcmk__str_none)) { pe_fence_node(scheduler, this_node, "peer failed Pacemaker membership criteria", FALSE); } else if (termination_requested) { if ((when_member <= 0) && (when_online <= 0) && pcmk__str_eq(join, CRMD_JOINSTATE_DOWN, pcmk__str_none)) { crm_info("%s was fenced as requested", pcmk__node_name(this_node)); return false; } pe_fence_node(scheduler, this_node, "fencing was requested", false); } else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_DOWN, pcmk__str_null_matches)) { if (pending_too_long(scheduler, this_node, when_member, when_online)) { pe_fence_node(scheduler, this_node, "peer pending timed out on joining the process group", FALSE); } else if ((when_member > 0) || (when_online > 0)) { crm_info("- %s is not ready to run resources", pcmk__node_name(this_node)); pcmk__set_node_flags(this_node, pcmk__node_standby); this_node->details->pending = TRUE; } else { crm_trace("%s is down or still coming up", pcmk__node_name(this_node)); } } else if (when_member <= 0) { // Consider PCMK_OPT_PRIORITY_FENCING_DELAY for lost nodes pe_fence_node(scheduler, this_node, "peer is no longer part of the cluster", TRUE); } else if (when_online <= 0) { pe_fence_node(scheduler, this_node, "peer process is no longer available", FALSE); /* Everything is running at this point, now check join state */ } else if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_none)) { crm_info("%s is active", pcmk__node_name(this_node)); } else if (pcmk__str_any_of(join, CRMD_JOINSTATE_PENDING, CRMD_JOINSTATE_DOWN, NULL)) { crm_info("%s is not ready to run resources", pcmk__node_name(this_node)); pcmk__set_node_flags(this_node, pcmk__node_standby); this_node->details->pending = TRUE; } else { pe_fence_node(scheduler, this_node, "peer was in an unknown state", FALSE); } return (when_member > 0); } static void determine_remote_online_status(pcmk_scheduler_t *scheduler, pcmk_node_t *this_node) { pcmk_resource_t *rsc = this_node->priv->remote; pcmk_resource_t *launcher = NULL; pcmk_node_t *host = NULL; const char *node_type = "Remote"; if (rsc == NULL) { /* This is a leftover node state entry for a former Pacemaker Remote * node whose connection resource was removed. Consider it offline. */ crm_trace("Pacemaker Remote node %s is considered OFFLINE because " "its connection resource has been removed from the CIB", this_node->priv->id); this_node->details->online = FALSE; return; } launcher = rsc->priv->launcher; if (launcher != NULL) { node_type = "Guest"; if (pcmk__list_of_1(rsc->priv->active_nodes)) { host = rsc->priv->active_nodes->data; } } /* If the resource is currently started, mark it online. */ if (rsc->priv->orig_role == pcmk_role_started) { this_node->details->online = TRUE; } /* consider this node shutting down if transitioning start->stop */ if ((rsc->priv->orig_role == pcmk_role_started) && (rsc->priv->next_role == pcmk_role_stopped)) { crm_trace("%s node %s shutting down because connection resource is stopping", node_type, this_node->priv->id); this_node->details->shutdown = TRUE; } /* Now check all the failure conditions. */ if ((launcher != NULL) && pcmk_is_set(launcher->flags, pcmk__rsc_failed)) { crm_trace("Guest node %s UNCLEAN because guest resource failed", this_node->priv->id); this_node->details->online = FALSE; pcmk__set_node_flags(this_node, pcmk__node_remote_reset); } else if (pcmk_is_set(rsc->flags, pcmk__rsc_failed)) { crm_trace("%s node %s OFFLINE because connection resource failed", node_type, this_node->priv->id); this_node->details->online = FALSE; } else if ((rsc->priv->orig_role == pcmk_role_stopped) || ((launcher != NULL) && (launcher->priv->orig_role == pcmk_role_stopped))) { crm_trace("%s node %s OFFLINE because its resource is stopped", node_type, this_node->priv->id); this_node->details->online = FALSE; pcmk__clear_node_flags(this_node, pcmk__node_remote_reset); } else if (host && (host->details->online == FALSE) && host->details->unclean) { crm_trace("Guest node %s UNCLEAN because host is unclean", this_node->priv->id); this_node->details->online = FALSE; pcmk__set_node_flags(this_node, pcmk__node_remote_reset); } else { crm_trace("%s node %s is %s", node_type, this_node->priv->id, this_node->details->online? "ONLINE" : "OFFLINE"); } } static void determine_online_status(const xmlNode *node_state, pcmk_node_t *this_node, pcmk_scheduler_t *scheduler) { gboolean online = FALSE; const char *exp_state = pcmk__xe_get(node_state, PCMK_XA_EXPECTED); CRM_CHECK(this_node != NULL, return); this_node->details->shutdown = FALSE; if (pe__shutdown_requested(this_node)) { this_node->details->shutdown = TRUE; } else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { pcmk__set_node_flags(this_node, pcmk__node_expected_up); } if (!pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { online = determine_online_status_no_fencing(scheduler, node_state, this_node); } else { online = determine_online_status_fencing(scheduler, node_state, this_node); } if (online) { this_node->details->online = TRUE; } else { /* remove node from contention */ this_node->assign->score = -PCMK_SCORE_INFINITY; } if (online && this_node->details->shutdown) { /* don't run resources here */ this_node->assign->score = -PCMK_SCORE_INFINITY; } if (this_node->details->unclean) { pcmk__sched_warn(scheduler, "%s is unclean", pcmk__node_name(this_node)); } else if (!this_node->details->online) { crm_trace("%s is offline", pcmk__node_name(this_node)); } else if (this_node->details->shutdown) { crm_info("%s is shutting down", pcmk__node_name(this_node)); } else if (this_node->details->pending) { crm_info("%s is pending", pcmk__node_name(this_node)); } else if (pcmk_is_set(this_node->priv->flags, pcmk__node_standby)) { crm_info("%s is in standby", pcmk__node_name(this_node)); } else if (this_node->details->maintenance) { crm_info("%s is in maintenance", pcmk__node_name(this_node)); } else { crm_info("%s is online", pcmk__node_name(this_node)); } } /*! * \internal * \brief Find the end of a resource's name, excluding any clone suffix * * \param[in] id Resource ID to check * * \return Pointer to last character of resource's base name */ const char * pe_base_name_end(const char *id) { if (!pcmk__str_empty(id)) { const char *end = id + strlen(id) - 1; for (const char *s = end; s > id; --s) { switch (*s) { case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': break; case ':': return (s == end)? s : (s - 1); default: return end; } } return end; } return NULL; } /*! * \internal * \brief Get a resource name excluding any clone suffix * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_strip(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); char *basename = NULL; pcmk__assert(end != NULL); basename = strndup(last_rsc_id, end - last_rsc_id + 1); pcmk__assert(basename != NULL); return basename; } /*! * \internal * \brief Get the name of the first instance of a cloned resource * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name plus :0 * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_zero(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); size_t base_name_len = end - last_rsc_id + 1; char *zero = NULL; pcmk__assert(end != NULL); zero = pcmk__assert_alloc(base_name_len + 3, sizeof(char)); memcpy(zero, last_rsc_id, base_name_len); zero[base_name_len] = ':'; zero[base_name_len + 1] = '0'; return zero; } static pcmk_resource_t * create_fake_resource(const char *rsc_id, const xmlNode *rsc_entry, pcmk_scheduler_t *scheduler) { pcmk_resource_t *rsc = NULL; xmlNode *xml_rsc = pcmk__xe_create(NULL, PCMK_XE_PRIMITIVE); pcmk__xe_copy_attrs(xml_rsc, rsc_entry, pcmk__xaf_none); crm_xml_add(xml_rsc, PCMK_XA_ID, rsc_id); crm_log_xml_debug(xml_rsc, "Orphan resource"); if (pe__unpack_resource(xml_rsc, &rsc, NULL, scheduler) != pcmk_rc_ok) { return NULL; } if (xml_contains_remote_node(xml_rsc)) { pcmk_node_t *node; crm_debug("Detected orphaned remote node %s", rsc_id); node = pcmk_find_node(scheduler, rsc_id); if (node == NULL) { node = pe_create_node(rsc_id, rsc_id, PCMK_VALUE_REMOTE, 0, scheduler); } link_rsc2remotenode(scheduler, rsc); if (node) { crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id); node->details->shutdown = TRUE; } } if (pcmk__xe_get(rsc_entry, PCMK__META_CONTAINER)) { // This removed resource needs to be mapped to a launcher crm_trace("Launched resource %s was removed from the configuration", rsc_id); pcmk__set_rsc_flags(rsc, pcmk__rsc_removed_launched); } pcmk__set_rsc_flags(rsc, pcmk__rsc_removed); scheduler->priv->resources = g_list_append(scheduler->priv->resources, rsc); return rsc; } /*! * \internal * \brief Create orphan instance for anonymous clone resource history * * \param[in,out] parent Clone resource that orphan will be added to * \param[in] rsc_id Orphan's resource ID * \param[in] node Where orphan is active (for logging only) * \param[in,out] scheduler Scheduler data * * \return Newly added orphaned instance of \p parent */ static pcmk_resource_t * create_anonymous_orphan(pcmk_resource_t *parent, const char *rsc_id, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { pcmk_resource_t *top = pe__create_clone_child(parent, scheduler); pcmk_resource_t *orphan = NULL; // find_rsc() because we might be a cloned group orphan = top->priv->fns->find_rsc(top, rsc_id, NULL, pcmk_rsc_match_clone_only); pcmk__rsc_debug(parent, "Created orphan %s for %s: %s on %s", top->id, parent->id, rsc_id, pcmk__node_name(node)); return orphan; } /*! * \internal * \brief Check a node for an instance of an anonymous clone * * Return a child instance of the specified anonymous clone, in order of * preference: (1) the instance running on the specified node, if any; * (2) an inactive instance (i.e. within the total of \c PCMK_META_CLONE_MAX * instances); (3) a newly created orphan (that is, \c PCMK_META_CLONE_MAX * instances are already active). * * \param[in,out] scheduler Scheduler data * \param[in] node Node on which to check for instance * \param[in,out] parent Clone to check * \param[in] rsc_id Name of cloned resource in history (no instance) */ static pcmk_resource_t * find_anonymous_clone(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, pcmk_resource_t *parent, const char *rsc_id) { GList *rIter = NULL; pcmk_resource_t *rsc = NULL; pcmk_resource_t *inactive_instance = NULL; gboolean skip_inactive = FALSE; pcmk__assert(pcmk__is_anonymous_clone(parent)); // Check for active (or partially active, for cloned groups) instance pcmk__rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, pcmk__node_name(node), parent->id); for (rIter = parent->priv->children; (rIter != NULL) && (rsc == NULL); rIter = rIter->next) { GList *locations = NULL; pcmk_resource_t *child = rIter->data; /* Check whether this instance is already known to be active or pending * anywhere, at this stage of unpacking. Because this function is called * for a resource before the resource's individual operation history * entries are unpacked, locations will generally not contain the * desired node. * * However, there are three exceptions: * (1) when child is a cloned group and we have already unpacked the * history of another member of the group on the same node; * (2) when we've already unpacked the history of another numbered * instance on the same node (which can happen if * PCMK_META_GLOBALLY_UNIQUE was flipped from true to false); and * (3) when we re-run calculations on the same scheduler data as part of * a simulation. */ child->priv->fns->location(child, &locations, pcmk__rsc_node_current |pcmk__rsc_node_pending); if (locations) { /* We should never associate the same numbered anonymous clone * instance with multiple nodes, and clone instances can't migrate, * so there must be only one location, regardless of history. */ CRM_LOG_ASSERT(locations->next == NULL); if (pcmk__same_node((pcmk_node_t *) locations->data, node)) { /* This child instance is active on the requested node, so check * for a corresponding configured resource. We use find_rsc() * instead of child because child may be a cloned group, and we * need the particular member corresponding to rsc_id. * * If the history entry is orphaned, rsc will be NULL. */ rsc = parent->priv->fns->find_rsc(child, rsc_id, NULL, pcmk_rsc_match_clone_only); if (rsc) { /* If there are multiple instance history entries for an * anonymous clone in a single node's history (which can * happen if PCMK_META_GLOBALLY_UNIQUE is switched from true * to false), we want to consider the instances beyond the * first as orphans, even if there are inactive instance * numbers available. */ if (rsc->priv->active_nodes != NULL) { crm_notice("Active (now-)anonymous clone %s has " "multiple (orphan) instance histories on %s", parent->id, pcmk__node_name(node)); skip_inactive = TRUE; rsc = NULL; } else { pcmk__rsc_trace(parent, "Resource %s, active", rsc->id); } } } g_list_free(locations); } else { pcmk__rsc_trace(parent, "Resource %s, skip inactive", child->id); if (!skip_inactive && !inactive_instance && !pcmk_is_set(child->flags, pcmk__rsc_blocked)) { // Remember one inactive instance in case we don't find active inactive_instance = parent->priv->fns->find_rsc(child, rsc_id, NULL, pcmk_rsc_match_clone_only); /* ... but don't use it if it was already associated with a * pending action on another node */ if (inactive_instance != NULL) { const pcmk_node_t *pending_node = NULL; pending_node = inactive_instance->priv->pending_node; if ((pending_node != NULL) && !pcmk__same_node(pending_node, node)) { inactive_instance = NULL; } } } } } if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) { pcmk__rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id); rsc = inactive_instance; } /* If the resource has PCMK_META_REQUIRES set to PCMK_VALUE_QUORUM or * PCMK_VALUE_NOTHING, and we don't have a clone instance for every node, we * don't want to consume a valid instance number for unclean nodes. Such * instances may appear to be active according to the history, but should be * considered inactive, so we can start an instance elsewhere. Treat such * instances as orphans. * * An exception is instances running on guest nodes -- since guest node * "fencing" is actually just a resource stop, requires shouldn't apply. * * @TODO Ideally, we'd use an inactive instance number if it is not needed * for any clean instances. However, we don't know that at this point. */ if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk__rsc_needs_fencing) && (!node->details->online || node->details->unclean) && !pcmk__is_guest_or_bundle_node(node) && !pe__is_universal_clone(parent, scheduler)) { rsc = NULL; } if (rsc == NULL) { rsc = create_anonymous_orphan(parent, rsc_id, node, scheduler); pcmk__rsc_trace(parent, "Resource %s, orphan", rsc->id); } return rsc; } static pcmk_resource_t * unpack_find_resource(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, const char *rsc_id) { pcmk_resource_t *rsc = NULL; pcmk_resource_t *parent = NULL; crm_trace("looking for %s", rsc_id); rsc = pe_find_resource(scheduler->priv->resources, rsc_id); if (rsc == NULL) { /* If we didn't find the resource by its name in the operation history, * check it again as a clone instance. Even when PCMK_META_CLONE_MAX=0, * we create a single :0 orphan to match against here. */ char *clone0_id = clone_zero(rsc_id); pcmk_resource_t *clone0 = pe_find_resource(scheduler->priv->resources, clone0_id); if (clone0 && !pcmk_is_set(clone0->flags, pcmk__rsc_unique)) { rsc = clone0; parent = uber_parent(clone0); crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id); } else { crm_trace("%s is not known as %s either (orphan)", rsc_id, clone0_id); } free(clone0_id); } else if (rsc->priv->variant > pcmk__rsc_variant_primitive) { crm_trace("Resource history for %s is orphaned " "because it is no longer primitive", rsc_id); return NULL; } else { parent = uber_parent(rsc); } if (pcmk__is_anonymous_clone(parent)) { if (pcmk__is_bundled(parent)) { rsc = pe__find_bundle_replica(parent->priv->parent, node); } else { char *base = clone_strip(rsc_id); rsc = find_anonymous_clone(scheduler, node, parent, base); free(base); pcmk__assert(rsc != NULL); } } if (rsc && !pcmk__str_eq(rsc_id, rsc->id, pcmk__str_none) && !pcmk__str_eq(rsc_id, rsc->priv->history_id, pcmk__str_none)) { pcmk__str_update(&(rsc->priv->history_id), rsc_id); pcmk__rsc_debug(rsc, "Internally renamed %s on %s to %s%s", rsc_id, pcmk__node_name(node), rsc->id, pcmk_is_set(rsc->flags, pcmk__rsc_removed)? " (ORPHAN)" : ""); } return rsc; } static pcmk_resource_t * process_orphan_resource(const xmlNode *rsc_entry, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { pcmk_resource_t *rsc = NULL; const char *rsc_id = pcmk__xe_get(rsc_entry, PCMK_XA_ID); crm_debug("Detected orphan resource %s on %s", rsc_id, pcmk__node_name(node)); rsc = create_fake_resource(rsc_id, rsc_entry, scheduler); if (rsc == NULL) { return NULL; } if (!pcmk_is_set(scheduler->flags, pcmk__sched_stop_removed_resources)) { pcmk__clear_rsc_flags(rsc, pcmk__rsc_managed); } else { CRM_CHECK(rsc != NULL, return NULL); pcmk__rsc_trace(rsc, "Added orphan %s", rsc->id); resource_location(rsc, NULL, -PCMK_SCORE_INFINITY, "__orphan_do_not_run__", scheduler); } return rsc; } static void process_rsc_state(pcmk_resource_t *rsc, pcmk_node_t *node, enum pcmk__on_fail on_fail) { pcmk_node_t *tmpnode = NULL; char *reason = NULL; enum pcmk__on_fail save_on_fail = pcmk__on_fail_ignore; pcmk_scheduler_t *scheduler = NULL; bool known_active = false; pcmk__assert(rsc != NULL); scheduler = rsc->priv->scheduler; known_active = (rsc->priv->orig_role > pcmk_role_stopped); pcmk__rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s", rsc->id, pcmk_role_text(rsc->priv->orig_role), pcmk__node_name(node), pcmk__on_fail_text(on_fail)); /* process current state */ if (rsc->priv->orig_role != pcmk_role_unknown) { pcmk_resource_t *iter = rsc; while (iter) { if (g_hash_table_lookup(iter->priv->probed_nodes, node->priv->id) == NULL) { pcmk_node_t *n = pe__copy_node(node); pcmk__rsc_trace(rsc, "%s (%s in history) known on %s", rsc->id, pcmk__s(rsc->priv->history_id, "the same"), pcmk__node_name(n)); g_hash_table_insert(iter->priv->probed_nodes, (gpointer) n->priv->id, n); } if (pcmk_is_set(iter->flags, pcmk__rsc_unique)) { break; } iter = iter->priv->parent; } } /* If a managed resource is believed to be running, but node is down ... */ if (known_active && !node->details->online && !node->details->maintenance && pcmk_is_set(rsc->flags, pcmk__rsc_managed)) { gboolean should_fence = FALSE; /* If this is a guest node, fence it (regardless of whether fencing is * enabled, because guest node fencing is done by recovery of the * container resource rather than by the fencer). Mark the resource * we're processing as failed. When the guest comes back up, its * operation history in the CIB will be cleared, freeing the affected * resource to run again once we are sure we know its state. */ if (pcmk__is_guest_or_bundle_node(node)) { pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); should_fence = TRUE; } else if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { if (pcmk__is_remote_node(node) && (node->priv->remote != NULL) && !pcmk_is_set(node->priv->remote->flags, pcmk__rsc_failed)) { /* Setting unseen means that fencing of the remote node will * occur only if the connection resource is not going to start * somewhere. This allows connection resources on a failed * cluster node to move to another node without requiring the * remote nodes to be fenced as well. */ pcmk__clear_node_flags(node, pcmk__node_seen); reason = crm_strdup_printf("%s is active there (fencing will be" " revoked if remote connection can " "be re-established elsewhere)", rsc->id); } should_fence = TRUE; } if (should_fence) { if (reason == NULL) { reason = crm_strdup_printf("%s is thought to be active there", rsc->id); } pe_fence_node(scheduler, node, reason, FALSE); } free(reason); } /* In order to calculate priority_fencing_delay correctly, save the failure information and pass it to native_add_running(). */ save_on_fail = on_fail; if (node->details->unclean) { /* No extra processing needed * Also allows resources to be started again after a node is shot */ on_fail = pcmk__on_fail_ignore; } switch (on_fail) { case pcmk__on_fail_ignore: /* nothing to do */ break; case pcmk__on_fail_demote: pcmk__set_rsc_flags(rsc, pcmk__rsc_failed); demote_action(rsc, node, FALSE); break; case pcmk__on_fail_fence_node: /* treat it as if it is still running * but also mark the node as unclean */ reason = crm_strdup_printf("%s failed there", rsc->id); pe_fence_node(scheduler, node, reason, FALSE); free(reason); break; case pcmk__on_fail_standby_node: pcmk__set_node_flags(node, pcmk__node_standby|pcmk__node_fail_standby); break; case pcmk__on_fail_block: /* is_managed == FALSE will prevent any * actions being sent for the resource */ pcmk__clear_rsc_flags(rsc, pcmk__rsc_managed); pcmk__set_rsc_flags(rsc, pcmk__rsc_blocked); break; case pcmk__on_fail_ban: /* make sure it comes up somewhere else * or not at all */ resource_location(rsc, node, -PCMK_SCORE_INFINITY, "__action_migration_auto__", scheduler); break; case pcmk__on_fail_stop: pe__set_next_role(rsc, pcmk_role_stopped, PCMK_META_ON_FAIL "=" PCMK_VALUE_STOP); break; case pcmk__on_fail_restart: if (known_active) { pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); stop_action(rsc, node, FALSE); } break; case pcmk__on_fail_restart_container: pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); if ((rsc->priv->launcher != NULL) && pcmk__is_bundled(rsc)) { /* A bundle's remote connection can run on a different node than * the bundle's container. We don't necessarily know where the * container is running yet, so remember it and add a stop * action for it later. */ scheduler->priv->stop_needed = g_list_prepend(scheduler->priv->stop_needed, rsc->priv->launcher); } else if (rsc->priv->launcher != NULL) { stop_action(rsc->priv->launcher, node, FALSE); } else if (known_active) { stop_action(rsc, node, FALSE); } break; case pcmk__on_fail_reset_remote: pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { tmpnode = NULL; if (pcmk_is_set(rsc->flags, pcmk__rsc_is_remote_connection)) { tmpnode = pcmk_find_node(scheduler, rsc->id); } if (pcmk__is_remote_node(tmpnode) && !pcmk_is_set(tmpnode->priv->flags, pcmk__node_remote_fenced)) { /* The remote connection resource failed in a way that * should result in fencing the remote node. */ pe_fence_node(scheduler, tmpnode, "remote connection is unrecoverable", FALSE); } } /* require the stop action regardless if fencing is occurring or not. */ if (known_active) { stop_action(rsc, node, FALSE); } /* if reconnect delay is in use, prevent the connection from exiting the * "STOPPED" role until the failure is cleared by the delay timeout. */ if (rsc->priv->remote_reconnect_ms > 0U) { pe__set_next_role(rsc, pcmk_role_stopped, "remote reset"); } break; } /* Ensure a remote connection failure forces an unclean Pacemaker Remote * node to be fenced. By marking the node as seen, the failure will result * in a fencing operation regardless if we're going to attempt to reconnect * in this transition. */ if (pcmk_all_flags_set(rsc->flags, pcmk__rsc_failed|pcmk__rsc_is_remote_connection)) { tmpnode = pcmk_find_node(scheduler, rsc->id); if (tmpnode && tmpnode->details->unclean) { pcmk__set_node_flags(tmpnode, pcmk__node_seen); } } if (known_active) { if (pcmk_is_set(rsc->flags, pcmk__rsc_removed)) { if (pcmk_is_set(rsc->flags, pcmk__rsc_managed)) { crm_notice("Removed resource %s is active on %s and will be " "stopped when possible", rsc->id, pcmk__node_name(node)); } else { crm_notice("Removed resource %s must be stopped manually on %s " "because " PCMK_OPT_STOP_ORPHAN_RESOURCES " is set to false", rsc->id, pcmk__node_name(node)); } } native_add_running(rsc, node, scheduler, (save_on_fail != pcmk__on_fail_ignore)); switch (on_fail) { case pcmk__on_fail_ignore: break; case pcmk__on_fail_demote: case pcmk__on_fail_block: pcmk__set_rsc_flags(rsc, pcmk__rsc_failed); break; default: pcmk__set_rsc_flags(rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); break; } } else if ((rsc->priv->history_id != NULL) && (strchr(rsc->priv->history_id, ':') != NULL)) { /* @COMPAT This is for older (<1.1.8) status sections that included * instance numbers, otherwise stopped instances are considered orphans. * * @TODO We should be able to drop this, but some old regression tests * will need to be updated. Double-check that this is not still needed * for unique clones (which may have been later converted to anonymous). */ pcmk__rsc_trace(rsc, "Clearing history ID %s for %s (stopped)", rsc->priv->history_id, rsc->id); free(rsc->priv->history_id); rsc->priv->history_id = NULL; } else { GList *possible_matches = pe__resource_actions(rsc, node, PCMK_ACTION_STOP, FALSE); GList *gIter = possible_matches; for (; gIter != NULL; gIter = gIter->next) { pcmk_action_t *stop = (pcmk_action_t *) gIter->data; pcmk__set_action_flags(stop, pcmk__action_optional); } g_list_free(possible_matches); } /* A successful stop after migrate_to on the migration source doesn't make * the partially migrated resource stopped on the migration target. */ if ((rsc->priv->orig_role == pcmk_role_stopped) && (rsc->priv->active_nodes != NULL) && (rsc->priv->partial_migration_target != NULL) && pcmk__same_node(rsc->priv->partial_migration_source, node)) { rsc->priv->orig_role = pcmk_role_started; } } /* create active recurring operations as optional */ static void process_recurring(pcmk_node_t *node, pcmk_resource_t *rsc, int start_index, int stop_index, GList *sorted_op_list, pcmk_scheduler_t *scheduler) { int counter = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; pcmk__assert(rsc != NULL); pcmk__rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index); for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; guint interval_ms = 0; char *key = NULL; const char *id = pcmk__xe_id(rsc_op); counter++; if (node->details->online == FALSE) { pcmk__rsc_trace(rsc, "Skipping %s on %s: node is offline", rsc->id, pcmk__node_name(node)); break; /* Need to check if there's a monitor for role="Stopped" */ } else if (start_index < stop_index && counter <= stop_index) { pcmk__rsc_trace(rsc, "Skipping %s on %s: resource is not active", id, pcmk__node_name(node)); continue; } else if (counter < start_index) { pcmk__rsc_trace(rsc, "Skipping %s on %s: old %d", id, pcmk__node_name(node), counter); continue; } pcmk__xe_get_guint(rsc_op, PCMK_META_INTERVAL, &interval_ms); if (interval_ms == 0) { pcmk__rsc_trace(rsc, "Skipping %s on %s: non-recurring", id, pcmk__node_name(node)); continue; } status = pcmk__xe_get(rsc_op, PCMK__XA_OP_STATUS); if (pcmk__str_eq(status, "-1", pcmk__str_casei)) { pcmk__rsc_trace(rsc, "Skipping %s on %s: status", id, pcmk__node_name(node)); continue; } task = pcmk__xe_get(rsc_op, PCMK_XA_OPERATION); /* create the action */ key = pcmk__op_key(rsc->id, task, interval_ms); pcmk__rsc_trace(rsc, "Creating %s on %s", key, pcmk__node_name(node)); custom_action(rsc, key, task, node, TRUE, scheduler); } } void calculate_active_ops(const GList *sorted_op_list, int *start_index, int *stop_index) { int counter = -1; int implied_monitor_start = -1; int implied_clone_start = -1; const char *task = NULL; const char *status = NULL; *stop_index = -1; *start_index = -1; for (const GList *iter = sorted_op_list; iter != NULL; iter = iter->next) { const xmlNode *rsc_op = (const xmlNode *) iter->data; counter++; task = pcmk__xe_get(rsc_op, PCMK_XA_OPERATION); status = pcmk__xe_get(rsc_op, PCMK__XA_OP_STATUS); if (pcmk__str_eq(task, PCMK_ACTION_STOP, pcmk__str_casei) && pcmk__str_eq(status, "0", pcmk__str_casei)) { *stop_index = counter; } else if (pcmk__strcase_any_of(task, PCMK_ACTION_START, PCMK_ACTION_MIGRATE_FROM, NULL)) { *start_index = counter; } else if ((implied_monitor_start <= *stop_index) && pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei)) { const char *rc = pcmk__xe_get(rsc_op, PCMK__XA_RC_CODE); if (pcmk__strcase_any_of(rc, "0", "8", NULL)) { implied_monitor_start = counter; } } else if (pcmk__strcase_any_of(task, PCMK_ACTION_PROMOTE, PCMK_ACTION_DEMOTE, NULL)) { implied_clone_start = counter; } } if (*start_index == -1) { if (implied_clone_start != -1) { *start_index = implied_clone_start; } else if (implied_monitor_start != -1) { *start_index = implied_monitor_start; } } } // If resource history entry has shutdown lock, remember lock node and time static void unpack_shutdown_lock(const xmlNode *rsc_entry, pcmk_resource_t *rsc, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { time_t lock_time = 0; // When lock started (i.e. node shutdown time) time_t sched_time = 0; guint shutdown_lock_ms = scheduler->priv->shutdown_lock_ms; pcmk__xe_get_time(rsc_entry, PCMK_OPT_SHUTDOWN_LOCK, &lock_time); if (lock_time == 0) { return; } sched_time = pcmk__scheduler_epoch_time(scheduler); if ((shutdown_lock_ms > 0U) && (sched_time > (lock_time + pcmk__timeout_ms2s(shutdown_lock_ms)))) { pcmk__rsc_info(rsc, "Shutdown lock for %s on %s expired", rsc->id, pcmk__node_name(node)); pe__clear_resource_history(rsc, node); } else { rsc->priv->lock_node = node; rsc->priv->lock_time = lock_time; } } /*! * \internal * \brief Unpack one \c PCMK__XE_LRM_RESOURCE entry from a node's CIB status * * \param[in,out] node Node whose status is being unpacked * \param[in] rsc_entry \c PCMK__XE_LRM_RESOURCE XML being unpacked * \param[in,out] scheduler Scheduler data * * \return Resource corresponding to the entry, or NULL if no operation history */ static pcmk_resource_t * unpack_lrm_resource(pcmk_node_t *node, const xmlNode *lrm_resource, pcmk_scheduler_t *scheduler) { GList *gIter = NULL; int stop_index = -1; int start_index = -1; enum rsc_role_e req_role = pcmk_role_unknown; const char *rsc_id = pcmk__xe_id(lrm_resource); pcmk_resource_t *rsc = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; xmlNode *rsc_op = NULL; xmlNode *last_failure = NULL; enum pcmk__on_fail on_fail = pcmk__on_fail_ignore; enum rsc_role_e saved_role = pcmk_role_unknown; if (rsc_id == NULL) { pcmk__config_err("Ignoring invalid " PCMK__XE_LRM_RESOURCE " entry: No " PCMK_XA_ID); crm_log_xml_info(lrm_resource, "missing-id"); return NULL; } crm_trace("Unpacking " PCMK__XE_LRM_RESOURCE " for %s on %s", rsc_id, pcmk__node_name(node)); /* Build a list of individual PCMK__XE_LRM_RSC_OP entries, so we can sort * them */ for (rsc_op = pcmk__xe_first_child(lrm_resource, PCMK__XE_LRM_RSC_OP, NULL, NULL); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op, PCMK__XE_LRM_RSC_OP)) { op_list = g_list_prepend(op_list, rsc_op); } if (!pcmk_is_set(scheduler->flags, pcmk__sched_shutdown_lock)) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } } /* find the resource */ rsc = unpack_find_resource(scheduler, node, rsc_id); if (rsc == NULL) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } else { rsc = process_orphan_resource(lrm_resource, node, scheduler); } } pcmk__assert(rsc != NULL); // Check whether the resource is "shutdown-locked" to this node if (pcmk_is_set(scheduler->flags, pcmk__sched_shutdown_lock)) { unpack_shutdown_lock(lrm_resource, rsc, node, scheduler); } /* process operations */ saved_role = rsc->priv->orig_role; rsc->priv->orig_role = pcmk_role_unknown; sorted_op_list = g_list_sort(op_list, sort_op_by_callid); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail); } /* create active recurring operations as optional */ calculate_active_ops(sorted_op_list, &start_index, &stop_index); process_recurring(node, rsc, start_index, stop_index, sorted_op_list, scheduler); /* no need to free the contents */ g_list_free(sorted_op_list); process_rsc_state(rsc, node, on_fail); if (get_target_role(rsc, &req_role)) { if ((rsc->priv->next_role == pcmk_role_unknown) || (req_role < rsc->priv->next_role)) { pe__set_next_role(rsc, req_role, PCMK_META_TARGET_ROLE); } else if (req_role > rsc->priv->next_role) { pcmk__rsc_info(rsc, "%s: Not overwriting calculated next role %s" " with requested next role %s", rsc->id, pcmk_role_text(rsc->priv->next_role), pcmk_role_text(req_role)); } } if (saved_role > rsc->priv->orig_role) { rsc->priv->orig_role = saved_role; } return rsc; } static void handle_removed_launched_resources(const xmlNode *lrm_rsc_list, pcmk_scheduler_t *scheduler) { for (const xmlNode *rsc_entry = pcmk__xe_first_child(lrm_rsc_list, PCMK__XE_LRM_RESOURCE, NULL, NULL); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry, PCMK__XE_LRM_RESOURCE)) { pcmk_resource_t *rsc; pcmk_resource_t *launcher = NULL; const char *rsc_id; const char *launcher_id = NULL; launcher_id = pcmk__xe_get(rsc_entry, PCMK__META_CONTAINER); rsc_id = pcmk__xe_get(rsc_entry, PCMK_XA_ID); if ((launcher_id == NULL) || (rsc_id == NULL)) { continue; } launcher = pe_find_resource(scheduler->priv->resources, launcher_id); if (launcher == NULL) { continue; } rsc = pe_find_resource(scheduler->priv->resources, rsc_id); if ((rsc == NULL) || (rsc->priv->launcher != NULL) || !pcmk_is_set(rsc->flags, pcmk__rsc_removed_launched)) { continue; } pcmk__rsc_trace(rsc, "Mapped launcher of removed resource %s to %s", rsc->id, launcher_id); rsc->priv->launcher = launcher; launcher->priv->launched = g_list_append(launcher->priv->launched, rsc); } } /*! * \internal * \brief Unpack one node's lrm status section * * \param[in,out] node Node whose status is being unpacked * \param[in] xml CIB node state XML * \param[in,out] scheduler Scheduler data */ static void unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml, pcmk_scheduler_t *scheduler) { bool found_removed_launched_resource = false; // Drill down to PCMK__XE_LRM_RESOURCES section xml = pcmk__xe_first_child(xml, PCMK__XE_LRM, NULL, NULL); if (xml == NULL) { return; } xml = pcmk__xe_first_child(xml, PCMK__XE_LRM_RESOURCES, NULL, NULL); if (xml == NULL) { return; } // Unpack each PCMK__XE_LRM_RESOURCE entry for (const xmlNode *rsc_entry = pcmk__xe_first_child(xml, PCMK__XE_LRM_RESOURCE, NULL, NULL); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry, PCMK__XE_LRM_RESOURCE)) { pcmk_resource_t *rsc = unpack_lrm_resource(node, rsc_entry, scheduler); if ((rsc != NULL) && pcmk_is_set(rsc->flags, pcmk__rsc_removed_launched)) { found_removed_launched_resource = true; } } /* Now that all resource state has been unpacked for this node, map any * removed launched resources to their launchers. */ if (found_removed_launched_resource) { handle_removed_launched_resources(xml, scheduler); } } static void set_active(pcmk_resource_t *rsc) { const pcmk_resource_t *top = pe__const_top_resource(rsc, false); if (top && pcmk_is_set(top->flags, pcmk__rsc_promotable)) { rsc->priv->orig_role = pcmk_role_unpromoted; } else { rsc->priv->orig_role = pcmk_role_started; } } static void set_node_score(gpointer key, gpointer value, gpointer user_data) { pcmk_node_t *node = value; int *score = user_data; node->assign->score = *score; } #define XPATH_NODE_STATE "/" PCMK_XE_CIB "/" PCMK_XE_STATUS \ "/" PCMK__XE_NODE_STATE #define SUB_XPATH_LRM_RESOURCE "/" PCMK__XE_LRM \ "/" PCMK__XE_LRM_RESOURCES \ "/" PCMK__XE_LRM_RESOURCE #define SUB_XPATH_LRM_RSC_OP "/" PCMK__XE_LRM_RSC_OP static xmlNode * find_lrm_op(const char *resource, const char *op, const char *node, const char *source, int target_rc, pcmk_scheduler_t *scheduler) { GString *xpath = NULL; xmlNode *xml = NULL; CRM_CHECK((resource != NULL) && (op != NULL) && (node != NULL), return NULL); xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node, "']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", resource, "']" SUB_XPATH_LRM_RSC_OP "[@" PCMK_XA_OPERATION "='", op, "'", NULL); /* Need to check against transition_magic too? */ if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_TO) == 0)) { pcmk__g_strcat(xpath, " and @" PCMK__META_MIGRATE_TARGET "='", source, "']", NULL); } else if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_FROM) == 0)) { pcmk__g_strcat(xpath, " and @" PCMK__META_MIGRATE_SOURCE "='", source, "']", NULL); } else { g_string_append_c(xpath, ']'); } xml = pcmk__xpath_find_one(scheduler->input->doc, xpath->str, LOG_DEBUG); g_string_free(xpath, TRUE); if (xml && target_rc >= 0) { int rc = PCMK_OCF_UNKNOWN_ERROR; int status = PCMK_EXEC_ERROR; pcmk__xe_get_int(xml, PCMK__XA_RC_CODE, &rc); pcmk__xe_get_int(xml, PCMK__XA_OP_STATUS, &status); if ((rc != target_rc) || (status != PCMK_EXEC_DONE)) { return NULL; } } return xml; } static xmlNode * find_lrm_resource(const char *rsc_id, const char *node_name, pcmk_scheduler_t *scheduler) { GString *xpath = NULL; xmlNode *xml = NULL; CRM_CHECK((rsc_id != NULL) && (node_name != NULL), return NULL); xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node_name, "']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", rsc_id, "']", NULL); xml = pcmk__xpath_find_one(scheduler->input->doc, xpath->str, LOG_DEBUG); g_string_free(xpath, TRUE); return xml; } /*! * \internal * \brief Check whether a resource has no completed action history on a node * * \param[in,out] rsc Resource to check * \param[in] node_name Node to check * * \return true if \p rsc_id is unknown on \p node_name, otherwise false */ static bool unknown_on_node(pcmk_resource_t *rsc, const char *node_name) { bool result = false; xmlXPathObject *search; char *xpath = NULL; xpath = crm_strdup_printf(XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='%s']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='%s']" SUB_XPATH_LRM_RSC_OP "[@" PCMK__XA_RC_CODE "!='%d']", node_name, rsc->id, PCMK_OCF_UNKNOWN); search = pcmk__xpath_search(rsc->priv->scheduler->input->doc, xpath); result = (pcmk__xpath_num_results(search) == 0); xmlXPathFreeObject(search); free(xpath); return result; } /*! * \internal * \brief Check whether a probe/monitor indicating the resource was not running * on a node happened after some event * * \param[in] rsc_id Resource being checked * \param[in] node_name Node being checked * \param[in] xml_op Event that monitor is being compared to * \param[in,out] scheduler Scheduler data * * \return true if such a monitor happened after event, false otherwise */ static bool monitor_not_running_after(const char *rsc_id, const char *node_name, const xmlNode *xml_op, pcmk_scheduler_t *scheduler) { /* Any probe/monitor operation on the node indicating it was not running * there */ xmlNode *monitor = find_lrm_op(rsc_id, PCMK_ACTION_MONITOR, node_name, NULL, PCMK_OCF_NOT_RUNNING, scheduler); return (monitor != NULL) && (pe__is_newer_op(monitor, xml_op) > 0); } /*! * \internal * \brief Check whether any non-monitor operation on a node happened after some * event * * \param[in] rsc_id Resource being checked * \param[in] node_name Node being checked * \param[in] xml_op Event that non-monitor is being compared to * \param[in,out] scheduler Scheduler data * * \return true if such a operation happened after event, false otherwise */ static bool non_monitor_after(const char *rsc_id, const char *node_name, const xmlNode *xml_op, pcmk_scheduler_t *scheduler) { xmlNode *lrm_resource = NULL; lrm_resource = find_lrm_resource(rsc_id, node_name, scheduler); if (lrm_resource == NULL) { return false; } for (xmlNode *op = pcmk__xe_first_child(lrm_resource, PCMK__XE_LRM_RSC_OP, NULL, NULL); op != NULL; op = pcmk__xe_next(op, PCMK__XE_LRM_RSC_OP)) { const char * task = NULL; if (op == xml_op) { continue; } task = pcmk__xe_get(op, PCMK_XA_OPERATION); if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_STOP, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL) && pe__is_newer_op(op, xml_op) > 0) { return true; } } return false; } /*! * \internal * \brief Check whether the resource has newer state on a node after a migration * attempt * * \param[in] rsc_id Resource being checked * \param[in] node_name Node being checked * \param[in] migrate_to Any migrate_to event that is being compared to * \param[in] migrate_from Any migrate_from event that is being compared to * \param[in,out] scheduler Scheduler data * * \return true if such a operation happened after event, false otherwise */ static bool newer_state_after_migrate(const char *rsc_id, const char *node_name, const xmlNode *migrate_to, const xmlNode *migrate_from, pcmk_scheduler_t *scheduler) { const xmlNode *xml_op = (migrate_from != NULL)? migrate_from : migrate_to; const char *source = pcmk__xe_get(xml_op, PCMK__META_MIGRATE_SOURCE); /* It's preferred to compare to the migrate event on the same node if * existing, since call ids are more reliable. */ if ((xml_op != migrate_to) && (migrate_to != NULL) && pcmk__str_eq(node_name, source, pcmk__str_casei)) { xml_op = migrate_to; } /* If there's any newer non-monitor operation on the node, or any newer * probe/monitor operation on the node indicating it was not running there, * the migration events potentially no longer matter for the node. */ return non_monitor_after(rsc_id, node_name, xml_op, scheduler) || monitor_not_running_after(rsc_id, node_name, xml_op, scheduler); } /*! * \internal * \brief Parse migration source and target node names from history entry * * \param[in] entry Resource history entry for a migration action * \param[in] source_node If not NULL, source must match this node * \param[in] target_node If not NULL, target must match this node * \param[out] source_name Where to store migration source node name * \param[out] target_name Where to store migration target node name * * \return Standard Pacemaker return code */ static int get_migration_node_names(const xmlNode *entry, const pcmk_node_t *source_node, const pcmk_node_t *target_node, const char **source_name, const char **target_name) { *source_name = pcmk__xe_get(entry, PCMK__META_MIGRATE_SOURCE); *target_name = pcmk__xe_get(entry, PCMK__META_MIGRATE_TARGET); if ((*source_name == NULL) || (*target_name == NULL)) { pcmk__config_err("Ignoring resource history entry %s without " PCMK__META_MIGRATE_SOURCE " and " PCMK__META_MIGRATE_TARGET, pcmk__xe_id(entry)); return pcmk_rc_unpack_error; } if ((source_node != NULL) && !pcmk__str_eq(*source_name, source_node->priv->name, pcmk__str_casei|pcmk__str_null_matches)) { pcmk__config_err("Ignoring resource history entry %s because " PCMK__META_MIGRATE_SOURCE "='%s' does not match %s", pcmk__xe_id(entry), *source_name, pcmk__node_name(source_node)); return pcmk_rc_unpack_error; } if ((target_node != NULL) && !pcmk__str_eq(*target_name, target_node->priv->name, pcmk__str_casei|pcmk__str_null_matches)) { pcmk__config_err("Ignoring resource history entry %s because " PCMK__META_MIGRATE_TARGET "='%s' does not match %s", pcmk__xe_id(entry), *target_name, pcmk__node_name(target_node)); return pcmk_rc_unpack_error; } return pcmk_rc_ok; } /* * \internal * \brief Add a migration source to a resource's list of dangling migrations * * If the migrate_to and migrate_from actions in a live migration both * succeeded, but there is no stop on the source, the migration is considered * "dangling." Add the source to the resource's dangling migration list, which * will be used to schedule a stop on the source without affecting the target. * * \param[in,out] rsc Resource involved in migration * \param[in] node Migration source */ static void add_dangling_migration(pcmk_resource_t *rsc, const pcmk_node_t *node) { pcmk__rsc_trace(rsc, "Dangling migration of %s requires stop on %s", rsc->id, pcmk__node_name(node)); rsc->priv->orig_role = pcmk_role_stopped; rsc->priv->dangling_migration_sources = g_list_prepend(rsc->priv->dangling_migration_sources, (gpointer) node); } /*! * \internal * \brief Update resource role etc. after a successful migrate_to action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_to_success(struct action_history *history) { /* A complete migration sequence is: * 1. migrate_to on source node (which succeeded if we get to this function) * 2. migrate_from on target node * 3. stop on source node * * If no migrate_from has happened, the migration is considered to be * "partial". If the migrate_from succeeded but no stop has happened, the * migration is considered to be "dangling". * * If a successful migrate_to and stop have happened on the source node, we * still need to check for a partial migration, due to scenarios (easier to * produce with batch-limit=1) like: * * - A resource is migrating from node1 to node2, and a migrate_to is * initiated for it on node1. * * - node2 goes into standby mode while the migrate_to is pending, which * aborts the transition. * * - Upon completion of the migrate_to, a new transition schedules a stop * on both nodes and a start on node1. * * - If the new transition is aborted for any reason while the resource is * stopping on node1, the transition after that stop completes will see * the migrate_to and stop on the source, but it's still a partial * migration, and the resource must be stopped on node2 because it is * potentially active there due to the migrate_to. * * We also need to take into account that either node's history may be * cleared at any point in the migration process. */ int from_rc = PCMK_OCF_OK; int from_status = PCMK_EXEC_PENDING; pcmk_node_t *target_node = NULL; xmlNode *migrate_from = NULL; const char *source = NULL; const char *target = NULL; bool source_newer_op = false; bool target_newer_state = false; bool active_on_target = false; pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; // Get source and target node names from XML if (get_migration_node_names(history->xml, history->node, NULL, &source, &target) != pcmk_rc_ok) { return; } // Check for newer state on the source source_newer_op = non_monitor_after(history->rsc->id, source, history->xml, scheduler); // Check for a migrate_from action from this source on the target migrate_from = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_FROM, target, source, -1, scheduler); if (migrate_from != NULL) { if (source_newer_op) { /* There's a newer non-monitor operation on the source and a * migrate_from on the target, so this migrate_to is irrelevant to * the resource's state. */ return; } pcmk__xe_get_int(migrate_from, PCMK__XA_RC_CODE, &from_rc); pcmk__xe_get_int(migrate_from, PCMK__XA_OP_STATUS, &from_status); } /* If the resource has newer state on both the source and target after the * migration events, this migrate_to is irrelevant to the resource's state. */ target_newer_state = newer_state_after_migrate(history->rsc->id, target, history->xml, migrate_from, scheduler); if (source_newer_op && target_newer_state) { return; } /* Check for dangling migration (migrate_from succeeded but stop not done). * We know there's no stop because we already returned if the target has a * migrate_from and the source has any newer non-monitor operation. */ if ((from_rc == PCMK_OCF_OK) && (from_status == PCMK_EXEC_DONE)) { add_dangling_migration(history->rsc, history->node); return; } /* Without newer state, this migrate_to implies the resource is active. * (Clones are not allowed to migrate, so role can't be promoted.) */ history->rsc->priv->orig_role = pcmk_role_started; target_node = pcmk_find_node(scheduler, target); active_on_target = !target_newer_state && (target_node != NULL) && target_node->details->online; if (from_status != PCMK_EXEC_PENDING) { // migrate_from failed on target if (active_on_target) { native_add_running(history->rsc, target_node, scheduler, TRUE); } else { // Mark resource as failed, require recovery, and prevent migration pcmk__set_rsc_flags(history->rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); pcmk__clear_rsc_flags(history->rsc, pcmk__rsc_migratable); } return; } // The migrate_from is pending, complete but erased, or to be scheduled /* If there is no history at all for the resource on an online target, then * it was likely cleaned. Just return, and we'll schedule a probe. Once we * have the probe result, it will be reflected in target_newer_state. */ if ((target_node != NULL) && target_node->details->online && unknown_on_node(history->rsc, target)) { return; } if (active_on_target) { pcmk_node_t *source_node = pcmk_find_node(scheduler, source); native_add_running(history->rsc, target_node, scheduler, FALSE); if ((source_node != NULL) && source_node->details->online) { /* This is a partial migration: the migrate_to completed * successfully on the source, but the migrate_from has not * completed. Remember the source and target; if the newly * chosen target remains the same when we schedule actions * later, we may continue with the migration. */ history->rsc->priv->partial_migration_target = target_node; history->rsc->priv->partial_migration_source = source_node; } } else if (!source_newer_op) { // Mark resource as failed, require recovery, and prevent migration pcmk__set_rsc_flags(history->rsc, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); pcmk__clear_rsc_flags(history->rsc, pcmk__rsc_migratable); } } /*! * \internal * \brief Update resource role etc. after a failed migrate_to action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_to_failure(struct action_history *history) { xmlNode *target_migrate_from = NULL; const char *source = NULL; const char *target = NULL; pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; // Get source and target node names from XML if (get_migration_node_names(history->xml, history->node, NULL, &source, &target) != pcmk_rc_ok) { return; } /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ history->rsc->priv->orig_role = pcmk_role_started; // Check for migrate_from on the target target_migrate_from = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_FROM, target, source, PCMK_OCF_OK, scheduler); if (/* If the resource state is unknown on the target, it will likely be * probed there. * Don't just consider it running there. We will get back here anyway in * case the probe detects it's running there. */ !unknown_on_node(history->rsc, target) /* If the resource has newer state on the target after the migration * events, this migrate_to no longer matters for the target. */ && !newer_state_after_migrate(history->rsc->id, target, history->xml, target_migrate_from, scheduler)) { /* The resource has no newer state on the target, so assume it's still * active there. * (if it is up). */ pcmk_node_t *target_node = pcmk_find_node(scheduler, target); if (target_node && target_node->details->online) { native_add_running(history->rsc, target_node, scheduler, FALSE); } } else if (!non_monitor_after(history->rsc->id, source, history->xml, scheduler)) { /* We know the resource has newer state on the target, but this * migrate_to still matters for the source as long as there's no newer * non-monitor operation there. */ // Mark node as having dangling migration so we can force a stop later history->rsc->priv->dangling_migration_sources = g_list_prepend(history->rsc->priv->dangling_migration_sources, (gpointer) history->node); } } /*! * \internal * \brief Update resource role etc. after a failed migrate_from action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_from_failure(struct action_history *history) { xmlNode *source_migrate_to = NULL; const char *source = NULL; const char *target = NULL; pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; // Get source and target node names from XML if (get_migration_node_names(history->xml, NULL, history->node, &source, &target) != pcmk_rc_ok) { return; } /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ history->rsc->priv->orig_role = pcmk_role_started; // Check for a migrate_to on the source source_migrate_to = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_TO, source, target, PCMK_OCF_OK, scheduler); if (/* If the resource state is unknown on the source, it will likely be * probed there. * Don't just consider it running there. We will get back here anyway in * case the probe detects it's running there. */ !unknown_on_node(history->rsc, source) /* If the resource has newer state on the source after the migration * events, this migrate_from no longer matters for the source. */ && !newer_state_after_migrate(history->rsc->id, source, source_migrate_to, history->xml, scheduler)) { /* The resource has no newer state on the source, so assume it's still * active there (if it is up). */ pcmk_node_t *source_node = pcmk_find_node(scheduler, source); if (source_node && source_node->details->online) { native_add_running(history->rsc, source_node, scheduler, TRUE); } } } /*! * \internal * \brief Add an action to cluster's list of failed actions * * \param[in,out] history Parsed action result history */ static void record_failed_op(struct action_history *history) { const pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; if (!(history->node->details->online)) { return; } for (const xmlNode *xIter = scheduler->priv->failed->children; xIter != NULL; xIter = xIter->next) { const char *key = pcmk__xe_history_key(xIter); const char *uname = pcmk__xe_get(xIter, PCMK_XA_UNAME); if (pcmk__str_eq(history->key, key, pcmk__str_none) && pcmk__str_eq(uname, history->node->priv->name, pcmk__str_casei)) { crm_trace("Skipping duplicate entry %s on %s", history->key, pcmk__node_name(history->node)); return; } } crm_trace("Adding entry for %s on %s to failed action list", history->key, pcmk__node_name(history->node)); crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->priv->name); crm_xml_add(history->xml, PCMK__XA_RSC_ID, history->rsc->id); pcmk__xml_copy(scheduler->priv->failed, history->xml); } static char * last_change_str(const xmlNode *xml_op) { time_t when; char *result = NULL; if (pcmk__xe_get_time(xml_op, PCMK_XA_LAST_RC_CHANGE, &when) == pcmk_rc_ok) { char *when_s = pcmk__epoch2str(&when, 0); const char *p = strchr(when_s, ' '); // Skip day of week to make message shorter if ((p != NULL) && (*(++p) != '\0')) { result = pcmk__str_copy(p); } free(when_s); } if (result == NULL) { result = pcmk__str_copy("unknown_time"); } return result; } /*! * \internal * \brief Ban a resource (or its clone if an anonymous instance) from all nodes * * \param[in,out] rsc Resource to ban */ static void ban_from_all_nodes(pcmk_resource_t *rsc) { int score = -PCMK_SCORE_INFINITY; const pcmk_scheduler_t *scheduler = rsc->priv->scheduler; if (rsc->priv->parent != NULL) { pcmk_resource_t *parent = uber_parent(rsc); if (pcmk__is_anonymous_clone(parent)) { /* For anonymous clones, if an operation with * PCMK_META_ON_FAIL=PCMK_VALUE_STOP fails for any instance, the * entire clone must stop. */ rsc = parent; } } // Ban the resource from all nodes crm_notice("%s will not be started under current conditions", rsc->id); if (rsc->priv->allowed_nodes != NULL) { g_hash_table_destroy(rsc->priv->allowed_nodes); } rsc->priv->allowed_nodes = pe__node_list2table(scheduler->nodes); g_hash_table_foreach(rsc->priv->allowed_nodes, set_node_score, &score); } /*! * \internal * \brief Get configured failure handling and role after failure for an action * * \param[in,out] history Unpacked action history entry * \param[out] on_fail Where to set configured failure handling * \param[out] fail_role Where to set to role after failure */ static void unpack_failure_handling(struct action_history *history, enum pcmk__on_fail *on_fail, enum rsc_role_e *fail_role) { xmlNode *config = pcmk__find_action_config(history->rsc, history->task, history->interval_ms, true); GHashTable *meta = pcmk__unpack_action_meta(history->rsc, history->node, history->task, history->interval_ms, config); const char *on_fail_str = g_hash_table_lookup(meta, PCMK_META_ON_FAIL); *on_fail = pcmk__parse_on_fail(history->rsc, history->task, history->interval_ms, on_fail_str); *fail_role = pcmk__role_after_failure(history->rsc, history->task, *on_fail, meta); g_hash_table_destroy(meta); } /*! * \internal * \brief Update resource role, failure handling, etc., after a failed action * * \param[in,out] history Parsed action result history * \param[in] config_on_fail Action failure handling from configuration * \param[in] fail_role Resource's role after failure of this action * \param[out] last_failure This will be set to the history XML * \param[in,out] on_fail Actual handling of action result */ static void unpack_rsc_op_failure(struct action_history *history, enum pcmk__on_fail config_on_fail, enum rsc_role_e fail_role, xmlNode **last_failure, enum pcmk__on_fail *on_fail) { bool is_probe = false; char *last_change_s = NULL; pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; *last_failure = history->xml; is_probe = pcmk_xe_is_probe(history->xml); last_change_s = last_change_str(history->xml); if (!pcmk_is_set(scheduler->flags, pcmk__sched_symmetric_cluster) && (history->exit_status == PCMK_OCF_NOT_INSTALLED)) { crm_trace("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " QB_XS " exit-status=%d id=%s", crm_exit_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), (is_probe? "probe" : history->task), history->rsc->id, pcmk__node_name(history->node), last_change_s, history->exit_status, history->id); } else { pcmk__sched_warn(scheduler, "Unexpected result (%s%s%s) was recorded for %s of " "%s on %s at %s " QB_XS " exit-status=%d id=%s", crm_exit_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), (is_probe? "probe" : history->task), history->rsc->id, pcmk__node_name(history->node), last_change_s, history->exit_status, history->id); if (is_probe && (history->exit_status != PCMK_OCF_OK) && (history->exit_status != PCMK_OCF_NOT_RUNNING) && (history->exit_status != PCMK_OCF_RUNNING_PROMOTED)) { /* A failed (not just unexpected) probe result could mean the user * didn't know resources will be probed even where they can't run. */ crm_notice("If it is not possible for %s to run on %s, see " "the " PCMK_XA_RESOURCE_DISCOVERY " option for location " "constraints", history->rsc->id, pcmk__node_name(history->node)); } record_failed_op(history); } free(last_change_s); if (*on_fail < config_on_fail) { pcmk__rsc_trace(history->rsc, "on-fail %s -> %s for %s", pcmk__on_fail_text(*on_fail), pcmk__on_fail_text(config_on_fail), history->key); *on_fail = config_on_fail; } if (strcmp(history->task, PCMK_ACTION_STOP) == 0) { resource_location(history->rsc, history->node, -PCMK_SCORE_INFINITY, "__stop_fail__", scheduler); } else if (strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0) { unpack_migrate_to_failure(history); } else if (strcmp(history->task, PCMK_ACTION_MIGRATE_FROM) == 0) { unpack_migrate_from_failure(history); } else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) { history->rsc->priv->orig_role = pcmk_role_promoted; } else if (strcmp(history->task, PCMK_ACTION_DEMOTE) == 0) { if (config_on_fail == pcmk__on_fail_block) { history->rsc->priv->orig_role = pcmk_role_promoted; pe__set_next_role(history->rsc, pcmk_role_stopped, "demote with " PCMK_META_ON_FAIL "=block"); } else if (history->exit_status == PCMK_OCF_NOT_RUNNING) { history->rsc->priv->orig_role = pcmk_role_stopped; } else { /* Staying in the promoted role would put the scheduler and * controller into a loop. Setting the role to unpromoted is not * dangerous because the resource will be stopped as part of * recovery, and any promotion will be ordered after that stop. */ history->rsc->priv->orig_role = pcmk_role_unpromoted; } } if (is_probe && (history->exit_status == PCMK_OCF_NOT_INSTALLED)) { /* leave stopped */ pcmk__rsc_trace(history->rsc, "Leaving %s stopped", history->rsc->id); history->rsc->priv->orig_role = pcmk_role_stopped; } else if (history->rsc->priv->orig_role < pcmk_role_started) { pcmk__rsc_trace(history->rsc, "Setting %s active", history->rsc->id); set_active(history->rsc); } pcmk__rsc_trace(history->rsc, "Resource %s: role=%s unclean=%s on_fail=%s fail_role=%s", history->rsc->id, pcmk_role_text(history->rsc->priv->orig_role), pcmk__btoa(history->node->details->unclean), pcmk__on_fail_text(config_on_fail), pcmk_role_text(fail_role)); if ((fail_role != pcmk_role_started) && (history->rsc->priv->next_role < fail_role)) { pe__set_next_role(history->rsc, fail_role, "failure"); } if (fail_role == pcmk_role_stopped) { ban_from_all_nodes(history->rsc); } } /*! * \internal * \brief Block a resource with a failed action if it cannot be recovered * * If resource action is a failed stop and fencing is not possible, mark the * resource as unmanaged and blocked, since recovery cannot be done. * * \param[in,out] history Parsed action history entry */ static void block_if_unrecoverable(struct action_history *history) { char *last_change_s = NULL; if (strcmp(history->task, PCMK_ACTION_STOP) != 0) { return; // All actions besides stop are always recoverable } if (pe_can_fence(history->node->priv->scheduler, history->node)) { return; // Failed stops are recoverable via fencing } last_change_s = last_change_str(history->xml); pcmk__sched_err(history->node->priv->scheduler, "No further recovery can be attempted for %s " "because %s on %s failed (%s%s%s) at %s " QB_XS " rc=%d id=%s", history->rsc->id, history->task, pcmk__node_name(history->node), crm_exit_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), last_change_s, history->exit_status, history->id); free(last_change_s); pcmk__clear_rsc_flags(history->rsc, pcmk__rsc_managed); pcmk__set_rsc_flags(history->rsc, pcmk__rsc_blocked); } /*! * \internal * \brief Update action history's execution status and why * * \param[in,out] history Parsed action history entry * \param[out] why Where to store reason for update * \param[in] value New value * \param[in] reason Description of