diff --git a/daemons/attrd/pacemaker-attrd.c b/daemons/attrd/pacemaker-attrd.c index 037825b313..3e6d7365b8 100644 --- a/daemons/attrd/pacemaker-attrd.c +++ b/daemons/attrd/pacemaker-attrd.c @@ -1,358 +1,360 @@ /* * Copyright 2013-2023 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 #include "pacemaker-attrd.h" #define SUMMARY "daemon for managing Pacemaker node attributes" gboolean stand_alone = FALSE; gchar **log_files = NULL; static GOptionEntry entries[] = { { "stand-alone", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &stand_alone, "(Advanced use only) Run in stand-alone mode", NULL }, { "logfile", 'l', G_OPTION_FLAG_NONE, G_OPTION_ARG_FILENAME_ARRAY, &log_files, "Send logs to the additional named logfile", NULL }, { NULL } }; static pcmk__output_t *out = NULL; static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; lrmd_t *the_lrmd = NULL; crm_cluster_t *attrd_cluster = NULL; crm_trigger_t *attrd_config_read = NULL; crm_exit_t attrd_exit_status = CRM_EX_OK; static void attrd_cib_destroy_cb(gpointer user_data) { cib_t *conn = user_data; conn->cmds->signoff(conn); /* Ensure IPC is cleaned up */ if (attrd_shutting_down()) { crm_info("Connection disconnection complete"); } else { /* eventually this should trigger a reconnect, not a shutdown */ crm_crit("Lost connection to the CIB manager, shutting down"); attrd_exit_status = CRM_EX_DISCONNECT; attrd_shutdown(0); } return; } static void attrd_erase_cb(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { do_crm_log_unlikely((rc? LOG_NOTICE : LOG_DEBUG), "Cleared transient attributes: %s " CRM_XS " xpath=%s rc=%d", pcmk_strerror(rc), (char *) user_data, rc); } #define XPATH_TRANSIENT "//node_state[@uname='%s']/" XML_TAG_TRANSIENT_NODEATTRS /*! * \internal * \brief Wipe all transient attributes for this node from the CIB * * Clear any previous transient node attributes from the CIB. This is * normally done by the DC's controller when this node leaves the cluster, but * this handles the case where the node restarted so quickly that the * cluster layer didn't notice. * * \todo If pacemaker-attrd respawns after crashing (see PCMK_respawned), * ideally we'd skip this and sync our attributes from the writer. * However, currently we reject any values for us that the writer has, in * attrd_peer_update(). */ static void attrd_erase_attrs(void) { int call_id; char *xpath = crm_strdup_printf(XPATH_TRANSIENT, attrd_cluster->uname); crm_info("Clearing transient attributes from CIB " CRM_XS " xpath=%s", xpath); call_id = the_cib->cmds->remove(the_cib, xpath, NULL, cib_xpath); the_cib->cmds->register_callback_full(the_cib, call_id, 120, FALSE, xpath, "attrd_erase_cb", attrd_erase_cb, free); } static int attrd_cib_connect(int max_retry) { static int attempts = 0; int rc = -ENOTCONN; the_cib = cib_new(); if (the_cib == NULL) { return -ENOTCONN; } do { if(attempts > 0) { sleep(attempts); } attempts++; crm_debug("Connection attempt %d to the CIB manager", attempts); rc = the_cib->cmds->signon(the_cib, T_ATTRD, cib_command); } while(rc != pcmk_ok && attempts < max_retry); if (rc != pcmk_ok) { crm_err("Connection to the CIB manager failed: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); goto cleanup; } crm_debug("Connected to the CIB manager after %d attempts", attempts); rc = the_cib->cmds->set_connection_dnotify(the_cib, attrd_cib_destroy_cb); if (rc != pcmk_ok) { crm_err("Could not set disconnection callback"); goto cleanup; } rc = the_cib->cmds->add_notify_callback(the_cib, T_CIB_REPLACE_NOTIFY, attrd_cib_replaced_cb); if(rc != pcmk_ok) { crm_err("Could not set CIB notification callback"); goto cleanup; } rc = the_cib->cmds->add_notify_callback(the_cib, T_CIB_DIFF_NOTIFY, attrd_cib_updated_cb); if (rc != pcmk_ok) { crm_err("Could not set CIB notification callback (update)"); goto cleanup; } return pcmk_ok; cleanup: cib__clean_up_connection(&the_cib); return -ENOTCONN; } /*! * \internal * \brief Prepare the CIB after cluster is connected */ static void attrd_cib_init(void) { // We have no attribute values in memory, wipe the CIB to match attrd_erase_attrs(); // Set a trigger for reading the CIB (for the alerts section) attrd_config_read = mainloop_add_trigger(G_PRIORITY_HIGH, attrd_read_options, NULL); // Always read the CIB at start-up mainloop_set_trigger(attrd_config_read); } static bool ipc_already_running(void) { pcmk_ipc_api_t *old_instance = NULL; int rc = pcmk_rc_ok; rc = pcmk_new_ipc_api(&old_instance, pcmk_ipc_attrd); if (rc != pcmk_rc_ok) { return false; } - rc = pcmk_connect_ipc(old_instance, pcmk_ipc_dispatch_sync); + rc = pcmk__connect_ipc(old_instance, pcmk_ipc_dispatch_sync, 2); if (rc != pcmk_rc_ok) { + crm_debug("No existing %s manager instance found: %s", + pcmk_ipc_name(old_instance, true), pcmk_rc_str(rc)); pcmk_free_ipc_api(old_instance); return false; } pcmk_disconnect_ipc(old_instance); pcmk_free_ipc_api(old_instance); return true; } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); pcmk__add_main_args(context, entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; GError *error = NULL; bool initialized = false; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, NULL); GOptionContext *context = build_arg_context(args, &output_group); attrd_init_mainloop(); crm_log_preinit(NULL, argc, argv); mainloop_add_signal(SIGTERM, attrd_shutdown); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { attrd_exit_status = CRM_EX_USAGE; goto done; } rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if ((rc != pcmk_rc_ok) || (out == NULL)) { attrd_exit_status = CRM_EX_ERROR; g_set_error(&error, PCMK__EXITC_ERROR, attrd_exit_status, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); goto done; } if (args->version) { out->version(out, false); goto done; } // Open additional log files pcmk__add_logfiles(log_files, out); crm_log_init(T_ATTRD, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); crm_notice("Starting Pacemaker node attribute manager%s", stand_alone ? " in standalone mode" : ""); if (ipc_already_running()) { const char *msg = "pacemaker-attrd is already active, aborting startup"; attrd_exit_status = CRM_EX_OK; g_set_error(&error, PCMK__EXITC_ERROR, attrd_exit_status, "%s", msg); crm_err(msg); goto done; } initialized = true; attributes = pcmk__strkey_table(NULL, attrd_free_attribute); /* Connect to the CIB before connecting to the cluster or listening for IPC. * This allows us to assume the CIB is connected whenever we process a * cluster or IPC message (which also avoids start-up race conditions). */ if (!stand_alone) { if (attrd_cib_connect(30) != pcmk_ok) { attrd_exit_status = CRM_EX_FATAL; g_set_error(&error, PCMK__EXITC_ERROR, attrd_exit_status, "Could not connect to the CIB"); goto done; } crm_info("CIB connection active"); } if (attrd_cluster_connect() != pcmk_ok) { attrd_exit_status = CRM_EX_FATAL; g_set_error(&error, PCMK__EXITC_ERROR, attrd_exit_status, "Could not connect to the cluster"); goto done; } crm_info("Cluster connection active"); // Initialization that requires the cluster to be connected attrd_election_init(); if (!stand_alone) { attrd_cib_init(); } /* Set a private attribute for ourselves with the protocol version we * support. This lets all nodes determine the minimum supported version * across all nodes. It also ensures that the writer learns our node name, * so it can send our attributes to the CIB. */ attrd_broadcast_protocol(); attrd_init_ipc(); crm_notice("Pacemaker node attribute manager successfully started and accepting connections"); attrd_run_mainloop(); done: if (initialized) { crm_info("Shutting down attribute manager"); attrd_election_fini(); attrd_ipc_fini(); attrd_lrmd_disconnect(); if (!stand_alone) { attrd_cib_disconnect(); } attrd_free_waitlist(); pcmk_cluster_free(attrd_cluster); g_hash_table_destroy(attributes); } g_strfreev(processed_args); pcmk__free_arg_context(context); g_strfreev(log_files); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, attrd_exit_status, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(attrd_exit_status); } diff --git a/daemons/based/pacemaker-based.c b/daemons/based/pacemaker-based.c index 86e53e3acb..8266d84400 100644 --- a/daemons/based/pacemaker-based.c +++ b/daemons/based/pacemaker-based.c @@ -1,443 +1,443 @@ /* * Copyright 2004-2023 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 #define SUMMARY "daemon for managing the configuration of a Pacemaker cluster" extern int init_remote_listener(int port, gboolean encrypted); gboolean cib_shutdown_flag = FALSE; int cib_status = pcmk_ok; crm_cluster_t *crm_cluster = NULL; GMainLoop *mainloop = NULL; gchar *cib_root = NULL; static gboolean preserve_status = FALSE; gboolean cib_writes_enabled = TRUE; int remote_fd = 0; int remote_tls_fd = 0; GHashTable *config_hash = NULL; GHashTable *local_notify_queue = NULL; pcmk__output_t *logger_out = NULL; static void cib_init(void); void cib_shutdown(int nsig); static bool startCib(const char *filename); extern int write_cib_contents(gpointer p); static crm_exit_t exit_code = CRM_EX_OK; static void cib_enable_writes(int nsig) { crm_info("(Re)enabling disk writes"); cib_writes_enabled = TRUE; } /*! * \internal * \brief Set up options, users, and groups for stand-alone mode * * \param[out] error GLib error object * * \return Standard Pacemaker return code */ static int setup_stand_alone(GError **error) { int rc = 0; struct passwd *pwentry = NULL; preserve_status = TRUE; cib_writes_enabled = FALSE; errno = 0; pwentry = getpwnam(CRM_DAEMON_USER); if (pwentry == NULL) { exit_code = CRM_EX_FATAL; if (errno != 0) { g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Error getting password DB entry for %s: %s", CRM_DAEMON_USER, strerror(errno)); return errno; } g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Password DB entry for '%s' not found", CRM_DAEMON_USER); return ENXIO; } rc = setgid(pwentry->pw_gid); if (rc < 0) { exit_code = CRM_EX_FATAL; g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not set group to %d: %s", pwentry->pw_gid, strerror(errno)); return errno; } rc = initgroups(CRM_DAEMON_USER, pwentry->pw_gid); if (rc < 0) { exit_code = CRM_EX_FATAL; g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not setup groups for user %d: %s", pwentry->pw_uid, strerror(errno)); return errno; } rc = setuid(pwentry->pw_uid); if (rc < 0) { exit_code = CRM_EX_FATAL; g_set_error(error, PCMK__EXITC_ERROR, exit_code, "Could not set user to %d: %s", pwentry->pw_uid, strerror(errno)); return errno; } return pcmk_rc_ok; } static GOptionEntry entries[] = { { "stand-alone", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &stand_alone, "(Advanced use only) Run in stand-alone mode", NULL }, { "disk-writes", 'w', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &cib_writes_enabled, "(Advanced use only) Enable disk writes (enabled by default unless in " "stand-alone mode)", NULL }, { "cib-root", 'r', G_OPTION_FLAG_NONE, G_OPTION_ARG_FILENAME, &cib_root, "(Advanced use only) Directory where the CIB XML file should be located " "(default: " CRM_CONFIG_DIR ")", NULL }, { NULL } }; static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, "[metadata]"); pcmk__add_main_args(context, entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_ipc_t *old_instance = NULL; 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, "r"); 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 (args->version) { out->version(out, false); goto done; } rc = pcmk__log_output_new(&logger_out); if (rc != pcmk_rc_ok) { exit_code = CRM_EX_ERROR; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format log: %s", pcmk_rc_str(rc)); goto done; } pcmk__output_set_log_level(logger_out, LOG_TRACE); mainloop_add_signal(SIGTERM, cib_shutdown); mainloop_add_signal(SIGPIPE, cib_enable_writes); cib_writer = mainloop_add_trigger(G_PRIORITY_LOW, write_cib_contents, NULL); if ((g_strv_length(processed_args) >= 2) && pcmk__str_eq(processed_args[1], "metadata", pcmk__str_none)) { cib_metadata(); goto done; } pcmk__cli_init_logging("pacemaker-based", args->verbosity); crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); crm_notice("Starting Pacemaker CIB manager"); old_instance = crm_ipc_new(PCMK__SERVER_BASED_RO, 0); if (old_instance == NULL) { /* crm_ipc_new() will have already logged an error message with * crm_err() */ exit_code = CRM_EX_FATAL; goto done; } - if (crm_ipc_connect(old_instance)) { + if (pcmk__connect_generic_ipc(old_instance) == pcmk_rc_ok) { /* IPC end-point already up */ crm_ipc_close(old_instance); crm_ipc_destroy(old_instance); crm_err("pacemaker-based is already active, aborting startup"); goto done; } else { /* not up or not authentic, we'll proceed either way */ crm_ipc_destroy(old_instance); old_instance = NULL; } if (stand_alone) { rc = setup_stand_alone(&error); if (rc != pcmk_rc_ok) { goto done; } } if (cib_root == NULL) { cib_root = g_strdup(CRM_CONFIG_DIR); } else { crm_notice("Using custom config location: %s", cib_root); } if (!pcmk__daemon_can_write(cib_root, NULL)) { exit_code = CRM_EX_FATAL; crm_err("Terminating due to bad permissions on %s", cib_root); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Bad permissions on %s (see logs for details)", cib_root); goto done; } crm_peer_init(); // Read initial CIB, connect to cluster, and start IPC servers cib_init(); // Run the main loop mainloop = g_main_loop_new(NULL, FALSE); crm_notice("Pacemaker CIB manager successfully started and accepting connections"); g_main_loop_run(mainloop); /* If main loop returned, clean up and exit. We disconnect in case * terminate_cib() was called with fast=-1. */ crm_cluster_disconnect(crm_cluster); pcmk__stop_based_ipc(ipcs_ro, ipcs_rw, ipcs_shm); done: g_strfreev(processed_args); pcmk__free_arg_context(context); crm_peer_destroy(); if (local_notify_queue != NULL) { g_hash_table_destroy(local_notify_queue); } if (config_hash != NULL) { g_hash_table_destroy(config_hash); } based_free_transaction_table(); pcmk__client_cleanup(); pcmk_cluster_free(crm_cluster); g_free(cib_root); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } #if SUPPORT_COROSYNC static void cib_cs_dispatch(cpg_handle_t handle, const struct cpg_name *groupName, uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len) { uint32_t kind = 0; xmlNode *xml = NULL; const char *from = NULL; char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from); if(data == NULL) { return; } if (kind == crm_class_cluster) { xml = string2xml(data); if (xml == NULL) { crm_err("Invalid XML: '%.120s'", data); free(data); return; } crm_xml_add(xml, F_ORIG, from); /* crm_xml_add_int(xml, F_SEQ, wrapper->id); */ cib_peer_callback(xml, NULL); } free_xml(xml); free(data); } static void cib_cs_destroy(gpointer user_data) { if (cib_shutdown_flag) { crm_info("Corosync disconnection complete"); } else { crm_crit("Lost connection to cluster layer, shutting down"); terminate_cib(__func__, CRM_EX_DISCONNECT); } } #endif static void cib_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data) { switch (type) { case crm_status_processes: if (cib_legacy_mode() && !pcmk_is_set(node->processes, crm_get_cluster_proc())) { uint32_t old = data? *(const uint32_t *)data : 0; if ((node->processes ^ old) & crm_proc_cpg) { crm_info("Attempting to disable legacy mode after %s left the cluster", node->uname); legacy_mode = FALSE; } } break; case crm_status_uname: case crm_status_nstate: if (cib_shutdown_flag && (crm_active_peers() < 2) && (pcmk__ipc_client_count() == 0)) { crm_info("No more peers"); terminate_cib(__func__, -1); } break; } } static void cib_init(void) { crm_cluster = pcmk_cluster_new(); #if SUPPORT_COROSYNC if (is_corosync_cluster()) { crm_cluster->destroy = cib_cs_destroy; crm_cluster->cpg.cpg_deliver_fn = cib_cs_dispatch; crm_cluster->cpg.cpg_confchg_fn = pcmk_cpg_membership; } #endif // SUPPORT_COROSYNC config_hash = pcmk__strkey_table(free, free); if (startCib("cib.xml") == FALSE) { crm_crit("Cannot start CIB... terminating"); crm_exit(CRM_EX_NOINPUT); } if (!stand_alone) { crm_set_status_callback(&cib_peer_update_callback); if (!crm_cluster_connect(crm_cluster)) { crm_crit("Cannot sign in to the cluster... terminating"); crm_exit(CRM_EX_FATAL); } } pcmk__serve_based_ipc(&ipcs_ro, &ipcs_rw, &ipcs_shm, &ipc_ro_callbacks, &ipc_rw_callbacks); if (stand_alone) { based_is_primary = true; } } static bool startCib(const char *filename) { gboolean active = FALSE; xmlNode *cib = readCibXmlFile(cib_root, filename, !preserve_status); if (activateCibXml(cib, TRUE, "start") == 0) { int port = 0; active = TRUE; cib_read_config(config_hash, cib); pcmk__scan_port(crm_element_value(cib, "remote-tls-port"), &port); if (port >= 0) { remote_tls_fd = init_remote_listener(port, TRUE); } pcmk__scan_port(crm_element_value(cib, "remote-clear-port"), &port); if (port >= 0) { remote_fd = init_remote_listener(port, FALSE); } } return active; } diff --git a/daemons/controld/controld_schedulerd.c b/daemons/controld/controld_schedulerd.c index 912f9a588c..d3a74891e5 100644 --- a/daemons/controld/controld_schedulerd.c +++ b/daemons/controld/controld_schedulerd.c @@ -1,506 +1,507 @@ /* * Copyright 2004-2023 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 /* pid_t, sleep, ssize_t */ #include #include #include #include #include #include #include #include #include static void handle_disconnect(void); static pcmk_ipc_api_t *schedulerd_api = NULL; /*! * \internal * \brief Close any scheduler connection and free associated memory */ void controld_shutdown_schedulerd_ipc(void) { controld_clear_fsa_input_flags(R_PE_REQUIRED); pcmk_disconnect_ipc(schedulerd_api); handle_disconnect(); pcmk_free_ipc_api(schedulerd_api); schedulerd_api = NULL; } /*! * \internal * \brief Save CIB query result to file, raising FSA error * * \param[in] msg Ignored * \param[in] call_id Call ID of CIB query * \param[in] rc Return code of CIB query * \param[in,out] output Result of CIB query * \param[in] user_data Unique identifier for filename * * \note This is intended to be called after a scheduler connection fails. */ static void save_cib_contents(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { const char *id = user_data; register_fsa_error_adv(C_FSA_INTERNAL, I_ERROR, NULL, NULL, __func__); CRM_CHECK(id != NULL, return); if (rc == pcmk_ok) { char *filename = crm_strdup_printf(PE_STATE_DIR "/pe-core-%s.bz2", id); if (write_xml_file(output, filename, TRUE) < 0) { crm_err("Could not save Cluster Information Base to %s after scheduler crash", filename); } else { crm_notice("Saved Cluster Information Base to %s after scheduler crash", filename); } free(filename); } } /*! * \internal * \brief Respond to scheduler connection failure */ static void handle_disconnect(void) { // If we aren't connected to the scheduler, we can't expect a reply controld_expect_sched_reply(NULL); if (pcmk_is_set(controld_globals.fsa_input_register, R_PE_REQUIRED)) { int rc = pcmk_ok; char *uuid_str = crm_generate_uuid(); crm_crit("Connection to the scheduler failed " CRM_XS " uuid=%s", uuid_str); /* * The scheduler died... * * Save the current CIB so that we have a chance of * figuring out what killed it. * * Delay raising the I_ERROR until the query below completes or * 5s is up, whichever comes first. * */ rc = controld_globals.cib_conn->cmds->query(controld_globals.cib_conn, NULL, NULL, cib_scope_local); fsa_register_cib_callback(rc, uuid_str, save_cib_contents); } else { crm_info("Connection to the scheduler released"); } controld_clear_fsa_input_flags(R_PE_CONNECTED); controld_trigger_fsa(); return; } static void handle_reply(pcmk_schedulerd_api_reply_t *reply) { const char *msg_ref = NULL; if (!AM_I_DC) { return; } msg_ref = reply->data.graph.reference; if (msg_ref == NULL) { crm_err("%s - Ignoring calculation with no reference", CRM_OP_PECALC); } else if (pcmk__str_eq(msg_ref, controld_globals.fsa_pe_ref, pcmk__str_none)) { ha_msg_input_t fsa_input; xmlNode *crm_data_node; controld_stop_sched_timer(); /* do_te_invoke (which will eventually process the fsa_input we are constructing * here) requires that fsa_input.xml be non-NULL. That will only happen if * copy_ha_msg_input (which is called by register_fsa_input_adv) sees the * fsa_input.msg that it is expecting. The scheduler's IPC dispatch function * gave us the values we need, we just need to put them into XML. * * The name of the top level element here is irrelevant. Nothing checks it. */ fsa_input.msg = create_xml_node(NULL, "dummy-reply"); crm_xml_add(fsa_input.msg, XML_ATTR_REFERENCE, msg_ref); crm_xml_add(fsa_input.msg, F_CRM_TGRAPH_INPUT, reply->data.graph.input); crm_data_node = create_xml_node(fsa_input.msg, F_CRM_DATA); add_node_copy(crm_data_node, reply->data.graph.tgraph); register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input); free_xml(fsa_input.msg); } else { crm_info("%s calculation %s is obsolete", CRM_OP_PECALC, msg_ref); } } static void scheduler_event_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk_schedulerd_api_reply_t *reply = event_data; switch (event_type) { case pcmk_ipc_event_disconnect: handle_disconnect(); break; case pcmk_ipc_event_reply: handle_reply(reply); break; default: break; } } static bool new_schedulerd_ipc_connection(void) { int rc; controld_set_fsa_input_flags(R_PE_REQUIRED); if (schedulerd_api == NULL) { rc = pcmk_new_ipc_api(&schedulerd_api, pcmk_ipc_schedulerd); if (rc != pcmk_rc_ok) { crm_err("Error connecting to the scheduler: %s", pcmk_rc_str(rc)); return false; } } pcmk_register_ipc_callback(schedulerd_api, scheduler_event_callback, NULL); - rc = pcmk_connect_ipc(schedulerd_api, pcmk_ipc_dispatch_main); + rc = pcmk__connect_ipc(schedulerd_api, pcmk_ipc_dispatch_main, 3); if (rc != pcmk_rc_ok) { - crm_err("Error connecting to the scheduler: %s", pcmk_rc_str(rc)); + crm_err("Error connecting to %s: %s", + pcmk_ipc_name(schedulerd_api, true), pcmk_rc_str(rc)); return false; } controld_set_fsa_input_flags(R_PE_CONNECTED); return true; } static void do_pe_invoke_callback(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data); /* A_PE_START, A_PE_STOP, O_PE_RESTART */ void do_pe_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { if (pcmk_is_set(action, A_PE_STOP)) { controld_clear_fsa_input_flags(R_PE_REQUIRED); pcmk_disconnect_ipc(schedulerd_api); handle_disconnect(); } if (pcmk_is_set(action, A_PE_START) && !pcmk_is_set(controld_globals.fsa_input_register, R_PE_CONNECTED)) { if (cur_state == S_STOPPING) { crm_info("Ignoring request to connect to scheduler while shutting down"); } else if (!new_schedulerd_ipc_connection()) { crm_warn("Could not connect to scheduler"); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } } } static int fsa_pe_query = 0; static mainloop_timer_t *controld_sched_timer = NULL; // @TODO Make this a configurable cluster option if there's demand for it #define SCHED_TIMEOUT_MS (120000) /*! * \internal * \brief Handle a timeout waiting for scheduler reply * * \param[in] user_data Ignored * * \return FALSE (indicating that timer should not be restarted) */ static gboolean controld_sched_timeout(gpointer user_data) { if (AM_I_DC) { /* If this node is the DC but can't communicate with the scheduler, just * exit (and likely get fenced) so this node doesn't interfere with any * further DC elections. * * @TODO We could try something less drastic first, like disconnecting * and reconnecting to the scheduler, but something is likely going * seriously wrong, so perhaps it's better to just fail as quickly as * possible. */ crmd_exit(CRM_EX_FATAL); } return FALSE; } void controld_stop_sched_timer(void) { if ((controld_sched_timer != NULL) && (controld_globals.fsa_pe_ref != NULL)) { crm_trace("Stopping timer for scheduler reply %s", controld_globals.fsa_pe_ref); } mainloop_timer_stop(controld_sched_timer); } /*! * \internal * \brief Set the scheduler request currently being waited on * * \param[in] ref Request to expect reply to (or NULL for none) * * \note This function takes ownership of \p ref. */ void controld_expect_sched_reply(char *ref) { if (ref) { if (controld_sched_timer == NULL) { controld_sched_timer = mainloop_timer_add("scheduler_reply_timer", SCHED_TIMEOUT_MS, FALSE, controld_sched_timeout, NULL); } mainloop_timer_start(controld_sched_timer); } else { controld_stop_sched_timer(); } free(controld_globals.fsa_pe_ref); controld_globals.fsa_pe_ref = ref; } /*! * \internal * \brief Free the scheduler reply timer */ void controld_free_sched_timer(void) { if (controld_sched_timer != NULL) { mainloop_timer_del(controld_sched_timer); controld_sched_timer = NULL; } } /* A_PE_INVOKE */ void do_pe_invoke(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { cib_t *cib_conn = controld_globals.cib_conn; if (AM_I_DC == FALSE) { crm_err("Not invoking scheduler because not DC: %s", fsa_action2string(action)); return; } if (!pcmk_is_set(controld_globals.fsa_input_register, R_PE_CONNECTED)) { if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { crm_err("Cannot shut down gracefully without the scheduler"); register_fsa_input_before(C_FSA_INTERNAL, I_TERMINATE, NULL); } else { crm_info("Waiting for the scheduler to connect"); crmd_fsa_stall(FALSE); controld_set_fsa_action_flags(A_PE_START); controld_trigger_fsa(); } return; } if (cur_state != S_POLICY_ENGINE) { crm_notice("Not invoking scheduler because in state %s", fsa_state2string(cur_state)); return; } if (!pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { crm_err("Attempted to invoke scheduler without consistent Cluster Information Base!"); /* start the join from scratch */ register_fsa_input_before(C_FSA_INTERNAL, I_ELECTION, NULL); return; } fsa_pe_query = cib_conn->cmds->query(cib_conn, NULL, NULL, cib_scope_local); crm_debug("Query %d: Requesting the current CIB: %s", fsa_pe_query, fsa_state2string(controld_globals.fsa_state)); controld_expect_sched_reply(NULL); fsa_register_cib_callback(fsa_pe_query, NULL, do_pe_invoke_callback); } static void force_local_option(xmlNode *xml, const char *attr_name, const char *attr_value) { int max = 0; int lpc = 0; const char *xpath_base = NULL; char *xpath_string = NULL; xmlXPathObjectPtr xpathObj = NULL; xpath_base = pcmk_cib_xpath_for(XML_CIB_TAG_CRMCONFIG); if (xpath_base == NULL) { crm_err(XML_CIB_TAG_CRMCONFIG " CIB element not known (bug?)"); return; } xpath_string = crm_strdup_printf("%s//%s//nvpair[@name='%s']", xpath_base, XML_CIB_TAG_PROPSET, attr_name); xpathObj = xpath_search(xml, xpath_string); max = numXpathResults(xpathObj); free(xpath_string); for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); crm_trace("Forcing %s/%s = %s", ID(match), attr_name, attr_value); crm_xml_add(match, XML_NVPAIR_ATTR_VALUE, attr_value); } if(max == 0) { xmlNode *configuration = NULL; xmlNode *crm_config = NULL; xmlNode *cluster_property_set = NULL; crm_trace("Creating %s-%s for %s=%s", CIB_OPTIONS_FIRST, attr_name, attr_name, attr_value); configuration = pcmk__xe_match(xml, XML_CIB_TAG_CONFIGURATION, NULL, NULL); if (configuration == NULL) { configuration = create_xml_node(xml, XML_CIB_TAG_CONFIGURATION); } crm_config = pcmk__xe_match(configuration, XML_CIB_TAG_CRMCONFIG, NULL, NULL); if (crm_config == NULL) { crm_config = create_xml_node(configuration, XML_CIB_TAG_CRMCONFIG); } cluster_property_set = pcmk__xe_match(crm_config, XML_CIB_TAG_PROPSET, NULL, NULL); if (cluster_property_set == NULL) { cluster_property_set = create_xml_node(crm_config, XML_CIB_TAG_PROPSET); crm_xml_add(cluster_property_set, XML_ATTR_ID, CIB_OPTIONS_FIRST); } xml = create_xml_node(cluster_property_set, XML_CIB_TAG_NVPAIR); crm_xml_set_id(xml, "%s-%s", CIB_OPTIONS_FIRST, attr_name); crm_xml_add(xml, XML_NVPAIR_ATTR_NAME, attr_name); crm_xml_add(xml, XML_NVPAIR_ATTR_VALUE, attr_value); } freeXpathObject(xpathObj); } static void do_pe_invoke_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *ref = NULL; pid_t watchdog = pcmk__locate_sbd(); if (rc != pcmk_ok) { crm_err("Could not retrieve the Cluster Information Base: %s " CRM_XS " rc=%d call=%d", pcmk_strerror(rc), rc, call_id); register_fsa_error_adv(C_FSA_INTERNAL, I_ERROR, NULL, NULL, __func__); return; } else if (call_id != fsa_pe_query) { crm_trace("Skipping superseded CIB query: %d (current=%d)", call_id, fsa_pe_query); return; } else if (!AM_I_DC || !pcmk_is_set(controld_globals.fsa_input_register, R_PE_CONNECTED)) { crm_debug("No need to invoke the scheduler anymore"); return; } else if (controld_globals.fsa_state != S_POLICY_ENGINE) { crm_debug("Discarding scheduler request in state: %s", fsa_state2string(controld_globals.fsa_state)); return; /* this callback counts as 1 */ } else if (num_cib_op_callbacks() > 1) { crm_debug("Re-asking for the CIB: %d other peer updates still pending", (num_cib_op_callbacks() - 1)); sleep(1); controld_set_fsa_action_flags(A_PE_INVOKE); controld_trigger_fsa(); return; } CRM_LOG_ASSERT(output != NULL); /* Refresh the remote node cache and the known node cache when the * scheduler is invoked */ pcmk__refresh_node_caches_from_cib(output); crm_xml_add(output, XML_ATTR_DC_UUID, controld_globals.our_uuid); pcmk__xe_set_bool_attr(output, XML_ATTR_HAVE_QUORUM, pcmk_is_set(controld_globals.flags, controld_has_quorum)); force_local_option(output, XML_ATTR_HAVE_WATCHDOG, pcmk__btoa(watchdog)); if (pcmk_is_set(controld_globals.flags, controld_ever_had_quorum) && !crm_have_quorum) { crm_xml_add_int(output, XML_ATTR_QUORUM_PANIC, 1); } rc = pcmk_rc2legacy(pcmk_schedulerd_api_graph(schedulerd_api, output, &ref)); if (rc < 0) { crm_err("Could not contact the scheduler: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); register_fsa_error_adv(C_FSA_INTERNAL, I_ERROR, NULL, NULL, __func__); } else { CRM_ASSERT(ref != NULL); controld_expect_sched_reply(ref); crm_debug("Invoking the scheduler: query=%d, ref=%s, seq=%llu, " "quorate=%s", fsa_pe_query, controld_globals.fsa_pe_ref, crm_peer_seq, pcmk__btoa(pcmk_is_set(controld_globals.flags, controld_has_quorum))); } } diff --git a/daemons/controld/pacemaker-controld.c b/daemons/controld/pacemaker-controld.c index 5858898c3a..e4a72c2445 100644 --- a/daemons/controld/pacemaker-controld.c +++ b/daemons/controld/pacemaker-controld.c @@ -1,205 +1,205 @@ /* * Copyright 2004-2023 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 #define SUMMARY "daemon for coordinating a Pacemaker cluster's response " \ "to events" _Noreturn void crmd_init(void); extern void init_dotfile(void); controld_globals_t controld_globals = { // Automatic initialization to 0, false, or NULL is fine for most members .fsa_state = S_STARTING, .fsa_actions = A_NOTHING, }; static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { return pcmk__build_arg_context(args, "text (default), xml", group, "[metadata]"); } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_exit_t exit_code = CRM_EX_OK; bool initialize = true; crm_ipc_t *old_instance = NULL; 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, NULL); 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 (args->version) { out->version(out, false); initialize = false; goto done; } if ((g_strv_length(processed_args) >= 2) && pcmk__str_eq(processed_args[1], "metadata", pcmk__str_none)) { crmd_metadata(); initialize = false; goto done; } pcmk__cli_init_logging("pacemaker-controld", args->verbosity); crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); crm_notice("Starting Pacemaker controller"); old_instance = crm_ipc_new(CRM_SYSTEM_CRMD, 0); if (old_instance == NULL) { /* crm_ipc_new will have already printed an error message with crm_err. */ exit_code = CRM_EX_FATAL; goto done; } - if (crm_ipc_connect(old_instance)) { + if (pcmk__connect_generic_ipc(old_instance) == pcmk_rc_ok) { /* IPC end-point already up */ crm_ipc_close(old_instance); crm_ipc_destroy(old_instance); crm_err("pacemaker-controld is already active, aborting startup"); initialize = false; goto done; } else { /* not up or not authentic, we'll proceed either way */ crm_ipc_destroy(old_instance); old_instance = NULL; } if (pcmk__daemon_can_write(PE_STATE_DIR, NULL) == FALSE) { exit_code = CRM_EX_FATAL; crm_err("Terminating due to bad permissions on " PE_STATE_DIR); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Bad permissions on " PE_STATE_DIR " (see logs for details)"); goto done; } else if (pcmk__daemon_can_write(CRM_CONFIG_DIR, NULL) == FALSE) { exit_code = CRM_EX_FATAL; crm_err("Terminating due to bad permissions on " CRM_CONFIG_DIR); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Bad permissions on " CRM_CONFIG_DIR " (see logs for details)"); goto done; } if (pcmk__log_output_new(&(controld_globals.logger_out)) != pcmk_rc_ok) { exit_code = CRM_EX_FATAL; goto done; } pcmk__output_set_log_level(controld_globals.logger_out, LOG_TRACE); done: g_strfreev(processed_args); pcmk__free_arg_context(context); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); if ((exit_code == CRM_EX_OK) && initialize) { // Does not return crmd_init(); } crm_exit(exit_code); } void crmd_init(void) { crm_exit_t exit_code = CRM_EX_OK; enum crmd_fsa_state state; init_dotfile(); register_fsa_input(C_STARTUP, I_STARTUP, NULL); crm_peer_init(); state = s_crmd_fsa(C_STARTUP); if (state == S_PENDING || state == S_STARTING) { /* Create the mainloop and run it... */ crm_trace("Starting %s's mainloop", crm_system_name); controld_globals.mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(controld_globals.mainloop); if (pcmk_is_set(controld_globals.fsa_input_register, R_STAYDOWN)) { crm_info("Inhibiting automated respawn"); exit_code = CRM_EX_FATAL; } } else { crm_err("Startup of %s failed. Current state: %s", crm_system_name, fsa_state2string(state)); exit_code = CRM_EX_ERROR; } crm_info("%s[%lu] exiting with status %d (%s)", crm_system_name, (unsigned long) getpid(), exit_code, crm_exit_str(exit_code)); crmd_fast_exit(exit_code); } diff --git a/daemons/fenced/pacemaker-fenced.c b/daemons/fenced/pacemaker-fenced.c index a199803501..99651d63b8 100644 --- a/daemons/fenced/pacemaker-fenced.c +++ b/daemons/fenced/pacemaker-fenced.c @@ -1,1752 +1,1752 @@ /* * Copyright 2009-2023 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 // PRIu32, PRIx32 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define SUMMARY "daemon for executing fencing devices in a Pacemaker cluster" char *stonith_our_uname = NULL; long stonith_watchdog_timeout_ms = 0; GList *stonith_watchdog_targets = NULL; static GMainLoop *mainloop = NULL; gboolean stand_alone = FALSE; static gboolean stonith_shutdown_flag = FALSE; static qb_ipcs_service_t *ipcs = NULL; static xmlNode *local_cib = NULL; static pe_working_set_t *fenced_data_set = NULL; static const unsigned long long data_set_flags = pe_flag_quick_location | pe_flag_no_compat | pe_flag_no_counts; static cib_t *cib_api = NULL; static pcmk__output_t *logger_out = NULL; static pcmk__output_t *out = NULL; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static struct { bool no_cib_connect; gchar **log_files; } options; static crm_exit_t exit_code = CRM_EX_OK; static void stonith_shutdown(int nsig); static void stonith_cleanup(void); static int32_t st_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { if (stonith_shutdown_flag) { crm_info("Ignoring new client [%d] during shutdown", pcmk__client_pid(c)); return -EPERM; } if (pcmk__new_client(c, uid, gid) == NULL) { return -EIO; } return 0; } /* Exit code means? */ static int32_t st_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; int call_options = 0; xmlNode *request = NULL; pcmk__client_t *c = pcmk__find_client(qbc); const char *op = NULL; if (c == NULL) { crm_info("Invalid client: %p", qbc); return 0; } request = pcmk__client_data2xml(c, data, &id, &flags); if (request == NULL) { pcmk__ipc_send_ack(c, id, flags, "nack", NULL, CRM_EX_PROTOCOL); return 0; } op = crm_element_value(request, F_CRM_TASK); if(pcmk__str_eq(op, CRM_OP_RM_NODE_CACHE, pcmk__str_casei)) { crm_xml_add(request, F_TYPE, T_STONITH_NG); crm_xml_add(request, F_STONITH_OPERATION, op); crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, pcmk__client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); send_cluster_message(NULL, crm_msg_stonith_ng, request, FALSE); free_xml(request); return 0; } if (c->name == NULL) { const char *value = crm_element_value(request, F_STONITH_CLIENTNAME); if (value == NULL) { value = "unknown"; } c->name = crm_strdup_printf("%s.%u", value, c->pid); } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_trace("Flags %#08" PRIx32 "/%#08x for command %" PRIu32 " from client %s", flags, call_options, id, pcmk__client_name(c)); if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(flags & crm_ipc_client_response); CRM_LOG_ASSERT(c->request_id == 0); /* This means the client has two synchronous events in-flight */ c->request_id = id; /* Reply only to the last one */ } crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, pcmk__client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); crm_log_xml_trace(request, "ipc-received"); stonith_command(c, id, flags, request, NULL); free_xml(request); return 0; } /* Error code means? */ static int32_t st_ipc_closed(qb_ipcs_connection_t * c) { pcmk__client_t *client = pcmk__find_client(c); if (client == NULL) { return 0; } crm_trace("Connection %p closed", c); pcmk__free_client(client); /* 0 means: yes, go ahead and destroy the connection */ return 0; } static void st_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p destroyed", c); st_ipc_closed(c); } static void stonith_peer_callback(xmlNode * msg, void *private_data) { const char *remote_peer = crm_element_value(msg, F_ORIG); const char *op = crm_element_value(msg, F_STONITH_OPERATION); if (pcmk__str_eq(op, "poke", pcmk__str_none)) { return; } crm_log_xml_trace(msg, "Peer[inbound]"); stonith_command(NULL, 0, 0, msg, remote_peer); } #if SUPPORT_COROSYNC static void stonith_peer_ais_callback(cpg_handle_t handle, const struct cpg_name *groupName, uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len) { uint32_t kind = 0; xmlNode *xml = NULL; const char *from = NULL; char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from); if(data == NULL) { return; } if (kind == crm_class_cluster) { xml = string2xml(data); if (xml == NULL) { crm_err("Invalid XML: '%.120s'", data); free(data); return; } crm_xml_add(xml, F_ORIG, from); /* crm_xml_add_int(xml, F_SEQ, wrapper->id); */ stonith_peer_callback(xml, NULL); } free_xml(xml); free(data); return; } static void stonith_peer_cs_destroy(gpointer user_data) { crm_crit("Lost connection to cluster layer, shutting down"); stonith_shutdown(0); } #endif void do_local_reply(xmlNode *notify_src, pcmk__client_t *client, int call_options) { /* send callback to originating child */ int local_rc = pcmk_rc_ok; int rid = 0; uint32_t ipc_flags = crm_ipc_server_event; if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_LOG_ASSERT(client->request_id); rid = client->request_id; client->request_id = 0; ipc_flags = crm_ipc_flags_none; } local_rc = pcmk__ipc_send_xml(client, rid, notify_src, ipc_flags); if (local_rc == pcmk_rc_ok) { crm_trace("Sent response %d to client %s", rid, pcmk__client_name(client)); } else { crm_warn("%synchronous reply to client %s failed: %s", (pcmk_is_set(call_options, st_opt_sync_call)? "S" : "As"), pcmk__client_name(client), pcmk_rc_str(local_rc)); } } uint64_t get_stonith_flag(const char *name) { if (pcmk__str_eq(name, T_STONITH_NOTIFY_FENCE, pcmk__str_casei)) { return st_callback_notify_fence; } else if (pcmk__str_eq(name, STONITH_OP_DEVICE_ADD, pcmk__str_casei)) { return st_callback_device_add; } else if (pcmk__str_eq(name, STONITH_OP_DEVICE_DEL, pcmk__str_casei)) { return st_callback_device_del; } else if (pcmk__str_eq(name, T_STONITH_NOTIFY_HISTORY, pcmk__str_casei)) { return st_callback_notify_history; } else if (pcmk__str_eq(name, T_STONITH_NOTIFY_HISTORY_SYNCED, pcmk__str_casei)) { return st_callback_notify_history_synced; } return st_callback_unknown; } static void stonith_notify_client(gpointer key, gpointer value, gpointer user_data) { xmlNode *update_msg = user_data; pcmk__client_t *client = value; const char *type = NULL; CRM_CHECK(client != NULL, return); CRM_CHECK(update_msg != NULL, return); type = crm_element_value(update_msg, F_SUBTYPE); CRM_CHECK(type != NULL, crm_log_xml_err(update_msg, "notify"); return); if (client->ipcs == NULL) { crm_trace("Skipping client with NULL channel"); return; } if (pcmk_is_set(client->flags, get_stonith_flag(type))) { int rc = pcmk__ipc_send_xml(client, 0, update_msg, crm_ipc_server_event); if (rc != pcmk_rc_ok) { crm_warn("%s notification of client %s failed: %s " CRM_XS " id=%.8s rc=%d", type, pcmk__client_name(client), pcmk_rc_str(rc), client->id, rc); } else { crm_trace("Sent %s notification to client %s", type, pcmk__client_name(client)); } } } void do_stonith_async_timeout_update(const char *client_id, const char *call_id, int timeout) { pcmk__client_t *client = NULL; xmlNode *notify_data = NULL; if (!timeout || !call_id || !client_id) { return; } client = pcmk__find_client_by_id(client_id); if (!client) { return; } notify_data = create_xml_node(NULL, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_TYPE, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_STONITH_CALLID, call_id); crm_xml_add_int(notify_data, F_STONITH_TIMEOUT, timeout); crm_trace("timeout update is %d for client %s and call id %s", timeout, client_id, call_id); if (client) { pcmk__ipc_send_xml(client, 0, notify_data, crm_ipc_server_event); } free_xml(notify_data); } /*! * \internal * \brief Notify relevant IPC clients of a fencing operation result * * \param[in] type Notification type * \param[in] result Result of fencing operation (assume success if NULL) * \param[in] data If not NULL, add to notification as call data */ void fenced_send_notification(const char *type, const pcmk__action_result_t *result, xmlNode *data) { /* TODO: Standardize the contents of data */ xmlNode *update_msg = create_xml_node(NULL, "notify"); CRM_LOG_ASSERT(type != NULL); crm_xml_add(update_msg, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(update_msg, F_SUBTYPE, type); crm_xml_add(update_msg, F_STONITH_OPERATION, type); stonith__xe_set_result(update_msg, result); if (data != NULL) { add_message_xml(update_msg, F_STONITH_CALLDATA, data); } crm_trace("Notifying clients"); pcmk__foreach_ipc_client(stonith_notify_client, update_msg); free_xml(update_msg); crm_trace("Notify complete"); } /*! * \internal * \brief Send notifications for a configuration change to subscribed clients * * \param[in] op Notification type (STONITH_OP_DEVICE_ADD, * STONITH_OP_DEVICE_DEL, STONITH_OP_LEVEL_ADD, or * STONITH_OP_LEVEL_DEL) * \param[in] result Operation result * \param[in] desc Description of what changed * \param[in] active Current number of devices or topologies in use */ static void send_config_notification(const char *op, const pcmk__action_result_t *result, const char *desc, int active) { xmlNode *notify_data = create_xml_node(NULL, op); CRM_CHECK(notify_data != NULL, return); crm_xml_add(notify_data, F_STONITH_DEVICE, desc); crm_xml_add_int(notify_data, F_STONITH_ACTIVE, active); fenced_send_notification(op, result, notify_data); free_xml(notify_data); } /*! * \internal * \brief Send notifications for a device change to subscribed clients * * \param[in] op Notification type (STONITH_OP_DEVICE_ADD or * STONITH_OP_DEVICE_DEL) * \param[in] result Operation result * \param[in] desc ID of device that changed */ void fenced_send_device_notification(const char *op, const pcmk__action_result_t *result, const char *desc) { send_config_notification(op, result, desc, g_hash_table_size(device_list)); } /*! * \internal * \brief Send notifications for a topology level change to subscribed clients * * \param[in] op Notification type (STONITH_OP_LEVEL_ADD or * STONITH_OP_LEVEL_DEL) * \param[in] result Operation result * \param[in] desc String representation of level ([]) */ void fenced_send_level_notification(const char *op, const pcmk__action_result_t *result, const char *desc) { send_config_notification(op, result, desc, g_hash_table_size(topology)); } static void topology_remove_helper(const char *node, int level) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; xmlNode *data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); fenced_unregister_level(data, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_DEL, &result, desc); pcmk__reset_result(&result); free_xml(data); free(desc); } static void remove_cib_device(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); } if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } rsc_id = crm_element_value(match, XML_ATTR_ID); stonith_device_remove(rsc_id, true); } } static void remove_topology_level(xmlNode *match) { int index = 0; char *key = NULL; CRM_CHECK(match != NULL, return); key = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); topology_remove_helper(key, index); free(key); } static void add_topology_level(xmlNode *match) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; CRM_CHECK(match != NULL, return); fenced_register_level(match, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_ADD, &result, desc); pcmk__reset_result(&result); free(desc); } static void remove_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if (match && crm_element_value(match, XML_DIFF_MARKER)) { /* Deletion */ int index = 0; char *target = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); if (target == NULL) { crm_err("Invalid fencing target in element %s", ID(match)); } else if (index <= 0) { crm_err("Invalid level for %s in element %s", target, ID(match)); } else { topology_remove_helper(target, index); } /* } else { Deal with modifications during the 'addition' stage */ } } } static void register_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); remove_topology_level(match); add_topology_level(match); } } /* Fencing */ static void fencing_topology_init(void) { xmlXPathObjectPtr xpathObj = NULL; const char *xpath = "//" XML_TAG_FENCING_LEVEL; crm_trace("Full topology refresh"); free_topology_list(); init_topology_list(); /* Grab everything */ xpathObj = xpath_search(local_cib, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } #define rsc_name(x) x->clone_name?x->clone_name:x->id /*! * \internal * \brief Check whether our uname is in a resource's allowed node list * * \param[in] rsc Resource to check * * \return Pointer to node object if found, NULL otherwise */ static pe_node_t * our_node_allowed_for(const pe_resource_t *rsc) { GHashTableIter iter; pe_node_t *node = NULL; if (rsc && stonith_our_uname) { g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node && strcmp(node->details->uname, stonith_our_uname) == 0) { break; } node = NULL; } } return node; } static void watchdog_device_update(void) { if (stonith_watchdog_timeout_ms > 0) { if (!g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) && !stonith_watchdog_targets) { /* getting here watchdog-fencing enabled, no device there yet and reason isn't stonith_watchdog_targets preventing that */ int rc; xmlNode *xml; xml = create_device_registration_xml( STONITH_WATCHDOG_ID, st_namespace_internal, STONITH_WATCHDOG_AGENT, NULL, /* stonith_device_register will add our own name as PCMK_STONITH_HOST_LIST param so we can skip that here */ NULL); rc = stonith_device_register(xml, TRUE); free_xml(xml); if (rc != pcmk_ok) { rc = pcmk_legacy2rc(rc); exit_code = CRM_EX_FATAL; crm_crit("Cannot register watchdog pseudo fence agent: %s", pcmk_rc_str(rc)); stonith_shutdown(0); } } } else if (g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) != NULL) { /* be silent if no device - todo parameter to stonith_device_remove */ stonith_device_remove(STONITH_WATCHDOG_ID, true); } } static void update_stonith_watchdog_timeout_ms(xmlNode *cib) { long timeout_ms = 0; xmlNode *stonith_watchdog_xml = NULL; const char *value = NULL; stonith_watchdog_xml = get_xpath_object("//nvpair[@name='stonith-watchdog-timeout']", cib, LOG_NEVER); if (stonith_watchdog_xml) { value = crm_element_value(stonith_watchdog_xml, XML_NVPAIR_ATTR_VALUE); } if (value) { timeout_ms = crm_get_msec(value); } if (timeout_ms < 0) { timeout_ms = pcmk__auto_watchdog_timeout(); } stonith_watchdog_timeout_ms = timeout_ms; } /*! * \internal * \brief If a resource or any of its children are STONITH devices, update their * definitions given a cluster working set. * * \param[in,out] rsc Resource to check * \param[in,out] data_set Cluster working set with device information */ static void cib_device_update(pe_resource_t *rsc, pe_working_set_t *data_set) { pe_node_t *node = NULL; const char *value = NULL; const char *rclass = NULL; pe_node_t *parent = NULL; /* If this is a complex resource, check children rather than this resource itself. */ if(rsc->children) { GList *gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { cib_device_update(gIter->data, data_set); if(pe_rsc_is_clone(rsc)) { crm_trace("Only processing one copy of the clone %s", rsc->id); break; } } return; } /* We only care about STONITH resources. */ rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); if (!pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { return; } /* If this STONITH resource is disabled, remove it. */ if (pe__resource_is_disabled(rsc)) { crm_info("Device %s has been disabled", rsc->id); return; } /* if watchdog-fencing is disabled handle any watchdog-fence resource as if it was disabled */ if ((stonith_watchdog_timeout_ms <= 0) && pcmk__str_eq(rsc->id, STONITH_WATCHDOG_ID, pcmk__str_none)) { crm_info("Watchdog-fencing disabled thus handling " "device %s as disabled", rsc->id); return; } /* Check whether our node is allowed for this resource (and its parent if in a group) */ node = our_node_allowed_for(rsc); if (rsc->parent && (rsc->parent->variant == pe_group)) { parent = our_node_allowed_for(rsc->parent); } if(node == NULL) { /* Our node is disallowed, so remove the device */ GHashTableIter iter; crm_info("Device %s has been disabled on %s: unknown", rsc->id, stonith_our_uname); g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { crm_trace("Available: %s = %d", pe__node_name(node), node->weight); } return; } else if(node->weight < 0 || (parent && parent->weight < 0)) { /* Our node (or its group) is disallowed by score, so remove the device */ int score = (node->weight < 0)? node->weight : parent->weight; crm_info("Device %s has been disabled on %s: score=%s", rsc->id, stonith_our_uname, pcmk_readable_score(score)); return; } else { /* Our node is allowed, so update the device information */ int rc; xmlNode *data; GHashTable *rsc_params = NULL; GHashTableIter gIter; stonith_key_value_t *params = NULL; const char *name = NULL; const char *agent = crm_element_value(rsc->xml, XML_EXPR_ATTR_TYPE); const char *rsc_provides = NULL; crm_debug("Device %s is allowed on %s: score=%d", rsc->id, stonith_our_uname, node->weight); rsc_params = pe_rsc_params(rsc, node, data_set); get_meta_attributes(rsc->meta, rsc, node, data_set); rsc_provides = g_hash_table_lookup(rsc->meta, PCMK_STONITH_PROVIDES); g_hash_table_iter_init(&gIter, rsc_params); while (g_hash_table_iter_next(&gIter, (gpointer *) & name, (gpointer *) & value)) { if (!name || !value) { continue; } params = stonith_key_value_add(params, name, value); crm_trace(" %s=%s", name, value); } data = create_device_registration_xml(rsc_name(rsc), st_namespace_any, agent, params, rsc_provides); stonith_key_value_freeall(params, 1, 1); rc = stonith_device_register(data, TRUE); CRM_ASSERT(rc == pcmk_ok); free_xml(data); } } /*! * \internal * \brief Update all STONITH device definitions based on current CIB */ static void cib_devices_update(void) { GHashTableIter iter; stonith_device_t *device = NULL; crm_info("Updating devices to version %s.