why value was changed */ static inline void remap_because(struct action_history *history, const char **why, int value, const char *reason) { if (history->execution_status != value) { history->execution_status = value; *why = reason; } } /*! * \internal * \brief Remap informational monitor results and operation status * * For the monitor results, certain OCF codes are for providing extended information * to the user about services that aren't yet failed but not entirely healthy either. * These must be treated as the "normal" result by Pacemaker. * * For operation status, the action result can be used to determine an appropriate * status for the purposes of responding to the action. The status provided by the * executor is not directly usable since the executor does not know what was expected. * * \param[in,out] history Parsed action history entry * \param[in,out] on_fail What should be done about the result * \param[in] expired Whether result is expired * * \note If the result is remapped and the node is not shutting down or failed, * the operation will be recorded in the scheduler data's list of failed * operations to highlight it for the user. * * \note This may update the resource's current and next role. */ static void remap_operation(struct action_history *history, enum pcmk__on_fail *on_fail, bool expired) { /* @TODO It would probably also be a good idea to map an exit status of * CRM_EX_PROMOTED or CRM_EX_DEGRADED_PROMOTED to CRM_EX_OK for promote * actions */ bool is_probe = false; int orig_exit_status = history->exit_status; int orig_exec_status = history->execution_status; const char *why = NULL; const char *task = history->task; // Remap degraded results to their successful counterparts history->exit_status = pcmk__effective_rc(history->exit_status); if (history->exit_status != orig_exit_status) { why = "degraded result"; if (!expired && (!history->node->details->shutdown || history->node->details->online)) { record_failed_op(history); } } if (!pcmk__is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml) && ((history->execution_status != PCMK_EXEC_DONE) || (history->exit_status != PCMK_OCF_NOT_RUNNING))) { history->execution_status = PCMK_EXEC_DONE; history->exit_status = PCMK_OCF_NOT_RUNNING; why = "equivalent probe result"; } /* If the executor reported an execution status of anything but done or * error, consider that final. But for done or error, we know better whether * it should be treated as a failure or not, because we know the expected * result. */ switch (history->execution_status) { case PCMK_EXEC_DONE: case PCMK_EXEC_ERROR: break; // These should be treated as node-fatal case PCMK_EXEC_NO_FENCE_DEVICE: case PCMK_EXEC_NO_SECRETS: remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "node-fatal error"); goto remap_done; default: goto remap_done; } is_probe = pcmk_xe_is_probe(history->xml); if (is_probe) { task = "probe"; } if (history->expected_exit_status < 0) { /* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with * Heartbeat 2.0.7 or earlier as the cluster layer, did not include the * expected exit status in the transition key, which (along with the * similar case of a corrupted transition key in the CIB) will be * reported to this function as -1. Pacemaker 2.0+ does not support * rolling upgrades from those versions or processing of saved CIB files * from those versions, so we do not need to care much about this case. */ remap_because(history, &why, PCMK_EXEC_ERROR, "obsolete history format"); pcmk__config_warn("Expected result not found for %s on %s " "(corrupt or obsolete CIB?)", history->key, pcmk__node_name(history->node)); } else if (history->exit_status == history->expected_exit_status) { remap_because(history, &why, PCMK_EXEC_DONE, "expected result"); } else { remap_because(history, &why, PCMK_EXEC_ERROR, "unexpected result"); pcmk__rsc_debug(history->rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)", history->key, pcmk__node_name(history->node), history->expected_exit_status, crm_exit_str(history->expected_exit_status), history->exit_status, crm_exit_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, "")); } switch (history->exit_status) { case PCMK_OCF_OK: if (is_probe && (history->expected_exit_status == PCMK_OCF_NOT_RUNNING)) { char *last_change_s = last_change_str(history->xml); remap_because(history, &why, PCMK_EXEC_DONE, "probe"); pcmk__rsc_info(history->rsc, "Probe found %s active on %s at %s", history->rsc->id, pcmk__node_name(history->node), last_change_s); free(last_change_s); } break; case PCMK_OCF_NOT_RUNNING: if (is_probe || (history->expected_exit_status == history->exit_status) || !pcmk_is_set(history->rsc->flags, pcmk__rsc_managed)) { /* For probes, recurring monitors for the Stopped role, and * unmanaged resources, "not running" is not considered a * failure. */ remap_because(history, &why, PCMK_EXEC_DONE, "exit status"); history->rsc->priv->orig_role = pcmk_role_stopped; *on_fail = pcmk__on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "not running"); } break; case PCMK_OCF_RUNNING_PROMOTED: if (is_probe && (history->exit_status != history->expected_exit_status)) { char *last_change_s = last_change_str(history->xml); remap_because(history, &why, PCMK_EXEC_DONE, "probe"); pcmk__rsc_info(history->rsc, "Probe found %s active and promoted on %s at %s", history->rsc->id, pcmk__node_name(history->node), last_change_s); free(last_change_s); } if (!expired || (history->exit_status == history->expected_exit_status)) { history->rsc->priv->orig_role = pcmk_role_promoted; } break; case PCMK_OCF_FAILED_PROMOTED: if (!expired) { history->rsc->priv->orig_role = pcmk_role_promoted; } remap_because(history, &why, PCMK_EXEC_ERROR, "exit status"); break; case PCMK_OCF_NOT_CONFIGURED: remap_because(history, &why, PCMK_EXEC_ERROR_FATAL, "exit status"); break; case PCMK_OCF_UNIMPLEMENT_FEATURE: { guint interval_ms = 0; pcmk__xe_get_guint(history->xml, PCMK_META_INTERVAL, &interval_ms); if (interval_ms == 0) { if (!expired) { block_if_unrecoverable(history); } remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "exit status"); } else { remap_because(history, &why, PCMK_EXEC_NOT_SUPPORTED, "exit status"); } } break; case PCMK_OCF_NOT_INSTALLED: case PCMK_OCF_INVALID_PARAM: case PCMK_OCF_INSUFFICIENT_PRIV: if (!expired) { block_if_unrecoverable(history); } remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "exit status"); break; default: if (history->execution_status == PCMK_EXEC_DONE) { char *last_change_s = last_change_str(history->xml); crm_info("Treating unknown exit status %d from %s of %s " "on %s at %s as failure", history->exit_status, task, history->rsc->id, pcmk__node_name(history->node), last_change_s); remap_because(history, &why, PCMK_EXEC_ERROR, "unknown exit status"); free(last_change_s); } break; } remap_done: if (why != NULL) { pcmk__rsc_trace(history->rsc, "Remapped %s result from [%s: %s] to [%s: %s] " "because of %s", history->key, pcmk_exec_status_str(orig_exec_status), crm_exit_str(orig_exit_status), pcmk_exec_status_str(history->execution_status), crm_exit_str(history->exit_status), why); } } // return TRUE if start or monitor last failure but parameters changed static bool should_clear_for_param_change(const xmlNode *xml_op, const char *task, pcmk_resource_t *rsc, pcmk_node_t *node) { if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_MONITOR, NULL)) { if (pe__bundle_needs_remote_name(rsc)) { /* We haven't allocated resources yet, so we can't reliably * substitute addr parameters for the REMOTE_CONTAINER_HACK. * When that's needed, defer the check until later. */ pcmk__add_param_check(xml_op, rsc, node, pcmk__check_last_failure); } else { pcmk__op_digest_t *digest_data = NULL; digest_data = rsc_action_digest_cmp(rsc, xml_op, node, rsc->priv->scheduler); switch (digest_data->rc) { case pcmk__digest_unknown: crm_trace("Resource %s history entry %s on %s" " has no digest to compare", rsc->id, pcmk__xe_history_key(xml_op), node->priv->id); break; case pcmk__digest_match: break; default: return TRUE; } } } return FALSE; } // Order action after fencing of remote node, given connection rsc static void order_after_remote_fencing(pcmk_action_t *action, pcmk_resource_t *remote_conn, pcmk_scheduler_t *scheduler) { pcmk_node_t *remote_node = pcmk_find_node(scheduler, remote_conn->id); if (remote_node) { pcmk_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL, FALSE, scheduler); order_actions(fence, action, pcmk__ar_first_implies_then); } } static bool should_ignore_failure_timeout(const pcmk_resource_t *rsc, const char *task, guint interval_ms, bool is_last_failure) { /* Clearing failures of recurring monitors has special concerns. The * executor reports only changes in the monitor result, so if the * monitor is still active and still getting the same failure result, * that will go undetected after the failure is cleared. * * Also, the operation history will have the time when the recurring * monitor result changed to the given code, not the time when the * result last happened. * * @TODO We probably should clear such failures only when the failure * timeout has passed since the last occurrence of the failed result. * However we don't record that information. We could maybe approximate * that by clearing only if there is a more recent successful monitor or * stop result, but we don't even have that information at this point * since we are still unpacking the resource's operation history. * * This is especially important for remote connection resources with a * reconnect interval, so in that case, we skip clearing failures * if the remote node hasn't been fenced. */ if ((rsc->priv->remote_reconnect_ms > 0U) && pcmk_is_set(rsc->priv->scheduler->flags, pcmk__sched_fencing_enabled) && (interval_ms != 0) && pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei)) { pcmk_node_t *remote_node = pcmk_find_node(rsc->priv->scheduler, rsc->id); if (remote_node && !pcmk_is_set(remote_node->priv->flags, pcmk__node_remote_fenced)) { if (is_last_failure) { crm_info("Waiting to clear monitor failure for remote node %s" " until fencing has occurred", rsc->id); } return TRUE; } } return FALSE; } /*! * \internal * \brief Check operation age and schedule failure clearing when appropriate * * This function has two distinct purposes. The first is to check whether an * operation history entry is expired (i.e. the resource has a failure timeout, * the entry is older than the timeout, and the resource either has no fail * count or its fail count is entirely older than the timeout). The second is to * schedule fail count clearing when appropriate (i.e. the operation is expired * and either the resource has an expired fail count or the operation is a * last_failure for a remote connection resource with a reconnect interval, * or the operation is a last_failure for a start or monitor operation and the * resource's parameters have changed since the operation). * * \param[in,out] history Parsed action result history * * \return true if operation history entry is expired, otherwise false */ static bool check_operation_expiry(struct action_history *history) { bool expired = false; - bool is_last_failure = pcmk__ends_with(history->id, "_last_failure_0"); + bool is_last_failure = (history->id != NULL) + && g_str_has_suffix(history->id, "_last_failure_0"); time_t last_run = 0; int unexpired_fail_count = 0; const char *clear_reason = NULL; const guint expiration_sec = pcmk__timeout_ms2s(history->rsc->priv->failure_expiration_ms); pcmk_scheduler_t *scheduler = history->rsc->priv->scheduler; if (history->execution_status == PCMK_EXEC_NOT_INSTALLED) { pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not expired: " "Not Installed does not expire", history->id, pcmk__node_name(history->node)); return false; // "Not installed" must always be cleared manually } if ((expiration_sec > 0) && (pcmk__xe_get_time(history->xml, PCMK_XA_LAST_RC_CHANGE, &last_run) == pcmk_rc_ok)) { /* Resource has a PCMK_META_FAILURE_TIMEOUT and history entry has a * timestamp */ time_t now = pcmk__scheduler_epoch_time(scheduler); time_t last_failure = 0; // Is this particular operation history older than the failure timeout? if ((now >= (last_run + expiration_sec)) && !should_ignore_failure_timeout(history->rsc, history->task, history->interval_ms, is_last_failure)) { expired = true; } // Does the resource as a whole have an unexpired fail count? unexpired_fail_count = pe_get_failcount(history->node, history->rsc, &last_failure, pcmk__fc_effective, history->xml); // Update scheduler recheck time according to *last* failure crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d " "expiration=%s last-failure@%lld", history->id, (long long) last_run, (expired? "" : "not "), (long long) now, unexpired_fail_count, pcmk__readable_interval(expiration_sec * 1000), (long long) last_failure); last_failure += expiration_sec + 1; if (unexpired_fail_count && (now < last_failure)) { pcmk__update_recheck_time(last_failure, scheduler, "fail count expiration"); } } if (expired) { if (pe_get_failcount(history->node, history->rsc, NULL, pcmk__fc_default, history->xml)) { // There is a fail count ignoring timeout if (unexpired_fail_count == 0) { // There is no fail count considering timeout clear_reason = "it expired"; } else { /* This operation is old, but there is an unexpired fail count. * In a properly functioning cluster, this should only be * possible if this operation is not a failure (otherwise the * fail count should be expired too), so this is really just a * failsafe. */ pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not " "expired: Unexpired fail count", history->id, pcmk__node_name(history->node)); expired = false; } } else if (is_last_failure && (history->rsc->priv->remote_reconnect_ms > 0U)) { /* Clear any expired last failure when reconnect interval is set, * even if there is no fail count. */ clear_reason = "reconnect interval is set"; } } if (!expired && is_last_failure && should_clear_for_param_change(history->xml, history->task, history->rsc, history->node)) { clear_reason = "resource parameters have changed"; } if (clear_reason != NULL) { pcmk_action_t *clear_op = NULL; // Schedule clearing of the fail count clear_op = pe__clear_failcount(history->rsc, history->node, clear_reason, scheduler); if (pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled) && (history->rsc->priv->remote_reconnect_ms > 0)) { /* If we're clearing a remote connection due to a reconnect * interval, we want to wait until any scheduled fencing * completes. * * We could limit this to remote_node->details->unclean, but at * this point, that's always true (it won't be reliable until * after unpack_node_history() is done). */ crm_info("Clearing %s failure will wait until any scheduled " "fencing of %s completes", history->task, history->rsc->id); order_after_remote_fencing(clear_op, history->rsc, scheduler); } } if (expired && (history->interval_ms == 0) && pcmk__str_eq(history->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { switch (history->exit_status) { case PCMK_OCF_OK: case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Don't expire probes that return these values pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not " "expired: Probe result", history->id, pcmk__node_name(history->node)); expired = false; break; } } return expired; } int pe__target_rc_from_xml(const xmlNode *xml_op) { int target_rc = 0; const char *key = pcmk__xe_get(xml_op, PCMK__XA_TRANSITION_KEY); if (key == NULL) { return -1; } decode_transition_key(key, NULL, NULL, NULL, &target_rc); return target_rc; } /*! * \internal * \brief Update a resource's state for an action result * * \param[in,out] history Parsed action history entry * \param[in] exit_status Exit status to base new state on * \param[in] last_failure Resource's last_failure entry, if known * \param[in,out] on_fail Resource's current failure handling */ static void update_resource_state(struct action_history *history, int exit_status, const xmlNode *last_failure, enum pcmk__on_fail *on_fail) { bool clear_past_failure = false; if ((exit_status == PCMK_OCF_NOT_INSTALLED) || (!pcmk__is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml))) { history->rsc->priv->orig_role = pcmk_role_stopped; } else if (exit_status == PCMK_OCF_NOT_RUNNING) { clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { if ((last_failure != NULL) && pcmk__str_eq(history->key, pcmk__xe_history_key(last_failure), pcmk__str_none)) { clear_past_failure = true; } if (history->rsc->priv->orig_role < pcmk_role_started) { set_active(history->rsc); } } else if (pcmk__str_eq(history->task, PCMK_ACTION_START, pcmk__str_none)) { history->rsc->priv->orig_role = pcmk_role_started; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_STOP, pcmk__str_none)) { history->rsc->priv->orig_role = pcmk_role_stopped; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_PROMOTE, pcmk__str_none)) { history->rsc->priv->orig_role = pcmk_role_promoted; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_DEMOTE, pcmk__str_none)) { if (*on_fail == pcmk__on_fail_demote) { /* Demote clears an error only if * PCMK_META_ON_FAIL=PCMK_VALUE_DEMOTE */ clear_past_failure = true; } history->rsc->priv->orig_role = pcmk_role_unpromoted; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_FROM, pcmk__str_none)) { history->rsc->priv->orig_role = pcmk_role_started; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_TO, pcmk__str_none)) { unpack_migrate_to_success(history); } else if (history->rsc->priv->orig_role < pcmk_role_started) { pcmk__rsc_trace(history->rsc, "%s active on %s", history->rsc->id, pcmk__node_name(history->node)); set_active(history->rsc); } if (!clear_past_failure) { return; } switch (*on_fail) { case pcmk__on_fail_stop: case pcmk__on_fail_ban: case pcmk__on_fail_standby_node: case pcmk__on_fail_fence_node: pcmk__rsc_trace(history->rsc, "%s (%s) is not cleared by a completed %s", history->rsc->id, pcmk__on_fail_text(*on_fail), history->task); break; case pcmk__on_fail_block: case pcmk__on_fail_ignore: case pcmk__on_fail_demote: case pcmk__on_fail_restart: case pcmk__on_fail_restart_container: *on_fail = pcmk__on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "clear past failures"); break; case pcmk__on_fail_reset_remote: if (history->rsc->priv->remote_reconnect_ms == 0U) { /* With no reconnect interval, the connection is allowed to * start again after the remote node is fenced and * completely stopped. (With a reconnect interval, we wait * for the failure to be cleared entirely before attempting * to reconnect.) */ *on_fail = pcmk__on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "clear past failures and reset remote"); } break; } } /*! * \internal * \brief Check whether a given history entry matters for resource state * * \param[in] history Parsed action history entry * * \return true if action can affect resource state, otherwise false */ static inline bool can_affect_state(struct action_history *history) { return pcmk__str_any_of(history->task, PCMK_ACTION_MONITOR, PCMK_ACTION_START, PCMK_ACTION_STOP, PCMK_ACTION_PROMOTE, PCMK_ACTION_DEMOTE, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, "asyncmon", NULL); } /*! * \internal * \brief Unpack execution/exit status and exit reason from a history entry * * \param[in,out] history Action history entry to unpack * * \return Standard Pacemaker return code */ static int unpack_action_result(struct action_history *history) { if ((pcmk__xe_get_int(history->xml, PCMK__XA_OP_STATUS, &(history->execution_status)) != pcmk_rc_ok) || (history->execution_status < PCMK_EXEC_PENDING) || (history->execution_status > PCMK_EXEC_MAX) || (history->execution_status == PCMK_EXEC_CANCELLED)) { pcmk__config_err("Ignoring resource history entry %s for %s on %s " "with invalid " PCMK__XA_OP_STATUS " '%s'", history->id, history->rsc->id, pcmk__node_name(history->node), pcmk__s(pcmk__xe_get(history->xml, PCMK__XA_OP_STATUS), "")); return pcmk_rc_unpack_error; } if ((pcmk__xe_get_int(history->xml, PCMK__XA_RC_CODE, &(history->exit_status)) != pcmk_rc_ok) || (history->exit_status < 0) || (history->exit_status > CRM_EX_MAX)) { pcmk__config_err("Ignoring resource history entry %s for %s on %s " "with invalid " PCMK__XA_RC_CODE " '%s'", history->id, history->rsc->id, pcmk__node_name(history->node), pcmk__s(pcmk__xe_get(history->xml, PCMK__XA_RC_CODE), "")); return pcmk_rc_unpack_error; } history->exit_reason = pcmk__xe_get(history->xml, PCMK_XA_EXIT_REASON); return pcmk_rc_ok; } /*! * \internal * \brief Process an action history entry whose result expired * * \param[in,out] history Parsed action history entry * \param[in] orig_exit_status Action exit status before remapping * * \return Standard Pacemaker return code (in particular, pcmk_rc_ok means the * entry needs no further processing) */ static int process_expired_result(struct action_history *history, int orig_exit_status) { if (!pcmk__is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml) && (orig_exit_status != history->expected_exit_status)) { if (history->rsc->priv->orig_role <= pcmk_role_stopped) { history->rsc->priv->orig_role = pcmk_role_unknown; } crm_trace("Ignoring resource history entry %s for probe of %s on %s: " "Masked failure expired", history->id, history->rsc->id, pcmk__node_name(history->node)); return pcmk_rc_ok; } if (history->exit_status == history->expected_exit_status) { return pcmk_rc_undetermined; // Only failures expire } if (history->interval_ms == 0) { crm_notice("Ignoring resource history entry %s for %s of %s on %s: " "Expired failure", history->id, history->task, history->rsc->id, pcmk__node_name(history->node)); return pcmk_rc_ok; } if (history->node->details->online && !history->node->details->unclean) { /* Reschedule the recurring action. schedule_cancel() won't work at * this stage, so as a hacky workaround, forcibly change the restart * digest so pcmk__check_action_config() does what we want later. * * @TODO We should skip this if there is a newer successful monitor. * Also, this causes rescheduling only if the history entry * has a PCMK__XA_OP_DIGEST (which the expire-non-blocked-failure * scheduler regression test doesn't, but that may not be a * realistic scenario in production). */ crm_notice("Rescheduling %s-interval %s of %s on %s " "after failure expired", pcmk__readable_interval(history->interval_ms), history->task, history->rsc->id, pcmk__node_name(history->node)); crm_xml_add(history->xml, PCMK__XA_OP_RESTART_DIGEST, "calculated-failure-timeout"); return pcmk_rc_ok; } return pcmk_rc_undetermined; } /*! * \internal * \brief Process a masked probe failure * * \param[in,out] history Parsed action history entry * \param[in] orig_exit_status Action exit status before remapping * \param[in] last_failure Resource's last_failure entry, if known * \param[in,out] on_fail Resource's current failure handling */ static void mask_probe_failure(struct action_history *history, int orig_exit_status, const xmlNode *last_failure, enum pcmk__on_fail *on_fail) { pcmk_resource_t *ban_rsc = history->rsc; if (!pcmk_is_set(history->rsc->flags, pcmk__rsc_unique)) { ban_rsc = uber_parent(history->rsc); } crm_notice("Treating probe result '%s' for %s on %s as 'not running'", crm_exit_str(orig_exit_status), history->rsc->id, pcmk__node_name(history->node)); update_resource_state(history, history->expected_exit_status, last_failure, on_fail); crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->priv->name); record_failed_op(history); resource_location(ban_rsc, history->node, -PCMK_SCORE_INFINITY, "masked-probe-failure", ban_rsc->priv->scheduler); } /*! * \internal Check whether a given failure is for a given pending action * * \param[in] history Parsed history entry for pending action * \param[in] last_failure Resource's last_failure entry, if known * * \return true if \p last_failure is failure of pending action in \p history, * otherwise false * \note Both \p history and \p last_failure must come from the same * \c PCMK__XE_LRM_RESOURCE block, as node and resource are assumed to be * the same. */ static bool failure_is_newer(const struct action_history *history, const xmlNode *last_failure) { guint failure_interval_ms = 0U; long long failure_change = 0LL; long long this_change = 0LL; if (last_failure == NULL) { return false; // Resource has no last_failure entry } if (!pcmk__str_eq(history->task, pcmk__xe_get(last_failure, PCMK_XA_OPERATION), pcmk__str_none)) { return false; // last_failure is for different action } if ((pcmk__xe_get_guint(last_failure, PCMK_META_INTERVAL, &failure_interval_ms) != pcmk_rc_ok) || (history->interval_ms != failure_interval_ms)) { return false; // last_failure is for action with different interval } if ((pcmk__scan_ll(pcmk__xe_get(history->xml, PCMK_XA_LAST_RC_CHANGE), &this_change, 0LL) != pcmk_rc_ok) || (pcmk__scan_ll(pcmk__xe_get(last_failure, PCMK_XA_LAST_RC_CHANGE), &failure_change, 0LL) != pcmk_rc_ok) || (failure_change < this_change)) { return false; // Failure is not known to be newer } return true; } /*! * \internal * \brief Update a resource's role etc. for a pending action * * \param[in,out] history Parsed history entry for pending action * \param[in] last_failure Resource's last_failure entry, if known */ static void process_pending_action(struct action_history *history, const xmlNode *last_failure) { /* For recurring monitors, a failure is recorded only in RSC_last_failure_0, * and there might be a RSC_monitor_INTERVAL entry with the last successful * or pending result. * * If last_failure contains the failure of the pending recurring monitor * we're processing here, and is newer, the action is no longer pending. * (Pending results have call ID -1, which sorts last, so the last failure * if any should be known.) */ if (failure_is_newer(history, last_failure)) { return; } if (strcmp(history->task, PCMK_ACTION_START) == 0) { pcmk__set_rsc_flags(history->rsc, pcmk__rsc_start_pending); set_active(history->rsc); } else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) { history->rsc->priv->orig_role = pcmk_role_promoted; } else if ((strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0) && history->node->details->unclean) { /* A migrate_to action is pending on a unclean source, so force a stop * on the target. */ const char *migrate_target = NULL; pcmk_node_t *target = NULL; migrate_target = pcmk__xe_get(history->xml, PCMK__META_MIGRATE_TARGET); target = pcmk_find_node(history->rsc->priv->scheduler, migrate_target); if (target != NULL) { stop_action(history->rsc, target, FALSE); } } if (history->rsc->priv->pending_action != NULL) { /* There should never be multiple pending actions, but as a failsafe, * just remember the first one processed for display purposes. */ return; } if (pcmk_is_probe(history->task, history->interval_ms)) { /* Pending probes are currently never displayed, even if pending * operations are requested. If we ever want to change that, * enable the below and the corresponding part of * native.c:native_pending_action(). */ #if 0 history->rsc->private->pending_action = strdup("probe"); history->rsc->private->pending_node = history->node; #endif } else { history->rsc->priv->pending_action = strdup(history->task); history->rsc->priv->pending_node = history->node; } } static void unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum pcmk__on_fail *on_fail) { int old_rc = 0; bool expired = false; pcmk_resource_t *parent = rsc; enum rsc_role_e fail_role = pcmk_role_unknown; enum pcmk__on_fail failure_strategy = pcmk__on_fail_restart; struct action_history history = { .rsc = rsc, .node = node, .xml = xml_op, .execution_status = PCMK_EXEC_UNKNOWN, }; CRM_CHECK(rsc && node && xml_op, return); history.id = pcmk__xe_id(xml_op); if (history.id == NULL) { pcmk__config_err("Ignoring resource history entry for %s on %s " "without ID", rsc->id, pcmk__node_name(node)); return; } // Task and interval history.task = pcmk__xe_get(xml_op, PCMK_XA_OPERATION); if (history.task == NULL) { pcmk__config_err("Ignoring resource history entry %s for %s on %s " "without " PCMK_XA_OPERATION, history.id, rsc->id, pcmk__node_name(node)); return; } pcmk__xe_get_guint(xml_op, PCMK_META_INTERVAL, &(history.interval_ms)); if (!can_affect_state(&history)) { pcmk__rsc_trace(rsc, "Ignoring resource history entry %s for %s on %s " "with irrelevant action '%s'", history.id, rsc->id, pcmk__node_name(node), history.task); return; } if (unpack_action_result(&history) != pcmk_rc_ok) { return; // Error already logged } history.expected_exit_status = pe__target_rc_from_xml(xml_op); history.key = pcmk__xe_history_key(xml_op); pcmk__xe_get_int(xml_op, PCMK__XA_CALL_ID, &(history.call_id)); pcmk__rsc_trace(rsc, "Unpacking %s (%s call %d on %s): %s (%s)", history.id, history.task, history.call_id, pcmk__node_name(node), pcmk_exec_status_str(history.execution_status), crm_exit_str(history.exit_status)); if (node->details->unclean) { pcmk__rsc_trace(rsc, "%s is running on %s, which is unclean (further action " "depends on value of stop's on-fail attribute)", rsc->id, pcmk__node_name(node)); } expired = check_operation_expiry(&history); old_rc = history.exit_status; remap_operation(&history, on_fail, expired); if (expired && (process_expired_result(&history, old_rc) == pcmk_rc_ok)) { goto done; } if (!pcmk__is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op)) { mask_probe_failure(&history, old_rc, *last_failure, on_fail); goto done; } if (!pcmk_is_set(rsc->flags, pcmk__rsc_unique)) { parent = uber_parent(rsc); } switch (history.execution_status) { case PCMK_EXEC_PENDING: process_pending_action(&history, *last_failure); goto done; case PCMK_EXEC_DONE: update_resource_state(&history, history.exit_status, *last_failure, on_fail); goto done; case PCMK_EXEC_NOT_INSTALLED: unpack_failure_handling(&history, &failure_strategy, &fail_role); if (failure_strategy == pcmk__on_fail_ignore) { crm_warn("Cannot ignore failed %s of %s on %s: " "Resource agent doesn't exist " QB_XS " status=%d rc=%d id=%s", history.task, rsc->id, pcmk__node_name(node), history.execution_status, history.exit_status, history.id); /* Also for printing it as "FAILED" by marking it as * pcmk__rsc_failed later */ *on_fail = pcmk__on_fail_ban; } resource_location(parent, node, -PCMK_SCORE_INFINITY, "hard-error", rsc->priv->scheduler); unpack_rsc_op_failure(&history, failure_strategy, fail_role, last_failure, on_fail); goto done; case PCMK_EXEC_NOT_CONNECTED: if (pcmk__is_pacemaker_remote_node(node) && pcmk_is_set(node->priv->remote->flags, pcmk__rsc_managed)) { /* We should never get into a situation where a managed remote * connection resource is considered OK but a resource action * behind the connection gets a "not connected" status. But as a * fail-safe in case a bug or unusual circumstances do lead to * that, ensure the remote connection is considered failed. */ pcmk__set_rsc_flags(node->priv->remote, pcmk__rsc_failed|pcmk__rsc_stop_if_failed); } break; // Not done, do error handling case PCMK_EXEC_ERROR: case PCMK_EXEC_ERROR_HARD: case PCMK_EXEC_ERROR_FATAL: case PCMK_EXEC_TIMEOUT: case PCMK_EXEC_NOT_SUPPORTED: case PCMK_EXEC_INVALID: break; // Not done, do error handling default: // No other value should be possible at this point break; } unpack_failure_handling(&history, &failure_strategy, &fail_role); if ((failure_strategy == pcmk__on_fail_ignore) || ((failure_strategy == pcmk__on_fail_restart_container) && (strcmp(history.task, PCMK_ACTION_STOP) == 0))) { char *last_change_s = last_change_str(xml_op); crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s succeeded " QB_XS " %s", history.task, crm_exit_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), rsc->id, pcmk__node_name(node), last_change_s, history.id); free(last_change_s); update_resource_state(&history, history.expected_exit_status, *last_failure, on_fail); crm_xml_add(xml_op, PCMK_XA_UNAME, node->priv->name); pcmk__set_rsc_flags(rsc, pcmk__rsc_ignore_failure); record_failed_op(&history); if ((failure_strategy == pcmk__on_fail_restart_container) && (*on_fail <= pcmk__on_fail_restart)) { *on_fail = failure_strategy; } } else { unpack_rsc_op_failure(&history, failure_strategy, fail_role, last_failure, on_fail); if (history.execution_status == PCMK_EXEC_ERROR_HARD) { uint8_t log_level = LOG_ERR; if (history.exit_status == PCMK_OCF_NOT_INSTALLED) { log_level = LOG_NOTICE; } do_crm_log(log_level, "Preventing %s from restarting on %s because " "of hard failure (%s%s%s) " QB_XS " %s", parent->id, pcmk__node_name(node), crm_exit_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), history.id); resource_location(parent, node, -PCMK_SCORE_INFINITY, "hard-error", rsc->priv->scheduler); } else if (history.execution_status == PCMK_EXEC_ERROR_FATAL) { pcmk__sched_err(rsc->priv->scheduler, "Preventing %s from restarting anywhere because " "of fatal failure (%s%s%s) " QB_XS " %s", parent->id, crm_exit_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), history.id); resource_location(parent, NULL, -PCMK_SCORE_INFINITY, "fatal-error", rsc->priv->scheduler); } } done: pcmk__rsc_trace(rsc, "%s role on %s after %s is %s (next %s)", rsc->id, pcmk__node_name(node), history.id, pcmk_role_text(rsc->priv->orig_role), pcmk_role_text(rsc->priv->next_role)); } /*! * \internal * \brief Insert a node attribute with value into a \c GHashTable * * \param[in,out] key Key to insert (either freed or owned by * \p user_data upon return) * \param[in] value Value to insert (owned by \p user_data upon return) * \param[in] user_data \c GHashTable to insert into */ static gboolean insert_attr(gpointer key, gpointer value, gpointer user_data) { GHashTable *table = user_data; g_hash_table_insert(table, key, value); return TRUE; } static void add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node, bool overwrite, pcmk_scheduler_t *scheduler) { const char *cluster_name = NULL; const char *dc_id = pcmk__xe_get(scheduler->input, PCMK_XA_DC_UUID); const pcmk_rule_input_t rule_input = { .now = scheduler->priv->now, }; pcmk__insert_dup(node->priv->attrs, CRM_ATTR_UNAME, node->priv->name); pcmk__insert_dup(node->priv->attrs, CRM_ATTR_ID, node->priv->id); if ((scheduler->dc_node == NULL) && pcmk__str_eq(node->priv->id, dc_id, pcmk__str_casei)) { scheduler->dc_node = node; pcmk__insert_dup(node->priv->attrs, CRM_ATTR_IS_DC, PCMK_VALUE_TRUE); } else if (!pcmk__same_node(node, scheduler->dc_node)) { pcmk__insert_dup(node->priv->attrs, CRM_ATTR_IS_DC, PCMK_VALUE_FALSE); } cluster_name = g_hash_table_lookup(scheduler->priv->options, PCMK_OPT_CLUSTER_NAME); if (cluster_name) { pcmk__insert_dup(node->priv->attrs, CRM_ATTR_CLUSTER_NAME, cluster_name); } if (overwrite) { /* @TODO Try to reorder some unpacking so that we don't need the * overwrite argument or to unpack into a temporary table */ GHashTable *unpacked = pcmk__strkey_table(free, free); pe__unpack_dataset_nvpairs(xml_obj, PCMK_XE_INSTANCE_ATTRIBUTES, &rule_input, unpacked, NULL, scheduler); g_hash_table_foreach_steal(unpacked, insert_attr, node->priv->attrs); g_hash_table_destroy(unpacked); } else { pe__unpack_dataset_nvpairs(xml_obj, PCMK_XE_INSTANCE_ATTRIBUTES, &rule_input, node->priv->attrs, NULL, scheduler); } pe__unpack_dataset_nvpairs(xml_obj, PCMK_XE_UTILIZATION, &rule_input, node->priv->utilization, NULL, scheduler); if (pcmk__node_attr(node, CRM_ATTR_SITE_NAME, NULL, pcmk__rsc_node_current) == NULL) { const char *site_name = pcmk__node_attr(node, "site-name", NULL, pcmk__rsc_node_current); if (site_name) { pcmk__insert_dup(node->priv->attrs, CRM_ATTR_SITE_NAME, site_name); } else if (cluster_name) { /* Default to cluster-name if unset */ pcmk__insert_dup(node->priv->attrs, CRM_ATTR_SITE_NAME, cluster_name); } } } static GList * extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter) { int counter = -1; int stop_index = -1; int start_index = -1; xmlNode *rsc_op = NULL; GList *gIter = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; /* extract operations */ op_list = NULL; sorted_op_list = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry, PCMK__XE_LRM_RSC_OP, NULL, NULL); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op, PCMK__XE_LRM_RSC_OP)) { crm_xml_add(rsc_op, PCMK_XA_RESOURCE, rsc); crm_xml_add(rsc_op, PCMK_XA_UNAME, node); op_list = g_list_prepend(op_list, rsc_op); } if (op_list == NULL) { /* if there are no operations, there is nothing to do */ return NULL; } sorted_op_list = g_list_sort(op_list, sort_op_by_callid); /* create active recurring operations as optional */ if (active_filter == FALSE) { return sorted_op_list; } op_list = NULL; calculate_active_ops(sorted_op_list, &start_index, &stop_index); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; if (start_index < stop_index) { crm_trace("Skipping %s: not active", pcmk__xe_id(rsc_entry)); break; } else if (counter < start_index) { crm_trace("Skipping %s: old", pcmk__xe_id(rsc_op)); continue; } op_list = g_list_append(op_list, rsc_op); } g_list_free(sorted_op_list); return op_list; } GList * find_operations(const char *rsc, const char *node, gboolean active_filter, pcmk_scheduler_t *scheduler) { GList *output = NULL; GList *intermediate = NULL; xmlNode *tmp = NULL; xmlNode *status = pcmk__xe_first_child(scheduler->input, PCMK_XE_STATUS, NULL, NULL); pcmk_node_t *this_node = NULL; xmlNode *node_state = NULL; CRM_CHECK(status != NULL, return NULL); for (node_state = pcmk__xe_first_child(status, PCMK__XE_NODE_STATE, NULL, NULL); node_state != NULL; node_state = pcmk__xe_next(node_state, PCMK__XE_NODE_STATE)) { const char *uname = pcmk__xe_get(node_state, PCMK_XA_UNAME); if (node != NULL && !pcmk__str_eq(uname, node, pcmk__str_casei)) { continue; } this_node = pcmk_find_node(scheduler, uname); if(this_node == NULL) { CRM_LOG_ASSERT(this_node != NULL); continue; } else if (pcmk__is_pacemaker_remote_node(this_node)) { determine_remote_online_status(scheduler, this_node); } else { determine_online_status(node_state, this_node, scheduler); } if (this_node->details->online || pcmk_is_set(scheduler->flags, pcmk__sched_fencing_enabled)) { /* offline nodes run no resources... * unless stonith is enabled in which case we need to * make sure rsc start events happen after the stonith */ xmlNode *lrm_rsc = NULL; tmp = pcmk__xe_first_child(node_state, PCMK__XE_LRM, NULL, NULL); tmp = pcmk__xe_first_child(tmp, PCMK__XE_LRM_RESOURCES, NULL, NULL); for (lrm_rsc = pcmk__xe_first_child(tmp, PCMK__XE_LRM_RESOURCE, NULL, NULL); lrm_rsc != NULL; lrm_rsc = pcmk__xe_next(lrm_rsc, PCMK__XE_LRM_RESOURCE)) { const char *rsc_id = pcmk__xe_get(lrm_rsc, PCMK_XA_ID); if ((rsc != NULL) && !pcmk__str_eq(rsc_id, rsc, pcmk__str_none)) { continue; } intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter); output = g_list_concat(output, intermediate); } } } return output; } diff --git a/lib/services/services_lsb.c b/lib/services/services_lsb.c index 920f35bece..3bc10bc73c 100644 --- a/lib/services/services_lsb.c +++ b/lib/services/services_lsb.c @@ -1,320 +1,322 @@ /* * Copyright 2010-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include +#include // g_str_has_prefix() + #include #include #include #include "services_private.h" #include "services_lsb.h" // @TODO Use XML string constants and maybe a real XML object #define lsb_metadata_template \ "\n" \ "<" PCMK_XE_RESOURCE_AGENT " " \ PCMK_XA_NAME "='%s' " \ PCMK_XA_VERSION "='" PCMK_DEFAULT_AGENT_VERSION "'>\n" \ " <" PCMK_XE_VERSION ">1.1\n" \ " <" PCMK_XE_LONGDESC " " PCMK_XA_LANG "='" PCMK__VALUE_EN "'>\n" \ "%s" \ " \n" \ " <" PCMK_XE_SHORTDESC " " PCMK_XA_LANG "='" PCMK__VALUE_EN "'>" \ "%s" \ "\n" \ " <" PCMK_XE_PARAMETERS "/>\n" \ " <" PCMK_XE_ACTIONS ">\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='" PCMK_ACTION_META_DATA "'" \ " " PCMK_META_TIMEOUT "='5s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='" PCMK_ACTION_START "'" \ " " PCMK_META_TIMEOUT "='15s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='" PCMK_ACTION_STOP "'" \ " " PCMK_META_TIMEOUT "='15s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='" PCMK_ACTION_STATUS "'" \ " " PCMK_META_TIMEOUT "='15s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='restart'" \ " " PCMK_META_TIMEOUT "='15s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='force-reload'" \ " " PCMK_META_TIMEOUT "='15s' />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "='" PCMK_ACTION_MONITOR "'" \ " " PCMK_META_TIMEOUT "='15s'" \ " " PCMK_META_INTERVAL "='15s' />\n" \ " \n" \ " <" PCMK_XE_SPECIAL " " PCMK_XA_TAG "='LSB'>\n" \ " %s\n" \ " %s\n" \ " %s\n" \ " %s\n" \ " %s\n" \ " %s\n" \ " %s\n" \ " \n" \ "\n" /* See "Comment Conventions for Init Scripts" in the LSB core specification at: * http://refspecs.linuxfoundation.org/lsb.shtml */ #define LSB_INITSCRIPT_INFOBEGIN_TAG "### BEGIN INIT INFO" #define LSB_INITSCRIPT_INFOEND_TAG "### END INIT INFO" #define PROVIDES "# Provides:" #define REQUIRED_START "# Required-Start:" #define REQUIRED_STOP "# Required-Stop:" #define SHOULD_START "# Should-Start:" #define SHOULD_STOP "# Should-Stop:" #define DEFAULT_START "# Default-Start:" #define DEFAULT_STOP "# Default-Stop:" #define SHORT_DESC "# Short-Description:" #define DESCRIPTION "# Description:" /*! * \internal * \brief Grab an LSB header value * * \param[in] line Line read from LSB init script * \param[in,out] value If not set, will be set to XML-safe copy of value * \param[in] prefix Set value if line starts with this pattern * * \return TRUE if value was set, FALSE otherwise */ static inline gboolean lsb_meta_helper_get_value(const char *line, gchar **value, const char *prefix) { /* @TODO Perhaps update later to use pcmk__xml_needs_escape(). Involves many * extra variables in the caller. */ - if ((*value == NULL) && pcmk__starts_with(line, prefix)) { + if ((*value == NULL) && g_str_has_prefix(line, prefix)) { *value = pcmk__xml_escape(line + strlen(prefix), pcmk__xml_escape_text); return TRUE; } return FALSE; } int services__get_lsb_metadata(const char *type, char **output) { char *ra_pathname = NULL; FILE *fp = NULL; char buffer[1024] = { 0, }; gchar *provides = NULL; gchar *required_start = NULL; gchar *required_stop = NULL; gchar *should_start = NULL; gchar *should_stop = NULL; gchar *default_start = NULL; gchar *default_stop = NULL; gchar *short_desc = NULL; gchar *long_desc = NULL; bool in_header = FALSE; if (type[0] == '/') { ra_pathname = pcmk__str_copy(type); } else { ra_pathname = crm_strdup_printf(PCMK__LSB_INIT_DIR "/%s", type); } crm_trace("Looking into %s", ra_pathname); fp = fopen(ra_pathname, "r"); free(ra_pathname); if (fp == NULL) { return -errno; } /* Enter into the LSB-compliant comment block */ while (fgets(buffer, sizeof(buffer), fp)) { // Ignore lines up to and including the block delimiter - if (pcmk__starts_with(buffer, LSB_INITSCRIPT_INFOBEGIN_TAG)) { + if (g_str_has_prefix(buffer, LSB_INITSCRIPT_INFOBEGIN_TAG)) { in_header = TRUE; continue; } if (!in_header) { continue; } /* Assume each of the following eight arguments contain one line */ if (lsb_meta_helper_get_value(buffer, &provides, PROVIDES)) { continue; } if (lsb_meta_helper_get_value(buffer, &required_start, REQUIRED_START)) { continue; } if (lsb_meta_helper_get_value(buffer, &required_stop, REQUIRED_STOP)) { continue; } if (lsb_meta_helper_get_value(buffer, &should_start, SHOULD_START)) { continue; } if (lsb_meta_helper_get_value(buffer, &should_stop, SHOULD_STOP)) { continue; } if (lsb_meta_helper_get_value(buffer, &default_start, DEFAULT_START)) { continue; } if (lsb_meta_helper_get_value(buffer, &default_stop, DEFAULT_STOP)) { continue; } if (lsb_meta_helper_get_value(buffer, &short_desc, SHORT_DESC)) { continue; } /* Long description may cross multiple lines */ if ((long_desc == NULL) // Haven't already found long description - && pcmk__starts_with(buffer, DESCRIPTION)) { + && g_str_has_prefix(buffer, DESCRIPTION)) { bool processed_line = TRUE; GString *desc = g_string_sized_new(2048); // Get remainder of description line itself g_string_append(desc, buffer + sizeof(DESCRIPTION) - 1); // Read any continuation lines of the description buffer[0] = '\0'; while (fgets(buffer, sizeof(buffer), fp)) { - if (pcmk__starts_with(buffer, "# ") - || pcmk__starts_with(buffer, "#\t")) { + if (g_str_has_prefix(buffer, "# ") + || g_str_has_prefix(buffer, "#\t")) { /* '#' followed by a tab or more than one space indicates a * continuation of the long description. */ g_string_append(desc, buffer + 1); } else { /* This line is not part of the long description, * so continue with normal processing. */ processed_line = FALSE; break; } } // Make long description safe to use in XML long_desc = pcmk__xml_escape(desc->str, pcmk__xml_escape_text); g_string_free(desc, TRUE); if (processed_line) { // We grabbed the line into the long description continue; } } // Stop if we leave the header block - if (pcmk__starts_with(buffer, LSB_INITSCRIPT_INFOEND_TAG)) { + if (g_str_has_prefix(buffer, LSB_INITSCRIPT_INFOEND_TAG)) { break; } if (buffer[0] != '#') { break; } } fclose(fp); *output = crm_strdup_printf(lsb_metadata_template, type, pcmk__s(long_desc, type), pcmk__s(short_desc, type), pcmk__s(provides, ""), pcmk__s(required_start, ""), pcmk__s(required_stop, ""), pcmk__s(should_start, ""), pcmk__s(should_stop, ""), pcmk__s(default_start, ""), pcmk__s(default_stop, "")); g_free(long_desc); g_free(short_desc); g_free(provides); g_free(required_start); g_free(required_stop); g_free(should_start); g_free(should_stop); g_free(default_start); g_free(default_stop); return pcmk_ok; } GList * services__list_lsb_agents(void) { return services_os_get_directory_list(PCMK__LSB_INIT_DIR, TRUE, TRUE); } bool services__lsb_agent_exists(const char *agent) { bool rc = FALSE; struct stat st; char *path = pcmk__full_path(agent, PCMK__LSB_INIT_DIR); rc = (stat(path, &st) == 0); free(path); return rc; } /*! * \internal * \brief Prepare an LSB action * * \param[in,out] op Action to prepare * * \return Standard Pacemaker return code */ int services__lsb_prepare(svc_action_t *op) { op->opaque->exec = pcmk__full_path(op->agent, PCMK__LSB_INIT_DIR); op->opaque->args[0] = strdup(op->opaque->exec); op->opaque->args[1] = strdup(op->action); if ((op->opaque->args[0] == NULL) || (op->opaque->args[1] == NULL)) { return ENOMEM; } return pcmk_rc_ok; } /*! * \internal * \brief Map an LSB result to a standard OCF result * * \param[in] action Action that result is for * \param[in] exit_status LSB agent exit status * * \return Standard OCF result */ enum ocf_exitcode services__lsb2ocf(const char *action, int exit_status) { // For non-status actions, LSB and OCF share error codes <= 7 if (!pcmk__str_any_of(action, PCMK_ACTION_STATUS, PCMK_ACTION_MONITOR, NULL)) { if ((exit_status < 0) || (exit_status > PCMK_LSB_NOT_RUNNING)) { return PCMK_OCF_UNKNOWN_ERROR; } return (enum ocf_exitcode) exit_status; } // LSB status actions have their own codes switch (exit_status) { case PCMK_LSB_STATUS_OK: return PCMK_OCF_OK; case PCMK_LSB_STATUS_NOT_INSTALLED: return PCMK_OCF_NOT_INSTALLED; case PCMK_LSB_STATUS_INSUFFICIENT_PRIV: return PCMK_OCF_INSUFFICIENT_PRIV; case PCMK_LSB_STATUS_VAR_PID: case PCMK_LSB_STATUS_VAR_LOCK: case PCMK_LSB_STATUS_NOT_RUNNING: return PCMK_OCF_NOT_RUNNING; default: return PCMK_OCF_UNKNOWN_ERROR; } } diff --git a/lib/services/systemd.c b/lib/services/systemd.c index 3b513a7c29..9b1ebd7a03 100644 --- a/lib/services/systemd.c +++ b/lib/services/systemd.c @@ -1,1432 +1,1435 @@ /* * Copyright 2012-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include +#include #include // PRIu32 #include #include // uint32_t #include // fopen(), NULL, etc. #include + #include +#include // g_str_has_suffix() + #include #include -#include #include static void invoke_unit_by_path(svc_action_t *op, const char *unit); /* Systemd D-Bus interface * https://www.freedesktop.org/software/systemd/man/latest/org.freedesktop.systemd1.html */ #define BUS_NAME "org.freedesktop.systemd1" #define BUS_NAME_MANAGER BUS_NAME ".Manager" #define BUS_NAME_UNIT BUS_NAME ".Unit" #define BUS_PATH "/org/freedesktop/systemd1" /*! * \internal * \brief Prepare a systemd action * * \param[in,out] op Action to prepare * * \return Standard Pacemaker return code */ int services__systemd_prepare(svc_action_t *op) { op->opaque->exec = strdup("systemd-dbus"); if (op->opaque->exec == NULL) { return ENOMEM; } return pcmk_rc_ok; } /*! * \internal * \brief Map a systemd result to a standard OCF result * * \param[in] exit_status Systemd result * * \return Standard OCF result */ enum ocf_exitcode services__systemd2ocf(int exit_status) { // This library uses OCF codes for systemd actions return (enum ocf_exitcode) exit_status; } static inline DBusMessage * systemd_new_method(const char *method) { crm_trace("Calling: %s on " BUS_NAME_MANAGER, method); return dbus_message_new_method_call(BUS_NAME, BUS_PATH, BUS_NAME_MANAGER, method); } /* * Functions to manage a static DBus connection */ static DBusConnection* systemd_proxy = NULL; static inline DBusPendingCall * systemd_send(DBusMessage *msg, void(*done)(DBusPendingCall *pending, void *user_data), void *user_data, int timeout) { return pcmk_dbus_send(msg, systemd_proxy, done, user_data, timeout); } static inline DBusMessage * systemd_send_recv(DBusMessage *msg, DBusError *error, int timeout) { return pcmk_dbus_send_recv(msg, systemd_proxy, error, timeout); } /*! * \internal * \brief Send a method to systemd without arguments, and wait for reply * * \param[in] method Method to send * * \return Systemd reply on success, NULL (and error will be logged) otherwise * * \note The caller must call dbus_message_unref() on the reply after * handling it. */ static DBusMessage * systemd_call_simple_method(const char *method) { DBusMessage *msg = NULL; DBusMessage *reply = NULL; DBusError error; /* Don't call systemd_init() here, because that calls this */ CRM_CHECK(systemd_proxy, return NULL); msg = systemd_new_method(method); if (msg == NULL) { crm_err("Could not create message to send %s to systemd", method); return NULL; } dbus_error_init(&error); reply = systemd_send_recv(msg, &error, DBUS_TIMEOUT_USE_DEFAULT); dbus_message_unref(msg); if (dbus_error_is_set(&error)) { crm_err("Could not send %s to systemd: %s (%s)", method, error.message, error.name); dbus_error_free(&error); return NULL; } else if (reply == NULL) { crm_err("Could not send %s to systemd: no reply received", method); return NULL; } return reply; } /*! * \internal * \brief Subscribe to D-Bus signals from systemd * * Systemd does not broadcast signal messages unless at least one client has * called the \c Subscribe() method. Also, a D-Bus client ignores broadcast * messages unless an appropriate match rule is set, so we set one here. * * \return Standard Pacemaker return code */ static int subscribe_to_signals(void) { const char *match_rule = "type='signal'," "sender='" BUS_NAME "'," "interface='" BUS_NAME_MANAGER "'," "path='" BUS_PATH "'"; DBusMessage *reply = NULL; DBusError error; /* Tell D-Bus to accept signal messages from systemd. * https://dbus.freedesktop.org/doc/dbus-specification.html#message-bus-routing-match-rules */ dbus_error_init(&error); dbus_bus_add_match(systemd_proxy, match_rule, &error); if (dbus_error_is_set(&error)) { crm_err("Could not listen for systemd DBus signals: %s " QB_XS " (%s)", error.message, error.name); dbus_error_free(&error); return ECOMM; } // Tell systemd to broadcast signals reply = systemd_call_simple_method("Subscribe"); if (reply == NULL) { dbus_bus_remove_match(systemd_proxy, match_rule, &error); return ECOMM; } dbus_message_unref(reply); return pcmk_rc_ok; } static bool systemd_init(void) { static int need_init = 1; // https://dbus.freedesktop.org/doc/api/html/group__DBusConnection.html if (systemd_proxy && dbus_connection_get_is_connected(systemd_proxy) == FALSE) { crm_warn("Connection to System DBus is closed. Reconnecting..."); pcmk_dbus_disconnect(systemd_proxy); systemd_proxy = NULL; need_init = 1; } if (need_init) { need_init = 0; systemd_proxy = pcmk_dbus_connect(); if (subscribe_to_signals() != pcmk_rc_ok) { pcmk_dbus_disconnect(systemd_proxy); systemd_proxy = NULL; } } return (systemd_proxy != NULL); } static inline char * systemd_get_property(const char *unit, const char *name, void (*callback)(const char *name, const char *value, void *userdata), void *userdata, DBusPendingCall **pending, int timeout) { return systemd_proxy? pcmk_dbus_get_property(systemd_proxy, BUS_NAME, unit, BUS_NAME_UNIT, name, callback, userdata, pending, timeout) : NULL; } void systemd_cleanup(void) { if (systemd_proxy) { pcmk_dbus_disconnect(systemd_proxy); systemd_proxy = NULL; } } /* * end of systemd_proxy functions */ /*! * \internal * \brief Check whether a file name represents a manageable systemd unit * * \param[in] name File name to check * * \return Pointer to "dot" before filename extension if so, NULL otherwise */ static const char * systemd_unit_extension(const char *name) { if (name) { const char *dot = strrchr(name, '.'); if (dot && (!strcmp(dot, ".service") || !strcmp(dot, ".socket") || !strcmp(dot, ".mount") || !strcmp(dot, ".timer") || !strcmp(dot, ".path"))) { return dot; } } return NULL; } static char * systemd_unit_name(const char *name, bool add_instance_name) { const char *dot = NULL; if (pcmk__str_empty(name)) { return NULL; } /* Services that end with an @ sign are systemd templates. They expect an * instance name to follow the service name. If no instance name was * provided, just add "pacemaker" to the string as the instance name. It * doesn't seem to matter for purposes of looking up whether a service * exists or not. * * A template can be specified either with or without the unit extension, * so this block handles both cases. */ dot = systemd_unit_extension(name); if (dot) { if (dot != name && *(dot-1) == '@') { return crm_strdup_printf("%.*spacemaker%s", (int) (dot - name), name, dot); } else { return pcmk__str_copy(name); } } else if (add_instance_name && *(name+strlen(name)-1) == '@') { return crm_strdup_printf("%spacemaker.service", name); } else { return crm_strdup_printf("%s.service", name); } } static void systemd_daemon_reload_complete(DBusPendingCall *pending, void *user_data) { DBusError error; DBusMessage *reply = NULL; unsigned int reload_count = GPOINTER_TO_UINT(user_data); dbus_error_init(&error); if(pending) { reply = dbus_pending_call_steal_reply(pending); } if (pcmk_dbus_find_error(pending, reply, &error)) { crm_warn("Could not issue systemd reload %d: %s", reload_count, error.message); dbus_error_free(&error); } else { crm_trace("Reload %d complete", reload_count); } if(pending) { dbus_pending_call_unref(pending); } if(reply) { dbus_message_unref(reply); } } static bool systemd_daemon_reload(int timeout) { static unsigned int reload_count = 0; DBusMessage *msg = systemd_new_method("Reload"); reload_count++; pcmk__assert(msg != NULL); systemd_send(msg, systemd_daemon_reload_complete, GUINT_TO_POINTER(reload_count), timeout); dbus_message_unref(msg); return TRUE; } /*! * \internal * \brief Set an action result based on a method error * * \param[in,out] op Action to set result for * \param[in] error Method error */ static void set_result_from_method_error(svc_action_t *op, const DBusError *error) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to invoke systemd DBus method"); if (strstr(error->name, "org.freedesktop.systemd1.InvalidName") || strstr(error->name, "org.freedesktop.systemd1.LoadFailed") || strstr(error->name, "org.freedesktop.systemd1.NoSuchUnit")) { if (pcmk__str_eq(op->action, PCMK_ACTION_STOP, pcmk__str_casei)) { crm_trace("Masking systemd stop failure (%s) for %s " "because unknown service can be considered stopped", error->name, pcmk__s(op->rsc, "unknown resource")); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); return; } services__format_result(op, PCMK_OCF_NOT_INSTALLED, PCMK_EXEC_NOT_INSTALLED, "systemd unit %s not found", op->agent); /* If systemd happens to be re-executing by `systemctl daemon-reexec` at the * same time, dbus gives an error with the name * `org.freedesktop.DBus.Error.NoReply` and the message "Message recipient * disconnected from message bus without replying". * Consider the monitor pending rather than return an error yet, so that it * can retry with another iteration. */ } else if (pcmk__str_any_of(op->action, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL) && strstr(error->name, DBUS_ERROR_NO_REPLY) && strstr(error->message, "disconnected")) { services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); } crm_info("DBus request for %s of systemd unit %s%s%s failed: %s", op->action, op->agent, ((op->rsc == NULL)? "" : " for resource "), pcmk__s(op->rsc, ""), error->message); } /*! * \internal * \brief Extract unit path from LoadUnit reply, and execute action * * \param[in] reply LoadUnit reply * \param[in,out] op Action to execute (or NULL to just return path) * * \return DBus object path for specified unit if successful (only valid for * lifetime of \p reply), otherwise NULL */ static const char * execute_after_loadunit(DBusMessage *reply, svc_action_t *op) { const char *path = NULL; DBusError error; /* path here is not used other than as a non-NULL flag to indicate that a * request was indeed sent */ if (pcmk_dbus_find_error((void *) &path, reply, &error)) { if (op != NULL) { set_result_from_method_error(op, &error); } dbus_error_free(&error); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { if (op != NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "systemd DBus method had unexpected reply"); crm_info("Could not load systemd unit %s for %s: " "DBus reply has unexpected type", op->agent, op->id); } else { crm_info("Could not load systemd unit: " "DBus reply has unexpected type"); } } else { dbus_message_get_args (reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); } if (op != NULL) { if (path != NULL) { invoke_unit_by_path(op, path); } else if (!