%s.%s", crm_element_value(local_cib, XML_ATTR_GENERATION_ADMIN), crm_element_value(local_cib, XML_ATTR_GENERATION), crm_element_value(local_cib, XML_ATTR_NUMUPDATES)); if (fenced_data_set->now != NULL) { crm_time_free(fenced_data_set->now); fenced_data_set->now = NULL; } fenced_data_set->localhost = stonith_our_uname; pcmk__schedule_actions(local_cib, data_set_flags, fenced_data_set); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->cib_registered) { device->dirty = TRUE; } } /* have list repopulated if cib has a watchdog-fencing-resource TODO: keep a cached list for queries happening while we are refreshing */ g_list_free_full(stonith_watchdog_targets, free); stonith_watchdog_targets = NULL; g_list_foreach(fenced_data_set->resources, (GFunc) cib_device_update, fenced_data_set); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->dirty) { g_hash_table_iter_remove(&iter); } } fenced_data_set->input = NULL; // Wasn't a copy, so don't let API free it pe_reset_working_set(fenced_data_set); } static void update_cib_stonith_devices_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; char *reason = NULL; bool needs_update = FALSE; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); const char *shortpath = NULL; if ((op == NULL) || (strcmp(op, "move") == 0) || strstr(xpath, "/"XML_CIB_TAG_STATUS)) { continue; } else if (pcmk__str_eq(op, "delete", pcmk__str_casei) && strstr(xpath, "/"XML_CIB_TAG_RESOURCE)) { const char *rsc_id = NULL; char *search = NULL; char *mutable = NULL; if (strstr(xpath, XML_TAG_ATTR_SETS) || strstr(xpath, XML_TAG_META_SETS)) { needs_update = TRUE; pcmk__str_update(&reason, "(meta) attribute deleted from resource"); break; } pcmk__str_update(&mutable, xpath); rsc_id = strstr(mutable, "primitive[@" XML_ATTR_ID "=\'"); if (rsc_id != NULL) { rsc_id += strlen("primitive[@" XML_ATTR_ID "=\'"); search = strchr(rsc_id, '\''); } if (search != NULL) { *search = 0; stonith_device_remove(rsc_id, true); /* watchdog_device_update called afterwards to fall back to implicit definition if needed */ } else { crm_warn("Ignoring malformed CIB update (resource deletion)"); } free(mutable); } else if (strstr(xpath, "/"XML_CIB_TAG_RESOURCES) || strstr(xpath, "/"XML_CIB_TAG_CONSTRAINTS) || strstr(xpath, "/"XML_CIB_TAG_RSCCONFIG)) { shortpath = strrchr(xpath, '/'); CRM_ASSERT(shortpath); reason = crm_strdup_printf("%s %s", op, shortpath+1); needs_update = TRUE; break; } } if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } else { crm_trace("No updates for device list found in CIB"); } free(reason); } static void update_cib_stonith_devices_v1(const char *event, xmlNode * msg) { const char *reason = "none"; gboolean needs_update = FALSE; xmlXPathObjectPtr xpath_obj = NULL; /* process new constraints */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_CONS_TAG_RSC_LOCATION); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; /* Safest and simplest to always recompute */ needs_update = TRUE; reason = "new location constraint"; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpath_obj, lpc); crm_log_xml_trace(match, "new constraint"); } } freeXpathObject(xpath_obj); /* process deletions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { remove_cib_device(xpath_obj); } freeXpathObject(xpath_obj); /* process additions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpath_obj, lpc); rsc_id = crm_element_value(match, XML_ATTR_ID); standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } crm_trace("Fencing resource %s was added or modified", rsc_id); reason = "new resource"; needs_update = TRUE; } } freeXpathObject(xpath_obj); if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } } static void update_cib_stonith_devices(const char *event, xmlNode * msg) { int format = 1; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); switch(format) { case 1: update_cib_stonith_devices_v1(event, msg); break; case 2: update_cib_stonith_devices_v2(event, msg); break; default: crm_warn("Unknown patch format: %d", format); } } /*! * \internal * \brief Check whether a node has a specific attribute name/value * * \param[in] node Name of node to check * \param[in] name Name of an attribute to look for * \param[in] value The value the named attribute needs to be set to in order to be considered a match * * \return TRUE if the locally cached CIB has the specified node attribute */ gboolean node_has_attr(const char *node, const char *name, const char *value) { GString *xpath = NULL; xmlNode *match; CRM_CHECK((local_cib != NULL) && (node != NULL) && (name != NULL) && (value != NULL), return FALSE); /* Search for the node's attributes in the CIB. While the schema allows * multiple sets of instance attributes, and allows instance attributes to * use id-ref to reference values elsewhere, that is intended for resources, * so we ignore that here. */ xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, "//" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE "[@" XML_ATTR_UNAME "='", node, "']/" XML_TAG_ATTR_SETS "/" XML_CIB_TAG_NVPAIR "[@" XML_NVPAIR_ATTR_NAME "='", name, "' " "and @" XML_NVPAIR_ATTR_VALUE "='", value, "']", NULL); match = get_xpath_object((const char *) xpath->str, local_cib, LOG_NEVER); g_string_free(xpath, TRUE); return (match != NULL); } /*! * \internal * \brief Check whether a node does watchdog-fencing * * \param[in] node Name of node to check * * \return TRUE if node found in stonith_watchdog_targets * or stonith_watchdog_targets is empty indicating * all nodes are doing watchdog-fencing */ gboolean node_does_watchdog_fencing(const char *node) { return ((stonith_watchdog_targets == NULL) || pcmk__str_in_list(node, stonith_watchdog_targets, pcmk__str_casei)); } static void update_fencing_topology(const char *event, xmlNode * msg) { int format = 1; const char *xpath; xmlXPathObjectPtr xpathObj = NULL; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); if(format == 1) { /* Process deletions (only) */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); remove_fencing_topology(xpathObj); freeXpathObject(xpathObj); /* Process additions and changes */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } else if(format == 2) { xmlNode *change = NULL; int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; xml_patch_versions(patchset, add, del); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); if(op == NULL) { continue; } else if(strstr(xpath, "/" XML_TAG_FENCING_LEVEL) != NULL) { /* Change to a specific entry */ crm_trace("Handling %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "create") == 0) { add_topology_level(change->children); } else if(strcmp(op, "modify") == 0) { xmlNode *match = first_named_child(change, XML_DIFF_RESULT); if(match) { remove_topology_level(match->children); add_topology_level(match->children); } } else if(strcmp(op, "delete") == 0) { /* Nuclear option, all we have is the path and an id... not enough to remove a specific entry */ crm_info("Re-initializing fencing topology after %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else if (strstr(xpath, "/" XML_TAG_FENCING_TOPOLOGY) != NULL) { /* Change to the topology in general */ crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } else if (strstr(xpath, "/" XML_CIB_TAG_CONFIGURATION)) { /* Changes to the whole config section, possibly including the topology as a whild */ if(first_named_child(change, XML_TAG_FENCING_TOPOLOGY) == NULL) { crm_trace("Nothing for us in %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); } else if(strcmp(op, "delete") == 0 || strcmp(op, "create") == 0) { crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else { crm_trace("Nothing for us in %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); } } } else { crm_warn("Unknown patch format: %d", format); } } static bool have_cib_devices = FALSE; static void update_cib_cache_cb(const char *event, xmlNode * msg) { int rc = pcmk_ok; long timeout_ms_saved = stonith_watchdog_timeout_ms; bool need_full_refresh = false; if(!have_cib_devices) { crm_trace("Skipping updates until we get a full dump"); return; } else if(msg == NULL) { crm_trace("Missing %s update", event); return; } /* Maintain a local copy of the CIB so that we have full access * to device definitions, location constraints, and node attributes */ if (local_cib != NULL) { int rc = pcmk_ok; xmlNode *patchset = NULL; crm_element_value_int(msg, F_CIB_RC, &rc); if (rc != pcmk_ok) { return; } patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); pcmk__output_set_log_level(logger_out, LOG_TRACE); out->message(out, "xml-patchset", patchset); rc = xml_apply_patchset(local_cib, patchset, TRUE); switch (rc) { case pcmk_ok: case -pcmk_err_old_data: break; case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; break; default: crm_warn("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; } } if (local_cib == NULL) { crm_trace("Re-requesting full CIB"); rc = cib_api->cmds->query(cib_api, NULL, &local_cib, cib_scope_local | cib_sync_call); if(rc != pcmk_ok) { crm_err("Couldn't retrieve the CIB: %s (%d)", pcmk_strerror(rc), rc); return; } CRM_ASSERT(local_cib != NULL); need_full_refresh = true; } pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); if (timeout_ms_saved != stonith_watchdog_timeout_ms) { need_full_refresh = true; } if (need_full_refresh) { fencing_topology_init(); cib_devices_update(); } else { // Partial refresh update_fencing_topology(event, msg); update_cib_stonith_devices(event, msg); } watchdog_device_update(); } static void init_cib_cache_cb(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { crm_info("Updating device list from CIB"); have_cib_devices = TRUE; local_cib = copy_xml(output); pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); fencing_topology_init(); cib_devices_update(); watchdog_device_update(); } static void stonith_shutdown(int nsig) { crm_info("Terminating with %d clients", pcmk__ipc_client_count()); stonith_shutdown_flag = TRUE; if (mainloop != NULL && g_main_loop_is_running(mainloop)) { g_main_loop_quit(mainloop); } } static void cib_connection_destroy(gpointer user_data) { if (stonith_shutdown_flag) { crm_info("Connection to the CIB manager closed"); return; } else { crm_crit("Lost connection to the CIB manager, shutting down"); } if (cib_api) { cib_api->cmds->signoff(cib_api); } stonith_shutdown(0); } static void stonith_cleanup(void) { if (cib_api) { cib_api->cmds->del_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb); cib_api->cmds->signoff(cib_api); } if (ipcs) { qb_ipcs_destroy(ipcs); } crm_peer_destroy(); pcmk__client_cleanup(); free_stonith_remote_op_list(); free_topology_list(); free_device_list(); free_metadata_cache(); fenced_unregister_handlers(); free(stonith_our_uname); stonith_our_uname = NULL; free_xml(local_cib); local_cib = NULL; } static gboolean stand_alone_cpg_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { stand_alone = FALSE; options.no_cib_connect = true; return TRUE; } static void setup_cib(void) { int rc, retries = 0; cib_api = cib_new(); if (cib_api == NULL) { crm_err("No connection to the CIB manager"); return; } do { sleep(retries); rc = cib_api->cmds->signon(cib_api, CRM_SYSTEM_STONITHD, cib_command); } while (rc == -ENOTCONN && ++retries < 5); if (rc != pcmk_ok) { crm_err("Could not connect to the CIB manager: %s (%d)", pcmk_strerror(rc), rc); } else if (pcmk_ok != cib_api->cmds->add_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb)) { crm_err("Could not set CIB notification callback"); } else { rc = cib_api->cmds->query(cib_api, NULL, NULL, cib_scope_local); cib_api->cmds->register_callback(cib_api, rc, 120, FALSE, NULL, "init_cib_cache_cb", init_cib_cache_cb); cib_api->cmds->set_connection_dnotify(cib_api, cib_connection_destroy); crm_info("Watching for fencing topology changes"); } } struct qb_ipcs_service_handlers ipc_callbacks = { .connection_accept = st_ipc_accept, .connection_created = NULL, .msg_process = st_ipc_dispatch, .connection_closed = st_ipc_closed, .connection_destroyed = st_ipc_destroy }; /*! * \internal * \brief Callback for peer status changes * * \param[in] type What changed * \param[in] node What peer had the change * \param[in] data Previous value of what changed */ static void st_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data) { if ((type != crm_status_processes) && !pcmk_is_set(node->flags, crm_remote_node)) { /* * This is a hack until we can send to a nodeid and/or we fix node name lookups * These messages are ignored in stonith_peer_callback() */ xmlNode *query = create_xml_node(NULL, "stonith_command"); crm_xml_add(query, F_XML_TAGNAME, "stonith_command"); crm_xml_add(query, F_TYPE, T_STONITH_NG); crm_xml_add(query, F_STONITH_OPERATION, "poke"); crm_debug("Broadcasting our uname because of node %u", node->id); send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); } } static pcmk__cluster_option_t fencer_options[] = { /* name, old name, type, allowed values, * default value, validator, * short description, * long description */ { PCMK_STONITH_HOST_ARGUMENT, NULL, "string", NULL, "port", NULL, N_("Advanced use only: An alternate parameter to supply instead of 'port'"), N_("some devices do not support the " "standard 'port' parameter or may provide additional ones. Use " "this to specify an alternate, device-specific, parameter " "that should indicate the machine to be fenced. A value of " "none can be used to tell the cluster not to supply any " "additional parameters.") }, { PCMK_STONITH_HOST_MAP,NULL, "string", NULL, "", NULL, N_("A mapping of host names to ports numbers for devices that do not support host names."), N_("Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and 3 for node2") }, { PCMK_STONITH_HOST_LIST,NULL, "string", NULL, "", NULL, N_("Eg. node1,node2,node3"), N_("A list of machines controlled by " "this device (Optional unless pcmk_host_list=static-list)") }, { PCMK_STONITH_HOST_CHECK,NULL, "string", NULL, "dynamic-list", NULL, N_("How to determine which machines are controlled by the device."), N_("Allowed values: dynamic-list " "(query the device via the 'list' command), static-list " "(check the pcmk_host_list attribute), status " "(query the device via the 'status' command), " "none (assume every device can fence every " "machine)") }, { PCMK_STONITH_DELAY_MAX,NULL, "time", NULL, "0s", NULL, N_("Enable a base delay for fencing actions and specify base delay value."), N_("Enable a delay of no more than the " "time specified before executing fencing actions. Pacemaker " "derives the overall delay by taking the value of " "pcmk_delay_base and adding a random delay value such " "that the sum is kept below this maximum.") }, { PCMK_STONITH_DELAY_BASE,NULL, "string", NULL, "0s", NULL, N_("Enable a base delay for " "fencing actions and specify base delay value."), N_("This enables a static delay for " "fencing actions, which can help avoid \"death matches\" where " "two nodes try to fence each other at the same time. If " "pcmk_delay_max is also used, a random delay will be " "added such that the total delay is kept below that value." "This can be set to a single time value to apply to any node " "targeted by this device (useful if a separate device is " "configured for each target), or to a node map (for example, " "\"node1:1s;node2:5\") to set a different value per target.") }, { PCMK_STONITH_ACTION_LIMIT,NULL, "integer", NULL, "1", NULL, N_("The maximum number of actions can be performed in parallel on this device"), N_("Cluster property concurrent-fencing=true needs to be configured first." "Then use this to specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.") }, { "pcmk_reboot_action",NULL, "string", NULL, "reboot", NULL, N_("Advanced use only: An alternate command to run instead of 'reboot'"), N_("Some devices do not support the standard commands or may provide additional ones.\n" "Use this to specify an alternate, device-specific, command that implements the \'reboot\' action.") }, { "pcmk_reboot_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for reboot actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'reboot\' actions.") }, { "pcmk_reboot_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'reboot' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'reboot\' actions before giving up.") }, { "pcmk_off_action",NULL, "string", NULL, "off", NULL, N_("Advanced use only: An alternate command to run instead of \'off\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'off\' action.") }, { "pcmk_off_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for off actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'off\' actions.") }, { "pcmk_off_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'off' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'off\' actions before giving up.") }, { "pcmk_on_action",NULL, "string", NULL, "on", NULL, N_("Advanced use only: An alternate command to run instead of 'on'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'on\' action.") }, { "pcmk_on_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for on actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'on\' actions.") }, { "pcmk_on_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'on' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'on\' actions before giving up.") }, { "pcmk_list_action",NULL, "string", NULL, "list", NULL, N_("Advanced use only: An alternate command to run instead of \'list\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'list\' action.") }, { "pcmk_list_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for list actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'list\' actions.") }, { "pcmk_list_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'list\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'list\' actions before giving up.") }, { "pcmk_monitor_action",NULL, "string", NULL, "monitor", NULL, N_("Advanced use only: An alternate command to run instead of \'monitor\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'monitor\' action.") }, { "pcmk_monitor_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for monitor actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal.\n" "Use this to specify an alternate, device-specific, timeout for \'monitor\' actions.") }, { "pcmk_monitor_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'monitor\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'monitor\' actions before giving up.") }, { "pcmk_status_action",NULL, "string", NULL, "status", NULL, N_("Advanced use only: An alternate command to run instead of \'status\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'status\' action.") }, { "pcmk_status_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for status actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'status\' actions.") }, { "pcmk_status_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'status\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'status\' actions before giving up.") }, }; void fencer_metadata(void) { const char *desc_short = N_("Instance attributes available for all " "\"stonith\"-class resources"); const char *desc_long = N_("Instance attributes available for all \"stonith\"-" "class resources and used by Pacemaker's fence " "daemon, formerly known as stonithd"); gchar *s = pcmk__format_option_metadata("pacemaker-fenced", desc_short, desc_long, fencer_options, PCMK__NELEM(fencer_options)); printf("%s", s); g_free(s); } static GOptionEntry entries[] = { { "stand-alone", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &stand_alone, N_("Deprecated (will be removed in a future release)"), NULL }, { "stand-alone-w-cpg", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, stand_alone_cpg_cb, N_("Intended for use in regression testing only"), NULL }, { "logfile", 'l', G_OPTION_FLAG_NONE, G_OPTION_ARG_FILENAME_ARRAY, &options.log_files, N_("Send logs to the additional named logfile"), NULL }, { NULL } }; static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, "[metadata]"); pcmk__add_main_args(context, entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_cluster_t *cluster = NULL; crm_ipc_t *old_instance = 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, "l"); 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 (args->version) { out->version(out, false); goto done; } if ((g_strv_length(processed_args) >= 2) && pcmk__str_eq(processed_args[1], "metadata", pcmk__str_none)) { fencer_metadata(); goto done; } // Open additional log files pcmk__add_logfiles(options.log_files, out); crm_log_init(NULL, LOG_INFO + args->verbosity, TRUE, (args->verbosity > 0), argc, argv, FALSE); crm_notice("Starting Pacemaker fencer"); old_instance = crm_ipc_new("stonith-ng", 0); if (old_instance == NULL) { /* crm_ipc_new() will have already logged an error message with * crm_err() */ exit_code = CRM_EX_FATAL; goto done; } - if (crm_ipc_connect(old_instance)) { + if (pcmk__connect_generic_ipc(old_instance) == pcmk_rc_ok) { // IPC endpoint already up crm_ipc_close(old_instance); crm_ipc_destroy(old_instance); crm_err("pacemaker-fenced is already active, aborting startup"); goto done; } else { // Not up or not authentic, we'll proceed either way crm_ipc_destroy(old_instance); old_instance = NULL; } mainloop_add_signal(SIGTERM, stonith_shutdown); crm_peer_init(); fenced_data_set = pe_new_working_set(); CRM_ASSERT(fenced_data_set != NULL); cluster = pcmk_cluster_new(); /* Initialize the logger prior to setup_cib(). update_cib_cache_cb() may * call the "xml-patchset" message function, which needs the logger, after * setup_cib() has run. */ rc = pcmk__log_output_new(&logger_out) != pcmk_rc_ok; if (rc != pcmk_rc_ok) { exit_code = CRM_EX_FATAL; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format log: %s", pcmk_rc_str(rc)); goto done; } pe__register_messages(logger_out); pcmk__register_lib_messages(logger_out); pcmk__output_set_log_level(logger_out, LOG_TRACE); fenced_data_set->priv = logger_out; if (!stand_alone) { #if SUPPORT_COROSYNC if (is_corosync_cluster()) { cluster->destroy = stonith_peer_cs_destroy; cluster->cpg.cpg_deliver_fn = stonith_peer_ais_callback; cluster->cpg.cpg_confchg_fn = pcmk_cpg_membership; } #endif // SUPPORT_COROSYNC crm_set_status_callback(&st_peer_update_callback); if (crm_cluster_connect(cluster) == FALSE) { exit_code = CRM_EX_FATAL; crm_crit("Cannot sign in to the cluster... terminating"); goto done; } pcmk__str_update(&stonith_our_uname, cluster->uname); if (!options.no_cib_connect) { setup_cib(); } } else { pcmk__str_update(&stonith_our_uname, "localhost"); crm_warn("Stand-alone mode is deprecated and will be removed " "in a future release"); } init_device_list(); init_topology_list(); pcmk__serve_fenced_ipc(&ipcs, &ipc_callbacks); // Create the mainloop and run it... mainloop = g_main_loop_new(NULL, FALSE); crm_notice("Pacemaker fencer successfully started and accepting connections"); g_main_loop_run(mainloop); done: g_strfreev(processed_args); pcmk__free_arg_context(context); g_strfreev(options.log_files); stonith_cleanup(); pcmk_cluster_free(cluster); pe_free_working_set(fenced_data_set); pcmk__output_and_clear_error(&error, out); if (logger_out != NULL) { logger_out->finish(logger_out, exit_code, true, NULL); pcmk__output_free(logger_out); } if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } diff --git a/daemons/pacemakerd/pacemakerd.c b/daemons/pacemakerd/pacemakerd.c index 9f77ccca25..6a14fa4871 100644 --- a/daemons/pacemakerd/pacemakerd.c +++ b/daemons/pacemakerd/pacemakerd.c @@ -1,483 +1,487 @@ /* * Copyright 2010-2023 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 "pacemakerd.h" #include #include #include #include #include #include #include #include #include #include /* indirectly: CRM_EX_* */ #include #include #include #include #include #include #include #define SUMMARY "pacemakerd - primary Pacemaker daemon that launches and monitors all subsidiary Pacemaker daemons" struct { gboolean features; gboolean foreground; gboolean shutdown; gboolean standby; } options; static pcmk__output_t *out = NULL; static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; PCMK__OUTPUT_ARGS("features") static int pacemakerd_features(pcmk__output_t *out, va_list args) { out->info(out, "Pacemaker %s (Build: %s)\n Supporting v%s: %s", PACEMAKER_VERSION, BUILD_VERSION, CRM_FEATURE_SET, CRM_FEATURES); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("features") static int pacemakerd_features_xml(pcmk__output_t *out, va_list args) { gchar **feature_list = g_strsplit(CRM_FEATURES, " ", 0); pcmk__output_xml_create_parent(out, "pacemakerd", "version", PACEMAKER_VERSION, "build", BUILD_VERSION, "feature_set", CRM_FEATURE_SET, NULL); out->begin_list(out, NULL, NULL, "features"); for (char **s = feature_list; *s != NULL; s++) { pcmk__output_create_xml_text_node(out, "feature", *s); } out->end_list(out); g_strfreev(feature_list); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "features", "default", pacemakerd_features }, { "features", "xml", pacemakerd_features_xml }, { NULL, NULL, NULL } }; static gboolean pid_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return TRUE; } static gboolean standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { options.standby = TRUE; pcmk__set_env_option("node_start_state", "standby"); return TRUE; } static GOptionEntry entries[] = { { "features", 'F', 0, G_OPTION_ARG_NONE, &options.features, "Display full version and list of features Pacemaker was built with", NULL }, { "foreground", 'f', 0, G_OPTION_ARG_NONE, &options.foreground, "(Ignored) Pacemaker always runs in the foreground", NULL }, { "pid-file", 'p', 0, G_OPTION_ARG_CALLBACK, pid_cb, "(Ignored) Daemon pid file location", "FILE" }, { "shutdown", 'S', 0, G_OPTION_ARG_NONE, &options.shutdown, "Instruct Pacemaker to shutdown on this machine", NULL }, { "standby", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, standby_cb, "Start node in standby state", NULL }, { NULL } }; static void pcmk_ignore(int nsig) { crm_info("Ignoring signal %s (%d)", strsignal(nsig), nsig); } static void pcmk_sigquit(int nsig) { pcmk__panic(__func__); } static void mcp_chown(const char *path, uid_t uid, gid_t gid) { int rc = chown(path, uid, gid); if (rc < 0) { crm_warn("Cannot change the ownership of %s to user %s and gid %d: %s", path, CRM_DAEMON_USER, gid, pcmk_rc_str(errno)); } } static void create_pcmk_dirs(void) { uid_t pcmk_uid = 0; gid_t pcmk_gid = 0; const char *dirs[] = { CRM_PACEMAKER_DIR, // core/blackbox/scheduler/CIB files CRM_CORE_DIR, // core files CRM_BLACKBOX_DIR, // blackbox dumps PE_STATE_DIR, // scheduler inputs CRM_CONFIG_DIR, // the Cluster Information Base (CIB) // Don't build CRM_RSCTMP_DIR, pacemaker-execd will do it NULL }; if (pcmk_daemon_user(&pcmk_uid, &pcmk_gid) < 0) { crm_err("Cluster user %s does not exist, aborting Pacemaker startup", CRM_DAEMON_USER); crm_exit(CRM_EX_NOUSER); } // Used by some resource agents if ((mkdir(CRM_STATE_DIR, 0750) < 0) && (errno != EEXIST)) { crm_warn("Could not create directory " CRM_STATE_DIR ": %s", pcmk_rc_str(errno)); } else { mcp_chown(CRM_STATE_DIR, pcmk_uid, pcmk_gid); } for (int i = 0; dirs[i] != NULL; ++i) { int rc = pcmk__build_path(dirs[i], 0750); if (rc != pcmk_rc_ok) { crm_warn("Could not create directory %s: %s", dirs[i], pcmk_rc_str(rc)); } else { mcp_chown(dirs[i], pcmk_uid, pcmk_gid); } } } static void remove_core_file_limit(void) { struct rlimit cores; // Get current limits if (getrlimit(RLIMIT_CORE, &cores) < 0) { crm_notice("Unable to check system core file limits " "(consider ensuring the size is unlimited): %s", strerror(errno)); return; } // Check whether core dumps are disabled if (cores.rlim_max == 0) { if (geteuid() != 0) { // Yes, and there's nothing we can do about it crm_notice("Core dumps are disabled (consider enabling them)"); return; } cores.rlim_max = RLIM_INFINITY; // Yes, but we're root, so enable them } // Raise soft limit to hard limit (if not already done) if (cores.rlim_cur != cores.rlim_max) { cores.rlim_cur = cores.rlim_max; if (setrlimit(RLIMIT_CORE, &cores) < 0) { crm_notice("Unable to raise system limit on core file size " "(consider doing so manually): %s", strerror(errno)); return; } } if (cores.rlim_cur == RLIM_INFINITY) { crm_trace("Core file size is unlimited"); } else { crm_trace("Core file size is limited to %llu bytes", (unsigned long long) cores.rlim_cur); } } static void pacemakerd_event_cb(pcmk_ipc_api_t *pacemakerd_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk_pacemakerd_api_reply_t *reply = event_data; switch (event_type) { case pcmk_ipc_event_reply: break; default: return; } if (status != CRM_EX_OK) { out->err(out, "Bad reply from pacemakerd: %s", crm_exit_str(status)); return; } if (reply->reply_type != pcmk_pacemakerd_reply_shutdown) { out->err(out, "Unknown reply type %d from pacemakerd", reply->reply_type); } } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); pcmk__add_main_args(context, entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_exit_t exit_code = CRM_EX_OK; GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "p"); GOptionContext *context = build_arg_context(args, &output_group); bool old_instance_connected = false; pcmk_ipc_api_t *old_instance = NULL; qb_ipcs_service_t *ipcs = NULL; subdaemon_check_progress = time(NULL); setenv("LC_ALL", "C", 1); // Ensure logs are in a common language crm_log_preinit(NULL, argc, argv); mainloop_add_signal(SIGHUP, pcmk_ignore); mainloop_add_signal(SIGQUIT, pcmk_sigquit); 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) || (out == NULL)) { 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; } pcmk__force_args(context, &error, "%s --xml-simple-list", g_get_prgname()); pcmk__register_messages(out, fmt_functions); if (options.features) { out->message(out, "features"); exit_code = CRM_EX_OK; goto done; } if (args->version) { out->version(out, false); goto done; } pcmk__set_env_option("mcp", "true"); if (options.shutdown) { pcmk__cli_init_logging("pacemakerd", args->verbosity); } else { crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); } crm_debug("Checking for existing Pacemaker instance"); rc = pcmk_new_ipc_api(&old_instance, pcmk_ipc_pacemakerd); if (old_instance == NULL) { out->err(out, "Could not check for existing pacemakerd: %s", pcmk_rc_str(rc)); exit_code = pcmk_rc2exitc(rc); goto done; } pcmk_register_ipc_callback(old_instance, pacemakerd_event_cb, NULL); - rc = pcmk_connect_ipc(old_instance, pcmk_ipc_dispatch_sync); + rc = pcmk__connect_ipc(old_instance, pcmk_ipc_dispatch_sync, 2); + if (rc != pcmk_rc_ok) { + crm_debug("No existing %s instance found: %s", + pcmk_ipc_name(old_instance, true), pcmk_rc_str(rc)); + } old_instance_connected = pcmk_ipc_is_connected(old_instance); if (options.shutdown) { if (old_instance_connected) { rc = pcmk_pacemakerd_api_shutdown(old_instance, crm_system_name); pcmk_dispatch_ipc(old_instance); exit_code = pcmk_rc2exitc(rc); if (exit_code != CRM_EX_OK) { pcmk_free_ipc_api(old_instance); goto done; } /* We get the ACK immediately, and the response right after that, * but it might take a while for pacemakerd to get around to * shutting down. Wait for that to happen (with 30-minute timeout). */ for (int i = 0; i < 900; i++) { if (!pcmk_ipc_is_connected(old_instance)) { exit_code = CRM_EX_OK; pcmk_free_ipc_api(old_instance); goto done; } sleep(2); } exit_code = CRM_EX_TIMEOUT; pcmk_free_ipc_api(old_instance); goto done; } else { out->err(out, "Could not request shutdown " "of existing Pacemaker instance: %s", pcmk_rc_str(rc)); pcmk_free_ipc_api(old_instance); exit_code = CRM_EX_DISCONNECT; goto done; } } else if (old_instance_connected) { pcmk_free_ipc_api(old_instance); crm_err("Aborting start-up because active Pacemaker instance found"); exit_code = CRM_EX_FATAL; goto done; } pcmk_free_ipc_api(old_instance); /* Don't allow any accidental output after this point. */ if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); out = NULL; } #ifdef SUPPORT_COROSYNC if (mcp_read_config() == FALSE) { crm_exit(CRM_EX_UNAVAILABLE); } #endif // OCF shell functions and cluster-glue need facility under different name { const char *facility = pcmk__env_option(PCMK__ENV_LOGFACILITY); if (!pcmk__str_eq(facility, PCMK__VALUE_NONE, pcmk__str_casei|pcmk__str_null_matches)) { setenv("HA_LOGFACILITY", facility, 1); } } crm_notice("Starting Pacemaker %s "CRM_XS" build=%s features:%s", PACEMAKER_VERSION, BUILD_VERSION, CRM_FEATURES); mainloop = g_main_loop_new(NULL, FALSE); remove_core_file_limit(); create_pcmk_dirs(); pcmk__serve_pacemakerd_ipc(&ipcs, &mcp_ipc_callbacks); #ifdef SUPPORT_COROSYNC /* Allows us to block shutdown */ if (!cluster_connect_cfg()) { exit_code = CRM_EX_PROTOCOL; goto done; } #endif if (pcmk__locate_sbd() > 0) { setenv("PCMK_watchdog", "true", 1); running_with_sbd = TRUE; } else { setenv("PCMK_watchdog", "false", 1); } switch (find_and_track_existing_processes()) { case pcmk_rc_ok: break; case pcmk_rc_ipc_unauthorized: exit_code = CRM_EX_CANTCREAT; goto done; default: exit_code = CRM_EX_FATAL; goto done; }; mainloop_add_signal(SIGTERM, pcmk_shutdown); mainloop_add_signal(SIGINT, pcmk_shutdown); if ((running_with_sbd) && pcmk__get_sbd_sync_resource_startup()) { crm_notice("Waiting for startup-trigger from SBD."); pacemakerd_state = XML_PING_ATTR_PACEMAKERDSTATE_WAITPING; startup_trigger = mainloop_add_trigger(G_PRIORITY_HIGH, init_children_processes, NULL); } else { if (running_with_sbd) { crm_warn("Enabling SBD_SYNC_RESOURCE_STARTUP would (if supported " "by your SBD version) improve reliability of " "interworking between SBD & pacemaker."); } pacemakerd_state = XML_PING_ATTR_PACEMAKERDSTATE_STARTINGDAEMONS; init_children_processes(NULL); } crm_notice("Pacemaker daemon successfully started and accepting connections"); g_main_loop_run(mainloop); if (ipcs) { crm_trace("Closing IPC server"); mainloop_del_ipc_server(ipcs); ipcs = NULL; } g_main_loop_unref(mainloop); #ifdef SUPPORT_COROSYNC cluster_disconnect_cfg(); #endif done: g_strfreev(processed_args); pcmk__free_arg_context(context); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } diff --git a/include/crm/common/ipc_internal.h b/include/crm/common/ipc_internal.h index d20392495b..e71fe415ad 100644 --- a/include/crm/common/ipc_internal.h +++ b/include/crm/common/ipc_internal.h @@ -1,294 +1,298 @@ /* * Copyright 2013-2023 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__IPC_INTERNAL_H #define PCMK__IPC_INTERNAL_H #ifdef __cplusplus extern "C" { #endif #include // bool #include // uint32_t, uint64_t, UINT64_C() #include // struct iovec #include // uid_t, gid_t, pid_t, size_t #ifdef HAVE_GNUTLS_GNUTLS_H # include // gnutls_session_t #endif #include // guint, gpointer, GQueue, ... #include // xmlNode #include // qb_ipcs_connection_t, ... #include // HAVE_GETPEEREID #include #include // pcmk_controld_api_reply #include // pcmk_pacemakerd_{api_reply,state} #include // mainloop_io_t /* * XML attribute names used only by internal code */ #define PCMK__XA_IPC_PROTO_VERSION "ipc-protocol-version" /* denotes "non yieldable PID" on FreeBSD, or actual PID1 in scenarios that require a delicate handling anyway (socket-based activation with systemd); we can be reasonably sure that this PID is never possessed by the actual child daemon, as it gets taken either by the proper init, or by pacemakerd itself (i.e. this precludes anything else); note that value of zero is meant to carry "unset" meaning, and better not to bet on/conditionalize over signedness of pid_t */ #define PCMK__SPECIAL_PID 1 // Timeout (in seconds) to use for IPC client sends, reply waits, etc. #define PCMK__IPC_TIMEOUT 120 #if defined(HAVE_GETPEEREID) /* on FreeBSD, we don't want to expose "non-yieldable PID" (leading to "IPC liveness check only") as its nominal representation, which could cause confusion -- this is unambiguous as long as there's no socket-based activation like with systemd (very improbable) */ #define PCMK__SPECIAL_PID_AS_0(p) (((p) == PCMK__SPECIAL_PID) ? 0 : (p)) #else #define PCMK__SPECIAL_PID_AS_0(p) (p) #endif /*! * \internal * \brief Check the authenticity and liveness of the process via IPC end-point * * When IPC daemon under given IPC end-point (name) detected, its authenticity * is verified by the means of comparing against provided referential UID and * GID, and the result of this check can be deduced from the return value. * As an exception, referential UID of 0 (~ root) satisfies arbitrary * detected daemon's credentials. * * \param[in] name IPC name to base the search on * \param[in] refuid referential UID to check against * \param[in] refgid referential GID to check against * \param[out] gotpid to optionally store obtained PID of the found process * upon returning 1 or -2 * (not available on FreeBSD, special value of 1, * see PCMK__SPECIAL_PID, used instead, and the caller * is required to special case this value respectively) * * \return Standard Pacemaker return code * * \note Return codes of particular interest include pcmk_rc_ipc_unresponsive * indicating that no trace of IPC liveness was detected, and * pcmk_rc_ipc_unauthorized indicating that the IPC endpoint is blocked by * an unauthorized process. * \note This function emits a log message for return codes other than * pcmk_rc_ok and pcmk_rc_ipc_unresponsive, and when there isn't a perfect * match in respect to \p reguid and/or \p refgid, for a possible * least privilege principle violation. * * \see crm_ipc_is_authentic_process */ int pcmk__ipc_is_authentic_process_active(const char *name, uid_t refuid, gid_t refgid, pid_t *gotpid); +int pcmk__connect_generic_ipc(crm_ipc_t *ipc); +int pcmk__ipc_fd(crm_ipc_t *ipc, int *fd); +int pcmk__connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type, + int attempts); /* * Server-related */ typedef struct pcmk__client_s pcmk__client_t; struct pcmk__remote_s { /* Shared */ char *buffer; size_t buffer_size; size_t buffer_offset; int auth_timeout; int tcp_socket; mainloop_io_t *source; time_t uptime; char *start_state; /* CIB-only */ char *token; /* TLS only */ # ifdef HAVE_GNUTLS_GNUTLS_H gnutls_session_t *tls_session; # endif }; enum pcmk__client_flags { // Lower 32 bits are reserved for server (not library) use // Next 8 bits are reserved for client type (sort of a cheap enum) //! Client uses plain IPC pcmk__client_ipc = (UINT64_C(1) << 32), //! Client uses TCP connection pcmk__client_tcp = (UINT64_C(1) << 33), # ifdef HAVE_GNUTLS_GNUTLS_H //! Client uses TCP with TLS pcmk__client_tls = (UINT64_C(1) << 34), # endif // The rest are client attributes //! Client IPC is proxied pcmk__client_proxied = (UINT64_C(1) << 40), //! Client is run by root or cluster user pcmk__client_privileged = (UINT64_C(1) << 41), //! Local client to be proxied pcmk__client_to_proxy = (UINT64_C(1) << 42), /*! * \brief Client IPC connection accepted * * Used only for remote CIB connections via \c remote-tls-port. */ pcmk__client_authenticated = (UINT64_C(1) << 43), # ifdef HAVE_GNUTLS_GNUTLS_H //! Client TLS handshake is complete pcmk__client_tls_handshake_complete = (UINT64_C(1) << 44), # endif }; #define PCMK__CLIENT_TYPE(client) ((client)->flags & UINT64_C(0xff00000000)) struct pcmk__client_s { unsigned int pid; char *id; char *name; char *user; uint64_t flags; // Group of pcmk__client_flags int request_id; void *userdata; int event_timer; GQueue *event_queue; /* Depending on the client type, only some of the following will be * populated/valid. @TODO Maybe convert to a union. */ qb_ipcs_connection_t *ipcs; /* IPC */ struct pcmk__remote_s *remote; /* TCP/TLS */ unsigned int queue_backlog; /* IPC queue length after last flush */ unsigned int queue_max; /* Evict client whose queue grows this big */ }; #define pcmk__set_client_flags(client, flags_to_set) do { \ (client)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Client", pcmk__client_name(client), \ (client)->flags, (flags_to_set), #flags_to_set); \ } while (0) #define pcmk__clear_client_flags(client, flags_to_clear) do { \ (client)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Client", pcmk__client_name(client), \ (client)->flags, (flags_to_clear), #flags_to_clear); \ } while (0) #define pcmk__set_ipc_flags(ipc_flags, ipc_name, flags_to_set) do { \ ipc_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "IPC", (ipc_name), \ (ipc_flags), (flags_to_set), \ #flags_to_set); \ } while (0) #define pcmk__clear_ipc_flags(ipc_flags, ipc_name, flags_to_clear) do { \ ipc_flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "IPC", (ipc_name), \ (ipc_flags), (flags_to_clear), \ #flags_to_clear); \ } while (0) guint pcmk__ipc_client_count(void); void pcmk__foreach_ipc_client(GHFunc func, gpointer user_data); void pcmk__client_cleanup(void); pcmk__client_t *pcmk__find_client(const qb_ipcs_connection_t *c); pcmk__client_t *pcmk__find_client_by_id(const char *id); const char *pcmk__client_name(const pcmk__client_t *c); const char *pcmk__client_type_str(uint64_t client_type); pcmk__client_t *pcmk__new_unauth_client(void *key); pcmk__client_t *pcmk__new_client(qb_ipcs_connection_t *c, uid_t uid, gid_t gid); void pcmk__free_client(pcmk__client_t *c); void pcmk__drop_all_clients(qb_ipcs_service_t *s); bool pcmk__set_client_queue_max(pcmk__client_t *client, const char *qmax); xmlNode *pcmk__ipc_create_ack_as(const char *function, int line, uint32_t flags, const char *tag, const char *ver, crm_exit_t status); #define pcmk__ipc_create_ack(flags, tag, ver, st) \ pcmk__ipc_create_ack_as(__func__, __LINE__, (flags), (tag), (ver), (st)) int pcmk__ipc_send_ack_as(const char *function, int line, pcmk__client_t *c, uint32_t request, uint32_t flags, const char *tag, const char *ver, crm_exit_t status); #define pcmk__ipc_send_ack(c, req, flags, tag, ver, st) \ pcmk__ipc_send_ack_as(__func__, __LINE__, (c), (req), (flags), (tag), (ver), (st)) int pcmk__ipc_prepare_iov(uint32_t request, xmlNode *message, uint32_t max_send_size, struct iovec **result, ssize_t *bytes); int pcmk__ipc_send_xml(pcmk__client_t *c, uint32_t request, xmlNode *message, uint32_t flags); int pcmk__ipc_send_iov(pcmk__client_t *c, struct iovec *iov, uint32_t flags); xmlNode *pcmk__client_data2xml(pcmk__client_t *c, void *data, uint32_t *id, uint32_t *flags); int pcmk__client_pid(qb_ipcs_connection_t *c); void pcmk__serve_attrd_ipc(qb_ipcs_service_t **ipcs, struct qb_ipcs_service_handlers *cb); void pcmk__serve_fenced_ipc(qb_ipcs_service_t **ipcs, struct qb_ipcs_service_handlers *cb); void pcmk__serve_pacemakerd_ipc(qb_ipcs_service_t **ipcs, struct qb_ipcs_service_handlers *cb); qb_ipcs_service_t *pcmk__serve_schedulerd_ipc(struct qb_ipcs_service_handlers *cb); qb_ipcs_service_t *pcmk__serve_controld_ipc(struct qb_ipcs_service_handlers *cb); void pcmk__serve_based_ipc(qb_ipcs_service_t **ipcs_ro, qb_ipcs_service_t **ipcs_rw, qb_ipcs_service_t **ipcs_shm, struct qb_ipcs_service_handlers *ro_cb, struct qb_ipcs_service_handlers *rw_cb); void pcmk__stop_based_ipc(qb_ipcs_service_t *ipcs_ro, qb_ipcs_service_t *ipcs_rw, qb_ipcs_service_t *ipcs_shm); static inline const char * pcmk__ipc_sys_name(const char *ipc_name, const char *fallback) { return ipc_name ? ipc_name : ((crm_system_name ? crm_system_name : fallback)); } const char *pcmk__pcmkd_state_enum2friendly(enum pcmk_pacemakerd_state state); const char *pcmk__controld_api_reply2str(enum pcmk_controld_api_reply reply); const char *pcmk__pcmkd_api_reply2str(enum pcmk_pacemakerd_api_reply reply); #ifdef __cplusplus } #endif #endif diff --git a/lib/cib/cib_native.c b/lib/cib/cib_native.c index 4a87f56ab6..2368b8f8e7 100644 --- a/lib/cib/cib_native.c +++ b/lib/cib/cib_native.c @@ -1,502 +1,506 @@ /* * Copyright 2004 International Business Machines * Later changes copyright 2004-2023 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #include #include typedef struct cib_native_opaque_s { char *token; crm_ipc_t *ipc; void (*dnotify_fn) (gpointer user_data); mainloop_io_t *source; } cib_native_opaque_t; static int cib_native_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; int reply_id = 0; enum crm_ipc_flags ipc_flags = crm_ipc_flags_none; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; cib_native_opaque_t *native = cib->variant_opaque; if (cib->state == cib_disconnected) { return -ENOTCONN; } if (output_data != NULL) { *output_data = NULL; } if (op == NULL) { crm_err("No operation specified"); return -EINVAL; } if (call_options & cib_sync_call) { pcmk__set_ipc_flags(ipc_flags, "client", crm_ipc_client_response); } cib->call_id++; if (cib->call_id < 1) { cib->call_id = 1; } op_msg = cib_create_op(cib->call_id, op, host, section, data, call_options, user_name); if (op_msg == NULL) { return -EPROTO; } crm_trace("Sending %s message to the CIB manager (timeout=%ds)", op, cib->call_timeout); rc = crm_ipc_send(native->ipc, op_msg, ipc_flags, cib->call_timeout * 1000, &op_reply); free_xml(op_msg); if (rc < 0) { crm_err("Couldn't perform %s operation (timeout=%ds): %s (%d)", op, cib->call_timeout, pcmk_strerror(rc), rc); rc = -ECOMM; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (!(call_options & cib_sync_call)) { crm_trace("Async call, returning %d", cib->call_id); CRM_CHECK(cib->call_id != 0, return -ENOMSG); free_xml(op_reply); return cib->call_id; } rc = pcmk_ok; crm_element_value_int(op_reply, F_CIB_CALLID, &reply_id); if (reply_id == cib->call_id) { xmlNode *tmp = get_message_xml(op_reply, F_CIB_CALLDATA); crm_trace("Synchronous reply %d received", reply_id); if (crm_element_value_int(op_reply, F_CIB_RC, &rc) != 0) { rc = -EPROTO; } if (output_data == NULL || (call_options & cib_discard_reply)) { crm_trace("Discarding reply"); } else if (tmp != NULL) { *output_data = copy_xml(tmp); } } else if (reply_id <= 0) { crm_err("Received bad reply: No id set"); crm_log_xml_err(op_reply, "Bad reply"); rc = -ENOMSG; goto done; } else { crm_err("Received bad reply: %d (wanted %d)", reply_id, cib->call_id); crm_log_xml_err(op_reply, "Old reply"); rc = -ENOMSG; goto done; } if (op_reply == NULL && cib->state == cib_disconnected) { rc = -ENOTCONN; } else if (rc == pcmk_ok && op_reply == NULL) { rc = -ETIME; } switch (rc) { case pcmk_ok: case -EPERM: break; /* This is an internal value that clients do not and should not care about */ case -pcmk_err_diff_resync: rc = pcmk_ok; break; /* These indicate internal problems */ case -EPROTO: case -ENOMSG: crm_err("Call failed: %s", pcmk_strerror(rc)); if (op_reply) { crm_log_xml_err(op_reply, "Invalid reply"); } break; default: if (!pcmk__str_eq(op, PCMK__CIB_REQUEST_QUERY, pcmk__str_none)) { crm_warn("Call failed: %s", pcmk_strerror(rc)); } } done: if (!crm_ipc_connected(native->ipc)) { crm_err("The CIB manager disconnected"); cib->state = cib_disconnected; } free_xml(op_reply); return rc; } static int cib_native_dispatch_internal(const char *buffer, ssize_t length, gpointer userdata) { const char *type = NULL; xmlNode *msg = NULL; cib_t *cib = userdata; crm_trace("dispatching %p", userdata); if (cib == NULL) { crm_err("No CIB!"); return 0; } msg = string2xml(buffer); if (msg == NULL) { crm_warn("Received a NULL message from the CIB manager"); return 0; } /* do callbacks */ type = crm_element_value(msg, F_TYPE); crm_trace("Activating %s callbacks...", type); crm_log_xml_explicit(msg, "cib-reply"); if (pcmk__str_eq(type, T_CIB, pcmk__str_casei)) { cib_native_callback(cib, msg, 0, 0); } else if (pcmk__str_eq(type, T_CIB_NOTIFY, pcmk__str_casei)) { g_list_foreach(cib->notify_list, cib_native_notify, msg); } else { crm_err("Unknown message type: %s", type); } free_xml(msg); return 0; } static void cib_native_destroy(void *userdata) { cib_t *cib = userdata; cib_native_opaque_t *native = cib->variant_opaque; crm_trace("destroying %p", userdata); cib->state = cib_disconnected; native->source = NULL; native->ipc = NULL; if (native->dnotify_fn) { native->dnotify_fn(userdata); } } static int cib_native_signoff(cib_t *cib) { cib_native_opaque_t *native = cib->variant_opaque; crm_debug("Disconnecting from the CIB manager"); cib_free_notify(cib); remove_cib_op_callback(0, TRUE); if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } cib->state = cib_disconnected; cib->type = cib_no_connection; return pcmk_ok; } static int cib_native_signon_raw(cib_t *cib, const char *name, enum cib_conn_type type, int *async_fd) { int rc = pcmk_ok; const char *channel = NULL; cib_native_opaque_t *native = cib->variant_opaque; struct ipc_client_callbacks cib_callbacks = { .dispatch = cib_native_dispatch_internal, .destroy = cib_native_destroy }; cib->call_timeout = PCMK__IPC_TIMEOUT; if (type == cib_command) { cib->state = cib_connected_command; channel = PCMK__SERVER_BASED_RW; } else if (type == cib_command_nonblocking) { cib->state = cib_connected_command; channel = PCMK__SERVER_BASED_SHM; } else if (type == cib_query) { cib->state = cib_connected_query; channel = PCMK__SERVER_BASED_RO; } else { return -ENOTCONN; } crm_trace("Connecting %s channel", channel); if (async_fd != NULL) { native->ipc = crm_ipc_new(channel, 0); - - if (native->ipc && crm_ipc_connect(native->ipc)) { - *async_fd = crm_ipc_get_fd(native->ipc); - - } else if (native->ipc) { - rc = -ENOTCONN; + if (native->ipc != NULL) { + rc = pcmk__connect_generic_ipc(native->ipc); + if (rc == pcmk_rc_ok) { + rc = pcmk__ipc_fd(native->ipc, async_fd); + if (rc != pcmk_rc_ok) { + crm_info("Couldn't get file descriptor for %s IPC", + channel); + } + } + rc = pcmk_rc2legacy(rc); } } else { native->source = mainloop_add_ipc_client(channel, G_PRIORITY_HIGH, 512 * 1024 /* 512k */ , cib, &cib_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (rc != pcmk_ok || native->ipc == NULL || !crm_ipc_connected(native->ipc)) { crm_info("Could not connect to CIB manager for %s", name); rc = -ENOTCONN; } if (rc == pcmk_ok) { xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "cib_command"); crm_xml_add(hello, F_TYPE, T_CIB); crm_xml_add(hello, F_CIB_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_CIB_CLIENTNAME, name); crm_xml_add_int(hello, F_CIB_CALLOPTS, cib_sync_call); if (crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply) > 0) { const char *msg_type = crm_element_value(reply, F_CIB_OPERATION); rc = pcmk_ok; crm_log_xml_trace(reply, "reg-reply"); if (!pcmk__str_eq(msg_type, CRM_OP_REGISTER, pcmk__str_casei)) { crm_info("Reply to CIB registration message has " "unknown type '%s'", msg_type); rc = -EPROTO; } else { native->token = crm_element_value_copy(reply, F_CIB_CLIENTID); if (native->token == NULL) { rc = -EPROTO; } } free_xml(reply); } else { rc = -ECOMM; } free_xml(hello); } if (rc == pcmk_ok) { crm_info("Successfully connected to CIB manager for %s", name); return pcmk_ok; } crm_info("Connection to CIB manager for %s failed: %s", name, pcmk_strerror(rc)); cib_native_signoff(cib); return rc; } static int cib_native_signon(cib_t *cib, const char *name, enum cib_conn_type type) { return cib_native_signon_raw(cib, name, type, NULL); } static int cib_native_free(cib_t *cib) { int rc = pcmk_ok; if (cib->state != cib_disconnected) { rc = cib_native_signoff(cib); } if (cib->state == cib_disconnected) { cib_native_opaque_t *native = cib->variant_opaque; free(native->token); free(cib->variant_opaque); free(cib->cmds); free(cib); } return rc; } static int cib_native_register_notification(cib_t *cib, const char *callback, int enabled) { int rc = pcmk_ok; xmlNode *notify_msg = create_xml_node(NULL, "cib-callback"); cib_native_opaque_t *native = cib->variant_opaque; if (cib->state != cib_disconnected) { crm_xml_add(notify_msg, F_CIB_OPERATION, T_CIB_NOTIFY); crm_xml_add(notify_msg, F_CIB_NOTIFY_TYPE, callback); crm_xml_add_int(notify_msg, F_CIB_NOTIFY_ACTIVATE, enabled); rc = crm_ipc_send(native->ipc, notify_msg, crm_ipc_client_response, 1000 * cib->call_timeout, NULL); if (rc <= 0) { crm_trace("Notification not registered: %d", rc); rc = -ECOMM; } } free_xml(notify_msg); return rc; } static int cib_native_set_connection_dnotify(cib_t *cib, void (*dnotify) (gpointer user_data)) { cib_native_opaque_t *native = NULL; if (cib == NULL) { crm_err("No CIB!"); return FALSE; } native = cib->variant_opaque; native->dnotify_fn = dnotify; return pcmk_ok; } /*! * \internal * \brief Get the given CIB connection's unique client identifier * * These can be used to check whether this client requested the action that * triggered a CIB notification. * * \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 (specifically, \p pcmk_ok) * * \note This is the \p cib_native variant implementation of * \p cib_api_operations_t:client_id(). * \note For \p cib_native objects, \p async_id and \p sync_id are the same. * \note The client ID is assigned during CIB sign-on. */ static int cib_native_client_id(const cib_t *cib, const char **async_id, const char **sync_id) { cib_native_opaque_t *native = cib->variant_opaque; if (async_id != NULL) { *async_id = native->token; } if (sync_id != NULL) { *sync_id = native->token; } return pcmk_ok; } cib_t * cib_native_new(void) { cib_native_opaque_t *native = NULL; cib_t *cib = cib_new_variant(); if (cib == NULL) { return NULL; } native = calloc(1, sizeof(cib_native_opaque_t)); if (native == NULL) { free(cib); return NULL; } cib->variant = cib_native; cib->variant_opaque = native; native->ipc = NULL; native->source = NULL; native->dnotify_fn = NULL; /* assign variant specific ops */ cib->delegate_fn = cib_native_perform_op_delegate; cib->cmds->signon = cib_native_signon; cib->cmds->signon_raw = cib_native_signon_raw; cib->cmds->signoff = cib_native_signoff; cib->cmds->free = cib_native_free; cib->cmds->register_notification = cib_native_register_notification; cib->cmds->set_connection_dnotify = cib_native_set_connection_dnotify; cib->cmds->client_id = cib_native_client_id; return cib; } diff --git a/lib/common/crmcommon_private.h b/lib/common/crmcommon_private.h index dbf924f2bb..3020be377c 100644 --- a/lib/common/crmcommon_private.h +++ b/lib/common/crmcommon_private.h @@ -1,327 +1,286 @@ /* * Copyright 2018-2023 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 CRMCOMMON_PRIVATE__H # define CRMCOMMON_PRIVATE__H /* This header is for the sole use of libcrmcommon, so that functions can be * declared with G_GNUC_INTERNAL for efficiency. */ #include // uint8_t, uint32_t #include // bool #include // size_t #include // GList #include // xmlNode, xmlAttr #include // struct qb_ipc_response_header // Decent chunk size for processing large amounts of data #define PCMK__BUFFER_SIZE 4096 #if defined(PCMK__UNIT_TESTING) #undef G_GNUC_INTERNAL #define G_GNUC_INTERNAL #endif /* When deleting portions of an XML tree, we keep a record so we can know later * (e.g. when checking differences) that something was deleted. */ typedef struct pcmk__deleted_xml_s { char *path; int position; } pcmk__deleted_xml_t; typedef struct xml_node_private_s { long check; uint32_t flags; } xml_node_private_t; typedef struct xml_doc_private_s { long check; uint32_t flags; char *user; GList *acls; GList *deleted_objs; // List of pcmk__deleted_xml_t } xml_doc_private_t; #define pcmk__set_xml_flags(xml_priv, flags_to_set) do { \ (xml_priv)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_NEVER, "XML", "XML node", (xml_priv)->flags, \ (flags_to_set), #flags_to_set); \ } while (0) #define pcmk__clear_xml_flags(xml_priv, flags_to_clear) do { \ (xml_priv)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_NEVER, "XML", "XML node", (xml_priv)->flags, \ (flags_to_clear), #flags_to_clear); \ } while (0) G_GNUC_INTERNAL void pcmk__xml2text(xmlNodePtr data, uint32_t options, GString *buffer, int depth); G_GNUC_INTERNAL bool pcmk__tracking_xml_changes(xmlNode *xml, bool lazy); G_GNUC_INTERNAL void pcmk__mark_xml_created(xmlNode *xml); G_GNUC_INTERNAL int pcmk__xml_position(const xmlNode *xml, enum xml_private_flags ignore_if_set); G_GNUC_INTERNAL xmlNode *pcmk__xml_match(const xmlNode *haystack, const xmlNode *needle, bool exact); G_GNUC_INTERNAL void pcmk__xml_update(xmlNode *parent, xmlNode *target, xmlNode *update, bool as_diff); G_GNUC_INTERNAL xmlNode *pcmk__xc_match(const xmlNode *root, const xmlNode *search_comment, bool exact); G_GNUC_INTERNAL void pcmk__xc_update(xmlNode *parent, xmlNode *target, xmlNode *update); G_GNUC_INTERNAL void pcmk__free_acls(GList *acls); G_GNUC_INTERNAL void pcmk__unpack_acl(xmlNode *source, xmlNode *target, const char *user); G_GNUC_INTERNAL bool pcmk__is_user_in_group(const char *user, const char *group); G_GNUC_INTERNAL void pcmk__apply_acl(xmlNode *xml); G_GNUC_INTERNAL void pcmk__apply_creation_acl(xmlNode *xml, bool check_top); G_GNUC_INTERNAL void pcmk__mark_xml_attr_dirty(xmlAttr *a); G_GNUC_INTERNAL bool pcmk__xa_filterable(const char *name); G_GNUC_INTERNAL void pcmk__log_xmllib_err(void *ctx, const char *fmt, ...) G_GNUC_PRINTF(2, 3); G_GNUC_INTERNAL void pcmk__mark_xml_node_dirty(xmlNode *xml); G_GNUC_INTERNAL bool pcmk__marked_as_deleted(xmlAttrPtr a, void *user_data); G_GNUC_INTERNAL void pcmk__dump_xml_attr(const xmlAttr *attr, GString *buffer); /* * IPC */ #define PCMK__IPC_VERSION 1 #define PCMK__CONTROLD_API_MAJOR "1" #define PCMK__CONTROLD_API_MINOR "0" // IPC behavior that varies by daemon typedef struct pcmk__ipc_methods_s { /*! * \internal * \brief Allocate any private data needed by daemon IPC * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code */ int (*new_data)(pcmk_ipc_api_t *api); /*! * \internal * \brief Free any private data used by daemon IPC * * \param[in,out] api_data Data allocated by new_data() method */ void (*free_data)(void *api_data); /*! * \internal * \brief Perform daemon-specific handling after successful connection * * Some daemons require clients to register before sending any other * commands. The controller requires a CRM_OP_HELLO (with no reply), and * the CIB manager, executor, and fencer require a CRM_OP_REGISTER (with a * reply). Ideally this would be consistent across all daemons, but for now * this allows each to do its own authorization. * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code */ int (*post_connect)(pcmk_ipc_api_t *api); /*! * \internal * \brief Check whether an IPC request results in a reply * * \param[in,out] api IPC API connection * \param[in,out] request IPC request XML * * \return true if request would result in an IPC reply, false otherwise */ bool (*reply_expected)(pcmk_ipc_api_t *api, xmlNode *request); /*! * \internal * \brief Perform daemon-specific handling of an IPC message * * \param[in,out] api IPC API connection * \param[in,out] msg Message read from IPC connection * * \return true if more IPC reply messages should be expected */ bool (*dispatch)(pcmk_ipc_api_t *api, xmlNode *msg); /*! * \internal * \brief Perform daemon-specific handling of an IPC disconnect * * \param[in,out] api IPC API connection */ void (*post_disconnect)(pcmk_ipc_api_t *api); } pcmk__ipc_methods_t; // Implementation of pcmk_ipc_api_t struct pcmk_ipc_api_s { enum pcmk_ipc_server server; // Daemon this IPC API instance is for enum pcmk_ipc_dispatch dispatch_type; // How replies should be dispatched size_t ipc_size_max; // maximum IPC buffer size crm_ipc_t *ipc; // IPC connection mainloop_io_t *mainloop_io; // If using mainloop, I/O source for IPC bool free_on_disconnect; // Whether disconnect should free object pcmk_ipc_callback_t cb; // Caller-registered callback (if any) void *user_data; // Caller-registered data (if any) void *api_data; // For daemon-specific use pcmk__ipc_methods_t *cmds; // Behavior that varies by daemon }; typedef struct pcmk__ipc_header_s { struct qb_ipc_response_header qb; uint32_t size_uncompressed; uint32_t size_compressed; uint32_t flags; uint8_t version; } pcmk__ipc_header_t; G_GNUC_INTERNAL int pcmk__send_ipc_request(pcmk_ipc_api_t *api, xmlNode *request); G_GNUC_INTERNAL void pcmk__call_ipc_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data); G_GNUC_INTERNAL unsigned int pcmk__ipc_buffer_size(unsigned int max); G_GNUC_INTERNAL bool pcmk__valid_ipc_header(const pcmk__ipc_header_t *header); G_GNUC_INTERNAL pcmk__ipc_methods_t *pcmk__attrd_api_methods(void); G_GNUC_INTERNAL pcmk__ipc_methods_t *pcmk__controld_api_methods(void); G_GNUC_INTERNAL pcmk__ipc_methods_t *pcmk__pacemakerd_api_methods(void); G_GNUC_INTERNAL pcmk__ipc_methods_t *pcmk__schedulerd_api_methods(void); /* * Logging */ //! XML is newly created #define PCMK__XML_PREFIX_CREATED "++" //! XML has been deleted #define PCMK__XML_PREFIX_DELETED "--" //! XML has been modified #define PCMK__XML_PREFIX_MODIFIED "+ " //! XML has been moved #define PCMK__XML_PREFIX_MOVED "+~" -/*! - * \brief Check the authenticity of the IPC socket peer process - * - * If everything goes well, peer's authenticity is verified by the means - * of comparing against provided referential UID and GID (either satisfies), - * and the result of this check can be deduced from the return value. - * As an exception, detected UID of 0 ("root") satisfies arbitrary - * provided referential daemon's credentials. - * - * \param[in] qb_ipc libqb client connection if available - * \param[in] sock IPC related, connected Unix socket to check peer of - * \param[in] refuid referential UID to check against - * \param[in] refgid referential GID to check against - * \param[out] gotpid to optionally store obtained PID of the peer - * (not available on FreeBSD, special value of 1 - * used instead, and the caller is required to - * special case this value respectively) - * \param[out] gotuid to optionally store obtained UID of the peer - * \param[out] gotgid to optionally store obtained GID of the peer - * - * \return Standard Pacemaker return code - * ie: 0 if it the connection is authentic - * pcmk_rc_ipc_unauthorized if the connection is not authentic, - * standard errors. - * - * \note While this function is tolerant on what constitutes authorized - * IPC daemon process (its effective user matches UID=0 or \p refuid, - * or at least its group matches \p refgid), either or both (in case - * of UID=0) mismatches on the expected credentials of such peer - * process \e shall be investigated at the caller when value of 1 - * gets returned there, since higher-than-expected privileges in - * respect to the expected/intended credentials possibly violate - * the least privilege principle and may pose an additional risk - * (i.e. such accidental inconsistency shall be eventually fixed). - */ -int pcmk__crm_ipc_is_authentic_process(qb_ipcc_connection_t *qb_ipc, int sock, - uid_t refuid, gid_t refgid, - pid_t *gotpid, uid_t *gotuid, - gid_t *gotgid); - - /* * Output */ G_GNUC_INTERNAL int pcmk__bare_output_new(pcmk__output_t **out, const char *fmt_name, const char *filename, char **argv); G_GNUC_INTERNAL void pcmk__register_patchset_messages(pcmk__output_t *out); /* * Utils */ #define PCMK__PW_BUFFER_LEN 500 #endif // CRMCOMMON_PRIVATE__H diff --git a/lib/common/ipc_attrd.c b/lib/common/ipc_attrd.c index 7c40aa7d5c..016db58f64 100644 --- a/lib/common/ipc_attrd.c +++ b/lib/common/ipc_attrd.c @@ -1,590 +1,587 @@ /* * Copyright 2011-2023 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 _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include "crmcommon_private.h" static void set_pairs_data(pcmk__attrd_api_reply_t *data, xmlNode *msg_data) { const char *name = NULL; pcmk__attrd_query_pair_t *pair; name = crm_element_value(msg_data, PCMK__XA_ATTR_NAME); for (xmlNode *node = first_named_child(msg_data, XML_CIB_TAG_NODE); node != NULL; node = crm_next_same_xml(node)) { pair = calloc(1, sizeof(pcmk__attrd_query_pair_t)); CRM_ASSERT(pair != NULL); pair->node = crm_element_value(node, PCMK__XA_ATTR_NODE_NAME); pair->name = name; pair->value = crm_element_value(node, PCMK__XA_ATTR_VALUE); data->data.pairs = g_list_prepend(data->data.pairs, pair); } } static bool reply_expected(pcmk_ipc_api_t *api, xmlNode *request) { const char *command = crm_element_value(request, PCMK__XA_TASK); return pcmk__str_any_of(command, PCMK__ATTRD_CMD_CLEAR_FAILURE, PCMK__ATTRD_CMD_QUERY, PCMK__ATTRD_CMD_REFRESH, PCMK__ATTRD_CMD_UPDATE, PCMK__ATTRD_CMD_UPDATE_BOTH, PCMK__ATTRD_CMD_UPDATE_DELAY, NULL); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { const char *value = NULL; crm_exit_t status = CRM_EX_OK; pcmk__attrd_api_reply_t reply_data = { pcmk__attrd_reply_unknown }; if (pcmk__str_eq((const char *) reply->name, "ack", pcmk__str_none)) { return false; } /* Do some basic validation of the reply */ value = crm_element_value(reply, F_TYPE); if (pcmk__str_empty(value) || !pcmk__str_eq(value, T_ATTRD, pcmk__str_none)) { crm_info("Unrecognizable message from attribute manager: " "message type '%s' not '" T_ATTRD "'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } value = crm_element_value(reply, F_SUBTYPE); /* Only the query command gets a reply for now. NULL counts as query for * backward compatibility with attribute managers <2.1.3 that didn't set it. */ if (pcmk__str_eq(value, PCMK__ATTRD_CMD_QUERY, pcmk__str_null_matches)) { if (!xmlHasProp(reply, (pcmkXmlStr) PCMK__XA_ATTR_NAME)) { status = ENXIO; // Most likely, the attribute doesn't exist goto done; } reply_data.reply_type = pcmk__attrd_reply_query; set_pairs_data(&reply_data, reply); } else { crm_info("Unrecognizable message from attribute manager: " "message subtype '%s' unknown", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); /* Free any reply data that was allocated */ if (reply_data.data.pairs) { g_list_free_full(reply_data.data.pairs, free); } return false; } pcmk__ipc_methods_t * pcmk__attrd_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = NULL; cmds->free_data = NULL; cmds->post_connect = NULL; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; } return cmds; } /*! * \internal * \brief Create a generic pacemaker-attrd operation * * \param[in] user_name If not NULL, ACL user to set for operation * * \return XML of pacemaker-attrd operation */ static xmlNode * create_attrd_op(const char *user_name) { xmlNode *attrd_op = create_xml_node(NULL, __func__); crm_xml_add(attrd_op, F_TYPE, T_ATTRD); crm_xml_add(attrd_op, F_ORIG, (crm_system_name? crm_system_name: "unknown")); crm_xml_add(attrd_op, PCMK__XA_ATTR_USER, user_name); return attrd_op; } static int create_api(pcmk_ipc_api_t **api) { int rc = pcmk_new_ipc_api(api, pcmk_ipc_attrd); if (rc != pcmk_rc_ok) { crm_err("Could not connect to attrd: %s", pcmk_rc_str(rc)); } return rc; } static void destroy_api(pcmk_ipc_api_t *api) { pcmk_disconnect_ipc(api); pcmk_free_ipc_api(api); api = NULL; } static int connect_and_send_attrd_request(pcmk_ipc_api_t *api, xmlNode *request) { int rc = pcmk_rc_ok; - int max = 5; - - while (max > 0) { - crm_info("Connecting to cluster... %d retries remaining", max); - rc = pcmk_connect_ipc(api, pcmk_ipc_dispatch_sync); - - if (rc == pcmk_rc_ok) { - rc = pcmk__send_ipc_request(api, request); - break; - } else if (rc == EAGAIN || rc == EALREADY) { - sleep(5 - max); - max--; - } else { - crm_err("Could not connect to attrd: %s", pcmk_rc_str(rc)); - break; - } + + rc = pcmk__connect_ipc(api, pcmk_ipc_dispatch_sync, 5); + if (rc != pcmk_rc_ok) { + crm_err("Could not connect to %s: %s", + pcmk_ipc_name(api, true), pcmk_rc_str(rc)); + return rc; } - return rc; + rc = pcmk__send_ipc_request(api, request); + if (rc != pcmk_rc_ok) { + crm_err("Could not send request to %s: %s", + pcmk_ipc_name(api, true), pcmk_rc_str(rc)); + return rc; + } + + return pcmk_rc_ok; } static int send_attrd_request(pcmk_ipc_api_t *api, xmlNode *request) { return pcmk__send_ipc_request(api, request); } int pcmk__attrd_api_clear_failures(pcmk_ipc_api_t *api, const char *node, const char *resource, const char *operation, const char *interval_spec, const char *user_name, uint32_t options) { int rc = pcmk_rc_ok; xmlNode *request = create_attrd_op(user_name); const char *interval_desc = NULL; const char *op_desc = NULL; const char *target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_CLEAR_FAILURE); pcmk__xe_add_node(request, node, 0); crm_xml_add(request, PCMK__XA_ATTR_RESOURCE, resource); crm_xml_add(request, PCMK__XA_ATTR_OPERATION, operation); crm_xml_add(request, PCMK__XA_ATTR_INTERVAL, interval_spec); crm_xml_add_int(request, PCMK__XA_ATTR_IS_REMOTE, pcmk_is_set(options, pcmk__node_attr_remote)); if (api == NULL) { rc = create_api(&api); if (rc != pcmk_rc_ok) { return rc; } rc = connect_and_send_attrd_request(api, request); destroy_api(api); } else if (!pcmk_ipc_is_connected(api)) { rc = connect_and_send_attrd_request(api, request); } else { rc = send_attrd_request(api, request); } free_xml(request); if (operation) { interval_desc = interval_spec? interval_spec : "nonrecurring"; op_desc = operation; } else { interval_desc = "all"; op_desc = "operations"; } crm_debug("Asked pacemaker-attrd to clear failure of %s %s for %s on %s: %s (%d)", interval_desc, op_desc, (resource? resource : "all resources"), (node? node : "all nodes"), pcmk_rc_str(rc), rc); return rc; } int pcmk__attrd_api_delete(pcmk_ipc_api_t *api, const char *node, const char *name, uint32_t options) { const char *target = NULL; if (name == NULL) { return EINVAL; } target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } /* Make sure the right update option is set. */ options &= ~pcmk__node_attr_delay; options |= pcmk__node_attr_value; return pcmk__attrd_api_update(api, node, name, NULL, NULL, NULL, NULL, options); } int pcmk__attrd_api_purge(pcmk_ipc_api_t *api, const char *node) { int rc = pcmk_rc_ok; xmlNode *request = NULL; const char *display_host = (node ? node : "localhost"); const char *target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } request = create_attrd_op(NULL); crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_PEER_REMOVE); pcmk__xe_add_node(request, node, 0); if (api == NULL) { rc = create_api(&api); if (rc != pcmk_rc_ok) { return rc; } rc = connect_and_send_attrd_request(api, request); destroy_api(api); } else if (!pcmk_ipc_is_connected(api)) { rc = connect_and_send_attrd_request(api, request); } else { rc = send_attrd_request(api, request); } free_xml(request); crm_debug("Asked pacemaker-attrd to purge %s: %s (%d)", display_host, pcmk_rc_str(rc), rc); return rc; } int pcmk__attrd_api_query(pcmk_ipc_api_t *api, const char *node, const char *name, uint32_t options) { int rc = pcmk_rc_ok; xmlNode *request = NULL; const char *target = NULL; if (name == NULL) { return EINVAL; } if (pcmk_is_set(options, pcmk__node_attr_query_all)) { node = NULL; } else { target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } } request = create_attrd_op(NULL); crm_xml_add(request, PCMK__XA_ATTR_NAME, name); crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_QUERY); pcmk__xe_add_node(request, node, 0); rc = send_attrd_request(api, request); free_xml(request); if (node) { crm_debug("Queried pacemaker-attrd