(op->synchronous)) { if (!pcmk__str_any_of(op->action, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL) || op->status != PCMK_EXEC_PENDING) { services__format_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus object found for systemd unit %s", op->agent); } services__finalize_async_op(op); } } return path; } /*! * \internal * \brief Execute a systemd action after its LoadUnit completes * * \param[in,out] pending If not NULL, DBus call associated with LoadUnit * \param[in,out] user_data Action to execute */ static void loadunit_completed(DBusPendingCall *pending, void *user_data) { DBusMessage *reply = NULL; svc_action_t *op = user_data; crm_trace("LoadUnit result for %s arrived", op->id); // Grab the reply if (pending != NULL) { reply = dbus_pending_call_steal_reply(pending); } // The call is no longer pending CRM_LOG_ASSERT(pending == op->opaque->pending); services_set_op_pending(op, NULL); // Execute the desired action based on the reply execute_after_loadunit(reply, user_data); if (reply != NULL) { dbus_message_unref(reply); } } /*! * \internal * \brief Execute a systemd action, given the unit name * * \param[in] arg_name Unit name (possibly without ".service" extension) * \param[in,out] op Action to execute (if NULL, just get object path) * \param[out] path If non-NULL and \p op is NULL or synchronous, where * to store DBus object path for specified unit * * \return Standard Pacemaker return code (for NULL \p op, pcmk_rc_ok means unit * was found; for synchronous actions, pcmk_rc_ok means unit was * executed, with the actual result stored in \p op; for asynchronous * actions, pcmk_rc_ok means action was initiated) * \note It is the caller's responsibility to free the path. */ static int invoke_unit_by_name(const char *arg_name, svc_action_t *op, char **path) { DBusMessage *msg; DBusMessage *reply = NULL; DBusPendingCall *pending = NULL; char *name = NULL; if (pcmk__str_empty(arg_name)) { return EINVAL; } if (!systemd_init()) { if (op != NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus connection"); } return ENOTCONN; } /* Create a LoadUnit DBus method (equivalent to GetUnit if already loaded), * which makes the unit usable via further DBus methods. * * * * * */ msg = systemd_new_method("LoadUnit"); pcmk__assert(msg != NULL); // Add the (expanded) unit name as the argument name = systemd_unit_name(arg_name, (op == NULL) || pcmk__str_eq(op->action, PCMK_ACTION_META_DATA, pcmk__str_none)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &name, DBUS_TYPE_INVALID)); free(name); if ((op == NULL) || op->synchronous) { // For synchronous ops, wait for a reply and extract the result const char *unit = NULL; int rc = pcmk_rc_ok; reply = systemd_send_recv(msg, NULL, (op? op->timeout : DBUS_TIMEOUT_USE_DEFAULT)); dbus_message_unref(msg); unit = execute_after_loadunit(reply, op); if (unit == NULL) { rc = ENOENT; if (path != NULL) { *path = NULL; } } else if (path != NULL) { *path = strdup(unit); if (*path == NULL) { rc = ENOMEM; } } if (reply != NULL) { dbus_message_unref(reply); } return rc; } // For asynchronous ops, initiate the LoadUnit call and return pending = systemd_send(msg, loadunit_completed, op, op->timeout); if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to send DBus message"); dbus_message_unref(msg); return ECOMM; } // LoadUnit was successfully initiated services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); services_set_op_pending(op, pending); dbus_message_unref(msg); return pcmk_rc_ok; } /*! * \internal * \brief Compare two strings alphabetically (case-insensitive) * * \param[in] a First string to compare * \param[in] b Second string to compare * * \return 0 if strings are equal, -1 if a < b, 1 if a > b * * \note Usable as a GCompareFunc with g_list_sort(). * NULL is considered less than non-NULL. */ static gint sort_str(gconstpointer a, gconstpointer b) { if (!a && !b) { return 0; } else if (!a) { return -1; } else if (!b) { return 1; } return strcasecmp(a, b); } GList * systemd_unit_listall(void) { int nfiles = 0; GList *units = NULL; DBusMessageIter args; DBusMessageIter unit; DBusMessageIter elem; DBusMessage *reply = NULL; if (!systemd_init()) { return NULL; } /* " \n" \ " \n" \ " \n" \ */ reply = systemd_call_simple_method("ListUnitFiles"); if (reply == NULL) { return NULL; } if (!dbus_message_iter_init(reply, &args)) { crm_err("Could not list systemd unit files: systemd reply has no arguments"); dbus_message_unref(reply); return NULL; } if (!pcmk_dbus_type_check(reply, &args, DBUS_TYPE_ARRAY, __func__, __LINE__)) { crm_err("Could not list systemd unit files: systemd reply has invalid arguments"); dbus_message_unref(reply); return NULL; } dbus_message_iter_recurse(&args, &unit); for (; dbus_message_iter_get_arg_type(&unit) != DBUS_TYPE_INVALID; dbus_message_iter_next(&unit)) { DBusBasicValue value; const char *match = NULL; char *unit_name = NULL; char *basename = NULL; if(!pcmk_dbus_type_check(reply, &unit, DBUS_TYPE_STRUCT, __func__, __LINE__)) { crm_warn("Skipping systemd reply argument with unexpected type"); continue; } dbus_message_iter_recurse(&unit, &elem); if(!pcmk_dbus_type_check(reply, &elem, DBUS_TYPE_STRING, __func__, __LINE__)) { crm_warn("Skipping systemd reply argument with no string"); continue; } dbus_message_iter_get_basic(&elem, &value); if (value.str == NULL) { crm_debug("ListUnitFiles reply did not provide a string"); continue; } crm_trace("DBus ListUnitFiles listed: %s", value.str); match = systemd_unit_extension(value.str); if (match == NULL) { // This is not a unit file type we know how to manage crm_debug("ListUnitFiles entry '%s' is not supported as resource", value.str); continue; } // ListUnitFiles returns full path names, we just want base name basename = strrchr(value.str, '/'); if (basename) { basename = basename + 1; } else { basename = value.str; } if (!strcmp(match, ".service")) { // Service is the "default" unit type, so strip it unit_name = strndup(basename, match - basename); } else { unit_name = strdup(basename); } nfiles++; units = g_list_prepend(units, unit_name); } dbus_message_unref(reply); crm_trace("Found %d manageable systemd unit files", nfiles); units = g_list_sort(units, sort_str); return units; } bool systemd_unit_exists(const char *name) { char *path = NULL; char *state = NULL; int rc = false; /* Note: Makes a blocking dbus calls * Used by resources_find_service_class() when resource class=service */ if ((invoke_unit_by_name(name, NULL, &path) != pcmk_rc_ok) || (path == NULL)) { goto done; } /* A successful LoadUnit is not sufficient to determine the unit's * existence; it merely means the LoadUnit request received a reply. * We must make another blocking call to check the LoadState property. */ state = systemd_get_property(path, "LoadState", NULL, NULL, NULL, DBUS_TIMEOUT_USE_DEFAULT); rc = pcmk__str_any_of(state, "loaded", "masked", NULL); done: free(path); free(state); return rc; } // @TODO Use XML string constants and maybe a real XML object #define METADATA_FORMAT \ "\n" \ "<" PCMK_XE_RESOURCE_AGENT " " \ PCMK_XA_NAME "=\"%s\" " \ PCMK_XA_VERSION "=\"" PCMK_DEFAULT_AGENT_VERSION "\">\n" \ " <" PCMK_XE_VERSION ">1.1\n" \ " <" PCMK_XE_LONGDESC " " PCMK_XA_LANG "=\"" PCMK__VALUE_EN "\">\n" \ " %s\n" \ " \n" \ " <" PCMK_XE_SHORTDESC " " PCMK_XA_LANG "=\"" PCMK__VALUE_EN "\">" \ "systemd unit file for %s" \ "\n" \ " <" PCMK_XE_PARAMETERS "/>\n" \ " <" PCMK_XE_ACTIONS ">\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "=\"" PCMK_ACTION_START "\"" \ " " PCMK_META_TIMEOUT "=\"100s\" />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "=\"" PCMK_ACTION_STOP "\"" \ " " PCMK_META_TIMEOUT "=\"100s\" />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "=\"" PCMK_ACTION_STATUS "\"" \ " " PCMK_META_TIMEOUT "=\"100s\" />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "=\"" PCMK_ACTION_MONITOR "\"" \ " " PCMK_META_TIMEOUT "=\"100s\"" \ " " PCMK_META_INTERVAL "=\"60s\" />\n" \ " <" PCMK_XE_ACTION " " PCMK_XA_NAME "=\"" PCMK_ACTION_META_DATA "\"" \ " " PCMK_META_TIMEOUT "=\"5s\" />\n" \ " \n" \ " <" PCMK_XE_SPECIAL " " PCMK_XA_TAG "=\"systemd\"/>\n" \ "\n" static char * systemd_unit_metadata(const char *name, int timeout) { char *meta = NULL; char *desc = NULL; char *path = NULL; if (invoke_unit_by_name(name, NULL, &path) == pcmk_rc_ok) { /* TODO: Worth a making blocking call for? Probably not. Possibly if cached. */ desc = systemd_get_property(path, "Description", NULL, NULL, NULL, timeout); } else { desc = crm_strdup_printf("Systemd unit file for %s", name); } if (pcmk__xml_needs_escape(desc, pcmk__xml_escape_text)) { gchar *escaped = pcmk__xml_escape(desc, pcmk__xml_escape_text); meta = crm_strdup_printf(METADATA_FORMAT, name, escaped, name); g_free(escaped); } else { meta = crm_strdup_printf(METADATA_FORMAT, name, desc, name); } free(desc); free(path); return meta; } /*! * \internal * \brief Determine result of method from reply * * \param[in] reply Reply to start, stop, or restart request * \param[in,out] op Action that was executed */ static void process_unit_method_reply(DBusMessage *reply, svc_action_t *op) { bool start_stop = pcmk__strcase_any_of(op->action, PCMK_ACTION_START, PCMK_ACTION_STOP, NULL); DBusError error; dbus_error_init(&error); /* The first use of error here is not used other than as a non-NULL flag to * indicate that a request was indeed sent */ if (pcmk_dbus_find_error((void *) &error, reply, &error)) { set_result_from_method_error(op, &error); dbus_error_free(&error); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { const char *reason = "systemd D-Bus method had unexpected reply"; crm_info("DBus request for %s of %s succeeded but " "return type was unexpected", op->action, pcmk__s(op->rsc, "unknown resource")); if (!op->synchronous && start_stop) { /* The start or stop job is enqueued but is not complete. We need a * job path to detect completion in job_removed_filter(). */ services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, reason); } else { /* Something weird happened, but the action is finished and there * was no D-Bus error. So call it a success. */ services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, reason); } } else { const char *path = NULL; dbus_message_get_args(reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); crm_debug("DBus request for %s of %s using %s succeeded", op->action, pcmk__s(op->rsc, "unknown resource"), path); if (!op->synchronous && start_stop) { // Should be set to unknown/pending already services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); pcmk__str_update(&(op->opaque->job_path), path); } else { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } } } /*! * \internal * \brief Process a systemd \c JobRemoved signal for a given service action * * This filter is expected to be added with \c finalize_async_action_dbus() as * the \c free_data_function. Then if \p message is a \c JobRemoved signal for * the action specified by \p user_data, the action's result is set, the filter * is removed, and the action is finalized. * * \param[in,out] connection D-Bus connection * \param[in] message D-Bus message * \param[in,out] user_data Service action (\c svc_action_t) * * \retval \c DBUS_HANDLER_RESULT_HANDLED if \p message is a \c JobRemoved * signal for \p user_data * \retval \c DBUS_HANDLER_RESULT_NOT_YET_HANDLED otherwise (on error, if * \p message is not a \c JobRemoved signal, or if the signal is for * some other action's job) */ static DBusHandlerResult job_removed_filter(DBusConnection *connection, DBusMessage *message, void *user_data) { svc_action_t *action = user_data; const char *action_name = NULL; uint32_t job_id = 0; const char *bus_path = NULL; const char *unit_name = NULL; const char *result = NULL; DBusError error; CRM_CHECK((connection != NULL) && (message != NULL), return DBUS_HANDLER_RESULT_NOT_YET_HANDLED); // action should always be set when the filter is added if ((action == NULL) || !dbus_message_is_signal(message, BUS_NAME_MANAGER, "JobRemoved")) { return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; } dbus_error_init(&error); if (!dbus_message_get_args(message, &error, DBUS_TYPE_UINT32, &job_id, DBUS_TYPE_OBJECT_PATH, &bus_path, DBUS_TYPE_STRING, &unit_name, DBUS_TYPE_STRING, &result, DBUS_TYPE_INVALID)) { crm_err("Could not interpret systemd DBus signal: %s " QB_XS " (%s)", error.message, error.name); dbus_error_free(&error); return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; } if (!pcmk__str_eq(bus_path, action->opaque->job_path, pcmk__str_none)) { // This filter is not for this job return DBUS_HANDLER_RESULT_NOT_YET_HANDLED; } action_name = pcmk__s(action->action, "(unknown)"); crm_trace("Setting %s result for %s (JobRemoved id=%" PRIu32 ", result=%s", action_name, unit_name, job_id, result); if (pcmk__str_eq(result, "done", pcmk__str_none)) { services__set_result(action, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (pcmk__str_eq(result, "timeout", pcmk__str_none)) { services__format_result(action, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, "systemd %s job for %s timed out", action_name, unit_name); } else { services__format_result(action, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "systemd %s job for %s failed with result '%s'", action_name, unit_name, result); } /* This instance of the filter was specifically for the given action. * * The action gets finalized by services__finalize_async_op() via the * filter's free_data_function. */ dbus_connection_remove_filter(systemd_proxy, job_removed_filter, action); return DBUS_HANDLER_RESULT_HANDLED; } /*! * \internal * \brief \c DBusFreeFunction wrapper for \c services__finalize_async_op() * * \param[in,out] action Asynchronous service action to finalize */ static void finalize_async_action_dbus(void *action) { services__finalize_async_op((svc_action_t *) action); } /*! * \internal * \brief Process the completion of an asynchronous unit start, stop, or restart * * \param[in,out] pending If not NULL, DBus call associated with request * \param[in,out] user_data Action that was executed */ static void unit_method_complete(DBusPendingCall *pending, void *user_data) { DBusMessage *reply = NULL; svc_action_t *op = user_data; crm_trace("Result for %s arrived", op->id); // Grab the reply if (pending != NULL) { reply = dbus_pending_call_steal_reply(pending); } // The call is no longer pending CRM_LOG_ASSERT(pending == op->opaque->pending); services_set_op_pending(op, NULL); process_unit_method_reply(reply, op); if (reply != NULL) { dbus_message_unref(reply); } if ((op->status == PCMK_EXEC_PENDING) && pcmk__strcase_any_of(op->action, PCMK_ACTION_START, PCMK_ACTION_STOP, NULL)) { /* Start and stop method calls return when the job is enqueued, not when * it's complete. Start and stop actions must be finalized after the job * is complete, because the action callback function may use it. We add * a message filter to process the JobRemoved signal, which indicates * completion. * * The filter takes ownership of op, which will be finalized when the * filter is later removed. */ if (dbus_connection_add_filter(systemd_proxy, job_removed_filter, op, finalize_async_action_dbus)) { return; } crm_err("Could not add D-Bus filter for systemd JobRemoved signals"); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Failed to add D-Bus filter for systemd " "JobRemoved signal"); } services__finalize_async_op(op); } /* When the cluster manages a systemd resource, we create a unit file override * to order the service "before" pacemaker. The "before" relationship won't * actually be used, since systemd won't ever start the resource -- we're * interested in the reverse shutdown ordering it creates, to ensure that * systemd doesn't stop the resource at shutdown while pacemaker is still * running. * * @TODO Add start timeout */ #define SYSTEMD_UNIT_OVERRIDE_TEMPLATE \ "[Unit]\n" \ "Description=Cluster Controlled %s\n" \ "Before=pacemaker.service pacemaker_remote.service\n" #define SYSTEMD_SERVICE_OVERRIDE \ "\n" \ "[Service]\n" \ "Restart=no\n" /*! * \internal * \brief Get runtime drop-in directory path for a systemd unit * * \param[in] unit_name Systemd unit (with extension) * * \return Drop-in directory path */ static GString * get_override_dir(const char *unit_name) { GString *buf = g_string_sized_new(128); pcmk__g_strcat(buf, "/run/systemd/system/", unit_name, ".d", NULL); return buf; } /*! * \internal * \brief Append systemd override filename to a directory path * * \param[in,out] buf Buffer containing directory path to append to */ static inline void append_override_basename(GString *buf) { g_string_append(buf, "/50-pacemaker.conf"); } /*! * \internal * \brief Create a runtime override file for a systemd unit * * The systemd daemon is then reloaded. This file does not survive a reboot. * * \param[in] agent Systemd resource agent * \param[in] timeout Timeout for systemd daemon reload * * \return Standard Pacemaker return code * * \note Any configuration in \c /etc takes precedence over our drop-in. * \todo Document this in Pacemaker Explained or Administration? */ static int systemd_create_override(const char *agent, int timeout) { char *unit_name = NULL; GString *filename = NULL; GString *override = NULL; FILE *fp = NULL; int fd = 0; int rc = pcmk_rc_ok; unit_name = systemd_unit_name(agent, false); CRM_CHECK(!pcmk__str_empty(unit_name), rc = EINVAL; goto done); filename = get_override_dir(unit_name); rc = pcmk__build_path(filename->str, 0755); if (rc != pcmk_rc_ok) { crm_err("Could not create systemd override directory %s: %s", filename->str, pcmk_rc_str(rc)); goto done; } append_override_basename(filename); fp = fopen(filename->str, "w"); if (fp == NULL) { rc = errno; crm_err("Cannot open systemd override file %s for writing: %s", filename->str, pcmk_rc_str(rc)); goto done; } // Ensure the override file is world-readable (avoid systemd warning in log) fd = fileno(fp); if ((fd < 0) || (fchmod(fd, S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH) < 0)) { rc = errno; crm_err("Failed to set permissions on systemd override file %s: %s", filename->str, pcmk_rc_str(rc)); goto done; } override = g_string_sized_new(2 * sizeof(SYSTEMD_UNIT_OVERRIDE_TEMPLATE)); g_string_printf(override, SYSTEMD_UNIT_OVERRIDE_TEMPLATE, unit_name); - if (pcmk__ends_with_ext(unit_name, ".service")) { + if (g_str_has_suffix(unit_name, ".service")) { g_string_append(override, SYSTEMD_SERVICE_OVERRIDE); } if (fputs(override->str, fp) == EOF) { rc = EIO; crm_err("Cannot write to systemd override file %s", filename->str); } done: if (fp != NULL) { fclose(fp); } if (rc == pcmk_rc_ok) { // @TODO Make sure the reload succeeds systemd_daemon_reload(timeout); } else if (fp != NULL) { // File was created, so remove it unlink(filename->str); } free(unit_name); // coverity[check_after_deref : FALSE] if (filename != NULL) { g_string_free(filename, TRUE); } if (override != NULL) { g_string_free(override, TRUE); } return rc; } static void systemd_remove_override(const char *agent, int timeout) { char *unit_name = systemd_unit_name(agent, false); GString *filename = NULL; CRM_CHECK(!pcmk__str_empty(unit_name), goto done); filename = get_override_dir(unit_name); append_override_basename(filename); if (unlink(filename->str) < 0) { int rc = errno; if (rc != ENOENT) { // Stop may be called when already stopped, which is fine crm_warn("Cannot remove systemd override file %s: %s", filename->str, pcmk_rc_str(rc)); } } else { systemd_daemon_reload(timeout); } done: free(unit_name); // coverity[check_after_deref : FALSE] if (filename != NULL) { g_string_free(filename, TRUE); } } /*! * \internal * \brief Parse result of systemd status check * * Set a status action's exit status and execution status based on a DBus * property check result, and finalize the action if asynchronous. * * \param[in] name DBus interface name for property that was checked * \param[in] state Property value * \param[in,out] userdata Status action that check was done for */ static void parse_status_result(const char *name, const char *state, void *userdata) { svc_action_t *op = userdata; crm_trace("Resource %s has %s='%s'", pcmk__s(op->rsc, "(unspecified)"), name, pcmk__s(state, "")); if (pcmk__str_eq(state, "active", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (pcmk__str_eq(state, "reloading", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (pcmk__str_eq(state, "activating", pcmk__str_none)) { services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); } else if (pcmk__str_eq(state, "deactivating", pcmk__str_none)) { services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); } else { services__set_result(op, PCMK_OCF_NOT_RUNNING, PCMK_EXEC_DONE, state); } if (!(op->synchronous)) { services_set_op_pending(op, NULL); services__finalize_async_op(op); } } /*! * \internal * \brief Invoke a systemd unit, given its DBus object path * * \param[in,out] op Action to execute * \param[in] unit DBus object path of systemd unit to invoke */ static void invoke_unit_by_path(svc_action_t *op, const char *unit) { const char *method = NULL; DBusMessage *msg = NULL; DBusMessage *reply = NULL; if (pcmk__str_any_of(op->action, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL)) { DBusPendingCall *pending = NULL; char *state; state = systemd_get_property(unit, "ActiveState", (op->synchronous? NULL : parse_status_result), op, (op->synchronous? NULL : &pending), op->timeout); if (op->synchronous) { parse_status_result("ActiveState", state, op); free(state); } else if (pending == NULL) { // Could not get ActiveState property services__format_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Could not get state for unit %s from DBus", op->agent); services__finalize_async_op(op); } else { services_set_op_pending(op, pending); } return; } else if (pcmk__str_eq(op->action, PCMK_ACTION_START, pcmk__str_none)) { int rc = pcmk_rc_ok; method = "StartUnit"; rc = systemd_create_override(op->agent, op->timeout); if (rc != pcmk_rc_ok) { services__format_result(op, pcmk_rc2ocf(rc), PCMK_EXEC_ERROR, "Failed to create systemd override file " "for %s", pcmk__s(op->agent, "(unspecified)")); if (!(op->synchronous)) { services__finalize_async_op(op); } return; } } else if (pcmk__str_eq(op->action, PCMK_ACTION_STOP, pcmk__str_none)) { method = "StopUnit"; systemd_remove_override(op->agent, op->timeout); } else if (pcmk__str_eq(op->action, "restart", pcmk__str_none)) { method = "RestartUnit"; } else { services__format_result(op, PCMK_OCF_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR, "Action %s not implemented " "for systemd resources", pcmk__s(op->action, "(unspecified)")); if (!(op->synchronous)) { services__finalize_async_op(op); } return; } crm_trace("Calling %s for unit path %s%s%s", method, unit, ((op->rsc == NULL)? "" : " for resource "), pcmk__s(op->rsc, "")); msg = systemd_new_method(method); pcmk__assert(msg != NULL); /* (ss) */ { const char *replace_s = "replace"; char *name = systemd_unit_name(op->agent, pcmk__str_eq(op->action, PCMK_ACTION_META_DATA, pcmk__str_none)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &name, DBUS_TYPE_INVALID)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &replace_s, DBUS_TYPE_INVALID)); free(name); } if (op->synchronous) { reply = systemd_send_recv(msg, NULL, op->timeout); dbus_message_unref(msg); process_unit_method_reply(reply, op); if (reply != NULL) { dbus_message_unref(reply); } } else { DBusPendingCall *pending = systemd_send(msg, unit_method_complete, op, op->timeout); dbus_message_unref(msg); if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to send DBus message"); services__finalize_async_op(op); } else { services_set_op_pending(op, pending); } } } static gboolean systemd_timeout_callback(gpointer p) { svc_action_t * op = p; op->opaque->timerid = 0; crm_info("%s action for systemd unit %s named '%s' timed out", op->action, op->agent, op->rsc); services__format_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, "%s action for systemd unit %s " "did not complete in time", op->action, op->agent); if (op->opaque->job_path != NULL) { // A filter owns this op dbus_connection_remove_filter(systemd_proxy, job_removed_filter, op); } else { services__finalize_async_op(op); } return FALSE; } /*! * \internal * \brief Execute a systemd action * * \param[in,out] op Action to execute * * \return Standard Pacemaker return code * \retval EBUSY Recurring operation could not be initiated * \retval pcmk_rc_error Synchronous action failed * \retval pcmk_rc_ok Synchronous action succeeded, or asynchronous action * should not be freed (because it's pending or because * it failed to execute and was already freed) * * \note If the return value for an asynchronous action is not pcmk_rc_ok, the * caller is responsible for freeing the action. */ int services__execute_systemd(svc_action_t *op) { pcmk__assert(op != NULL); if (pcmk__str_empty(op->action) || pcmk__str_empty(op->agent)) { services__set_result(op, PCMK_OCF_NOT_CONFIGURED, PCMK_EXEC_ERROR_FATAL, "Bug in action caller"); goto done; } if (!systemd_init()) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus connection"); goto done; } crm_debug("Performing %ssynchronous %s op on systemd unit %s%s%s", (op->synchronous? "" : "a"), op->action, op->agent, ((op->rsc == NULL)? "" : " for resource "), pcmk__s(op->rsc, "")); if (pcmk__str_eq(op->action, PCMK_ACTION_META_DATA, pcmk__str_casei)) { op->stdout_data = systemd_unit_metadata(op->agent, op->timeout); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); goto done; } /* invoke_unit_by_name() should always override these values, which are here * just as a fail-safe in case there are any code paths that neglect to */ services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Bug in service library"); if (invoke_unit_by_name(op->agent, op, NULL) == pcmk_rc_ok) { // @TODO Why plus 5000? No explanation in fccd046. op->opaque->timerid = pcmk__create_timer(op->timeout + 5000, systemd_timeout_callback, op); services_add_inflight_op(op); return pcmk_rc_ok; } done: if (op->synchronous) { return (op->rc == PCMK_OCF_OK)? pcmk_rc_ok : pcmk_rc_error; } else { return services__finalize_async_op(op); } } diff --git a/tools/crm_mon.c b/tools/crm_mon.c index 19a5580a5f..8fd69ec08a 100644 --- a/tools/crm_mon.c +++ b/tools/crm_mon.c @@ -1,2191 +1,2193 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include // g_str_has_prefix() + #include #include #include -#include // pcmk__ends_with_ext() #include #include #include #include #include #include #include #include #include #include #include #include #include #include // stonith__* #include "crm_mon.h" #define SUMMARY "Provides a summary of cluster's current state.\n\n" \ "Outputs varying levels of detail in a number of different formats." /* * Definitions indicating which items to print */ static uint32_t show; static uint32_t show_opts = pcmk_show_pending; /* * Definitions indicating how to output */ static mon_output_format_t output_format = mon_output_unset; /* other globals */ static GIOChannel *io_channel = NULL; static GMainLoop *mainloop = NULL; static guint reconnect_timer = 0; static mainloop_timer_t *refresh_timer = NULL; static enum pcmk_pacemakerd_state pcmkd_state = pcmk_pacemakerd_state_invalid; static cib_t *cib = NULL; static stonith_t *st = NULL; static xmlNode *current_cib = NULL; static GError *error = NULL; static pcmk__common_args_t *args = NULL; static pcmk__output_t *out = NULL; static GOptionContext *context = NULL; static gchar **processed_args = NULL; static time_t last_refresh = 0; volatile crm_trigger_t *refresh_trigger = NULL; static pcmk_scheduler_t *scheduler = NULL; static enum pcmk__fence_history fence_history = pcmk__fence_history_none; int interactive_fence_level = 0; static pcmk__supported_format_t formats[] = { #if PCMK__ENABLE_CURSES CRM_MON_SUPPORTED_FORMAT_CURSES, #endif PCMK__SUPPORTED_FORMAT_HTML, PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_default(pcmk__output_t *out, va_list args) { return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_html(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); if (out->dest != stdout) { out->reset(out); } pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, "Not connected to CIB"); if (desc != NULL) { pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, ": "); pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, desc); } if (state != pcmk_pacemakerd_state_invalid) { const char *state_s = pcmk__pcmkd_state_enum2friendly(state); pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, " ("); pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, state_s); pcmk__output_create_xml_text_node(out, PCMK__XE_SPAN, ")"); } out->finish(out, CRM_EX_DISCONNECT, true, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_text(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); int rc = pcmk_rc_ok; if (out->dest != stdout) { out->reset(out); } if (state != pcmk_pacemakerd_state_invalid) { rc = out->info(out, "Not connected to CIB%s%s (%s)", (desc != NULL)? ": " : "", pcmk__s(desc, ""), pcmk__pcmkd_state_enum2friendly(state)); } else { rc = out->info(out, "Not connected to CIB%s%s", (desc != NULL)? ": " : "", pcmk__s(desc, "")); } out->finish(out, CRM_EX_DISCONNECT, true, NULL); return rc; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_xml(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = NULL; if (out->dest != stdout) { out->reset(out); } if (state != pcmk_pacemakerd_state_invalid) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } pcmk__output_create_xml_node(out, PCMK_XE_CRM_MON_DISCONNECTED, PCMK_XA_DESCRIPTION, desc, PCMK_XA_PACEMAKERD_STATE, state_s, NULL); out->finish(out, CRM_EX_DISCONNECT, true, NULL); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "crm-mon-disconnected", "default", crm_mon_disconnected_default }, { "crm-mon-disconnected", "html", crm_mon_disconnected_html }, { "crm-mon-disconnected", "text", crm_mon_disconnected_text }, { "crm-mon-disconnected", "xml", crm_mon_disconnected_xml }, { NULL, NULL, NULL }, }; #define RECONNECT_MSECS 5000 struct { guint reconnect_ms; enum mon_exec_mode exec_mode; gboolean fence_connect; gboolean print_pending; gboolean show_bans; gboolean watch_fencing; char *pid_file; char *external_agent; char *external_recipient; char *neg_location_prefix; char *only_node; char *only_rsc; GSList *user_includes_excludes; GSList *includes_excludes; } options = { .reconnect_ms = RECONNECT_MSECS, .exec_mode = mon_exec_unset, .fence_connect = TRUE, }; static crm_exit_t clean_up(crm_exit_t exit_code); static void crm_diff_update(const char *event, xmlNode * msg); static void clean_up_on_connection_failure(int rc); static int mon_refresh_display(gpointer user_data); static int setup_cib_connection(void); static int setup_fencer_connection(void); static int setup_api_connections(void); static void mon_st_callback_event(stonith_t * st, stonith_event_t * e); static void mon_st_callback_display(stonith_t * st, stonith_event_t * e); static void refresh_after_event(gboolean data_updated, gboolean enforce); static uint32_t all_includes(mon_output_format_t fmt) { if ((fmt == mon_output_plain) || (fmt == mon_output_console)) { return ~pcmk_section_options; } else { return pcmk_section_all; } } static uint32_t default_includes(mon_output_format_t fmt) { switch (fmt) { case mon_output_plain: case mon_output_console: case mon_output_html: return pcmk_section_summary |pcmk_section_nodes |pcmk_section_resources |pcmk_section_failures; case mon_output_xml: return all_includes(fmt); default: return 0; } } struct { const char *name; uint32_t bit; } sections[] = { { "attributes", pcmk_section_attributes }, { "bans", pcmk_section_bans }, { "counts", pcmk_section_counts }, { "dc", pcmk_section_dc }, { "failcounts", pcmk_section_failcounts }, { "failures", pcmk_section_failures }, { PCMK_VALUE_FENCING, pcmk_section_fencing_all }, { "fencing-failed", pcmk_section_fence_failed }, { "fencing-pending", pcmk_section_fence_pending }, { "fencing-succeeded", pcmk_section_fence_worked }, { "maint-mode", pcmk_section_maint_mode }, { "nodes", pcmk_section_nodes }, { "operations", pcmk_section_operations }, { "options", pcmk_section_options }, { "resources", pcmk_section_resources }, { "stack", pcmk_section_stack }, { "summary", pcmk_section_summary }, { "tickets", pcmk_section_tickets }, { "times", pcmk_section_times }, { NULL } }; static uint32_t find_section_bit(const char *name) { for (int i = 0; sections[i].name != NULL; i++) { if (pcmk__str_eq(sections[i].name, name, pcmk__str_casei)) { return sections[i].bit; } } return 0; } static gboolean apply_exclude(const gchar *excludes, GError **error) { char **parts = NULL; gboolean result = TRUE; parts = g_strsplit(excludes, ",", 0); for (char **s = parts; *s != NULL; s++) { uint32_t bit = find_section_bit(*s); if (pcmk__str_eq(*s, "all", pcmk__str_none)) { show = 0; } else if (pcmk__str_eq(*s, PCMK_VALUE_NONE, pcmk__str_none)) { show = all_includes(output_format); } else if (bit != 0) { show &= ~bit; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--exclude options: all, attributes, bans, counts, dc, " "failcounts, failures, fencing, fencing-failed, " "fencing-pending, fencing-succeeded, maint-mode, nodes, " PCMK_VALUE_NONE ", operations, options, resources, " "stack, summary, tickets, times"); result = FALSE; break; } } g_strfreev(parts); return result; } static gboolean apply_include(const gchar *includes, GError **error) { char **parts = NULL; gboolean result = TRUE; parts = g_strsplit(includes, ",", 0); for (char **s = parts; *s != NULL; s++) { uint32_t bit = find_section_bit(*s); if (pcmk__str_eq(*s, "all", pcmk__str_none)) { show = all_includes(output_format); - } else if (pcmk__starts_with(*s, "bans")) { + } else if (g_str_has_prefix(*s, "bans")) { show |= pcmk_section_bans; if (options.neg_location_prefix != NULL) { free(options.neg_location_prefix); options.neg_location_prefix = NULL; } if (strlen(*s) > 4 && (*s)[4] == ':') { options.neg_location_prefix = strdup(*s+5); } } else if (pcmk__str_any_of(*s, PCMK_VALUE_DEFAULT, "defaults", NULL)) { show |= default_includes(output_format); } else if (pcmk__str_eq(*s, PCMK_VALUE_NONE, pcmk__str_none)) { show = 0; } else if (bit != 0) { show |= bit; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--include options: all, attributes, bans[:PREFIX], counts, dc, " PCMK_VALUE_DEFAULT ", failcounts, failures, fencing, " "fencing-failed, fencing-pending, fencing-succeeded, " "maint-mode, nodes, " PCMK_VALUE_NONE ", operations, " "options, resources, stack, summary, tickets, times"); result = FALSE; break; } } g_strfreev(parts); return result; } static gboolean apply_include_exclude(GSList *lst, GError **error) { gboolean rc = TRUE; GSList *node = lst; while (node != NULL) { char *s = node->data; - if (pcmk__starts_with(s, "--include=")) { + if (s == NULL) { + } else if (g_str_has_prefix(s, "--include=")) { rc = apply_include(s+10, error); - } else if (pcmk__starts_with(s, "-I=")) { + } else if (g_str_has_prefix(s, "-I=")) { rc = apply_include(s+3, error); - } else if (pcmk__starts_with(s, "--exclude=")) { + } else if (g_str_has_prefix(s, "--exclude=")) { rc = apply_exclude(s+10, error); - } else if (pcmk__starts_with(s, "-U=")) { + } else if (g_str_has_prefix(s, "-U=")) { rc = apply_exclude(s+3, error); } if (rc != TRUE) { break; } node = node->next; } return rc; } static gboolean user_include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { char *s = crm_strdup_printf("%s=%s", option_name, optarg); options.user_includes_excludes = g_slist_append(options.user_includes_excludes, s); return TRUE; } static gboolean include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { char *s = crm_strdup_printf("%s=%s", option_name, optarg); options.includes_excludes = g_slist_append(options.includes_excludes, s); return TRUE; } static gboolean as_xml_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&args->output_ty, "xml"); output_format = mon_output_legacy_xml; return TRUE; } static gboolean fence_history_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (optarg == NULL) { interactive_fence_level = 2; } else { pcmk__scan_min_int(optarg, &interactive_fence_level, 0); } switch (interactive_fence_level) { case 3: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; return include_exclude_cb("--include", PCMK_VALUE_FENCING, data, err); case 2: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; return include_exclude_cb("--include", PCMK_VALUE_FENCING, data, err); case 1: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; return include_exclude_cb("--include", "fencing-failed,fencing-pending", data, err); case 0: options.fence_connect = FALSE; fence_history = pcmk__fence_history_none; return include_exclude_cb("--exclude", PCMK_VALUE_FENCING, data, err); default: g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Fence history must be 0-3"); return FALSE; } } static gboolean group_by_node_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_rscs_by_node; return TRUE; } static gboolean hide_headers_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--exclude", "summary", data, err); } static gboolean inactive_resources_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_inactive_rscs; return TRUE; } static gboolean print_brief_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_brief; return TRUE; } static gboolean print_detail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_details; return TRUE; } static gboolean print_description_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_description; return TRUE; } static gboolean print_timing_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_timing; return user_include_exclude_cb("--include", "operations", data, err); } static gboolean reconnect_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { int rc = crm_get_msec(optarg); if (rc == -1) { g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Invalid value for -i: %s", optarg); return FALSE; } else { pcmk_parse_interval_spec(optarg, &options.reconnect_ms); if (options.exec_mode != mon_exec_daemonized) { // Reconnect interval applies to daemonized too, so don't override options.exec_mode = mon_exec_update; } } return TRUE; } /*! * \internal * \brief Enable one-shot mode * * \param[in] option_name Name of option being parsed (ignored) * \param[in] optarg Value to be parsed (ignored) * \param[in] data User data (ignored) * \param[out] err Where to store error (ignored) */ static gboolean one_shot_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { options.exec_mode = mon_exec_one_shot; return TRUE; } /*! * \internal * \brief Enable daemonized mode * * \param[in] option_name Name of option being parsed (ignored) * \param[in] optarg Value to be parsed (ignored) * \param[in] data User data (ignored) * \param[out] err Where to store error (ignored) */ static gboolean daemonize_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { options.exec_mode = mon_exec_daemonized; return TRUE; } static gboolean show_attributes_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "attributes", data, err); } static gboolean show_bans_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (optarg != NULL) { char *s = crm_strdup_printf("bans:%s", optarg); gboolean rc = user_include_exclude_cb("--include", s, data, err); free(s); return rc; } else { return user_include_exclude_cb("--include", "bans", data, err); } } static gboolean show_failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "failcounts", data, err); } static gboolean show_operations_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "failcounts,operations", data, err); } static gboolean show_tickets_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "tickets", data, err); } static gboolean use_cib_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { setenv("CIB_file", optarg, 1); options.exec_mode = mon_exec_one_shot; return TRUE; } #define INDENT " " /* *INDENT-OFF* */ static GOptionEntry addl_entries[] = { { "interval", 'i', 0, G_OPTION_ARG_CALLBACK, reconnect_cb, "Update frequency (default is 5 seconds). Note: When run interactively\n" INDENT "on a live cluster, the display will be updated automatically\n" INDENT "whenever the cluster configuration or status changes.", "TIMESPEC" }, { "one-shot", '1', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, one_shot_cb, "Display the cluster status once and exit", NULL }, { "daemonize", 'd', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, daemonize_cb, "Run in the background as a daemon.\n" INDENT "Requires at least one of --output-to and --external-agent.", NULL }, { "pid-file", 'p', 0, G_OPTION_ARG_FILENAME, &options.pid_file, "(Advanced) Daemon pid file location", "FILE" }, { "external-agent", 'E', 0, G_OPTION_ARG_FILENAME, &options.external_agent, "A program to run when resource operations take place", "FILE" }, { "external-recipient", 'e', 0, G_OPTION_ARG_STRING, &options.external_recipient, "A recipient for your program (assuming you want the program to send something to someone).", "RCPT" }, { "watch-fencing", 'W', 0, G_OPTION_ARG_NONE, &options.watch_fencing, "Listen for fencing events. For use with --external-agent.", NULL }, { "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, use_cib_file_cb, NULL, NULL }, { NULL } }; static GOptionEntry display_entries[] = { { "include", 'I', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb, "A list of sections to include in the output.\n" INDENT "See `Output Control` help for more information.", "SECTION(s)" }, { "exclude", 'U', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb, "A list of sections to exclude from the output.\n" INDENT "See `Output Control` help for more information.", "SECTION(s)" }, { "node", 0, 0, G_OPTION_ARG_STRING, &options.only_node, "When displaying information about nodes, show only what's related to the given\n" INDENT "node, or to all nodes tagged with the given tag", "NODE" }, { "resource", 0, 0, G_OPTION_ARG_STRING, &options.only_rsc, "When displaying information about resources, show only what's related to the given\n" INDENT "resource, or to all resources tagged with the given tag", "RSC" }, { "group-by-node", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, group_by_node_cb, "Group resources by node", NULL }, { "inactive", 'r', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, inactive_resources_cb, "Display inactive resources", NULL }, { "failcounts", 'f', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_failcounts_cb, "Display resource fail counts", NULL }, { "operations", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_operations_cb, "Display resource operation history", NULL }, { "timing-details", 't', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_timing_cb, "Display resource operation history with timing details", NULL }, { "tickets", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_tickets_cb, "Display cluster tickets", NULL }, { "fence-history", 'm', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, fence_history_cb, "Show fence history:\n" INDENT "0=off, 1=failures and pending (default without option),\n" INDENT "2=add successes (default without value for option),\n" INDENT "3=show full history without reduction to most recent of each flavor", "LEVEL" }, { "neg-locations", 'L', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, show_bans_cb, "Display negative location constraints [optionally filtered by id prefix]", NULL }, { "show-node-attributes", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_attributes_cb, "Display node attributes", NULL }, { "hide-headers", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, hide_headers_cb, "Hide all headers", NULL }, { "show-detail", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_detail_cb, "Show more details (node IDs, individual clone instances)", NULL }, { "show-description", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_description_cb, "Show resource descriptions", NULL }, { "brief", 'b', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_brief_cb, "Brief output", NULL }, { "pending", 'j', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.print_pending, "Display pending state if '" PCMK_META_RECORD_PENDING "' is enabled", NULL }, { NULL } }; static GOptionEntry deprecated_entries[] = { /* @COMPAT resource-agents <4.15.0 uses --as-xml, so removing this option * must wait until we no longer support building on any platforms that ship * the older agents. */ { "as-xml", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_xml_cb, "Write cluster status as XML to stdout. This will enable one-shot mode.\n" INDENT "Use --output-as=xml instead.", NULL }, { NULL } }; /* *INDENT-ON* */ /* Reconnect to the CIB and fencing agent after reconnect_ms has passed. This sounds * like it would be more broadly useful, but only ever happens after a disconnect via * mon_cib_connection_destroy. */ static gboolean reconnect_after_timeout(gpointer data) { #if PCMK__ENABLE_CURSES if (output_format == mon_output_console) { clear(); refresh(); } #endif out->transient(out, "Reconnecting..."); if (setup_api_connections() == pcmk_rc_ok) { // Trigger redrawing the screen (needs reconnect_timer == 0) reconnect_timer = 0; refresh_after_event(FALSE, TRUE); return G_SOURCE_REMOVE; } out->message(out, "crm-mon-disconnected", "Latest connection attempt failed", pcmkd_state); reconnect_timer = pcmk__create_timer(options.reconnect_ms, reconnect_after_timeout, NULL); return G_SOURCE_REMOVE; } /* Called from various places when we are disconnected from the CIB or from the * fencing agent. If the CIB connection is still valid, this function will also * attempt to sign off and reconnect. */ static void mon_cib_connection_destroy(gpointer user_data) { const char *msg = "Connection to the cluster lost"; pcmkd_state = pcmk_pacemakerd_state_invalid; /* No crm-mon-disconnected message for console; a working implementation * is not currently worth the effort */ out->transient(out, "%s", msg); out->message(out, "crm-mon-disconnected", msg, pcmkd_state); if (refresh_timer != NULL) { /* we'll trigger a refresh after reconnect */ mainloop_timer_stop(refresh_timer); } if (reconnect_timer) { /* we'll trigger a new reconnect-timeout at the end */ g_source_remove(reconnect_timer); reconnect_timer = 0; } /* the client API won't properly reconnect notifications if they are still * in the table - so remove them */ if (st != NULL) { if (st->state != stonith_disconnected) { st->cmds->disconnect(st); } st->cmds->remove_notification(st, NULL); } if (cib) { cib->cmds->signoff(cib); reconnect_timer = pcmk__create_timer(options.reconnect_ms, reconnect_after_timeout, NULL); } } /* Signal handler installed into the mainloop for normal program shutdown */ static void mon_shutdown(int nsig) { clean_up(CRM_EX_OK); } #if PCMK__ENABLE_CURSES static volatile sighandler_t ncurses_winch_handler; /* Signal handler installed the regular way (not into the main loop) for when * the screen is resized. Commonly, this happens when running in an xterm and * the user changes its size. */ static void mon_winresize(int nsig) { static int not_done; int lines = 0, cols = 0; if (!not_done++) { if (ncurses_winch_handler) /* the original ncurses WINCH signal handler does the * magic of retrieving the new window size; * otherwise, we'd have to use ioctl or tgetent */ (*ncurses_winch_handler) (SIGWINCH); getmaxyx(stdscr, lines, cols); resizeterm(lines, cols); /* Alert the mainloop code we'd like the refresh_trigger to run next * time the mainloop gets around to checking. */ mainloop_set_trigger((crm_trigger_t *) refresh_trigger); } not_done--; } #endif static int setup_fencer_connection(void) { int rc = pcmk_ok; if (options.fence_connect && st == NULL) { st = stonith__api_new(); } if (!options.fence_connect || st == NULL || st->state != stonith_disconnected) { return rc; } rc = st->cmds->connect(st, crm_system_name, NULL); if (rc == pcmk_ok) { crm_trace("Setting up stonith callbacks"); if (options.watch_fencing) { st->cmds->register_notification(st, PCMK__VALUE_ST_NOTIFY_DISCONNECT, mon_st_callback_event); st->cmds->register_notification(st, PCMK__VALUE_ST_NOTIFY_FENCE, mon_st_callback_event); } else { st->cmds->register_notification(st, PCMK__VALUE_ST_NOTIFY_DISCONNECT, mon_st_callback_display); st->cmds->register_notification(st, PCMK__VALUE_ST_NOTIFY_HISTORY, mon_st_callback_display); } } else { stonith__api_free(st); st = NULL; } return rc; } static int setup_cib_connection(void) { int rc = pcmk_rc_ok; CRM_CHECK(cib != NULL, return EINVAL); if (cib->state != cib_disconnected) { // Already connected with notifications registered for CIB updates return rc; } rc = cib__signon_query(out, &cib, ¤t_cib); if (rc == pcmk_rc_ok) { rc = pcmk_legacy2rc(cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy)); if (rc == EPROTONOSUPPORT) { out->err(out, "CIB client does not support connection loss " "notifications; crm_mon will be unable to reconnect after " "connection loss"); rc = pcmk_rc_ok; } if (rc == pcmk_rc_ok) { cib->cmds->del_notify_callback(cib, PCMK__VALUE_CIB_DIFF_NOTIFY, crm_diff_update); rc = cib->cmds->add_notify_callback(cib, PCMK__VALUE_CIB_DIFF_NOTIFY, crm_diff_update); rc = pcmk_legacy2rc(rc); } if (rc != pcmk_rc_ok) { if (rc == EPROTONOSUPPORT) { out->err(out, "CIB client does not support CIB diff " "notifications"); } else { out->err(out, "CIB diff notification setup failed"); } out->err(out, "Cannot monitor CIB changes; exiting"); cib__clean_up_connection(&cib); stonith__api_free(st); st = NULL; } } return rc; } /* This is used to set up the fencing options after the interactive UI has been stared. * fence_history_cb can't be used because it builds up a list of includes/excludes that * then have to be processed with apply_include_exclude and that could affect other * things. */ static void set_fencing_options(int level) { switch (level) { case 3: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fencing_all; break; case 2: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fencing_all; break; case 1: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fence_failed | pcmk_section_fence_pending; break; default: interactive_fence_level = 0; options.fence_connect = FALSE; fence_history = pcmk__fence_history_none; show &= ~pcmk_section_fencing_all; break; } } static int setup_api_connections(void) { int rc = pcmk_rc_ok; CRM_CHECK(cib != NULL, return EINVAL); if (cib->state != cib_disconnected) { return rc; } if (cib->variant == cib_native) { rc = pcmk__pacemakerd_status(out, crm_system_name, options.reconnect_ms / 2, false, &pcmkd_state); if (rc != pcmk_rc_ok) { return rc; } switch (pcmkd_state) { case pcmk_pacemakerd_state_running: case pcmk_pacemakerd_state_remote: case pcmk_pacemakerd_state_shutting_down: /* Fencer and CIB may still be available while shutting down or * running on a Pacemaker Remote node */ break; default: // Fencer and CIB are definitely unavailable return ENOTCONN; } setup_fencer_connection(); } rc = setup_cib_connection(); return rc; } #if PCMK__ENABLE_CURSES static const char * get_option_desc(char c) { const char *desc = "No help available"; for (GOptionEntry *entry = display_entries; entry != NULL; entry++) { if (entry->short_name == c) { desc = entry->description; break; } } return desc; } #define print_option_help(out, option, condition) \ curses_formatted_printf(out, "%c %c: \t%s\n", ((condition)? '*': ' '), option, get_option_desc(option)); /* This function is called from the main loop when there is something to be read * on stdin, like an interactive user's keystroke. All it does is read the keystroke, * set flags (or show the page showing which keystrokes are valid), and redraw the * screen. It does not do anything with connections to the CIB or fencing agent * agent what would happen in mon_refresh_display. */ static gboolean detect_user_input(GIOChannel *channel, GIOCondition condition, gpointer user_data) { int c; gboolean config_mode = FALSE; gboolean rc = G_SOURCE_CONTINUE; /* If the attached pty device (pseudo-terminal) has been closed/deleted, * the condition (G_IO_IN | G_IO_ERR | G_IO_HUP) occurs. * Exit with an error, otherwise the process would persist in the * background and significantly raise the CPU usage. */ if ((condition & G_IO_ERR) && (condition & G_IO_HUP)) { rc = G_SOURCE_REMOVE; clean_up(CRM_EX_IOERR); } /* The connection/fd has been closed. Refresh the screen and remove this * event source hence ignore stdin. */ if (condition & (G_IO_HUP | G_IO_NVAL)) { rc = G_SOURCE_REMOVE; } if ((condition & G_IO_IN) == 0) { return rc; } while (1) { /* Get user input */ c = getchar(); switch (c) { case 'm': interactive_fence_level++; if (interactive_fence_level > 3) { interactive_fence_level = 0; } set_fencing_options(interactive_fence_level); break; case 'c': show ^= pcmk_section_tickets; break; case 'f': show ^= pcmk_section_failcounts; break; case 'n': show_opts ^= pcmk_show_rscs_by_node; break; case 'o': show ^= pcmk_section_operations; if (!pcmk_is_set(show, pcmk_section_operations)) { show_opts &= ~pcmk_show_timing; } break; case 'r': show_opts ^= pcmk_show_inactive_rscs; break; case 'R': show_opts ^= pcmk_show_details; break; case 't': show_opts ^= pcmk_show_timing; if (pcmk_is_set(show_opts, pcmk_show_timing)) { show |= pcmk_section_operations; } break; case 'A': show ^= pcmk_section_attributes; break; case 'L': show ^= pcmk_section_bans; break; case 'D': /* If any header is shown, clear them all, otherwise set them all */ if (pcmk_any_flags_set(show, pcmk_section_summary)) { show &= ~pcmk_section_summary; } else { show |= pcmk_section_summary; } /* Regardless, we don't show options in console mode. */ show &= ~pcmk_section_options; break; case 'b': show_opts ^= pcmk_show_brief; break; case 'j': show_opts ^= pcmk_show_pending; break; case '?': config_mode = TRUE; break; default: /* All other keys just redraw the screen. */ goto refresh; } if (!config_mode) goto refresh; clear(); refresh(); curses_formatted_printf(out, "%s", "Display option change mode\n"); print_option_help(out, 'c', pcmk_is_set(show, pcmk_section_tickets)); print_option_help(out, 'f', pcmk_is_set(show, pcmk_section_failcounts)); print_option_help(out, 'n', pcmk_is_set(show_opts, pcmk_show_rscs_by_node)); print_option_help(out, 'o', pcmk_is_set(show, pcmk_section_operations)); print_option_help(out, 'r', pcmk_is_set(show_opts, pcmk_show_inactive_rscs)); print_option_help(out, 't', pcmk_is_set(show_opts, pcmk_show_timing)); print_option_help(out, 'A', pcmk_is_set(show, pcmk_section_attributes)); print_option_help(out, 'L', pcmk_is_set(show, pcmk_section_bans)); print_option_help(out, 'D', !pcmk_is_set(show, pcmk_section_summary)); print_option_help(out, 'R', pcmk_any_flags_set(show_opts, pcmk_show_details)); print_option_help(out, 'b', pcmk_is_set(show_opts, pcmk_show_brief)); print_option_help(out, 'j', pcmk_is_set(show_opts, pcmk_show_pending)); curses_formatted_printf(out, "%d m: \t%s\n", interactive_fence_level, get_option_desc('m')); curses_formatted_printf(out, "%s", "\nToggle fields via field letter, type any other key to return\n"); } refresh: refresh_after_event(FALSE, TRUE); return rc; } #endif // PCMK__ENABLE_CURSES // Basically crm_signal_handler(SIGCHLD, SIG_IGN) plus the SA_NOCLDWAIT flag static void avoid_zombies(void) { struct sigaction sa; memset(&sa, 0, sizeof(struct sigaction)); if (sigemptyset(&sa.sa_mask) < 0) { crm_warn("Cannot avoid zombies: %s", pcmk_rc_str(errno)); return; } sa.sa_handler = SIG_IGN; sa.sa_flags = SA_RESTART|SA_NOCLDWAIT; if (sigaction(SIGCHLD, &sa, NULL) < 0) { crm_warn("Cannot avoid zombies: %s", pcmk_rc_str(errno)); } } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; GOptionEntry extra_prog_entries[] = { { "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet), "Be less descriptive in output.", NULL }, { NULL } }; #if PCMK__ENABLE_CURSES const char *fmts = "console (default), html, text, xml, none"; #else const char *fmts = "text (default), html, xml, none"; #endif // PCMK__ENABLE_CURSES const char *desc = NULL; desc = "Notes:\n\n" "Time Specification:\n\n" "The TIMESPEC in any command line option can be specified in many\n" "different formats. It can be an integer number of seconds, a\n" "number plus units (us/usec/ms/msec/s/sec/m/min/h/hr), or an ISO\n" "8601 period specification.\n\n" "Output Control:\n\n" "By default, a particular set of sections are written to the\n" "output destination. The default varies based on the output\n" "format: XML includes all sections by default, while other output\n" "formats include less. This set can be modified with the --include\n" "and --exclude command line options. Each option may be passed\n" "multiple times, and each can specify a comma-separated list of\n" "sections. The options are applied to the default set, in order\n" "from left to right as they are passed on the command line. For a\n" "list of valid sections, pass --include=list or --exclude=list.\n\n" "Interactive Use:\n\n" #if PCMK__ENABLE_CURSES "When run interactively, crm_mon can be told to hide and show\n" "various sections of output. To see a help screen explaining the\n" "options, press '?'. Any key stroke aside from those listed will\n" "cause the screen to refresh.\n\n" #else "The local installation of Pacemaker was built without support for\n" "interactive (console) mode. A curses library must be available at\n" "build time to support interactive mode.\n\n" #endif // PCMK__ENABLE_CURSES "Examples:\n\n" #if PCMK__ENABLE_CURSES "Display the cluster status on the console with updates as they\n" "occur:\n\n" "\tcrm_mon\n\n" #endif // PCMK__ENABLE_CURSES "Display the cluster status once and exit:\n\n" "\tcrm_mon -1\n\n" "Display the cluster status, group resources by node, and include\n" "inactive resources in the list:\n\n" "\tcrm_mon --group-by-node --inactive\n\n" "Start crm_mon as a background daemon and have it write the\n" "cluster status to an HTML file:\n\n" "\tcrm_mon --daemonize --output-as html " "--output-to /path/to/docroot/filename.html\n\n" "Display the cluster status as XML:\n\n" "\tcrm_mon --output-as xml\n\n"; context = pcmk__build_arg_context(args, fmts, group, NULL); pcmk__add_main_args(context, extra_prog_entries); g_option_context_set_description(context, desc); pcmk__add_arg_group(context, "display", "Display Options:", "Show display options", display_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); pcmk__add_arg_group(context, "deprecated", "Deprecated Options:", "Show deprecated options", deprecated_entries); return context; } /*! * \internal * \brief Set output format based on \c --output-as arguments and mode arguments * * When the deprecated \c --as-xml argument is parsed, a callback function sets * \c output_format. Otherwise, this function does the same based on the current * \c --output-as arguments and the \c --one-shot and \c --daemonize arguments. * * \param[in,out] args Command line arguments */ static void reconcile_output_format(pcmk__common_args_t *args) { if (output_format != mon_output_unset) { /* The deprecated --as-xml argument was used, and we're finished. Note * that this means the deprecated argument takes precedence. */ return; } if (pcmk__str_eq(args->output_ty, PCMK_VALUE_NONE, pcmk__str_none)) { output_format = mon_output_none; } else if (pcmk__str_eq(args->output_ty, "html", pcmk__str_none)) { output_format = mon_output_html; umask(S_IWGRP | S_IWOTH); // World-readable HTML } else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) { output_format = mon_output_xml; #if PCMK__ENABLE_CURSES } else if (pcmk__str_eq(args->output_ty, "console", pcmk__str_null_matches)) { /* Console is the default format if no conflicting options are given. * * Use text output instead if one of the following conditions is met: * * We've requested daemonized or one-shot mode (console output is * incompatible with modes other than mon_exec_update) * * We requested the version, which is effectively one-shot * * We specified a non-stdout output destination (console mode is * compatible only with stdout) */ if ((options.exec_mode == mon_exec_daemonized) || (options.exec_mode == mon_exec_one_shot) || args->version || !pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } else { pcmk__str_update(&args->output_ty, "console"); output_format = mon_output_console; crm_enable_stderr(FALSE); } #endif // PCMK__ENABLE_CURSES } else if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches)) { /* Text output was explicitly requested, or it's the default because * curses is not enabled */ pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } // Otherwise, invalid format. Let pcmk__output_new() throw an error. } /*! * \internal * \brief Set execution mode to the output format's default if appropriate * * \param[in,out] args Command line arguments */ static void set_default_exec_mode(const pcmk__common_args_t *args) { if (output_format == mon_output_console) { /* Update is the only valid mode for console, but set here instead of * reconcile_output_format() for isolation and consistency */ options.exec_mode = mon_exec_update; } else if (options.exec_mode == mon_exec_unset) { // Default to one-shot mode for all other formats options.exec_mode = mon_exec_one_shot; } else if ((options.exec_mode == mon_exec_update) && pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { // If not using console format, update mode cannot be used with stdout options.exec_mode = mon_exec_one_shot; } } static void clean_up_on_connection_failure(int rc) { if (rc == ENOTCONN) { if (pcmkd_state == pcmk_pacemakerd_state_remote) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error: remote-node not connected to cluster"); } else { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error: cluster is not available on this node"); } } else { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Connection to cluster failed: %s", pcmk_rc_str(rc)); } clean_up(pcmk_rc2exitc(rc)); } static void one_shot(void) { int rc = pcmk__status(out, cib, fence_history, show, show_opts, options.only_node, options.only_rsc, options.neg_location_prefix, 0); if (rc == pcmk_rc_ok) { clean_up(pcmk_rc2exitc(rc)); } else { clean_up_on_connection_failure(rc); } } static void exit_on_invalid_cib(void) { if (cib != NULL) { return; } // Shouldn't really be possible g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Invalid CIB source"); clean_up(CRM_EX_ERROR); } int main(int argc, char **argv) { int rc = pcmk_rc_ok; GOptionGroup *output_group = NULL; args = pcmk__new_common_args(SUMMARY); context = build_arg_context(args, &output_group); pcmk__register_formats(output_group, formats); options.pid_file = strdup("/tmp/ClusterMon.pid"); pcmk__cli_init_logging("crm_mon", 0); // Avoid needing to wait for subprocesses forked for -E/--external-agent avoid_zombies(); processed_args = pcmk__cmdline_preproc(argv, "eimpxEILU"); fence_history_cb("--fence-history", "1", NULL, NULL); /* Set an HTML title regardless of what format we will eventually use. * Doing this here means the user can give their own title on the command * line. */ if (!pcmk__force_args(context, &error, "%s --html-title \"Cluster Status\"", g_get_prgname())) { return clean_up(CRM_EX_USAGE); } if (!g_option_context_parse_strv(context, &processed_args, &error)) { return clean_up(CRM_EX_USAGE); } for (int i = 0; i < args->verbosity; i++) { crm_bump_log_level(argc, argv); } if (!args->version) { if (args->quiet) { include_exclude_cb("--exclude", "times", NULL, NULL); } if (options.watch_fencing) { fence_history_cb("--fence-history", "0", NULL, NULL); options.fence_connect = TRUE; } /* create the cib-object early to be able to do further * decisions based on the cib-source */ cib = cib_new(); exit_on_invalid_cib(); switch (cib->variant) { case cib_native: // Everything (fencer, CIB, pcmkd status) should be available break; case cib_file: // Live fence history is not meaningful fence_history_cb("--fence-history", "0", NULL, NULL); /* Notifications are unsupported; nothing to monitor * @COMPAT: Let setup_cib_connection() handle this by exiting? */ options.exec_mode = mon_exec_one_shot; break; case cib_remote: // We won't receive any fencing updates fence_history_cb("--fence-history", "0", NULL, NULL); break; default: /* something is odd */ exit_on_invalid_cib(); break; } if ((options.exec_mode == mon_exec_daemonized) && !options.external_agent && pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires at least one of --output-to " "(with value not set to '-') and --external-agent"); return clean_up(CRM_EX_USAGE); } } reconcile_output_format(args); set_default_exec_mode(args); rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); return