for %s on %s: %s (%d)", name, node, pcmk_rc_str(rc), rc); } else { crm_debug("Queried pacemaker-attrd for %s: %s (%d)", name, pcmk_rc_str(rc), rc); } return rc; } int pcmk__attrd_api_refresh(pcmk_ipc_api_t *api, const char *node) { int rc = pcmk_rc_ok; xmlNode *request = NULL; const char *display_host = (node ? node : "localhost"); const char *target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } request = create_attrd_op(NULL); crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_REFRESH); pcmk__xe_add_node(request, node, 0); if (api == NULL) { rc = create_api(&api); if (rc != pcmk_rc_ok) { return rc; } rc = connect_and_send_attrd_request(api, request); destroy_api(api); } else if (!pcmk_ipc_is_connected(api)) { rc = connect_and_send_attrd_request(api, request); } else { rc = send_attrd_request(api, request); } free_xml(request); crm_debug("Asked pacemaker-attrd to refresh %s: %s (%d)", display_host, pcmk_rc_str(rc), rc); return rc; } static void add_op_attr(xmlNode *op, uint32_t options) { if (pcmk_all_flags_set(options, pcmk__node_attr_value | pcmk__node_attr_delay)) { crm_xml_add(op, PCMK__XA_TASK, PCMK__ATTRD_CMD_UPDATE_BOTH); } else if (pcmk_is_set(options, pcmk__node_attr_value)) { crm_xml_add(op, PCMK__XA_TASK, PCMK__ATTRD_CMD_UPDATE); } else if (pcmk_is_set(options, pcmk__node_attr_delay)) { crm_xml_add(op, PCMK__XA_TASK, PCMK__ATTRD_CMD_UPDATE_DELAY); } } static void populate_update_op(xmlNode *op, const char *node, const char *name, const char *value, const char *dampen, const char *set, uint32_t options) { if (pcmk_is_set(options, pcmk__node_attr_pattern)) { crm_xml_add(op, PCMK__XA_ATTR_PATTERN, name); } else { crm_xml_add(op, PCMK__XA_ATTR_NAME, name); } if (pcmk_is_set(options, pcmk__node_attr_utilization)) { crm_xml_add(op, PCMK__XA_ATTR_SET_TYPE, XML_TAG_UTILIZATION); } else { crm_xml_add(op, PCMK__XA_ATTR_SET_TYPE, XML_TAG_ATTR_SETS); } add_op_attr(op, options); crm_xml_add(op, PCMK__XA_ATTR_VALUE, value); crm_xml_add(op, PCMK__XA_ATTR_DAMPENING, dampen); pcmk__xe_add_node(op, node, 0); crm_xml_add(op, PCMK__XA_ATTR_SET, set); crm_xml_add_int(op, PCMK__XA_ATTR_IS_REMOTE, pcmk_is_set(options, pcmk__node_attr_remote)); crm_xml_add_int(op, PCMK__XA_ATTR_IS_PRIVATE, pcmk_is_set(options, pcmk__node_attr_private)); if (pcmk_is_set(options, pcmk__node_attr_sync_local)) { crm_xml_add(op, PCMK__XA_ATTR_SYNC_POINT, PCMK__VALUE_LOCAL); } else if (pcmk_is_set(options, pcmk__node_attr_sync_cluster)) { crm_xml_add(op, PCMK__XA_ATTR_SYNC_POINT, PCMK__VALUE_CLUSTER); } } int pcmk__attrd_api_update(pcmk_ipc_api_t *api, const char *node, const char *name, const char *value, const char *dampen, const char *set, const char *user_name, uint32_t options) { int rc = pcmk_rc_ok; xmlNode *request = NULL; const char *display_host = (node ? node : "localhost"); const char *target = NULL; if (name == NULL) { return EINVAL; } target = pcmk__node_attr_target(node); if (target != NULL) { node = target; } request = create_attrd_op(user_name); populate_update_op(request, node, name, value, dampen, set, options); if (api == NULL) { rc = create_api(&api); if (rc != pcmk_rc_ok) { return rc; } rc = connect_and_send_attrd_request(api, request); destroy_api(api); } else if (!pcmk_ipc_is_connected(api)) { rc = connect_and_send_attrd_request(api, request); } else { rc = send_attrd_request(api, request); } free_xml(request); crm_debug("Asked pacemaker-attrd to update %s on %s: %s (%d)", name, display_host, pcmk_rc_str(rc), rc); return rc; } int pcmk__attrd_api_update_list(pcmk_ipc_api_t *api, GList *attrs, const char *dampen, const char *set, const char *user_name, uint32_t options) { int rc = pcmk_rc_ok; xmlNode *request = NULL; if (attrs == NULL) { return EINVAL; } /* There are two different ways of handling a list of attributes: * * (1) For messages originating from some command line tool, we have to send * them one at a time. In this loop, we just call pcmk__attrd_api_update * for each, letting it deal with creating the API object if it doesn't * already exist. * * The reason we can't use a single message in this case is that we can't * trust that the server supports it. Remote nodes could be involved * here, and there's no guarantee that a newer client running on a remote * node is talking to (or proxied through) a cluster node with a newer * attrd. We also can't just try sending a single message and then falling * back on multiple. There's no handshake with the attrd server to * determine its version. And then we would need to do that fallback in the * dispatch function for this to work for all connection types (mainloop in * particular), and at that point we won't know what the original message * was in order to break it apart and resend as individual messages. * * (2) For messages between daemons, we can be assured that the local attrd * will support the new message and that it can send to the other attrds * as one request or split up according to the minimum supported version. */ for (GList *iter = attrs; iter != NULL; iter = iter->next) { pcmk__attrd_query_pair_t *pair = (pcmk__attrd_query_pair_t *) iter->data; if (pcmk__is_daemon) { const char *target = NULL; xmlNode *child = NULL; /* First time through this loop - create the basic request. */ if (request == NULL) { request = create_attrd_op(user_name); add_op_attr(request, options); } /* Add a child node for this operation. We add the task to the top * level XML node so attrd_ipc_dispatch doesn't need changes. And * then we also add the task to each child node in populate_update_op * so attrd_client_update knows what form of update is taking place. */ child = create_xml_node(request, XML_ATTR_OP); target = pcmk__node_attr_target(pair->node); if (target != NULL) { pair->node = target; } populate_update_op(child, pair->node, pair->name, pair->value, dampen, set, options); } else { rc = pcmk__attrd_api_update(api, pair->node, pair->name, pair->value, dampen, set, user_name, options); } } /* If we were doing multiple attributes at once, we still need to send the * request. Do that now, creating and destroying the API object if needed. */ if (pcmk__is_daemon) { bool created_api = false; if (api == NULL) { rc = create_api(&api); if (rc != pcmk_rc_ok) { return rc; } created_api = true; } rc = connect_and_send_attrd_request(api, request); free_xml(request); if (created_api) { destroy_api(api); } } return rc; } diff --git a/lib/common/ipc_client.c b/lib/common/ipc_client.c index c6d1645048..5156a203f6 100644 --- a/lib/common/ipc_client.c +++ b/lib/common/ipc_client.c @@ -1,1576 +1,1688 @@ /* - * Copyright 2004-2022 the Pacemaker project contributors + * Copyright 2004-2023 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 #if defined(HAVE_UCRED) || defined(HAVE_SOCKPEERCRED) # ifdef HAVE_UCRED # ifndef _GNU_SOURCE # define _GNU_SOURCE # endif # endif # include #elif defined(HAVE_GETPEERUCRED) # include #endif #include #include #include #include #include /* indirectly: pcmk_err_generic */ #include #include #include #include "crmcommon_private.h" +static int is_ipc_provider_expected(qb_ipcc_connection_t *qb_ipc, int sock, + uid_t refuid, gid_t refgid, pid_t *gotpid, + uid_t *gotuid, gid_t *gotgid); + /*! * \brief Create a new object for using Pacemaker daemon IPC * * \param[out] api Where to store new IPC object * \param[in] server Which Pacemaker daemon the object is for * * \return Standard Pacemaker result code * * \note The caller is responsible for freeing *api using pcmk_free_ipc_api(). * \note This is intended to supersede crm_ipc_new() but currently only * supports the controller, pacemakerd, and schedulerd IPC API. */ int pcmk_new_ipc_api(pcmk_ipc_api_t **api, enum pcmk_ipc_server server) { if (api == NULL) { return EINVAL; } *api = calloc(1, sizeof(pcmk_ipc_api_t)); if (*api == NULL) { return errno; } (*api)->server = server; if (pcmk_ipc_name(*api, false) == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return EOPNOTSUPP; } (*api)->ipc_size_max = 0; // Set server methods and max_size (if not default) switch (server) { case pcmk_ipc_attrd: (*api)->cmds = pcmk__attrd_api_methods(); break; case pcmk_ipc_based: (*api)->ipc_size_max = 512 * 1024; // 512KB break; case pcmk_ipc_controld: (*api)->cmds = pcmk__controld_api_methods(); break; case pcmk_ipc_execd: break; case pcmk_ipc_fenced: break; case pcmk_ipc_pacemakerd: (*api)->cmds = pcmk__pacemakerd_api_methods(); break; case pcmk_ipc_schedulerd: (*api)->cmds = pcmk__schedulerd_api_methods(); // @TODO max_size could vary by client, maybe take as argument? (*api)->ipc_size_max = 5 * 1024 * 1024; // 5MB break; } if ((*api)->cmds == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } (*api)->ipc = crm_ipc_new(pcmk_ipc_name(*api, false), (*api)->ipc_size_max); if ((*api)->ipc == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } // If daemon API has its own data to track, allocate it if ((*api)->cmds->new_data != NULL) { if ((*api)->cmds->new_data(*api) != pcmk_rc_ok) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } } crm_trace("Created %s API IPC object", pcmk_ipc_name(*api, true)); return pcmk_rc_ok; } static void free_daemon_specific_data(pcmk_ipc_api_t *api) { if ((api != NULL) && (api->cmds != NULL)) { if ((api->cmds->free_data != NULL) && (api->api_data != NULL)) { api->cmds->free_data(api->api_data); api->api_data = NULL; } free(api->cmds); api->cmds = NULL; } } /*! * \internal * \brief Call an IPC API event callback, if one is registed * * \param[in,out] api IPC API connection * \param[in] event_type The type of event that occurred * \param[in] status Event status * \param[in,out] event_data Event-specific data */ void pcmk__call_ipc_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data) { if ((api != NULL) && (api->cb != NULL)) { api->cb(api, event_type, status, event_data, api->user_data); } } /*! * \internal * \brief Clean up after an IPC disconnect * * \param[in,out] user_data IPC API connection that disconnected * * \note This function can be used as a main loop IPC destroy callback. */ static void ipc_post_disconnect(gpointer user_data) { pcmk_ipc_api_t *api = user_data; crm_info("Disconnected from %s IPC API", pcmk_ipc_name(api, true)); // Perform any daemon-specific handling needed if ((api->cmds != NULL) && (api->cmds->post_disconnect != NULL)) { api->cmds->post_disconnect(api); } // Call client's registered event callback pcmk__call_ipc_callback(api, pcmk_ipc_event_disconnect, CRM_EX_DISCONNECT, NULL); /* If this is being called from a running main loop, mainloop_gio_destroy() * will free ipc and mainloop_io immediately after calling this function. * If this is called from a stopped main loop, these will leak, so the best * practice is to close the connection before stopping the main loop. */ api->ipc = NULL; api->mainloop_io = NULL; if (api->free_on_disconnect) { /* pcmk_free_ipc_api() has already been called, but did not free api * or api->cmds because this function needed them. Do that now. */ free_daemon_specific_data(api); crm_trace("Freeing IPC API object after disconnect"); free(api); } } /*! * \brief Free the contents of an IPC API object * * \param[in,out] api IPC API object to free */ void pcmk_free_ipc_api(pcmk_ipc_api_t *api) { bool free_on_disconnect = false; if (api == NULL) { return; } crm_debug("Releasing %s IPC API", pcmk_ipc_name(api, true)); if (api->ipc != NULL) { if (api->mainloop_io != NULL) { /* We need to keep the api pointer itself around, because it is the * user data for the IPC client destroy callback. That will be * triggered by the pcmk_disconnect_ipc() call below, but it might * happen later in the main loop (if still running). * * This flag tells the destroy callback to free the object. It can't * do that unconditionally, because the application might call this * function after a disconnect that happened by other means. */ free_on_disconnect = api->free_on_disconnect = true; } pcmk_disconnect_ipc(api); // Frees api if free_on_disconnect is true } if (!free_on_disconnect) { free_daemon_specific_data(api); crm_trace("Freeing IPC API object"); free(api); } } /*! * \brief Get the IPC name used with an IPC API connection * * \param[in] api IPC API connection * \param[in] for_log If true, return human-friendly name instead of IPC name * * \return IPC API's human-friendly or connection name, or if none is available, * "Pacemaker" if for_log is true and NULL if for_log is false */ const char * pcmk_ipc_name(const pcmk_ipc_api_t *api, bool for_log) { if (api == NULL) { return for_log? "Pacemaker" : NULL; } switch (api->server) { case pcmk_ipc_attrd: return for_log? "attribute manager" : T_ATTRD; case pcmk_ipc_based: return for_log? "CIB manager" : NULL /* PCMK__SERVER_BASED_RW */; case pcmk_ipc_controld: return for_log? "controller" : CRM_SYSTEM_CRMD; case pcmk_ipc_execd: return for_log? "executor" : NULL /* CRM_SYSTEM_LRMD */; case pcmk_ipc_fenced: return for_log? "fencer" : NULL /* "stonith-ng" */; case pcmk_ipc_pacemakerd: return for_log? "launcher" : CRM_SYSTEM_MCP; case pcmk_ipc_schedulerd: return for_log? "scheduler" : CRM_SYSTEM_PENGINE; default: return for_log? "Pacemaker" : NULL; } } /*! * \brief Check whether an IPC API connection is active * * \param[in,out] api IPC API connection * * \return true if IPC is connected, false otherwise */ bool pcmk_ipc_is_connected(pcmk_ipc_api_t *api) { return (api != NULL) && crm_ipc_connected(api->ipc); } /*! * \internal * \brief Call the daemon-specific API's dispatch function * * Perform daemon-specific handling of IPC reply dispatch. It is the daemon * method's responsibility to call the client's registered event callback, as * well as allocate and free any event data. * * \param[in,out] api IPC API connection * \param[in,out] message IPC reply XML to dispatch */ static bool call_api_dispatch(pcmk_ipc_api_t *api, xmlNode *message) { crm_log_xml_trace(message, "ipc-received"); if ((api->cmds != NULL) && (api->cmds->dispatch != NULL)) { return api->cmds->dispatch(api, message); } return false; } /*! * \internal * \brief Dispatch previously read IPC data * * \param[in] buffer Data read from IPC * \param[in,out] api IPC object * * \return Standard Pacemaker return code. In particular: * * pcmk_rc_ok: There are no more messages expected from the server. Quit * reading. * EINPROGRESS: There are more messages expected from the server. Keep reading. * * All other values indicate an error. */ static int dispatch_ipc_data(const char *buffer, pcmk_ipc_api_t *api) { bool more = false; xmlNode *msg; if (buffer == NULL) { crm_warn("Empty message received from %s IPC", pcmk_ipc_name(api, true)); return ENOMSG; } msg = string2xml(buffer); if (msg == NULL) { crm_warn("Malformed message received from %s IPC", pcmk_ipc_name(api, true)); return EPROTO; } more = call_api_dispatch(api, msg); free_xml(msg); if (more) { return EINPROGRESS; } else { return pcmk_rc_ok; } } /*! * \internal * \brief Dispatch data read from IPC source * * \param[in] buffer Data read from IPC * \param[in] length Number of bytes of data in buffer (ignored) * \param[in,out] user_data IPC object * * \return Always 0 (meaning connection is still required) * * \note This function can be used as a main loop IPC dispatch callback. */ static int dispatch_ipc_source_data(const char *buffer, ssize_t length, gpointer user_data) { pcmk_ipc_api_t *api = user_data; CRM_CHECK(api != NULL, return 0); dispatch_ipc_data(buffer, api); return 0; } /*! * \brief Check whether an IPC connection has data available (without main loop) * * \param[in] api IPC API connection * \param[in] timeout_ms If less than 0, poll indefinitely; if 0, poll once * and return immediately; otherwise, poll for up to * this many milliseconds * * \return Standard Pacemaker return code * * \note Callers of pcmk_connect_ipc() using pcmk_ipc_dispatch_poll should call * this function to check whether IPC data is available. Return values of * interest include pcmk_rc_ok meaning data is available, and EAGAIN * meaning no data is available; all other values indicate errors. * \todo This does not allow the caller to poll multiple file descriptors at * once. If there is demand for that, we could add a wrapper for - * crm_ipc_get_fd(api->ipc), so the caller can call poll() themselves. + * pcmk__ipc_fd(api->ipc), so the caller can call poll() themselves. */ int pcmk_poll_ipc(const pcmk_ipc_api_t *api, int timeout_ms) { int rc; struct pollfd pollfd = { 0, }; if ((api == NULL) || (api->dispatch_type != pcmk_ipc_dispatch_poll)) { return EINVAL; } - pollfd.fd = crm_ipc_get_fd(api->ipc); + + rc = pcmk__ipc_fd(api->ipc, &(pollfd.fd)); + if (rc != pcmk_rc_ok) { + crm_debug("Could not obtain file descriptor for %s IPC: %s", + pcmk_ipc_name(api, true), pcmk_rc_str(rc)); + return rc; + } + pollfd.events = POLLIN; rc = poll(&pollfd, 1, timeout_ms); if (rc < 0) { /* Some UNIX systems return negative and set EAGAIN for failure to * allocate memory; standardize the return code in that case */ return (errno == EAGAIN)? ENOMEM : errno; } else if (rc == 0) { return EAGAIN; } return pcmk_rc_ok; } /*! * \brief Dispatch available messages on an IPC connection (without main loop) * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code * * \note Callers of pcmk_connect_ipc() using pcmk_ipc_dispatch_poll should call * this function when IPC data is available. */ void pcmk_dispatch_ipc(pcmk_ipc_api_t *api) { if (api == NULL) { return; } while (crm_ipc_ready(api->ipc) > 0) { if (crm_ipc_read(api->ipc) > 0) { dispatch_ipc_data(crm_ipc_buffer(api->ipc), api); } } } // \return Standard Pacemaker return code static int connect_with_main_loop(pcmk_ipc_api_t *api) { int rc; struct ipc_client_callbacks callbacks = { .dispatch = dispatch_ipc_source_data, .destroy = ipc_post_disconnect, }; rc = pcmk__add_mainloop_ipc(api->ipc, G_PRIORITY_DEFAULT, api, &callbacks, &(api->mainloop_io)); if (rc != pcmk_rc_ok) { return rc; } crm_debug("Connected to %s IPC (attached to main loop)", pcmk_ipc_name(api, true)); /* After this point, api->mainloop_io owns api->ipc, so api->ipc * should not be explicitly freed. */ return pcmk_rc_ok; } // \return Standard Pacemaker return code static int connect_without_main_loop(pcmk_ipc_api_t *api) { - int rc; + int rc = pcmk__connect_generic_ipc(api->ipc); - if (!crm_ipc_connect(api->ipc)) { - rc = errno; + if (rc != pcmk_rc_ok) { crm_ipc_close(api->ipc); - return rc; + } else { + crm_debug("Connected to %s IPC (without main loop)", + pcmk_ipc_name(api, true)); } - crm_debug("Connected to %s IPC (without main loop)", - pcmk_ipc_name(api, true)); - return pcmk_rc_ok; + return rc; } /*! - * \brief Connect to a Pacemaker daemon via IPC + * \internal + * \brief Connect to a Pacemaker daemon via IPC (retrying after soft errors) * * \param[in,out] api IPC API instance * \param[in] dispatch_type How IPC replies should be dispatched + * \param[in] attempts How many times to try (in case of soft error) * * \return Standard Pacemaker return code */ int -pcmk_connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type) +pcmk__connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type, + int attempts) { - const int n_attempts = 2; int rc = pcmk_rc_ok; - if (api == NULL) { - crm_err("Cannot connect to uninitialized API object"); + if ((api == NULL) || (attempts < 1)) { return EINVAL; } if (api->ipc == NULL) { - api->ipc = crm_ipc_new(pcmk_ipc_name(api, false), - api->ipc_size_max); + api->ipc = crm_ipc_new(pcmk_ipc_name(api, false), api->ipc_size_max); if (api->ipc == NULL) { - crm_err("Failed to re-create IPC API"); return ENOMEM; } } if (crm_ipc_connected(api->ipc)) { - crm_trace("Already connected to %s IPC API", pcmk_ipc_name(api, true)); + crm_trace("Already connected to %s", pcmk_ipc_name(api, true)); return pcmk_rc_ok; } api->dispatch_type = dispatch_type; - for (int i = 0; i < n_attempts; i++) { + crm_debug("Attempting connection to %s (up to %d time%s)", + pcmk_ipc_name(api, true), attempts, pcmk__plural_s(attempts)); + for (int remaining = attempts - 1; remaining >= 0; --remaining) { switch (dispatch_type) { case pcmk_ipc_dispatch_main: rc = connect_with_main_loop(api); break; case pcmk_ipc_dispatch_sync: case pcmk_ipc_dispatch_poll: rc = connect_without_main_loop(api); break; } - if (rc != EAGAIN) { - break; + if ((remaining == 0) || ((rc != EAGAIN) && (rc != EALREADY))) { + break; // Result is final } - /* EAGAIN may occur due to interruption by a signal or due to some - * transient issue. Try one more time to be more resilient. - */ - if (i < (n_attempts - 1)) { - crm_trace("Connection to %s IPC API failed with EAGAIN, retrying", - pcmk_ipc_name(api, true)); - } + // Retry after soft error (interrupted by signal, etc.) + pcmk__sleep_ms((attempts - remaining) * 500); + crm_debug("Re-attempting connection to %s (%d attempt%s remaining)", + pcmk_ipc_name(api, true), remaining, + pcmk__plural_s(remaining)); } if (rc != pcmk_rc_ok) { return rc; } if ((api->cmds != NULL) && (api->cmds->post_connect != NULL)) { rc = api->cmds->post_connect(api); if (rc != pcmk_rc_ok) { crm_ipc_close(api->ipc); } } return rc; } +/*! + * \brief Connect to a Pacemaker daemon via IPC + * + * \param[in,out] api IPC API instance + * \param[in] dispatch_type How IPC replies should be dispatched + * + * \return Standard Pacemaker return code + */ +int +pcmk_connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type) +{ + int rc = pcmk__connect_ipc(api, dispatch_type, 2); + + if (rc != pcmk_rc_ok) { + crm_err("Connection to %s failed: %s", + pcmk_ipc_name(api, true), pcmk_rc_str(rc)); + } + return rc; +} + /*! * \brief Disconnect an IPC API instance * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code * * \note If the connection is attached to a main loop, this function should be * called before quitting the main loop, to ensure that all memory is * freed. */ void pcmk_disconnect_ipc(pcmk_ipc_api_t *api) { if ((api == NULL) || (api->ipc == NULL)) { return; } switch (api->dispatch_type) { case pcmk_ipc_dispatch_main: { mainloop_io_t *mainloop_io = api->mainloop_io; // Make sure no code with access to api can use these again api->mainloop_io = NULL; api->ipc = NULL; mainloop_del_ipc_client(mainloop_io); // After this point api might have already been freed } break; case pcmk_ipc_dispatch_poll: case pcmk_ipc_dispatch_sync: { crm_ipc_t *ipc = api->ipc; // Make sure no code with access to api can use ipc again api->ipc = NULL; // This should always be the case already, but to be safe api->free_on_disconnect = false; crm_ipc_close(ipc); crm_ipc_destroy(ipc); ipc_post_disconnect(api); } break; } } /*! * \brief Register a callback for IPC API events * * \param[in,out] api IPC API connection * \param[in] callback Callback to register * \param[in] userdata Caller data to pass to callback * * \note This function may be called multiple times to update the callback * and/or user data. The caller remains responsible for freeing * userdata in any case (after the IPC is disconnected, if the * user data is still registered with the IPC). */ void pcmk_register_ipc_callback(pcmk_ipc_api_t *api, pcmk_ipc_callback_t cb, void *user_data) { if (api == NULL) { return; } api->cb = cb; api->user_data = user_data; } /*! * \internal * \brief Send an XML request across an IPC API connection * * \param[in,out] api IPC API connection * \param[in,out] request XML request to send * * \return Standard Pacemaker return code * * \note Daemon-specific IPC API functions should call this function to send * requests, because it handles different dispatch types appropriately. */ int pcmk__send_ipc_request(pcmk_ipc_api_t *api, xmlNode *request) { int rc; xmlNode *reply = NULL; enum crm_ipc_flags flags = crm_ipc_flags_none; if ((api == NULL) || (api->ipc == NULL) || (request == NULL)) { return EINVAL; } crm_log_xml_trace(request, "ipc-sent"); // Synchronous dispatch requires waiting for a reply if ((api->dispatch_type == pcmk_ipc_dispatch_sync) && (api->cmds != NULL) && (api->cmds->reply_expected != NULL) && (api->cmds->reply_expected(api, request))) { flags = crm_ipc_client_response; } // The 0 here means a default timeout of 5 seconds rc = crm_ipc_send(api->ipc, request, flags, 0, &reply); if (rc < 0) { return pcmk_legacy2rc(rc); } else if (rc == 0) { return ENODATA; } // With synchronous dispatch, we dispatch any reply now if (reply != NULL) { bool more = call_api_dispatch(api, reply); free_xml(reply); while (more) { rc = crm_ipc_read(api->ipc); if (rc == -EAGAIN) { continue; } else if (rc == -ENOMSG || rc == pcmk_ok) { return