clean_up(CRM_EX_ERROR); } if (output_format == mon_output_legacy_xml) { output_format = mon_output_xml; pcmk__output_set_legacy_xml(out); } /* output_format MUST NOT BE CHANGED AFTER THIS POINT. */ /* If we had a valid format for pcmk__output_new(), output_format should be * set by now. */ pcmk__assert(output_format != mon_output_unset); if (output_format == mon_output_plain) { pcmk__output_text_set_fancy(out, true); } if (options.exec_mode == mon_exec_daemonized) { if (!options.external_agent && (output_format == mon_output_none)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires --external-agent if used with " "--output-as=none"); return clean_up(CRM_EX_USAGE); } crm_enable_stderr(FALSE); cib_delete(cib); cib = NULL; pcmk__daemonize(crm_system_name, options.pid_file); cib = cib_new(); exit_on_invalid_cib(); } show = default_includes(output_format); /* Apply --include/--exclude flags we used internally. There's no error reporting * here because this would be a programming error. */ apply_include_exclude(options.includes_excludes, &error); /* And now apply any --include/--exclude flags the user gave on the command line. * These are done in a separate pass from the internal ones because we want to * make sure whatever the user specifies overrides whatever we do. */ if (!apply_include_exclude(options.user_includes_excludes, &error)) { return clean_up(CRM_EX_USAGE); } /* Sync up the initial value of interactive_fence_level with whatever was set with * --include/--exclude= options. */ if (pcmk_all_flags_set(show, pcmk_section_fencing_all)) { interactive_fence_level = 3; } else if (pcmk_is_set(show, pcmk_section_fence_worked)) { interactive_fence_level = 2; } else if (pcmk_any_flags_set(show, pcmk_section_fence_failed | pcmk_section_fence_pending)) { interactive_fence_level = 1; } else { interactive_fence_level = 0; } pcmk__register_lib_messages(out); crm_mon_register_messages(out); pe__register_messages(out); stonith__register_messages(out); // Messages internal to this file, nothing curses-specific pcmk__register_messages(out, fmt_functions); if (args->version) { out->version(out); return clean_up(CRM_EX_OK); } if (output_format == mon_output_xml) { show_opts |= pcmk_show_inactive_rscs | pcmk_show_timing; } if ((output_format == mon_output_html) && (out->dest != stdout)) { char *content = pcmk__itoa(pcmk__timeout_ms2s(options.reconnect_ms)); pcmk__html_add_header(PCMK__XE_META, PCMK__XA_HTTP_EQUIV, PCMK__VALUE_REFRESH, PCMK__XA_CONTENT, content, NULL); free(content); } crm_info("Starting %s", crm_system_name); cib__set_output(cib, out); if (options.exec_mode == mon_exec_one_shot) { one_shot(); } scheduler = pcmk_new_scheduler(); pcmk__mem_assert(scheduler); scheduler->priv->out = out; if ((cib->variant == cib_native) && pcmk_is_set(show, pcmk_section_times)) { // Currently used only in the times section pcmk__query_node_name(out, 0, &(scheduler->priv->local_node_name), 0); } out->message(out, "crm-mon-disconnected", "Waiting for initial connection", pcmkd_state); do { out->transient(out, "Connecting to cluster..."); rc = setup_api_connections(); if (rc != pcmk_rc_ok) { if ((rc == ENOTCONN) || (rc == ECONNREFUSED)) { out->transient(out, "Connection failed. Retrying in %s...", pcmk__readable_interval(options.reconnect_ms)); } // Give some time to view all output even if we won't retry pcmk__sleep_ms(options.reconnect_ms); #if PCMK__ENABLE_CURSES if (output_format == mon_output_console) { clear(); refresh(); } #endif } } while ((rc == ENOTCONN) || (rc == ECONNREFUSED)); if (rc != pcmk_rc_ok) { clean_up_on_connection_failure(rc); } set_fencing_options(interactive_fence_level); mon_refresh_display(NULL); mainloop = g_main_loop_new(NULL, FALSE); mainloop_add_signal(SIGTERM, mon_shutdown); mainloop_add_signal(SIGINT, mon_shutdown); #if PCMK__ENABLE_CURSES if (output_format == mon_output_console) { ncurses_winch_handler = crm_signal_handler(SIGWINCH, mon_winresize); if (ncurses_winch_handler == SIG_DFL || ncurses_winch_handler == SIG_IGN || ncurses_winch_handler == SIG_ERR) ncurses_winch_handler = NULL; io_channel = g_io_channel_unix_new(STDIN_FILENO); g_io_add_watch(io_channel, (G_IO_IN | G_IO_ERR | G_IO_HUP | G_IO_NVAL), detect_user_input, NULL); } #endif /* When refresh_trigger->trigger is set to TRUE, call mon_refresh_display. In * this file, that is anywhere mainloop_set_trigger is called. */ refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_display, NULL); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); crm_info("Exiting %s", crm_system_name); return clean_up(CRM_EX_OK); } static int send_custom_trap(const char *node, const char *rsc, const char *task, int target_rc, int rc, int status, const char *desc) { pid_t pid; /*setenv needs chars, these are ints */ char *rc_s = pcmk__itoa(rc); char *status_s = pcmk__itoa(status); char *target_rc_s = pcmk__itoa(target_rc); crm_debug("Sending external notification to '%s' via '%s'", options.external_recipient, options.external_agent); if(rsc) { setenv("CRM_notify_rsc", rsc, 1); } if (options.external_recipient) { setenv("CRM_notify_recipient", options.external_recipient, 1); } setenv("CRM_notify_node", node, 1); setenv("CRM_notify_task", task, 1); setenv("CRM_notify_desc", desc, 1); setenv("CRM_notify_rc", rc_s, 1); setenv("CRM_notify_target_rc", target_rc_s, 1); setenv("CRM_notify_status", status_s, 1); pid = fork(); if (pid == -1) { out->err(out, "notification fork() failed: %s", strerror(errno)); } if (pid == 0) { /* crm_debug("notification: I am the child. Executing the nofitication program."); */ execl(options.external_agent, options.external_agent, NULL); crm_exit(CRM_EX_ERROR); } crm_trace("Finished running custom notification program '%s'.", options.external_agent); free(target_rc_s); free(status_s); free(rc_s); return 0; } static int handle_rsc_op(xmlNode *xml, void *userdata) { const char *node_id = (const char *) userdata; int rc = -1; int status = -1; int target_rc = -1; gboolean notify = TRUE; char *rsc = NULL; char *task = NULL; const char *desc = NULL; const char *magic = NULL; const char *id = NULL; const char *node = NULL; xmlNode *n = xml; xmlNode * rsc_op = xml; if(strcmp((const char*)xml->name, PCMK__XE_LRM_RSC_OP) != 0) { pcmk__xe_foreach_child(xml, NULL, handle_rsc_op, (void *) node_id); return pcmk_rc_ok; } id = pcmk__xe_history_key(rsc_op); magic = pcmk__xe_get(rsc_op, PCMK__XA_TRANSITION_MAGIC); if (magic == NULL) { /* non-change */ return pcmk_rc_ok; } if (!decode_transition_magic(magic, NULL, NULL, NULL, &status, &rc, &target_rc)) { crm_err("Invalid event %s detected for %s", magic, id); return pcmk_rc_ok; } if (parse_op_key(id, &rsc, &task, NULL) == FALSE) { crm_err("Invalid event detected for %s", id); goto bail; } node = pcmk__xe_get(rsc_op, PCMK__META_ON_NODE); while ((n != NULL) && !pcmk__xe_is(n, PCMK__XE_NODE_STATE)) { n = n->parent; } if(node == NULL && n) { node = pcmk__xe_get(n, PCMK_XA_UNAME); } if (node == NULL && n) { node = pcmk__xe_id(n); } if (node == NULL) { node = node_id; } if (node == NULL) { crm_err("No node detected for event %s (%s)", magic, id); goto bail; } /* look up where we expected it to be? */ desc = pcmk_rc_str(pcmk_rc_ok); if ((status == PCMK_EXEC_DONE) && (target_rc == rc)) { crm_notice("%s of %s on %s completed: %s", task, rsc, node, desc); if (rc == PCMK_OCF_NOT_RUNNING) { notify = FALSE; } } else if (status == PCMK_EXEC_DONE) { desc = crm_exit_str(rc); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } else { desc = pcmk_exec_status_str(status); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } if (notify && options.external_agent) { send_custom_trap(node, rsc, task, target_rc, rc, status, desc); } bail: free(rsc); free(task); return pcmk_rc_ok; } /* This function is just a wrapper around mainloop_set_trigger so that it can be * called from a mainloop directly. It's simply another way of ensuring the screen * gets redrawn. */ static gboolean mon_trigger_refresh(gpointer user_data) { mainloop_set_trigger((crm_trigger_t *) refresh_trigger); return FALSE; } static int handle_op_for_node(xmlNode *xml, void *userdata) { const char *node = pcmk__xe_get(xml, PCMK_XA_UNAME); if (node == NULL) { node = pcmk__xe_id(xml); } handle_rsc_op(xml, (void *) node); return pcmk_rc_ok; } static int crm_diff_update_element(xmlNode *change, void *userdata) { const char *name = NULL; const char *op = pcmk__xe_get(change, PCMK_XA_OPERATION); const char *xpath = pcmk__xe_get(change, PCMK_XA_PATH); xmlNode *match = NULL; const char *node = NULL; if (op == NULL) { return pcmk_rc_ok; } else if (strcmp(op, PCMK_VALUE_CREATE) == 0) { match = change->children; } else if (pcmk__str_any_of(op, PCMK_VALUE_MOVE, PCMK_VALUE_DELETE, NULL)) { return pcmk_rc_ok; } else if (strcmp(op, PCMK_VALUE_MODIFY) == 0) { match = pcmk__xe_first_child(change, PCMK_XE_CHANGE_RESULT, NULL, NULL); if(match) { match = match->children; } } if(match) { name = (const char *)match->name; } crm_trace("Handling %s operation for %s %p, %s", op, xpath, match, name); if(xpath == NULL) { /* Version field, ignore */ } else if(name == NULL) { crm_debug("No result for %s operation to %s", op, xpath); pcmk__assert(pcmk__str_any_of(op, PCMK_VALUE_MOVE, PCMK_VALUE_DELETE, NULL)); } else if (strcmp(name, PCMK_XE_CIB) == 0) { pcmk__xe_foreach_child(pcmk__xe_first_child(match, PCMK_XE_STATUS, NULL, NULL), NULL, handle_op_for_node, NULL); } else if (strcmp(name, PCMK_XE_STATUS) == 0) { pcmk__xe_foreach_child(match, NULL, handle_op_for_node, NULL); } else if (strcmp(name, PCMK__XE_NODE_STATE) == 0) { node = pcmk__xe_get(match, PCMK_XA_UNAME); if (node == NULL) { node = pcmk__xe_id(match); } handle_rsc_op(match, (void *) node); } else if (strcmp(name, PCMK__XE_LRM) == 0) { node = pcmk__xe_id(match); handle_rsc_op(match, (void *) node); } else if (strcmp(name, PCMK__XE_LRM_RESOURCES) == 0) { char *local_node = pcmk__xpath_node_id(xpath, PCMK__XE_LRM); handle_rsc_op(match, local_node); free(local_node); } else if (strcmp(name, PCMK__XE_LRM_RESOURCE) == 0) { char *local_node = pcmk__xpath_node_id(xpath, PCMK__XE_LRM); handle_rsc_op(match, local_node); free(local_node); } else if (strcmp(name, PCMK__XE_LRM_RSC_OP) == 0) { char *local_node = pcmk__xpath_node_id(xpath, PCMK__XE_LRM); handle_rsc_op(match, local_node); free(local_node); } else { crm_trace("Ignoring %s operation for %s %p, %s", op, xpath, match, name); } return pcmk_rc_ok; } static void crm_diff_update(const char *event, xmlNode * msg) { int rc = -1; static bool stale = FALSE; gboolean cib_updated = FALSE; xmlNode *wrapper = pcmk__xe_first_child(msg, PCMK__XE_CIB_UPDATE_RESULT, NULL, NULL); xmlNode *diff = pcmk__xe_first_child(wrapper, NULL, NULL, NULL); out->progress(out, false); if (current_cib != NULL) { rc = xml_apply_patchset(current_cib, diff, TRUE); switch (rc) { case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); pcmk__xml_free(current_cib); current_cib = NULL; break; case pcmk_ok: cib_updated = TRUE; break; default: crm_notice("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); pcmk__xml_free(current_cib); current_cib = NULL; } } if (current_cib == NULL) { crm_trace("Re-requesting the full cib"); cib->cmds->query(cib, NULL, ¤t_cib, cib_sync_call); } if (options.external_agent) { int format = 0; pcmk__xe_get_int(diff, PCMK_XA_FORMAT, &format); if (format == 2) { xmlNode *wrapper = pcmk__xe_first_child(msg, PCMK__XE_CIB_UPDATE_RESULT, NULL, NULL); xmlNode *diff = pcmk__xe_first_child(wrapper, NULL, NULL, NULL); pcmk__xe_foreach_child(diff, NULL, crm_diff_update_element, NULL); } else { crm_err("Unknown patch format: %d", format); } } if (current_cib == NULL) { if(!stale) { out->info(out, "--- Stale data ---"); } stale = TRUE; return; } stale = FALSE; refresh_after_event(cib_updated, FALSE); } static int mon_refresh_display(gpointer user_data) { int rc = pcmk_rc_ok; last_refresh = time(NULL); if (output_format == mon_output_none) { return G_SOURCE_REMOVE; } if (fence_history == pcmk__fence_history_full && !pcmk_all_flags_set(show, pcmk_section_fencing_all) && output_format != mon_output_xml) { fence_history = pcmk__fence_history_reduced; } // Get an up-to-date pacemakerd status for the cluster summary if (cib->variant == cib_native) { pcmk__pacemakerd_status(out, crm_system_name, options.reconnect_ms / 2, false, &pcmkd_state); } if (out->dest != stdout) { out->reset(out); } rc = pcmk__output_cluster_status(scheduler, st, cib, current_cib, pcmkd_state, fence_history, show, show_opts, options.only_node,options.only_rsc, options.neg_location_prefix); if (rc == pcmk_rc_schema_validation) { clean_up(CRM_EX_CONFIG); return G_SOURCE_REMOVE; } if (out->dest != stdout) { out->finish(out, CRM_EX_OK, true, NULL); } return G_SOURCE_CONTINUE; } /* This function is called for fencing events (see setup_fencer_connection() for * which ones) when --watch-fencing is used on the command line */ static void mon_st_callback_event(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else if (options.external_agent) { char *desc = stonith__event_description(e); send_custom_trap(e->target, NULL, e->operation, pcmk_ok, e->result, 0, desc); free(desc); } } /* Cause the screen to be redrawn (via mainloop_set_trigger) when various conditions are met: * * - If the last update occurred more than reconnect_ms ago (defaults to 5s, but * can be changed via the -i command line option), or * - After every 10 CIB updates, or * - If it's been 2s since the last update * * This function sounds like it would be more broadly useful, but it is only called when a * fencing event is received or a CIB diff occurrs. */ static void refresh_after_event(gboolean data_updated, gboolean enforce) { static int updates = 0; time_t now = time(NULL); if (data_updated) { updates++; } if(refresh_timer == NULL) { refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL); } if (reconnect_timer > 0) { /* we will receive a refresh request after successful reconnect */ mainloop_timer_stop(refresh_timer); return; } /* as we're not handling initial failure of fencer-connection as * fatal give it a retry here * not getting here if cib-reconnection is already on the way */ setup_fencer_connection(); if (enforce || ((now - last_refresh) > pcmk__timeout_ms2s(options.reconnect_ms)) || updates >= 10) { mainloop_set_trigger((crm_trigger_t *) refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else { mainloop_timer_start(refresh_timer); } } /* This function is called for fencing events (see setup_fencer_connection() for * which ones) when --watch-fencing is NOT used on the command line */ static void mon_st_callback_display(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else { out->progress(out, false); refresh_after_event(TRUE, FALSE); } } /* * De-init ncurses, disconnect from the CIB manager, disconnect fencing, * deallocate memory and show usage-message if requested. * * We don't actually return, but nominally returning crm_exit_t allows a usage * like "return clean_up(exit_code);" which helps static analysis understand the * code flow. */ static crm_exit_t clean_up(crm_exit_t exit_code) { /* Quitting crm_mon is much more complicated than it ought to be. */ /* (1) Close connections, free things, etc. */ if (io_channel != NULL) { g_io_channel_shutdown(io_channel, TRUE, NULL); } cib__clean_up_connection(&cib); stonith__api_free(st); free(options.neg_location_prefix); free(options.only_node); free(options.only_rsc); free(options.pid_file); g_slist_free_full(options.includes_excludes, free); g_strfreev(processed_args); pcmk_free_scheduler(scheduler); /* (2) If this is abnormal termination and we're in curses mode, shut down * curses first. Any messages displayed to the screen before curses is shut * down will be lost because doing the shut down will also restore the * screen to whatever it looked like before crm_mon was started. */ if (((error != NULL) || (exit_code == CRM_EX_USAGE)) && (output_format == mon_output_console) && (out != NULL)) { out->finish(out, exit_code, false, NULL); pcmk__output_free(out); out = NULL; } /* (3) If this is a command line usage related failure, print the usage * message. */ if (exit_code == CRM_EX_USAGE && (output_format == mon_output_console || output_format == mon_output_plain)) { char *help = g_option_context_get_help(context, TRUE, NULL); fprintf(stderr, "%s", help); g_free(help); } pcmk__free_arg_context(context); /* (4) If this is any kind of error, print the error out and exit. Make * sure to handle situations both before and after formatted output is * set up. We want errors to appear formatted if at all possible. */ if (error != NULL) { if (out != NULL) { out->err(out, "%s: %s", g_get_prgname(), error->message); out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } else { fprintf(stderr, "%s: %s\n", g_get_prgname(), error->message); } g_clear_error(&error); crm_exit(exit_code); } /* (5) Print formatted output to the screen if we made it far enough in * crm_mon to be able to do so. */ if (out != NULL) { if (options.exec_mode != mon_exec_daemonized) { out->finish(out, exit_code, true, NULL); } pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } diff --git a/tools/crm_resource_ban.c b/tools/crm_resource_ban.c index 9de379fbca..5edbae076e 100644 --- a/tools/crm_resource_ban.c +++ b/tools/crm_resource_ban.c @@ -1,523 +1,527 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include +#include // g_str_has_prefix() #include // xmlNode #include // xmlXPathObject, etc. #include static char * parse_cli_lifetime(pcmk__output_t *out, const char *move_lifetime) { char *later_s = NULL; crm_time_t *now = NULL; crm_time_t *later = NULL; crm_time_t *duration = NULL; if (move_lifetime == NULL) { return NULL; } duration = crm_time_parse_duration(move_lifetime); if (duration == NULL) { out->err(out, "Invalid duration specified: %s\n" "Please refer to https://en.wikipedia.org/wiki/ISO_8601#Durations " "for examples of valid durations", move_lifetime); return NULL; } now = crm_time_new(NULL); later = crm_time_add(now, duration); if (later == NULL) { out->err(out, "Unable to add %s to current time\n" "Please report to " PACKAGE_BUGREPORT " as possible bug", move_lifetime); crm_time_free(now); crm_time_free(duration); return NULL; } crm_time_log(LOG_INFO, "now ", now, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_INFO, "later ", later, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); crm_time_log(LOG_INFO, "duration", duration, crm_time_log_date | crm_time_log_timeofday); later_s = crm_time_as_string(later, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); out->info(out, "Migration will take effect until: %s", later_s); crm_time_free(duration); crm_time_free(later); crm_time_free(now); return later_s; } // \return Standard Pacemaker return code int cli_resource_ban(pcmk__output_t *out, const char *rsc_id, const char *host, const char *move_lifetime, cib_t *cib_conn, gboolean promoted_role_only, const char *promoted_role) { char *later_s = NULL; int rc = pcmk_rc_ok; xmlNode *fragment = NULL; xmlNode *location = NULL; later_s = parse_cli_lifetime(out, move_lifetime); if(move_lifetime && later_s == NULL) { return EINVAL; } fragment = pcmk__xe_create(NULL, PCMK_XE_CONSTRAINTS); location = pcmk__xe_create(fragment, PCMK_XE_RSC_LOCATION); pcmk__xe_set_id(location, "cli-ban-%s-on-%s", rsc_id, host); out->info(out, "WARNING: Creating " PCMK_XE_RSC_LOCATION " constraint '%s' with " "a score of " PCMK_VALUE_MINUS_INFINITY " for resource %s on %s." "\n\tThis will prevent %s from %s on %s until the constraint is " "removed using the clear option or by editing the CIB with an " "appropriate tool.\n" "\tThis will be the case even if %s is the last node in the " "cluster", pcmk__xe_id(location), rsc_id, host, rsc_id, (promoted_role_only? "being promoted" : "running"), host, host); crm_xml_add(location, PCMK_XA_RSC, rsc_id); if(promoted_role_only) { crm_xml_add(location, PCMK_XA_ROLE, promoted_role); } else { crm_xml_add(location, PCMK_XA_ROLE, PCMK_ROLE_STARTED); } if (later_s == NULL) { /* Short form */ crm_xml_add(location, PCMK_XE_NODE, host); crm_xml_add(location, PCMK_XA_SCORE, PCMK_VALUE_MINUS_INFINITY); } else { xmlNode *rule = pcmk__xe_create(location, PCMK_XE_RULE); xmlNode *expr = pcmk__xe_create(rule, PCMK_XE_EXPRESSION); pcmk__xe_set_id(rule, "cli-ban-%s-on-%s-rule", rsc_id, host); crm_xml_add(rule, PCMK_XA_SCORE, PCMK_VALUE_MINUS_INFINITY); crm_xml_add(rule, PCMK_XA_BOOLEAN_OP, PCMK_VALUE_AND); pcmk__xe_set_id(expr, "cli-ban-%s-on-%s-expr", rsc_id, host); crm_xml_add(expr, PCMK_XA_ATTRIBUTE, CRM_ATTR_UNAME); crm_xml_add(expr, PCMK_XA_OPERATION, PCMK_VALUE_EQ); crm_xml_add(expr, PCMK_XA_VALUE, host); crm_xml_add(expr, PCMK_XA_TYPE, PCMK_VALUE_STRING); expr = pcmk__xe_create(rule, PCMK_XE_DATE_EXPRESSION); pcmk__xe_set_id(expr, "cli-ban-%s-on-%s-lifetime", rsc_id, host); crm_xml_add(expr, PCMK_XA_OPERATION, PCMK_VALUE_LT); crm_xml_add(expr, PCMK_XA_END, later_s); } crm_log_xml_notice(fragment, "Modify"); rc = cib_conn->cmds->modify(cib_conn, PCMK_XE_CONSTRAINTS, fragment, cib_sync_call); rc = pcmk_legacy2rc(rc); pcmk__xml_free(fragment); free(later_s); if ((rc != pcmk_rc_ok) && promoted_role_only && (strcmp(promoted_role, PCMK_ROLE_PROMOTED) == 0)) { int banrc = cli_resource_ban(out, rsc_id, host, move_lifetime, cib_conn, promoted_role_only, PCMK__ROLE_PROMOTED_LEGACY); if (banrc == pcmk_rc_ok) { rc = banrc; } } return rc; } // \return Standard Pacemaker return code int cli_resource_prefer(pcmk__output_t *out,const char *rsc_id, const char *host, const char *move_lifetime, cib_t *cib_conn, gboolean promoted_role_only, const char *promoted_role) { char *later_s = parse_cli_lifetime(out, move_lifetime); int rc = pcmk_rc_ok; xmlNode *location = NULL; xmlNode *fragment = NULL; if(move_lifetime && later_s == NULL) { return EINVAL; } if(cib_conn == NULL) { free(later_s); return ENOTCONN; } fragment = pcmk__xe_create(NULL, PCMK_XE_CONSTRAINTS); location = pcmk__xe_create(fragment, PCMK_XE_RSC_LOCATION); pcmk__xe_set_id(location, "cli-prefer-%s", rsc_id); crm_xml_add(location, PCMK_XA_RSC, rsc_id); if(promoted_role_only) { crm_xml_add(location, PCMK_XA_ROLE, promoted_role); } else { crm_xml_add(location, PCMK_XA_ROLE, PCMK_ROLE_STARTED); } if (later_s == NULL) { /* Short form */ crm_xml_add(location, PCMK_XE_NODE, host); crm_xml_add(location, PCMK_XA_SCORE, PCMK_VALUE_INFINITY); } else { xmlNode *rule = pcmk__xe_create(location, PCMK_XE_RULE); xmlNode *expr = pcmk__xe_create(rule, PCMK_XE_EXPRESSION); pcmk__xe_set_id(rule, "cli-prefer-rule-%s", rsc_id); crm_xml_add(rule, PCMK_XA_SCORE, PCMK_VALUE_INFINITY); crm_xml_add(rule, PCMK_XA_BOOLEAN_OP, PCMK_VALUE_AND); pcmk__xe_set_id(expr, "cli-prefer-expr-%s", rsc_id); crm_xml_add(expr, PCMK_XA_ATTRIBUTE, CRM_ATTR_UNAME); crm_xml_add(expr, PCMK_XA_OPERATION, PCMK_VALUE_EQ); crm_xml_add(expr, PCMK_XA_VALUE, host); crm_xml_add(expr, PCMK_XA_TYPE, PCMK_VALUE_STRING); expr = pcmk__xe_create(rule, PCMK_XE_DATE_EXPRESSION); pcmk__xe_set_id(expr, "cli-prefer-lifetime-end-%s", rsc_id); crm_xml_add(expr, PCMK_XA_OPERATION, PCMK_VALUE_LT); crm_xml_add(expr, PCMK_XA_END, later_s); } crm_log_xml_info(fragment, "Modify"); rc = cib_conn->cmds->modify(cib_conn, PCMK_XE_CONSTRAINTS, fragment, cib_sync_call); rc = pcmk_legacy2rc(rc); pcmk__xml_free(fragment); free(later_s); if ((rc != pcmk_rc_ok) && promoted_role_only && (strcmp(promoted_role, PCMK_ROLE_PROMOTED) == 0)) { int preferrc = cli_resource_prefer(out, rsc_id, host, move_lifetime, cib_conn, promoted_role_only, PCMK__ROLE_PROMOTED_LEGACY); if (preferrc == pcmk_rc_ok) { rc = preferrc; } } return rc; } /* Nodes can be specified two different ways in the CIB, so we have two different * functions to try clearing out any constraints on them: * * (1) The node could be given by attribute=/value= in an expression XML node. * That's what resource_clear_node_in_expr handles. That XML looks like this: * * * * * * * * * (2) The node could be given by node= in a PCMK_XE_RSC_LOCATION XML node. * That's what resource_clear_node_in_location handles. That XML looks like * this: * * * * \return Standard Pacemaker return code */ static int resource_clear_node_in_expr(const char *rsc_id, const char *host, cib_t *cib_conn) { int rc = pcmk_rc_ok; char *xpath_string = NULL; #define XPATH_FMT \ "//" PCMK_XE_RSC_LOCATION "[@" PCMK_XA_ID "='cli-prefer-%s']" \ "[" PCMK_XE_RULE \ "[@" PCMK_XA_ID "='cli-prefer-rule-%s']" \ "/" PCMK_XE_EXPRESSION \ "[@" PCMK_XA_ATTRIBUTE "='" CRM_ATTR_UNAME "' " \ "and @" PCMK_XA_VALUE "='%s']" \ "]" xpath_string = crm_strdup_printf(XPATH_FMT, rsc_id, rsc_id, host); rc = cib_conn->cmds->remove(cib_conn, xpath_string, NULL, cib_xpath|cib_sync_call); if (rc == -ENXIO) { rc = pcmk_rc_ok; } else { rc = pcmk_legacy2rc(rc); } free(xpath_string); return rc; } // \return Standard Pacemaker return code static int resource_clear_node_in_location(const char *rsc_id, const char *host, cib_t * cib_conn, bool clear_ban_constraints, gboolean force) { int rc = pcmk_rc_ok; xmlNode *fragment = NULL; xmlNode *location = NULL; fragment = pcmk__xe_create(NULL, PCMK_XE_CONSTRAINTS); if (clear_ban_constraints == TRUE) { location = pcmk__xe_create(fragment, PCMK_XE_RSC_LOCATION); pcmk__xe_set_id(location, "cli-ban-%s-on-%s", rsc_id, host); } location = pcmk__xe_create(fragment, PCMK_XE_RSC_LOCATION); pcmk__xe_set_id(location, "cli-prefer-%s", rsc_id); if (force == FALSE) { crm_xml_add(location, PCMK_XE_NODE, host); } crm_log_xml_info(fragment, "Delete"); rc = cib_conn->cmds->remove(cib_conn, PCMK_XE_CONSTRAINTS, fragment, cib_sync_call); if (rc == -ENXIO) { rc = pcmk_rc_ok; } else { rc = pcmk_legacy2rc(rc); } pcmk__xml_free(fragment); return rc; } // \return Standard Pacemaker return code int cli_resource_clear(const char *rsc_id, const char *host, GList *allnodes, cib_t * cib_conn, bool clear_ban_constraints, gboolean force) { int rc = pcmk_rc_ok; if(cib_conn == NULL) { return ENOTCONN; } if (host) { rc = resource_clear_node_in_expr(rsc_id, host, cib_conn); /* rc does not tell us whether the previous operation did anything, only * whether it failed or not. Thus, as long as it did not fail, we need * to try the second clear method. */ if (rc == pcmk_rc_ok) { rc = resource_clear_node_in_location(rsc_id, host, cib_conn, clear_ban_constraints, force); } } else { GList *n = allnodes; /* Iterate over all nodes, attempting to clear the constraint from each. * On the first error, abort. */ for(; n; n = n->next) { pcmk_node_t *target = n->data; rc = cli_resource_clear(rsc_id, target->priv->name, NULL, cib_conn, clear_ban_constraints, force); if (rc != pcmk_rc_ok) { break; } } } return rc; } static void build_clear_xpath_string(GString *buf, const xmlNode *constraint_node, const char *rsc, const char *node, bool promoted_role_only) { const char *cons_id = pcmk__xe_id(constraint_node); const char *cons_rsc = pcmk__xe_get(constraint_node, PCMK_XA_RSC); GString *rsc_role_substr = NULL; const char *promoted_role_rule = "@" PCMK_XA_ROLE "='" PCMK_ROLE_PROMOTED "' or @" PCMK_XA_ROLE "='" PCMK__ROLE_PROMOTED_LEGACY "'"; pcmk__assert(buf != NULL); g_string_truncate(buf, 0); - if (!pcmk__starts_with(cons_id, "cli-ban-") - && !pcmk__starts_with(cons_id, "cli-prefer-")) { + if (cons_id == NULL) { + return; + } + if (!g_str_has_prefix(cons_id, "cli-ban-") + && !g_str_has_prefix(cons_id, "cli-prefer-")) { return; } g_string_append(buf, "//" PCMK_XE_RSC_LOCATION); if ((node != NULL) || (rsc != NULL) || promoted_role_only) { g_string_append_c(buf, '['); if (node != NULL) { pcmk__g_strcat(buf, "@" PCMK_XE_NODE "='", node, "'", NULL); if (promoted_role_only || (rsc != NULL)) { g_string_append(buf, " and "); } } if ((rsc != NULL) && promoted_role_only) { rsc_role_substr = g_string_sized_new(64); pcmk__g_strcat(rsc_role_substr, "@" PCMK_XA_RSC "='", rsc, "' " "and (" , promoted_role_rule, ")", NULL); } else if (rsc != NULL) { rsc_role_substr = g_string_sized_new(64); pcmk__g_strcat(rsc_role_substr, "@" PCMK_XA_RSC "='", rsc, "'", NULL); } else if (promoted_role_only) { rsc_role_substr = g_string_sized_new(64); g_string_append(rsc_role_substr, promoted_role_rule); } if (rsc_role_substr != NULL) { g_string_append(buf, rsc_role_substr->str); } g_string_append_c(buf, ']'); } if (node != NULL) { g_string_append(buf, "|//" PCMK_XE_RSC_LOCATION); if (rsc_role_substr != NULL) { pcmk__g_strcat(buf, "[", rsc_role_substr, "]", NULL); } pcmk__g_strcat(buf, "/" PCMK_XE_RULE "[" PCMK_XE_EXPRESSION "[@" PCMK_XA_ATTRIBUTE "='" CRM_ATTR_UNAME "' " "and @" PCMK_XA_VALUE "='", node, "']]", NULL); } g_string_append(buf, "//" PCMK_XE_DATE_EXPRESSION "[@" PCMK_XA_ID "='"); - if (pcmk__starts_with(cons_id, "cli-ban-")) { + if (g_str_has_prefix(cons_id, "cli-ban-")) { pcmk__g_strcat(buf, cons_id, "-lifetime']", NULL); } else { // starts with "cli-prefer-" pcmk__g_strcat(buf, "cli-prefer-lifetime-end-", cons_rsc, "']", NULL); } if (rsc_role_substr != NULL) { g_string_free(rsc_role_substr, TRUE); } } // \return Standard Pacemaker return code int cli_resource_clear_all_expired(xmlNode *root, cib_t *cib_conn, const char *rsc, const char *node, gboolean promoted_role_only) { GString *buf = NULL; xmlXPathObject *xpathObj = NULL; xmlNode *cib_constraints = NULL; crm_time_t *now = crm_time_new(NULL); int num_results = 0; int rc = pcmk_rc_ok; cib_constraints = pcmk_find_cib_element(root, PCMK_XE_CONSTRAINTS); xpathObj = pcmk__xpath_search(cib_constraints->doc, "//" PCMK_XE_RSC_LOCATION); num_results = pcmk__xpath_num_results(xpathObj); for (int i = 0; i < num_results; i++) { xmlNode *constraint_node = pcmk__xpath_result(xpathObj, i); xmlNode *date_expr_node = NULL; crm_time_t *end = NULL; int rc = pcmk_rc_ok; if (constraint_node == NULL) { continue; } if (buf == NULL) { buf = g_string_sized_new(1024); } build_clear_xpath_string(buf, constraint_node, rsc, node, promoted_role_only); if (buf->len == 0) { continue; } date_expr_node = pcmk__xpath_find_one(constraint_node->doc, buf->str, LOG_DEBUG); if (date_expr_node == NULL) { continue; } /* And then finally, see if the date expression is expired. If so, * clear the constraint. */ rc = pcmk__xe_get_datetime(date_expr_node, PCMK_XA_END, &end); if (rc != pcmk_rc_ok) { crm_trace("Date expression %s has invalid " PCMK_XA_END ": %s", pcmk__s(pcmk__xe_id(date_expr_node), "without ID"), pcmk_rc_str(rc)); continue; // Treat as unexpired } if (crm_time_compare(now, end) == 1) { xmlNode *fragment = NULL; xmlNode *location = NULL; fragment = pcmk__xe_create(NULL, PCMK_XE_CONSTRAINTS); location = pcmk__xe_create(fragment, PCMK_XE_RSC_LOCATION); pcmk__xe_set_id(location, "%s", pcmk__xe_id(constraint_node)); crm_log_xml_info(fragment, "Delete"); rc = cib_conn->cmds->remove(cib_conn, PCMK_XE_CONSTRAINTS, fragment, cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { goto done; } pcmk__xml_free(fragment); } crm_time_free(end); } done: if (buf != NULL) { g_string_free(buf, TRUE); } xmlXPathFreeObject(xpathObj); crm_time_free(now); return rc; } diff --git a/tools/crm_shadow.c b/tools/crm_shadow.c index 9f164508b1..a644b841b8 100644 --- a/tools/crm_shadow.c +++ b/tools/crm_shadow.c @@ -1,1305 +1,1307 @@ /* * Copyright 2004-2025 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include +#include // GOption, etc. + #include #include #include #include #include #include #define SUMMARY "perform Pacemaker configuration changes in a sandbox\n\n" \ "This command sets up an environment in which " \ "configuration tools (cibadmin,\n" \ "crm_resource, etc.) work offline instead of against a " \ "live cluster, allowing\n" \ "changes to be previewed and tested for side effects." #define INDENT " " enum shadow_command { shadow_cmd_none = 0, shadow_cmd_which, shadow_cmd_display, shadow_cmd_diff, shadow_cmd_file, shadow_cmd_create, shadow_cmd_create_empty, shadow_cmd_commit, shadow_cmd_delete, shadow_cmd_edit, shadow_cmd_reset, shadow_cmd_switch, }; /*! * \internal * \brief Bit flags to control which fields of shadow CIB info are displayed * * \note Ignored for XML output. */ enum shadow_disp_flags { shadow_disp_instance = (1 << 0), shadow_disp_file = (1 << 1), shadow_disp_content = (1 << 2), shadow_disp_diff = (1 << 3), }; static crm_exit_t exit_code = CRM_EX_OK; static struct { enum shadow_command cmd; int cmd_options; char *instance; gboolean force; gboolean batch; gboolean full_upload; gchar *validate_with; } options = { .cmd_options = cib_sync_call, }; /*! * \internal * \brief Display an instruction to the user * * \param[in,out] out Output object * \param[in] args Message-specific arguments * * \return Standard Pacemaker return code * * \note \p args should contain the following: * -# Instructional message */ PCMK__OUTPUT_ARGS("instruction", "const char *") static int instruction_default(pcmk__output_t *out, va_list args) { const char *msg = va_arg(args, const char *); if (msg == NULL) { return pcmk_rc_no_output; } return out->info(out, "%s", msg); } /*! * \internal * \brief Display an instruction to the user * * \param[in,out] out Output object * \param[in] args Message-specific arguments * * \return Standard Pacemaker return code * * \note \p args should contain the following: * -# Instructional message */ PCMK__OUTPUT_ARGS("instruction", "const char *") static int instruction_xml(pcmk__output_t *out, va_list args) { const char *msg = va_arg(args, const char *); if (msg == NULL) { return pcmk_rc_no_output; } pcmk__output_create_xml_text_node(out, "instruction", msg); return pcmk_rc_ok; } /*! * \internal * \brief Display information about a shadow CIB instance * * \param[in,out] out Output object * \param[in] args Message-specific arguments * * \return Standard Pacemaker return code * * \note \p args should contain the following: * -# Instance name (can be \p NULL) * -# Shadow file name (can be \p NULL) * -# Shadow file content (can be \p NULL) * -# Patchset containing the changes in the shadow CIB (can be \p NULL) * -# Group of \p shadow_disp_flags indicating which fields to display */ PCMK__OUTPUT_ARGS("shadow", "const char *", "const char *", "const xmlNode *", "const xmlNode *", "enum shadow_disp_flags") static int shadow_default(pcmk__output_t *out, va_list args) { const char *instance = va_arg(args, const char *); const char *filename = va_arg(args, const char *); const xmlNode *content = va_arg(args, const xmlNode *); const xmlNode *diff = va_arg(args, const xmlNode *); enum shadow_disp_flags flags = (enum shadow_disp_flags) va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(flags, shadow_disp_instance)) { rc = out->info(out, "Instance: %s", pcmk__s(instance, "")); } if (pcmk_is_set(flags, shadow_disp_file)) { rc = out->info(out, "File name: %s", pcmk__s(filename, "")); } if (pcmk_is_set(flags, shadow_disp_content)) { rc = out->info(out, "Content:"); if (content != NULL) { GString *buf = g_string_sized_new(1024); gchar *str = NULL; pcmk__xml_string(content, pcmk__xml_fmt_pretty|pcmk__xml_fmt_text, buf, 0); str = g_string_free(buf, FALSE); - str = pcmk__trim(str); + str = g_strchomp(str); if (!pcmk__str_empty(str)) { out->info(out, "%s", str); } g_free(str); } else { out->info(out, ""); } } if (pcmk_is_set(flags, shadow_disp_diff)) { rc = out->info(out, "Diff:"); if (diff != NULL) { out->message(out, "xml-patchset", diff); } else { out->info(out, ""); } } return rc; } /*! * \internal * \brief Display information about a shadow CIB instance * * \param[in,out] out Output object * \param[in] args Message-specific arguments * * \return Standard Pacemaker return code * * \note \p args should contain the following: * -# Instance name (can be \p NULL) * -# Shadow file name (can be \p NULL) * -# Shadow file content (can be \p NULL) * -# Patchset containing the changes in the shadow CIB (can be \p NULL) * -# Group of \p shadow_disp_flags indicating which fields to display */ PCMK__OUTPUT_ARGS("shadow", "const char *", "const char *", "const xmlNode *", "const xmlNode *", "enum shadow_disp_flags") static int shadow_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return shadow_default(out, args); } else { const char *instance = va_arg(args, const char *); const char *filename = va_arg(args, const char *); const xmlNode *content = va_arg(args, const xmlNode *); const xmlNode *diff = va_arg(args, const xmlNode *); enum shadow_disp_flags flags = (enum shadow_disp_flags) va_arg(args, int); int rc = pcmk_rc_no_output; bool quiet_orig = out->quiet; /* We have to disable quiet mode for the "xml-patchset" message if we * call it, so we might as well do so for this whole section. */ out->quiet = false; if (pcmk_is_set(flags, shadow_disp_instance) && (instance != NULL)) { rc = out->info(out, "%s", instance); } if (pcmk_is_set(flags, shadow_disp_file) && (filename != NULL)) { rc = out->info(out, "%s", filename); } if (pcmk_is_set(flags, shadow_disp_content) && (content != NULL)) { GString *buf = g_string_sized_new(1024); gchar *str = NULL; pcmk__xml_string(content, pcmk__xml_fmt_pretty|pcmk__xml_fmt_text, buf, 0); str = g_string_free(buf, FALSE); - str = pcmk__trim(str); + str = g_strchomp(str); rc = out->info(out, "%s", str); g_free(str); } if (pcmk_is_set(flags, shadow_disp_diff) && (diff != NULL)) { rc = out->message(out, "xml-patchset", diff); } out->quiet = quiet_orig; return rc; } } /*! * \internal * \brief Display information about a shadow CIB instance * * \param[in,out] out Output object * \param[in] args Message-specific arguments * * \return Standard Pacemaker return code * * \note \p args should contain the following: * -# Instance name (can be \p NULL) * -# Shadow file name (can be \p NULL) * -# Shadow file content (can be \p NULL) * -# Patchset containing the changes in the shadow CIB (can be \p NULL) * -# Group of \p shadow_disp_flags indicating which fields to display * (ignored) */ PCMK__OUTPUT_ARGS("shadow", "const char *", "const char *", "const xmlNode *", "const xmlNode *", "enum shadow_disp_flags") static int shadow_xml(pcmk__output_t *out, va_list args) { const char *instance = va_arg(args, const char *); const char *filename = va_arg(args, const char *); const xmlNode *content = va_arg(args, const xmlNode *); const xmlNode *diff = va_arg(args, const xmlNode *); enum shadow_disp_flags flags G_GNUC_UNUSED = (enum shadow_disp_flags) va_arg(args, int); pcmk__output_xml_create_parent(out, PCMK_XE_SHADOW, PCMK_XA_INSTANCE, instance, PCMK_XA_FILE, filename, NULL); if (content != NULL) { GString *buf = g_string_sized_new(1024); pcmk__xml_string(content, pcmk__xml_fmt_pretty|pcmk__xml_fmt_text, buf, 0); out->output_xml(out, PCMK_XE_CONTENT, buf->str); g_string_free(buf, TRUE); } if (diff != NULL) { out->message(out, "xml-patchset", diff); } pcmk__output_xml_pop_parent(out); return pcmk_rc_ok; } static const pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static const pcmk__message_entry_t fmt_functions[] = { { "instruction", "default", instruction_default }, { "instruction", "xml", instruction_xml }, { "shadow", "default", shadow_default }, { "shadow", "text", shadow_text }, { "shadow", "xml", shadow_xml }, { NULL, NULL, NULL } }; /*! * \internal * \brief Set the error when \p --force is not passed with a dangerous command * * \param[in] reason Why command is dangerous * \param[in] for_shadow If true, command is dangerous to the shadow file. * Otherwise, command is dangerous to the active * cluster. * \param[in] show_mismatch If true and the supplied shadow instance is not * the same as the active shadow instance, report * this * \param[out] error Where to store error */ static void set_danger_error(const char *reason, bool for_shadow, bool show_mismatch, GError **error) { const char *active = getenv("CIB_shadow"); char *full = NULL; if (show_mismatch && !pcmk__str_eq(active, options.instance, pcmk__str_null_matches)) { full = crm_strdup_printf("%s.\nAdditionally, the supplied shadow " "instance (%s) is not the same as the active " "one (%s)", reason, options.instance, active); reason = full; } g_set_error(error, PCMK__EXITC_ERROR, exit_code, "%s%sTo prevent accidental destruction of the %s, the --force " "flag is required in order to proceed.", pcmk__s(reason, ""), ((reason != NULL)? ".\n" : ""), (for_shadow? "shadow file" : "cluster")); free(full); } /*! * \internal * \brief Get the active shadow instance from the environment * * This sets \p options.instance to the value of the \p CIB_shadow env variable. * * \param[out] error Where to store error */ static int get_instance_from_env(GError **error) { int rc = pcmk_rc_ok; pcmk__str_update(&options.instance, getenv("CIB_shadow")); if (options.instance == NULL) { rc = ENXIO; exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "No active shadow configuration defined"); } return rc; } /*! * \internal * \brief Validate that the shadow file does or does not exist, as appropriate * * \param[in] filename Absolute path of shadow file * \param[in] should_exist Whether the shadow file is expected to exist * \param[out] error Where to store error * * \return Standard Pacemaker return code */ static int check_file_exists(const char *filename, bool should_exist, GError **error) { struct stat buf; if (!should_exist && (stat(filename, &buf) == 0)) { char *reason = crm_strdup_printf("A shadow instance '%s' already " "exists", options.instance); exit_code = CRM_EX_CANTCREAT; set_danger_error(reason, true, false, error); free(reason); return EEXIST; } if (should_exist && (stat(filename, &buf) < 0)) { int rc = errno; exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not access shadow instance '%s': %s", options.instance, strerror(rc)); return errno; } return pcmk_rc_ok; } /*! * \internal * \brief Connect to the "real" (non-shadow) CIB * * \param[out] real_cib Where to store CIB connection * \param[out] error Where to store error * * \return Standard Pacemaker return code */ static int connect_real_cib(cib_t **real_cib, GError **error) { const char *active = getenv("CIB_shadow"); int rc = pcmk_rc_ok; // Create a non-shadowed CIB connection object and then restore CIB_shadow unsetenv("CIB_shadow"); rc = cib__create_signon(real_cib); if (active != NULL) { setenv("CIB_shadow", active, 1); } if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not connect to CIB: %s", pcmk_rc_str(rc)); } return rc; } /*! * \internal * \brief Query the "real" (non-shadow) CIB and store the result * * \param[out] output Where to store query output * \param[out] error Where to store error * * \return Standard Pacemaker return code */ static int query_real_cib(xmlNode **output, GError **error) { cib_t *real_cib = NULL; int rc = connect_real_cib(&real_cib, error); if (rc != pcmk_rc_ok) { goto done; } rc = real_cib->cmds->query(real_cib, NULL, output, options.cmd_options); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not query the non-shadow CIB: %s", pcmk_rc_str(rc)); } done: cib_delete(real_cib); return rc; } /*! * \internal * \brief Read XML from the given file * * \param[in] filename Path of input file * \param[out] output Where to store XML read from \p filename * \param[out] error Where to store error * * \return Standard Pacemaker return code */ static int read_xml(const char *filename, xmlNode **output, GError **error) { int rc = pcmk_rc_ok; *output = pcmk__xml_read(filename); if (*output == NULL) { rc = pcmk_rc_no_input; exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not parse XML from input file '%s'", filename); } return rc; } /*! * \internal * \brief Write the shadow XML to a file * * \param[in] xml Shadow XML * \param[in] filename Name of destination file * \param[in] reset Whether the write is a reset (for logging only) * \param[out] error Where to store error */ static int write_shadow_file(const xmlNode *xml, const char *filename, bool reset, GError **error) { int rc = pcmk__xml_write_file(xml, filename, false); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not %s the shadow instance '%s': %s", reset? "reset" : "create", options.instance, pcmk_rc_str(rc)); } return rc; } /*! * \internal * \brief Create a shell prompt based on the given shadow instance name * * \return Newly created prompt * * \note The caller is responsible for freeing the return value using \p free(). */ static inline char * get_shadow_prompt(void) { return crm_strdup_printf("shadow[%.40s] # ", options.instance); } /*! * \internal * \brief Set up environment variables for a shadow instance * * \param[in,out] out Output object * \param[in] do_switch If true, switch to an existing instance (logging * only) * \param[out] error Where to store error */ static void shadow_setup(pcmk__output_t *out, bool do_switch, GError **error) { const char *active = getenv("CIB_shadow"); const char *prompt = getenv("PS1"); const char *shell = getenv("SHELL"); char *new_prompt = get_shadow_prompt(); if (pcmk__str_eq(active, options.instance, pcmk__str_none) && pcmk__str_eq(new_prompt, prompt, pcmk__str_none)) { // CIB_shadow and prompt environment variables are already set up goto done; } if (!options.batch && (shell != NULL)) { out->info(out, "Setting up shadow instance"); setenv("PS1", new_prompt, 1); setenv("CIB_shadow", options.instance, 1); out->message(out, PCMK_XE_INSTRUCTION, "Press Ctrl+D to exit the crm_shadow shell"); if (pcmk__str_eq(shell, "(^|/)bash$", pcmk__str_regex)) { execl(shell, shell, "--norc", "--noprofile", NULL); } else { execl(shell, shell, NULL); } exit_code = pcmk_rc2exitc(errno); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Failed to launch shell '%s': %s", shell, pcmk_rc_str(errno)); } else { char *msg = NULL; const char *prefix = "A new shadow instance was created. To begin " "using it"; if (do_switch) { prefix = "To switch to the named shadow instance"; } msg = crm_strdup_printf("%s, enter the following into your shell:\n" "\texport CIB_shadow=%s", prefix, options.instance); out->message(out, "instruction", msg); free(msg); } done: free(new_prompt); } /*! * \internal * \brief Remind the user to clean up the shadow environment * * \param[in,out] out Output object */ static void shadow_teardown(pcmk__output_t *out) { const char *active = getenv("CIB_shadow"); const char *prompt = getenv("PS1"); if (pcmk__str_eq(active, options.instance, pcmk__str_none)) { char *our_prompt = get_shadow_prompt(); if (pcmk__str_eq(prompt, our_prompt, pcmk__str_none)) { out->message(out, "instruction", "Press Ctrl+D to exit the crm_shadow shell"); } else { out->message(out, "instruction", "Remember to unset the CIB_shadow variable by " "entering the following into your shell:\n" "\tunset CIB_shadow"); } free(our_prompt); } } /*! * \internal * \brief Commit the shadow file contents to the active cluster * * \param[out] error Where to store error */ static void commit_shadow_file(GError **error) { char *filename = NULL; cib_t *real_cib = NULL; xmlNodePtr input = NULL; xmlNodePtr section_xml = NULL; const char *section = NULL; int rc = pcmk_rc_ok; if (!options.force) { const char *reason = "The commit command overwrites the active cluster " "configuration"; exit_code = CRM_EX_USAGE; set_danger_error(reason, false, true, error); return; } filename = get_shadow_file(options.instance); if (check_file_exists(filename, true, error) != pcmk_rc_ok) { goto done; } if (connect_real_cib(&real_cib, error) != pcmk_rc_ok) { goto done; } if (read_xml(filename, &input, error) != pcmk_rc_ok) { goto done; } section_xml = input; if (!options.full_upload) { section = PCMK_XE_CONFIGURATION; section_xml = pcmk__xe_first_child(input, section, NULL, NULL); } rc = real_cib->cmds->replace(real_cib, section, section_xml, options.cmd_options); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not commit shadow instance '%s' to the CIB: %s", options.instance, pcmk_rc_str(rc)); } done: free(filename); cib_delete(real_cib); pcmk__xml_free(input); } /*! * \internal * \brief Create a new empty shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error * * \note If \p --force is given, we try to write the file regardless of whether * it already exists. */ static void create_shadow_empty(pcmk__output_t *out, GError **error) { char *filename = get_shadow_file(options.instance); xmlNode *output = NULL; if (!options.force && (check_file_exists(filename, false, error) != pcmk_rc_ok)) { goto done; } output = createEmptyCib(0); crm_xml_add(output, PCMK_XA_VALIDATE_WITH, options.validate_with); out->info(out, "Created new %s configuration", pcmk__xe_get(output, PCMK_XA_VALIDATE_WITH)); if (write_shadow_file(output, filename, false, error) != pcmk_rc_ok) { goto done; } shadow_setup(out, false, error); done: free(filename); pcmk__xml_free(output); } /*! * \internal * \brief Create a shadow instance based on the active CIB * * \param[in,out] out Output object * \param[in] reset If true, overwrite the given existing shadow instance. * Otherwise, create a new shadow instance with the given * name. * \param[out] error Where to store error * * \note If \p --force is given, we try to write the file regardless of whether * it already exists. */ static void create_shadow_from_cib(pcmk__output_t *out, bool reset, GError **error) { char *filename = get_shadow_file(options.instance); xmlNode *output = NULL; if (!options.force) { if (reset) { const char *reason = "The reset command overwrites the active " "shadow configuration"; exit_code = CRM_EX_USAGE; set_danger_error(reason, true, true, error); goto done; } if (check_file_exists(filename, reset, error) != pcmk_rc_ok) { goto done; } } if (query_real_cib(&output, error) != pcmk_rc_ok) { goto done; } if (write_shadow_file(output, filename, reset, error) != pcmk_rc_ok) { goto done; } shadow_setup(out, false, error); done: free(filename); pcmk__xml_free(output); } /*! * \internal * \brief Delete the shadow file * * \param[in,out] out Output object * \param[out] error Where to store error */ static void delete_shadow_file(pcmk__output_t *out, GError **error) { char *filename = NULL; if (!options.force) { const char *reason = "The delete command removes the specified shadow " "file"; exit_code = CRM_EX_USAGE; set_danger_error(reason, true, true, error); return; } filename = get_shadow_file(options.instance); if ((unlink(filename) < 0) && (errno != ENOENT)) { exit_code = pcmk_rc2exitc(errno); g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not remove shadow instance '%s': %s", options.instance, strerror(errno)); } else { shadow_teardown(out); } free(filename); } /*! * \internal * \brief Open the shadow file in a text editor * * \param[out] error Where to store error * * \note The \p EDITOR environment variable must be set. */ static void edit_shadow_file(GError **error) { char *filename = NULL; const char *editor = NULL; if (get_instance_from_env(error) != pcmk_rc_ok) { return; } filename = get_shadow_file(options.instance); if (check_file_exists(filename, true, error) != pcmk_rc_ok) { goto done; } editor = getenv("EDITOR"); if (editor == NULL) { exit_code = CRM_EX_NOT_CONFIGURED; g_set_error(error, PCMK__EXITC_ERROR, exit_code, "No value for EDITOR defined"); goto done; } execlp(editor, "--", filename, NULL); exit_code = CRM_EX_OSFILE; g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not invoke EDITOR (%s %s): %s", editor, filename, strerror(errno)); done: free(filename); } /*! * \internal * \brief Show the contents of the active shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error */ static void show_shadow_contents(pcmk__output_t *out, GError **error) { char *filename = NULL; if (get_instance_from_env(error) != pcmk_rc_ok) { return; } filename = get_shadow_file(options.instance); if (check_file_exists(filename, true, error) == pcmk_rc_ok) { xmlNode *output = NULL; bool quiet_orig = out->quiet; if (read_xml(filename, &output, error) != pcmk_rc_ok) { goto done; } out->quiet = true; out->message(out, "shadow", options.instance, NULL, output, NULL, shadow_disp_content); out->quiet = quiet_orig; pcmk__xml_free(output); } done: free(filename); } /*! * \internal * \brief Show the changes in the active shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error */ static void show_shadow_diff(pcmk__output_t *out, GError **error) { char *filename = NULL; xmlNodePtr old_config = NULL; xmlNodePtr new_config = NULL; xmlNodePtr diff = NULL; bool quiet_orig = out->quiet; if (get_instance_from_env(error) != pcmk_rc_ok) { return; } filename = get_shadow_file(options.instance); if (check_file_exists(filename, true, error) != pcmk_rc_ok) { goto done; } if (query_real_cib(&old_config, error) != pcmk_rc_ok) { goto done; } if (read_xml(filename, &new_config, error) != pcmk_rc_ok) { goto done; } pcmk__xml_mark_changes(old_config, new_config); diff = xml_create_patchset(0, old_config, new_config, NULL, false); pcmk__log_xml_changes(LOG_INFO, new_config); pcmk__xml_commit_changes(new_config->doc); out->quiet = true; out->message(out, "shadow", options.instance, NULL, NULL, diff, shadow_disp_diff); out->quiet = quiet_orig; if (diff != NULL) { /* @COMPAT: Exit with CRM_EX_DIGEST? This is not really an error; we * just want to indicate that there are differences (as the diff command * does). */ exit_code = CRM_EX_ERROR; } done: free(filename); pcmk__xml_free(old_config); pcmk__xml_free(new_config); pcmk__xml_free(diff); } /*! * \internal * \brief Show the absolute path of the active shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error */ static void show_shadow_filename(pcmk__output_t *out, GError **error) { if (get_instance_from_env(error) == pcmk_rc_ok) { char *filename = get_shadow_file(options.instance); bool quiet_orig = out->quiet; out->quiet = true; out->message(out, "shadow", options.instance, filename, NULL, NULL, shadow_disp_file); out->quiet = quiet_orig; free(filename); } } /*! * \internal * \brief Show the active shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error */ static void show_shadow_instance(pcmk__output_t *out, GError **error) { if (get_instance_from_env(error) == pcmk_rc_ok) { bool quiet_orig = out->quiet; out->quiet = true; out->message(out, "shadow", options.instance, NULL, NULL, NULL, shadow_disp_instance); out->quiet = quiet_orig; } } /*! * \internal * \brief Switch to the given shadow instance * * \param[in,out] out Output object * \param[out] error Where to store error */ static void switch_shadow_instance(pcmk__output_t *out, GError **error) { char *filename = NULL; filename = get_shadow_file(options.instance); if (check_file_exists(filename, true, error) == pcmk_rc_ok) { shadow_setup(out, true, error); } free(filename); } static gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-w", "--which", NULL)) { options.cmd = shadow_cmd_which; } else if (pcmk__str_any_of(option_name, "-p", "--display", NULL)) { options.cmd = shadow_cmd_display; } else if (pcmk__str_any_of(option_name, "-d", "--diff", NULL)) { options.cmd = shadow_cmd_diff; } else if (pcmk__str_any_of(option_name, "-F", "--file", NULL)) { options.cmd = shadow_cmd_file; } else if (pcmk__str_any_of(option_name, "-c", "--create", NULL)) { options.cmd = shadow_cmd_create; } else if (pcmk__str_any_of(option_name, "-e", "--create-empty", NULL)) { options.cmd = shadow_cmd_create_empty; } else if (pcmk__str_any_of(option_name, "-C", "--commit", NULL)) { options.cmd = shadow_cmd_commit; } else if (pcmk__str_any_of(option_name, "-D", "--delete", NULL)) { options.cmd = shadow_cmd_delete; } else if (pcmk__str_any_of(option_name, "-E", "--edit", NULL)) { options.cmd = shadow_cmd_edit; } else if (pcmk__str_any_of(option_name, "-r", "--reset", NULL)) { options.cmd = shadow_cmd_reset; } else if (pcmk__str_any_of(option_name, "-s", "--switch", NULL)) { options.cmd = shadow_cmd_switch; } else { // Should be impossible return FALSE; } // optarg may be NULL and that's okay pcmk__str_update(&options.instance, optarg); return TRUE; } static GOptionEntry query_entries[] = { { "which", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Indicate the active shadow copy", NULL }, { "display", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the contents of the active shadow copy", NULL }, { "diff", 'd', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the changes in the active shadow copy", NULL }, { "file", 'F', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the location of the active shadow copy file", NULL }, { NULL } }; static GOptionEntry command_entries[] = { { "create", 'c', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "Create the named shadow copy of the active cluster configuration", "name" }, { "create-empty", 'e', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "Create the named shadow copy with an empty cluster configuration.\n" INDENT "Optional: --validate-with", "name" }, { "commit", 'C', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "Upload the contents of the named shadow copy to the cluster", "name" }, { "delete", 'D', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "Delete the contents of the named shadow copy", "name" }, { "edit", 'E', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Edit the contents of the active shadow copy with your favorite $EDITOR", NULL }, { "reset", 'r', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "Recreate named shadow copy from the active cluster configuration", "name. Required: --force." }, { "switch", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, command_cb, "(Advanced) Switch to the named shadow copy", "name" }, { NULL } }; static GOptionEntry addl_entries[] = { { "force", 'f', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.force, "(Advanced) Force the action to be performed", NULL }, { "batch", 'b', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.batch, "(Advanced) Don't spawn a new shell", NULL }, { "all", 'a', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.full_upload, "(Advanced) Upload entire CIB, including status, with --commit", NULL }, { "validate-with", 'v', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.validate_with, "(Advanced) Create an older configuration version", NULL }, { NULL } }; static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { const char *desc = NULL; GOptionContext *context = NULL; desc = "Examples:\n\n" "Create a blank shadow configuration:\n\n" "\t# crm_shadow --create-empty myShadow\n\n" "Create a shadow configuration from the running cluster\n\n" "\t# crm_shadow --create myShadow\n\n" "Display the current shadow configuration:\n\n" "\t# crm_shadow --display\n\n" "Discard the current shadow configuration (named myShadow):\n\n" "\t# crm_shadow --delete myShadow --force\n\n" "Upload current shadow configuration (named myShadow) to running " "cluster:\n\n" "\t# crm_shadow --commit myShadow\n\n"; context = pcmk__build_arg_context(args, "text (default), xml", group, "|"); g_option_context_set_description(context, desc); pcmk__add_arg_group(context, "queries", "Queries:", "Show query help", query_entries); pcmk__add_arg_group(context, "commands", "Commands:", "Show command help", command_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; pcmk__output_t *out = NULL; GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "CDcersv"); GOptionContext *context = build_arg_context(args, &output_group); crm_log_preinit(NULL, argc, argv); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { exit_code = CRM_EX_ERROR; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); goto done; } if (g_strv_length(processed_args) > 1) { gchar *help = g_option_context_get_help(context, TRUE, NULL); GString *extra = g_string_sized_new(128); for (int lpc = 1; processed_args[lpc] != NULL; lpc++) { if (extra->len > 0) { g_string_append_c(extra, ' '); } g_string_append(extra, processed_args[lpc]); } exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "non-option ARGV-elements: %s\n\n%s", extra->str, help); g_free(help); g_string_free(extra, TRUE); goto done; } if (args->version) { out->version(out); goto done; } pcmk__register_messages(out, fmt_functions); if (options.cmd == shadow_cmd_none) { // @COMPAT: Create a default command if other tools have one gchar *help = g_option_context_get_help(context, TRUE, NULL); exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must specify a query or command option\n\n%s", help); g_free(help); goto done; } pcmk__cli_init_logging("crm_shadow", args->verbosity); if (args->verbosity > 0) { cib__set_call_options(options.cmd_options, crm_system_name, cib_verbose); } // Run the command switch (options.cmd) { case shadow_cmd_commit: commit_shadow_file(&error); break; case shadow_cmd_create: create_shadow_from_cib(out, false, &error); break; case shadow_cmd_create_empty: create_shadow_empty(out, &error); break; case shadow_cmd_reset: create_shadow_from_cib(out, true, &error); break; case shadow_cmd_delete: delete_shadow_file(out, &error); break; case shadow_cmd_diff: show_shadow_diff(out, &error); break; case shadow_cmd_display: show_shadow_contents(out, &error); break; case shadow_cmd_edit: edit_shadow_file(&error); break; case shadow_cmd_file: show_shadow_filename(out, &error); break; case shadow_cmd_switch: switch_shadow_instance(out, &error); break; case shadow_cmd_which: show_shadow_instance(out, &error); break; default: // Should never reach this point break; } done: g_strfreev(processed_args); pcmk__free_arg_context(context); pcmk__output_and_clear_error(&error, out); free(options.instance); g_free(options.validate_with); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); }