pcmk_rc_ok; } else if (rc < 0) { return -rc; } rc = dispatch_ipc_data(crm_ipc_buffer(api->ipc), api); if (rc == pcmk_rc_ok) { more = false; } else if (rc == EINPROGRESS) { more = true; } else { continue; } } } return pcmk_rc_ok; } /*! * \internal * \brief Create the XML for an IPC request to purge a node from the peer cache * * \param[in] api IPC API connection * \param[in] node_name If not NULL, name of node to purge * \param[in] nodeid If not 0, node ID of node to purge * * \return Newly allocated IPC request XML * * \note The controller, fencer, and pacemakerd use the same request syntax, but * the attribute manager uses a different one. The CIB manager doesn't * have any syntax for it. The executor and scheduler don't connect to the * cluster layer and thus don't have or need any syntax for it. * * \todo Modify the attribute manager to accept the common syntax (as well * as its current one, for compatibility with older clients). Modify * the CIB manager to accept and honor the common syntax. Modify the * executor and scheduler to accept the syntax (immediately returning * success), just for consistency. Modify this function to use the * common syntax with all daemons if their version supports it. */ static xmlNode * create_purge_node_request(const pcmk_ipc_api_t *api, const char *node_name, uint32_t nodeid) { xmlNode *request = NULL; const char *client = crm_system_name? crm_system_name : "client"; switch (api->server) { case pcmk_ipc_attrd: request = create_xml_node(NULL, __func__); crm_xml_add(request, F_TYPE, T_ATTRD); crm_xml_add(request, F_ORIG, crm_system_name); crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_PEER_REMOVE); pcmk__xe_add_node(request, node_name, nodeid); break; case pcmk_ipc_controld: case pcmk_ipc_fenced: case pcmk_ipc_pacemakerd: request = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, pcmk_ipc_name(api, false), client, NULL); if (nodeid > 0) { crm_xml_set_id(request, "%lu", (unsigned long) nodeid); } crm_xml_add(request, XML_ATTR_UNAME, node_name); break; case pcmk_ipc_based: case pcmk_ipc_execd: case pcmk_ipc_schedulerd: break; } return request; } /*! * \brief Ask a Pacemaker daemon to purge a node from its peer cache * * \param[in,out] api IPC API connection * \param[in] node_name If not NULL, name of node to purge * \param[in] nodeid If not 0, node ID of node to purge * * \return Standard Pacemaker return code * * \note At least one of node_name or nodeid must be specified. */ int pcmk_ipc_purge_node(pcmk_ipc_api_t *api, const char *node_name, uint32_t nodeid) { int rc = 0; xmlNode *request = NULL; if (api == NULL) { return EINVAL; } if ((node_name == NULL) && (nodeid == 0)) { return EINVAL; } request = create_purge_node_request(api, node_name, nodeid); if (request == NULL) { return EOPNOTSUPP; } rc = pcmk__send_ipc_request(api, request); free_xml(request); crm_debug("%s peer cache purge of node %s[%lu]: rc=%d", pcmk_ipc_name(api, true), node_name, (unsigned long) nodeid, rc); return rc; } /* * Generic IPC API (to eventually be deprecated as public API and made internal) */ struct crm_ipc_s { struct pollfd pfd; unsigned int max_buf_size; // maximum bytes we can send or receive over IPC unsigned int buf_size; // size of allocated buffer int msg_size; int need_reply; char *buffer; char *server_name; // server IPC name being connected to qb_ipcc_connection_t *ipc; }; /*! * \brief Create a new (legacy) object for using Pacemaker daemon IPC * * \param[in] name IPC system name to connect to * \param[in] max_size Use a maximum IPC buffer size of at least this size * * \return Newly allocated IPC object on success, NULL otherwise * * \note The caller is responsible for freeing the result using * crm_ipc_destroy(). * \note This should be considered deprecated for use with daemons supported by * pcmk_new_ipc_api(). */ crm_ipc_t * crm_ipc_new(const char *name, size_t max_size) { crm_ipc_t *client = NULL; client = calloc(1, sizeof(crm_ipc_t)); if (client == NULL) { crm_err("Could not create IPC connection: %s", strerror(errno)); return NULL; } client->server_name = strdup(name); if (client->server_name == NULL) { crm_err("Could not create %s IPC connection: %s", name, strerror(errno)); free(client); return NULL; } client->buf_size = pcmk__ipc_buffer_size(max_size); client->buffer = malloc(client->buf_size); if (client->buffer == NULL) { crm_err("Could not create %s IPC connection: %s", name, strerror(errno)); free(client->server_name); free(client); return NULL; } /* Clients initiating connection pick the max buf size */ client->max_buf_size = client->buf_size; client->pfd.fd = -1; client->pfd.events = POLLIN; client->pfd.revents = 0; return client; } +/*! + * \internal + * \brief Connect a generic (not daemon-specific) IPC object + * + * \param[in,out] ipc Generic IPC object to connect + * + * \return Standard Pacemaker return code + */ +int +pcmk__connect_generic_ipc(crm_ipc_t *ipc) +{ + uid_t cl_uid = 0; + gid_t cl_gid = 0; + pid_t found_pid = 0; + uid_t found_uid = 0; + gid_t found_gid = 0; + int rc = pcmk_rc_ok; + + if (ipc == NULL) { + return EINVAL; + } + + ipc->need_reply = FALSE; + ipc->ipc = qb_ipcc_connect(ipc->server_name, ipc->buf_size); + if (ipc->ipc == NULL) { + return errno; + } + + rc = qb_ipcc_fd_get(ipc->ipc, &ipc->pfd.fd); + if (rc < 0) { // -errno + crm_ipc_close(ipc); + return -rc; + } + + rc = pcmk_daemon_user(&cl_uid, &cl_gid); + rc = pcmk_legacy2rc(rc); + if (rc != pcmk_rc_ok) { + crm_ipc_close(ipc); + return rc; + } + + rc = is_ipc_provider_expected(ipc->ipc, ipc->pfd.fd, cl_uid, cl_gid, + &found_pid, &found_uid, &found_gid); + if (rc != pcmk_rc_ok) { + if (rc == pcmk_rc_ipc_unauthorized) { + crm_info("%s IPC provider authentication failed: process %lld has " + "uid %lld (expected %lld) and gid %lld (expected %lld)", + ipc->server_name, + (long long) PCMK__SPECIAL_PID_AS_0(found_pid), + (long long) found_uid, (long long) cl_uid, + (long long) found_gid, (long long) cl_gid); + } + crm_ipc_close(ipc); + return rc; + } + + ipc->max_buf_size = qb_ipcc_get_buffer_size(ipc->ipc); + if (ipc->max_buf_size > ipc->buf_size) { + free(ipc->buffer); + ipc->buffer = calloc(ipc->max_buf_size, sizeof(char)); + if (ipc->buffer == NULL) { + rc = errno; + crm_ipc_close(ipc); + return rc; + } + ipc->buf_size = ipc->max_buf_size; + } + + return pcmk_rc_ok; +} + /*! * \brief Establish an IPC connection to a Pacemaker component * * \param[in,out] client Connection instance obtained from crm_ipc_new() * * \return true on success, false otherwise (in which case errno will be set; * specifically, in case of discovering the remote side is not * authentic, its value is set to ECONNABORTED). */ bool crm_ipc_connect(crm_ipc_t *client) { - uid_t cl_uid = 0; - gid_t cl_gid = 0; - pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; - int rv; + int rc = pcmk__connect_generic_ipc(client); - if (client == NULL) { - errno = EINVAL; - return false; + if (rc == pcmk_rc_ok) { + return true; } - - client->need_reply = FALSE; - client->ipc = qb_ipcc_connect(client->server_name, client->buf_size); - - if (client->ipc == NULL) { + if ((client != NULL) && (client->ipc == NULL)) { + errno = (rc > 0)? rc : ENOTCONN; crm_debug("Could not establish %s IPC connection: %s (%d)", client->server_name, pcmk_rc_str(errno), errno); - return false; - } - - client->pfd.fd = crm_ipc_get_fd(client); - if (client->pfd.fd < 0) { - rv = errno; - /* message already omitted */ - crm_ipc_close(client); - errno = rv; - return false; - } - - rv = pcmk_daemon_user(&cl_uid, &cl_gid); - if (rv < 0) { - /* message already omitted */ - crm_ipc_close(client); - errno = -rv; - return false; - } - - if ((rv = pcmk__crm_ipc_is_authentic_process(client->ipc, client->pfd.fd, cl_uid, cl_gid, - &found_pid, &found_uid, - &found_gid)) == pcmk_rc_ipc_unauthorized) { - crm_err("%s IPC provider authentication failed: process %lld has " - "uid %lld (expected %lld) and gid %lld (expected %lld)", - client->server_name, - (long long) PCMK__SPECIAL_PID_AS_0(found_pid), - (long long) found_uid, (long long) cl_uid, - (long long) found_gid, (long long) cl_gid); - crm_ipc_close(client); + } else if (rc == pcmk_rc_ipc_unauthorized) { + crm_err("%s IPC provider authentication failed", + (client == NULL)? "Pacemaker" : client->server_name); errno = ECONNABORTED; - return false; - - } else if (rv != pcmk_rc_ok) { - crm_perror(LOG_ERR, "Could not verify authenticity of %s IPC provider", - client->server_name); - crm_ipc_close(client); - if (rv > 0) { - errno = rv; - } else { - errno = ENOTCONN; - } - return false; - } - - qb_ipcc_context_set(client->ipc, client); - - client->max_buf_size = qb_ipcc_get_buffer_size(client->ipc); - if (client->max_buf_size > client->buf_size) { - free(client->buffer); - client->buffer = calloc(1, client->max_buf_size); - client->buf_size = client->max_buf_size; + } else { + crm_perror(LOG_ERR, + "Could not verify authenticity of %s IPC provider", + (client == NULL)? "Pacemaker" : client->server_name); + errno = ENOTCONN; } - return true; + return false; } void crm_ipc_close(crm_ipc_t * client) { if (client) { if (client->ipc) { qb_ipcc_connection_t *ipc = client->ipc; client->ipc = NULL; qb_ipcc_disconnect(ipc); } } } void crm_ipc_destroy(crm_ipc_t * client) { if (client) { if (client->ipc && qb_ipcc_is_connected(client->ipc)) { crm_notice("Destroying active %s IPC connection", client->server_name); /* The next line is basically unsafe * * If this connection was attached to mainloop and mainloop is active, * the 'disconnected' callback will end up back here and we'll end * up free'ing the memory twice - something that can still happen * even without this if we destroy a connection and it closes before * we call exit */ /* crm_ipc_close(client); */ } else { crm_trace("Destroying inactive %s IPC connection", client->server_name); } free(client->buffer); free(client->server_name); free(client); } } +/*! + * \internal + * \brief Get the file descriptor for a generic IPC object + * + * \param[in,out] ipc Generic IPC object to get file descriptor for + * \param[out] fd Where to store file descriptor + * + * \return Standard Pacemaker return code + */ +int +pcmk__ipc_fd(crm_ipc_t *ipc, int *fd) +{ + if ((ipc == NULL) || (fd == NULL)) { + return EINVAL; + } + if ((ipc->ipc == NULL) || (ipc->pfd.fd < 0)) { + return ENOTCONN; + } + *fd = ipc->pfd.fd; + return pcmk_rc_ok; +} + int crm_ipc_get_fd(crm_ipc_t * client) { - int fd = 0; + int fd = -1; - if (client && client->ipc && (qb_ipcc_fd_get(client->ipc, &fd) == 0)) { - return fd; + if (pcmk__ipc_fd(client, &fd) != pcmk_rc_ok) { + crm_err("Could not obtain file descriptor for %s IPC", + ((client == NULL)? "unspecified" : client->server_name)); + errno = EINVAL; + return -EINVAL; } - errno = EINVAL; - crm_perror(LOG_ERR, "Could not obtain file descriptor for %s IPC", - (client? client->server_name : "unspecified")); - return -errno; + return fd; } bool crm_ipc_connected(crm_ipc_t * client) { bool rc = FALSE; if (client == NULL) { crm_trace("No client"); return FALSE; } else if (client->ipc == NULL) { crm_trace("No connection"); return FALSE; } else if (client->pfd.fd < 0) { crm_trace("Bad descriptor"); return FALSE; } rc = qb_ipcc_is_connected(client->ipc); if (rc == FALSE) { client->pfd.fd = -EINVAL; } return rc; } /*! * \brief Check whether an IPC connection is ready to be read * * \param[in,out] client Connection to check * * \return Positive value if ready to be read, 0 if not ready, -errno on error */ int crm_ipc_ready(crm_ipc_t *client) { int rc; CRM_ASSERT(client != NULL); if (!crm_ipc_connected(client)) { return -ENOTCONN; } client->pfd.revents = 0; rc = poll(&(client->pfd), 1, 0); return (rc < 0)? -errno : rc; } // \return Standard Pacemaker return code static int crm_ipc_decompress(crm_ipc_t * client) { pcmk__ipc_header_t *header = (pcmk__ipc_header_t *)(void*)client->buffer; if (header->size_compressed) { int rc = 0; unsigned int size_u = 1 + header->size_uncompressed; /* never let buf size fall below our max size required for ipc reads. */ unsigned int new_buf_size = QB_MAX((sizeof(pcmk__ipc_header_t) + size_u), client->max_buf_size); char *uncompressed = calloc(1, new_buf_size); crm_trace("Decompressing message data %u bytes into %u bytes", header->size_compressed, size_u); rc = BZ2_bzBuffToBuffDecompress(uncompressed + sizeof(pcmk__ipc_header_t), &size_u, client->buffer + sizeof(pcmk__ipc_header_t), header->size_compressed, 1, 0); if (rc != BZ_OK) { crm_err("Decompression failed: %s " CRM_XS " bzerror=%d", bz2_strerror(rc), rc); free(uncompressed); return EILSEQ; } /* * This assert no longer holds true. For an identical msg, some clients may * require compression, and others may not. If that same msg (event) is sent * to multiple clients, it could result in some clients receiving a compressed * msg even though compression was not explicitly required for them. * * CRM_ASSERT((header->size_uncompressed + sizeof(pcmk__ipc_header_t)) >= ipc_buffer_max); */ CRM_ASSERT(size_u == header->size_uncompressed); memcpy(uncompressed, client->buffer, sizeof(pcmk__ipc_header_t)); /* Preserve the header */ header = (pcmk__ipc_header_t *)(void*)uncompressed; free(client->buffer); client->buf_size = new_buf_size; client->buffer = uncompressed; } CRM_ASSERT(client->buffer[sizeof(pcmk__ipc_header_t) + header->size_uncompressed - 1] == 0); return pcmk_rc_ok; } long crm_ipc_read(crm_ipc_t * client) { pcmk__ipc_header_t *header = NULL; CRM_ASSERT(client != NULL); CRM_ASSERT(client->ipc != NULL); CRM_ASSERT(client->buffer != NULL); client->buffer[0] = 0; client->msg_size = qb_ipcc_event_recv(client->ipc, client->buffer, client->buf_size, 0); if (client->msg_size >= 0) { int rc = crm_ipc_decompress(client); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } header = (pcmk__ipc_header_t *)(void*)client->buffer; if (!pcmk__valid_ipc_header(header)) { return -EBADMSG; } crm_trace("Received %s IPC event %d size=%u rc=%d text='%.100s'", client->server_name, header->qb.id, header->qb.size, client->msg_size, client->buffer + sizeof(pcmk__ipc_header_t)); } else { crm_trace("No message received from %s IPC: %s", client->server_name, pcmk_strerror(client->msg_size)); if (client->msg_size == -EAGAIN) { return -EAGAIN; } } if (!crm_ipc_connected(client) || client->msg_size == -ENOTCONN) { crm_err("Connection to %s IPC failed", client->server_name); } if (header) { /* Data excluding the header */ return header->size_uncompressed; } return -ENOMSG; } const char * crm_ipc_buffer(crm_ipc_t * client) { CRM_ASSERT(client != NULL); return client->buffer + sizeof(pcmk__ipc_header_t); } uint32_t crm_ipc_buffer_flags(crm_ipc_t * client) { pcmk__ipc_header_t *header = NULL; CRM_ASSERT(client != NULL); if (client->buffer == NULL) { return 0; } header = (pcmk__ipc_header_t *)(void*)client->buffer; return header->flags; } const char * crm_ipc_name(crm_ipc_t * client) { CRM_ASSERT(client != NULL); return client->server_name; } // \return Standard Pacemaker return code static int internal_ipc_get_reply(crm_ipc_t *client, int request_id, int ms_timeout, ssize_t *bytes) { time_t timeout = time(NULL) + 1 + (ms_timeout / 1000); int rc = pcmk_rc_ok; /* get the reply */ crm_trace("Waiting on reply to %s IPC message %d", client->server_name, request_id); do { *bytes = qb_ipcc_recv(client->ipc, client->buffer, client->buf_size, 1000); if (*bytes > 0) { pcmk__ipc_header_t *hdr = NULL; rc = crm_ipc_decompress(client); if (rc != pcmk_rc_ok) { return rc; } hdr = (pcmk__ipc_header_t *)(void*)client->buffer; if (hdr->qb.id == request_id) { /* Got it */ break; } else if (hdr->qb.id < request_id) { xmlNode *bad = string2xml(crm_ipc_buffer(client)); crm_err("Discarding old reply %d (need %d)", hdr->qb.id, request_id); crm_log_xml_notice(bad, "OldIpcReply"); } else { xmlNode *bad = string2xml(crm_ipc_buffer(client)); crm_err("Discarding newer reply %d (need %d)", hdr->qb.id, request_id); crm_log_xml_notice(bad, "ImpossibleReply"); CRM_ASSERT(hdr->qb.id <= request_id); } } else if (!crm_ipc_connected(client)) { crm_err("%s IPC provider disconnected while waiting for message %d", client->server_name, request_id); break; } } while (time(NULL) < timeout); if (*bytes < 0) { rc = (int) -*bytes; // System errno } return rc; } /*! * \brief Send an IPC XML message * * \param[in,out] client Connection to IPC server * \param[in,out] message XML message to send * \param[in] flags Bitmask of crm_ipc_flags * \param[in] ms_timeout Give up if not sent within this much time * (5 seconds if 0, or no timeout if negative) * \param[out] reply Reply from server (or NULL if none) * * \return Negative errno on error, otherwise size of reply received in bytes * if reply was needed, otherwise number of bytes sent */ int crm_ipc_send(crm_ipc_t * client, xmlNode * message, enum crm_ipc_flags flags, int32_t ms_timeout, xmlNode ** reply) { int rc = 0; ssize_t qb_rc = 0; ssize_t bytes = 0; struct iovec *iov; static uint32_t id = 0; static int factor = 8; pcmk__ipc_header_t *header; if (client == NULL) { crm_notice("Can't send IPC request without connection (bug?): %.100s", message); return -ENOTCONN; } else if (!crm_ipc_connected(client)) { /* Don't even bother */ crm_notice("Can't send %s IPC requests: Connection closed", client->server_name); return -ENOTCONN; } if (ms_timeout == 0) { ms_timeout = 5000; } if (client->need_reply) { qb_rc = qb_ipcc_recv(client->ipc, client->buffer, client->buf_size, ms_timeout); if (qb_rc < 0) { crm_warn("Sending %s IPC disabled until pending reply received", client->server_name); return -EALREADY; } else { crm_notice("Sending %s IPC re-enabled after pending reply received", client->server_name); client->need_reply = FALSE; } } id++; CRM_LOG_ASSERT(id != 0); /* Crude wrap-around detection */ rc = pcmk__ipc_prepare_iov(id, message, client->max_buf_size, &iov, &bytes); if (rc != pcmk_rc_ok) { crm_warn("Couldn't prepare %s IPC request: %s " CRM_XS " rc=%d", client->server_name, pcmk_rc_str(rc), rc); return pcmk_rc2legacy(rc); } header = iov[0].iov_base; pcmk__set_ipc_flags(header->flags, client->server_name, flags); if (pcmk_is_set(flags, crm_ipc_proxied)) { /* Don't look for a synchronous response */ pcmk__clear_ipc_flags(flags, "client", crm_ipc_client_response); } if(header->size_compressed) { if(factor < 10 && (client->max_buf_size / 10) < (bytes / factor)) { crm_notice("Compressed message exceeds %d0%% of configured IPC " "limit (%u bytes); consider setting PCMK_ipc_buffer to " "%u or higher", factor, client->max_buf_size, 2 * client->max_buf_size); factor++; } } crm_trace("Sending %s IPC request %d of %u bytes using %dms timeout", client->server_name, header->qb.id, header->qb.size, ms_timeout); if ((ms_timeout > 0) || !pcmk_is_set(flags, crm_ipc_client_response)) { time_t timeout = time(NULL) + 1 + (ms_timeout / 1000); do { /* @TODO Is this check really needed? Won't qb_ipcc_sendv() return * an error if it's not connected? */ if (!crm_ipc_connected(client)) { goto send_cleanup; } qb_rc = qb_ipcc_sendv(client->ipc, iov, 2); } while ((qb_rc == -EAGAIN) && (time(NULL) < timeout)); rc = (int) qb_rc; // Negative of system errno, or bytes sent if (qb_rc <= 0) { goto send_cleanup; } else if (!pcmk_is_set(flags, crm_ipc_client_response)) { crm_trace("Not waiting for reply to %s IPC request %d", client->server_name, header->qb.id); goto send_cleanup; } rc = internal_ipc_get_reply(client, header->qb.id, ms_timeout, &bytes); if (rc != pcmk_rc_ok) { /* We didn't get the reply in time, so disable future sends for now. * The only alternative would be to close the connection since we * don't know how to detect and discard out-of-sequence replies. * * @TODO Implement out-of-sequence detection */ client->need_reply = TRUE; } rc = (int) bytes; // Negative system errno, or size of reply received } else { // No timeout, and client response needed do { qb_rc = qb_ipcc_sendv_recv(client->ipc, iov, 2, client->buffer, client->buf_size, -1); } while ((qb_rc == -EAGAIN) && crm_ipc_connected(client)); rc = (int) qb_rc; // Negative system errno, or size of reply received } if (rc > 0) { pcmk__ipc_header_t *hdr = (pcmk__ipc_header_t *)(void*)client->buffer; crm_trace("Received %d-byte reply %d to %s IPC %d: %.100s", rc, hdr->qb.id, client->server_name, header->qb.id, crm_ipc_buffer(client)); if (reply) { *reply = string2xml(crm_ipc_buffer(client)); } } else { crm_trace("No reply to %s IPC %d: rc=%d", client->server_name, header->qb.id, rc); } send_cleanup: if (!crm_ipc_connected(client)) { crm_notice("Couldn't send %s IPC request %d: Connection closed " CRM_XS " rc=%d", client->server_name, header->qb.id, rc); } else if (rc == -ETIMEDOUT) { crm_warn("%s IPC request %d failed: %s after %dms " CRM_XS " rc=%d", client->server_name, header->qb.id, pcmk_strerror(rc), ms_timeout, rc); crm_write_blackbox(0, NULL); } else if (rc <= 0) { crm_warn("%s IPC request %d failed: %s " CRM_XS " rc=%d", client->server_name, header->qb.id, ((rc == 0)? "No bytes sent" : pcmk_strerror(rc)), rc); } pcmk_free_ipc_event(iov); return rc; } -int -pcmk__crm_ipc_is_authentic_process(qb_ipcc_connection_t *qb_ipc, int sock, uid_t refuid, gid_t refgid, - pid_t *gotpid, uid_t *gotuid, gid_t *gotgid) +/*! + * \brief Ensure an IPC provider has expected user or group + * + * \param[in] qb_ipc libqb client connection if available + * \param[in] sock Connected Unix socket for IPC + * \param[in] refuid Expected user ID + * \param[in] refgid Expected group ID + * \param[out] gotpid If not NULL, where to store provider's actual process ID + * (or 1 on platforms where ID is not available) + * \param[out] gotuid If not NULL, where to store provider's actual user ID + * \param[out] gotgid If not NULL, where to store provider's actual group ID + * + * \return Standard Pacemaker return code + * \note An actual user ID of 0 (root) will always be considered authorized, + * regardless of the expected values provided. The caller can use the + * output arguments to be stricter than this function. + */ +static int +is_ipc_provider_expected(qb_ipcc_connection_t *qb_ipc, int sock, + uid_t refuid, gid_t refgid, + pid_t *gotpid, uid_t *gotuid, gid_t *gotgid) { - int ret = 0; - pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; -#if defined(HAVE_UCRED) - struct ucred ucred; - socklen_t ucred_len = sizeof(ucred); -#endif + int rc = EOPNOTSUPP; + pid_t found_pid = 0; + uid_t found_uid = 0; + gid_t found_gid = 0; #ifdef HAVE_QB_IPCC_AUTH_GET - if (qb_ipc && !qb_ipcc_auth_get(qb_ipc, &found_pid, &found_uid, &found_gid)) { - goto do_checks; + if (qb_ipc != NULL) { + rc = qb_ipcc_auth_get(qb_ipc, &found_pid, &found_uid, &found_gid); + rc = -rc; // libqb returns 0 or -errno + if (rc == pcmk_rc_ok) { + goto found; + } } #endif -#if defined(HAVE_UCRED) - if (!getsockopt(sock, SOL_SOCKET, SO_PEERCRED, - &ucred, &ucred_len) - && ucred_len == sizeof(ucred)) { - found_pid = ucred.pid; found_uid = ucred.uid; found_gid = ucred.gid; +#ifdef HAVE_UCRED + { + struct ucred ucred; + socklen_t ucred_len = sizeof(ucred); -#elif defined(HAVE_SOCKPEERCRED) - struct sockpeercred sockpeercred; - socklen_t sockpeercred_len = sizeof(sockpeercred); - - if (!getsockopt(sock, SOL_SOCKET, SO_PEERCRED, - &sockpeercred, &sockpeercred_len) - && sockpeercred_len == sizeof(sockpeercred_len)) { - found_pid = sockpeercred.pid; - found_uid = sockpeercred.uid; found_gid = sockpeercred.gid; + if (getsockopt(sock, SOL_SOCKET, SO_PEERCRED, &ucred, &ucred_len) < 0) { + rc = errno; + } else if (ucred_len != sizeof(ucred)) { + rc = EOPNOTSUPP; + } else { + found_pid = ucred.pid; + found_uid = ucred.uid; + found_gid = ucred.gid; + goto found; + } + } +#endif -#elif defined(HAVE_GETPEEREID) - if (!getpeereid(sock, &found_uid, &found_gid)) { - found_pid = PCMK__SPECIAL_PID; /* cannot obtain PID (FreeBSD) */ +#ifdef HAVE_SOCKPEERCRED + { + struct sockpeercred sockpeercred; + socklen_t sockpeercred_len = sizeof(sockpeercred); -#elif defined(HAVE_GETPEERUCRED) - ucred_t *ucred; - if (!getpeerucred(sock, &ucred)) { - errno = 0; - found_pid = ucred_getpid(ucred); - found_uid = ucred_geteuid(ucred); found_gid = ucred_getegid(ucred); - ret = -errno; - ucred_free(ucred); - if (ret) { - return (ret < 0) ? ret : -pcmk_err_generic; + if (getsockopt(sock, SOL_SOCKET, SO_PEERCRED, + &sockpeercred, &sockpeercred_len) < 0) { + rc = errno; + } else if (sockpeercred_len != sizeof(sockpeercred)) { + rc = EOPNOTSUPP; + } else { + found_pid = sockpeercred.pid; + found_uid = sockpeercred.uid; + found_gid = sockpeercred.gid; + goto found; } - -#else -# error "No way to authenticate a Unix socket peer" - errno = 0; - if (0) { + } #endif -#ifdef HAVE_QB_IPCC_AUTH_GET - do_checks: + +#ifdef HAVE_GETPEEREID // For example, FreeBSD + if (getpeereid(sock, &found_uid, &found_gid) < 0) { + rc = errno; + } else { + found_pid = PCMK__SPECIAL_PID; + goto found; + } #endif - if (gotpid != NULL) { - *gotpid = found_pid; - } - if (gotuid != NULL) { - *gotuid = found_uid; - } - if (gotgid != NULL) { - *gotgid = found_gid; - } - if (found_uid == 0 || found_uid == refuid || found_gid == refgid) { - ret = 0; + +#ifdef HAVE_GETPEERUCRED + { + ucred_t *ucred = NULL; + + if (getpeerucred(sock, &ucred) < 0) { + rc = errno; } else { - ret = pcmk_rc_ipc_unauthorized; + found_pid = ucred_getpid(ucred); + found_uid = ucred_geteuid(ucred); + found_gid = ucred_getegid(ucred); + ucred_free(ucred); + goto found; } - } else { - ret = (errno > 0) ? errno : pcmk_rc_error; } - return ret; +#endif + + return rc; // If we get here, nothing succeeded + +found: + if (gotpid != NULL) { + *gotpid = found_pid; + } + if (gotuid != NULL) { + *gotuid = found_uid; + } + if (gotgid != NULL) { + *gotgid = found_gid; + } + if ((found_uid != 0) && (found_uid != refuid) && (found_gid != refgid)) { + return pcmk_rc_ipc_unauthorized; + } + return pcmk_rc_ok; } int crm_ipc_is_authentic_process(int sock, uid_t refuid, gid_t refgid, pid_t *gotpid, uid_t *gotuid, gid_t *gotgid) { - int ret = pcmk__crm_ipc_is_authentic_process(NULL, sock, refuid, refgid, - gotpid, gotuid, gotgid); + int ret = is_ipc_provider_expected(NULL, sock, refuid, refgid, + gotpid, gotuid, gotgid); /* The old function had some very odd return codes*/ if (ret == 0) { return 1; } else if (ret == pcmk_rc_ipc_unauthorized) { return 0; } else { return pcmk_rc2legacy(ret); } } int pcmk__ipc_is_authentic_process_active(const char *name, uid_t refuid, gid_t refgid, pid_t *gotpid) { static char last_asked_name[PATH_MAX / 2] = ""; /* log spam prevention */ int fd; int rc = pcmk_rc_ipc_unresponsive; int auth_rc = 0; int32_t qb_rc; pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; qb_ipcc_connection_t *c; #ifdef HAVE_QB_IPCC_CONNECT_ASYNC struct pollfd pollfd = { 0, }; int poll_rc; c = qb_ipcc_connect_async(name, 0, &(pollfd.fd)); #else c = qb_ipcc_connect(name, 0); #endif if (c == NULL) { crm_info("Could not connect to %s IPC: %s", name, strerror(errno)); rc = pcmk_rc_ipc_unresponsive; goto bail; } #ifdef HAVE_QB_IPCC_CONNECT_ASYNC pollfd.events = POLLIN; do { poll_rc = poll(&pollfd, 1, 2000); } while ((poll_rc == -1) && (errno == EINTR)); if ((poll_rc <= 0) || (qb_ipcc_connect_continue(c) != 0)) { crm_info("Could not connect to %s IPC: %s", name, (poll_rc == 0)?"timeout":strerror(errno)); rc = pcmk_rc_ipc_unresponsive; if (poll_rc > 0) { c = NULL; // qb_ipcc_connect_continue cleaned up for us } goto bail; } #endif qb_rc = qb_ipcc_fd_get(c, &fd); if (qb_rc != 0) { rc = (int) -qb_rc; // System errno crm_err("Could not get fd from %s IPC: %s " CRM_XS " rc=%d", name, pcmk_rc_str(rc), rc); goto bail; } - auth_rc = pcmk__crm_ipc_is_authentic_process(c, fd, refuid, refgid, &found_pid, - &found_uid, &found_gid); + auth_rc = is_ipc_provider_expected(c, fd, refuid, refgid, + &found_pid, &found_uid, &found_gid); if (auth_rc == pcmk_rc_ipc_unauthorized) { crm_err("Daemon (IPC %s) effectively blocked with unauthorized" " process %lld (uid: %lld, gid: %lld)", name, (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); rc = pcmk_rc_ipc_unauthorized; goto bail; } if (auth_rc != pcmk_rc_ok) { rc = auth_rc; crm_err("Could not get peer credentials from %s IPC: %s " CRM_XS " rc=%d", name, pcmk_rc_str(rc), rc); goto bail; } if (gotpid != NULL) { *gotpid = found_pid; } rc = pcmk_rc_ok; if ((found_uid != refuid || found_gid != refgid) && strncmp(last_asked_name, name, sizeof(last_asked_name))) { if ((found_uid == 0) && (refuid != 0)) { crm_warn("Daemon (IPC %s) runs as root, whereas the expected" " credentials are %lld:%lld, hazard of violating" " the least privilege principle", name, (long long) refuid, (long long) refgid); } else { crm_notice("Daemon (IPC %s) runs as %lld:%lld, whereas the" " expected credentials are %lld:%lld, which may" " mean a different set of privileges than expected", name, (long long) found_uid, (long long) found_gid, (long long) refuid, (long long) refgid); } memccpy(last_asked_name, name, '\0', sizeof(last_asked_name)); } bail: if (c != NULL) { qb_ipcc_disconnect(c); } return rc; } diff --git a/lib/common/mainloop.c b/lib/common/mainloop.c index 3124e43475..e883d39797 100644 --- a/lib/common/mainloop.c +++ b/lib/common/mainloop.c @@ -1,1480 +1,1493 @@ /* * Copyright 2004-2023 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include struct mainloop_child_s { pid_t pid; char *desc; unsigned timerid; gboolean timeout; void *privatedata; enum mainloop_child_flags flags; /* Called when a process dies */ void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode); }; struct trigger_s { GSource source; gboolean running; gboolean trigger; void *user_data; guint id; }; struct mainloop_timer_s { guint id; guint period_ms; bool repeat; char *name; GSourceFunc cb; void *userdata; }; static gboolean crm_trigger_prepare(GSource * source, gint * timeout) { crm_trigger_t *trig = (crm_trigger_t *) source; /* cluster-glue's FD and IPC related sources make use of * g_source_add_poll() but do not set a timeout in their prepare * functions * * This means mainloop's poll() will block until an event for one * of these sources occurs - any /other/ type of source, such as * this one or g_idle_*, that doesn't use g_source_add_poll() is * S-O-L and won't be processed until there is something fd-based * happens. * * Luckily the timeout we can set here affects all sources and * puts an upper limit on how long poll() can take. * * So unconditionally set a small-ish timeout, not too small that * we're in constant motion, which will act as an upper bound on * how long the signal handling might be delayed for. */ *timeout = 500; /* Timeout in ms */ return trig->trigger; } static gboolean crm_trigger_check(GSource * source) { crm_trigger_t *trig = (crm_trigger_t *) source; return trig->trigger; } /*! * \internal * \brief GSource dispatch function for crm_trigger_t * * \param[in] source crm_trigger_t being dispatched * \param[in] callback Callback passed at source creation * \param[in,out] userdata User data passed at source creation * * \return G_SOURCE_REMOVE to remove source, G_SOURCE_CONTINUE to keep it */ static gboolean crm_trigger_dispatch(GSource *source, GSourceFunc callback, gpointer userdata) { gboolean rc = G_SOURCE_CONTINUE; crm_trigger_t *trig = (crm_trigger_t *) source; if (trig->running) { /* Wait until the existing job is complete before starting the next one */ return G_SOURCE_CONTINUE; } trig->trigger = FALSE; if (callback) { int callback_rc = callback(trig->user_data); if (callback_rc < 0) { crm_trace("Trigger handler %p not yet complete", trig); trig->running = TRUE; } else if (callback_rc == 0) { rc = G_SOURCE_REMOVE; } } return rc; } static void crm_trigger_finalize(GSource * source) { crm_trace("Trigger %p destroyed", source); } static GSourceFuncs crm_trigger_funcs = { crm_trigger_prepare, crm_trigger_check, crm_trigger_dispatch, crm_trigger_finalize, }; static crm_trigger_t * mainloop_setup_trigger(GSource * source, int priority, int (*dispatch) (gpointer user_data), gpointer userdata) { crm_trigger_t *trigger = NULL; trigger = (crm_trigger_t *) source; trigger->id = 0; trigger->trigger = FALSE; trigger->user_data = userdata; if (dispatch) { g_source_set_callback(source, dispatch, trigger, NULL); } g_source_set_priority(source, priority); g_source_set_can_recurse(source, FALSE); trigger->id = g_source_attach(source, NULL); return trigger; } void mainloop_trigger_complete(crm_trigger_t * trig) { crm_trace("Trigger handler %p complete", trig); trig->running = FALSE; } /*! * \brief Create a trigger to be used as a mainloop source * * \param[in] priority Relative priority of source (lower number is higher priority) * \param[in] dispatch Trigger dispatch function (should return 0 to remove the * trigger from the mainloop, -1 if the trigger should be * kept but the job is still running and not complete, and * 1 if the trigger should be kept and the job is complete) * \param[in] userdata Pointer to pass to \p dispatch * * \return Newly allocated mainloop source for trigger */ crm_trigger_t * mainloop_add_trigger(int priority, int (*dispatch) (gpointer user_data), gpointer userdata) { GSource *source = NULL; CRM_ASSERT(sizeof(crm_trigger_t) > sizeof(GSource)); source = g_source_new(&crm_trigger_funcs, sizeof(crm_trigger_t)); CRM_ASSERT(source != NULL); return mainloop_setup_trigger(source, priority, dispatch, userdata); } void mainloop_set_trigger(crm_trigger_t * source) { if(source) { source->trigger = TRUE; } } gboolean mainloop_destroy_trigger(crm_trigger_t * source) { GSource *gs = NULL; if(source == NULL) { return TRUE; } gs = (GSource *)source; g_source_destroy(gs); /* Remove from mainloop, ref_count-- */ g_source_unref(gs); /* The caller no longer carries a reference to source * * At this point the source should be free'd, * unless we're currently processing said * source, in which case mainloop holds an * additional reference and it will be free'd * once our processing completes */ return TRUE; } // Define a custom glib source for signal handling // Data structure for custom glib source typedef struct signal_s { crm_trigger_t trigger; // trigger that invoked source (must be first) void (*handler) (int sig); // signal handler int signal; // signal that was received } crm_signal_t; // Table to associate signal handlers with signal numbers static crm_signal_t *crm_signals[NSIG]; /*! * \internal * \brief Dispatch an event from custom glib source for signals * * Given an signal event, clear the event trigger and call any registered * signal handler. * * \param[in] source glib source that triggered this dispatch * \param[in] callback (ignored) * \param[in] userdata (ignored) */ static gboolean crm_signal_dispatch(GSource *source, GSourceFunc callback, gpointer userdata) { crm_signal_t *sig = (crm_signal_t *) source; if(sig->signal != SIGCHLD) { crm_notice("Caught '%s' signal "CRM_XS" %d (%s handler)", strsignal(sig->signal), sig->signal, (sig->handler? "invoking" : "no")); } sig->trigger.trigger = FALSE; if (sig->handler) { sig->handler(sig->signal); } return TRUE; } /*! * \internal * \brief Handle a signal by setting a trigger for signal source * * \param[in] sig Signal number that was received * * \note This is the true signal handler for the mainloop signal source, and * must be async-safe. */ static void mainloop_signal_handler(int sig) { if (sig > 0 && sig < NSIG && crm_signals[sig] != NULL) { mainloop_set_trigger((crm_trigger_t *) crm_signals[sig]); } } // Functions implementing our custom glib source for signal handling static GSourceFuncs crm_signal_funcs = { crm_trigger_prepare, crm_trigger_check, crm_signal_dispatch, crm_trigger_finalize, }; /*! * \internal * \brief Set a true signal handler * * signal()-like interface to sigaction() * * \param[in] sig Signal number to register handler for * \param[in] dispatch Signal handler * * \return The previous value of the signal handler, or SIG_ERR on error * \note The dispatch function must be async-safe. */ sighandler_t crm_signal_handler(int sig, sighandler_t dispatch) { sigset_t mask; struct sigaction sa; struct sigaction old; if (sigemptyset(&mask) < 0) { crm_err("Could not set handler for signal %d: %s", sig, pcmk_rc_str(errno)); return SIG_ERR; } memset(&sa, 0, sizeof(struct sigaction)); sa.sa_handler = dispatch; sa.sa_flags = SA_RESTART; sa.sa_mask = mask; if (sigaction(sig, &sa, &old) < 0) { crm_err("Could not set handler for signal %d: %s", sig, pcmk_rc_str(errno)); return SIG_ERR; } return old.sa_handler; } static void mainloop_destroy_signal_entry(int sig) { crm_signal_t *tmp = crm_signals[sig]; crm_signals[sig] = NULL; crm_trace("Destroying signal %d", sig); mainloop_destroy_trigger((crm_trigger_t *) tmp); } /*! * \internal * \brief Add a signal handler to a mainloop * * \param[in] sig Signal number to handle * \param[in] dispatch Signal handler function * * \note The true signal handler merely sets a mainloop trigger to call this * dispatch function via the mainloop. Therefore, the dispatch function * does not need to be async-safe. */ gboolean mainloop_add_signal(int sig, void (*dispatch) (int sig)) { GSource *source = NULL; int priority = G_PRIORITY_HIGH - 1; if (sig == SIGTERM) { /* TERM is higher priority than other signals, * signals are higher priority than other ipc. * Yes, minus: smaller is "higher" */ priority--; } if (sig >= NSIG || sig < 0) { crm_err("Signal %d is out of range", sig); return FALSE; } else if (crm_signals[sig] != NULL && crm_signals[sig]->handler == dispatch) { crm_trace("Signal handler for %d is already installed", sig); return TRUE; } else if (crm_signals[sig] != NULL) { crm_err("Different signal handler for %d is already installed", sig); return FALSE; } CRM_ASSERT(sizeof(crm_signal_t) > sizeof(GSource)); source = g_source_new(&crm_signal_funcs, sizeof(crm_signal_t)); crm_signals[sig] = (crm_signal_t *) mainloop_setup_trigger(source, priority, NULL, NULL); CRM_ASSERT(crm_signals[sig] != NULL); crm_signals[sig]->handler = dispatch; crm_signals[sig]->signal = sig; if (crm_signal_handler(sig, mainloop_signal_handler) == SIG_ERR) { mainloop_destroy_signal_entry(sig); return FALSE; } #if 0 /* If we want signals to interrupt mainloop's poll(), instead of waiting for * the timeout, then we should call siginterrupt() below * * For now, just enforce a low timeout */ if (siginterrupt(sig, 1) < 0) { crm_perror(LOG_INFO, "Could not enable system call interruptions for signal %d", sig); } #endif return TRUE; } gboolean mainloop_destroy_signal(int sig) { if (sig >= NSIG || sig < 0) { crm_err("Signal %d is out of range", sig); return FALSE; } else if (crm_signal_handler(sig, NULL) == SIG_ERR) { crm_perror(LOG_ERR, "Could not uninstall signal handler for signal %d", sig); return FALSE; } else if (crm_signals[sig] == NULL) { return TRUE; } mainloop_destroy_signal_entry(sig); return TRUE; } static qb_array_t *gio_map = NULL; void mainloop_cleanup(void) { if (gio_map) { qb_array_free(gio_map); } for (int sig = 0; sig < NSIG; ++sig) { mainloop_destroy_signal_entry(sig); } } /* * libqb... */ struct gio_to_qb_poll { int32_t is_used; guint source; int32_t events; void *data; qb_ipcs_dispatch_fn_t fn; enum qb_loop_priority p; }; static gboolean gio_read_socket(GIOChannel * gio, GIOCondition condition, gpointer data) { struct gio_to_qb_poll *adaptor = (struct gio_to_qb_poll *)data; gint fd = g_io_channel_unix_get_fd(gio); crm_trace("%p.%d %d", data, fd, condition); /* if this assert get's hit, then there is a race condition between * when we destroy a fd and when mainloop actually gives it up */ CRM_ASSERT(adaptor->is_used > 0); return (adaptor->fn(fd, condition, adaptor->data) == 0); } static void gio_poll_destroy(gpointer data) { struct gio_to_qb_poll *adaptor = (struct gio_to_qb_poll *)data; adaptor->is_used--; CRM_ASSERT(adaptor->is_used >= 0); if (adaptor->is_used == 0) { crm_trace("Marking adaptor %p unused", adaptor); adaptor->source = 0; } } /*! * \internal * \brief Convert libqb's poll priority into GLib's one * * \param[in] prio libqb's poll priority (#QB_LOOP_MED assumed as fallback) * * \return best matching GLib's priority */ static gint conv_prio_libqb2glib(enum qb_loop_priority prio) { switch (prio) { case QB_LOOP_LOW: return G_PRIORITY_LOW; case QB_LOOP_HIGH: return G_PRIORITY_HIGH; default: return G_PRIORITY_DEFAULT; // QB_LOOP_MED } } /*! * \internal * \brief Convert libqb's poll priority to rate limiting spec * * \param[in] prio libqb's poll priority (#QB_LOOP_MED assumed as fallback) * * \return best matching rate limiting spec * \note This is the inverse of libqb's qb_ipcs_request_rate_limit(). */ static enum qb_ipcs_rate_limit conv_libqb_prio2ratelimit(enum qb_loop_priority prio) { switch (prio) { case QB_LOOP_LOW: return QB_IPCS_RATE_SLOW; case QB_LOOP_HIGH: return QB_IPCS_RATE_FAST; default: return QB_IPCS_RATE_NORMAL; // QB_LOOP_MED } } static int32_t gio_poll_dispatch_update(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn, int32_t add) { struct gio_to_qb_poll *adaptor; GIOChannel *channel; int32_t res = 0; res = qb_array_index(gio_map, fd, (void **)&adaptor); if (res < 0) { crm_err("Array lookup failed for fd=%d: %d", fd, res); return res; } crm_trace("Adding fd=%d to mainloop as adaptor %p", fd, adaptor); if (add && adaptor->source) { crm_err("Adaptor for descriptor %d is still in-use", fd); return -EEXIST; } if (!add && !adaptor->is_used) { crm_err("Adaptor for descriptor %d is not in-use", fd); return -ENOENT; } /* channel is created with ref_count = 1 */ channel = g_io_channel_unix_new(fd); if (!channel) { crm_err("No memory left to add fd=%d", fd); return -ENOMEM; } if (adaptor->source) { g_source_remove(adaptor->source); adaptor->source = 0; } /* Because unlike the poll() API, glib doesn't tell us about HUPs by default */ evts |= (G_IO_HUP | G_IO_NVAL | G_IO_ERR); adaptor->fn = fn; adaptor->events = evts; adaptor->data = data; adaptor->p = p; adaptor->is_used++; adaptor->source = g_io_add_watch_full(channel, conv_prio_libqb2glib(p), evts, gio_read_socket, adaptor, gio_poll_destroy); /* Now that mainloop now holds a reference to channel, * thanks to g_io_add_watch_full(), drop ours from g_io_channel_unix_new(). * * This means that channel will be free'd by: * g_main_context_dispatch() * -> g_source_destroy_internal() * -> g_source_callback_unref() * shortly after gio_poll_destroy() completes */ g_io_channel_unref(channel); crm_trace("Added to mainloop with gsource id=%d", adaptor->source); if (adaptor->source > 0) { return 0; } return -EINVAL; } static int32_t gio_poll_dispatch_add(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn) { return gio_poll_dispatch_update(p, fd, evts, data, fn, QB_TRUE); } static int32_t gio_poll_dispatch_mod(enum qb_loop_priority p, int32_t fd, int32_t evts, void *data, qb_ipcs_dispatch_fn_t fn) { return gio_poll_dispatch_update(p, fd, evts, data, fn, QB_FALSE); } static int32_t gio_poll_dispatch_del(int32_t fd) { struct gio_to_qb_poll *adaptor; crm_trace("Looking for fd=%d", fd); if (qb_array_index(gio_map, fd, (void **)&adaptor) == 0) { if (adaptor->source) { g_source_remove(adaptor->source); adaptor->source = 0; } } return 0; } struct qb_ipcs_poll_handlers gio_poll_funcs = { .job_add = NULL, .dispatch_add = gio_poll_dispatch_add, .dispatch_mod = gio_poll_dispatch_mod, .dispatch_del = gio_poll_dispatch_del, }; static enum qb_ipc_type pick_ipc_type(enum qb_ipc_type requested) { const char *env = getenv("PCMK_ipc_type"); if (env && strcmp("shared-mem", env) == 0) { return QB_IPC_SHM; } else if (env && strcmp("socket", env) == 0) { return QB_IPC_SOCKET; } else if (env && strcmp("posix", env) == 0) { return QB_IPC_POSIX_MQ; } else if (env && strcmp("sysv", env) == 0) { return QB_IPC_SYSV_MQ; } else if (requested == QB_IPC_NATIVE) { /* We prefer shared memory because the server never blocks on * send. If part of a message fits into the socket, libqb * needs to block until the remainder can be sent also. * Otherwise the client will wait forever for the remaining * bytes. */ return QB_IPC_SHM; } return requested; } qb_ipcs_service_t * mainloop_add_ipc_server(const char *name, enum qb_ipc_type type, struct qb_ipcs_service_handlers *callbacks) { return mainloop_add_ipc_server_with_prio(name, type, callbacks, QB_LOOP_MED); } qb_ipcs_service_t * mainloop_add_ipc_server_with_prio(const char *name, enum qb_ipc_type type, struct qb_ipcs_service_handlers *callbacks, enum qb_loop_priority prio) { int rc = 0; qb_ipcs_service_t *server = NULL; if (gio_map == NULL) { gio_map = qb_array_create_2(64, sizeof(struct gio_to_qb_poll), 1); } server = qb_ipcs_create(name, 0, pick_ipc_type(type), callbacks); if (server == NULL) { crm_err("Could not create %s IPC server: %s (%d)", name, pcmk_strerror(rc), rc); return NULL; } if (prio != QB_LOOP_MED) { qb_ipcs_request_rate_limit(server, conv_libqb_prio2ratelimit(prio)); } /* All clients should use at least ipc_buffer_max as their buffer size */ qb_ipcs_enforce_buffer_size(server, crm_ipc_default_buffer_size()); qb_ipcs_poll_handlers_set(server, &gio_poll_funcs); rc = qb_ipcs_run(server); if (rc < 0) { crm_err("Could not start %s IPC server: %s (%d)", name, pcmk_strerror(rc), rc); return NULL; // qb_ipcs_run() destroys server on failure } return server; } void mainloop_del_ipc_server(qb_ipcs_service_t * server) { if (server) { qb_ipcs_destroy(server); } } struct mainloop_io_s { char *name; void *userdata; int fd; guint source; crm_ipc_t *ipc; GIOChannel *channel; int (*dispatch_fn_ipc) (const char *buffer, ssize_t length, gpointer userdata); int (*dispatch_fn_io) (gpointer userdata); void (*destroy_fn) (gpointer userdata); }; /*! * \internal * \brief I/O watch callback function (GIOFunc) * * \param[in] gio I/O channel being watched * \param[in] condition I/O condition satisfied * \param[in] data User data passed when source was created * * \return G_SOURCE_REMOVE to remove source, G_SOURCE_CONTINUE to keep it */ static gboolean mainloop_gio_callback(GIOChannel *gio, GIOCondition condition, gpointer data) { gboolean rc = G_SOURCE_CONTINUE; mainloop_io_t *client = data; CRM_ASSERT(client->fd == g_io_channel_unix_get_fd(gio)); if (condition & G_IO_IN) { if (client->ipc) { long read_rc = 0L; int max = 10; do { read_rc = crm_ipc_read(client->ipc); if (read_rc <= 0) { crm_trace("Could not read IPC message from %s: %s (%ld)", client->name, pcmk_strerror(read_rc), read_rc); } else if (client->dispatch_fn_ipc) { const char *buffer = crm_ipc_buffer(client->ipc); crm_trace("New %ld-byte IPC message from %s " "after I/O condition %d", read_rc, client->name, (int) condition); if (client->dispatch_fn_ipc(buffer, read_rc, client->userdata) < 0) { crm_trace("Connection to %s no longer required", client->name); rc = G_SOURCE_REMOVE; } } } while ((rc == G_SOURCE_CONTINUE) && (read_rc > 0) && --max > 0); } else { crm_trace("New I/O event for %s after I/O condition %d", client->name, (int) condition); if (client->dispatch_fn_io) { if (client->dispatch_fn_io(client->userdata) < 0) { crm_trace("Connection to %s no longer required", client->name); rc = G_SOURCE_REMOVE; } } } } if (client->ipc && !crm_ipc_connected(client->ipc)) { crm_err("Connection to %s closed " CRM_XS "client=%p condition=%d", client->name, client, condition); rc = G_SOURCE_REMOVE; } else if (condition & (G_IO_HUP | G_IO_NVAL | G_IO_ERR)) { crm_trace("The connection %s[%p] has been closed (I/O condition=%d)", client->name, client, condition); rc = G_SOURCE_REMOVE; } else if ((condition & G_IO_IN) == 0) { /* #define GLIB_SYSDEF_POLLIN =1 #define GLIB_SYSDEF_POLLPRI =2 #define GLIB_SYSDEF_POLLOUT =4 #define GLIB_SYSDEF_POLLERR =8 #define GLIB_SYSDEF_POLLHUP =16 #define GLIB_SYSDEF_POLLNVAL =32 typedef enum { G_IO_IN GLIB_SYSDEF_POLLIN, G_IO_OUT GLIB_SYSDEF_POLLOUT, G_IO_PRI GLIB_SYSDEF_POLLPRI, G_IO_ERR GLIB_SYSDEF_POLLERR, G_IO_HUP GLIB_SYSDEF_POLLHUP, G_IO_NVAL GLIB_SYSDEF_POLLNVAL } GIOCondition; A bitwise combination representing a condition to watch for on an event source. G_IO_IN There is data to read. G_IO_OUT Data can be written (without blocking). G_IO_PRI There is urgent data to read. G_IO_ERR Error condition. G_IO_HUP Hung up (the connection has been broken, usually for pipes and sockets). G_IO_NVAL Invalid request. The file descriptor is not open. */ crm_err("Strange condition: %d", condition); } /* G_SOURCE_REMOVE results in mainloop_gio_destroy() being called * just before the source is removed from mainloop */ return rc; } static void mainloop_gio_destroy(gpointer c) { mainloop_io_t *client = c; char *c_name = strdup(client->name); /* client->source is valid but about to be destroyed (ref_count == 0) in gmain.c * client->channel will still have ref_count > 0... should be == 1 */ crm_trace("Destroying client %s[%p]", c_name, c); if (client->ipc) { crm_ipc_close(client->ipc); } if (client->destroy_fn) { void (*destroy_fn) (gpointer userdata) = client->destroy_fn; client->destroy_fn = NULL; destroy_fn(client->userdata); } if (client->ipc) { crm_ipc_t *ipc = client->ipc; client->ipc = NULL; crm_ipc_destroy(ipc); } crm_trace("Destroyed client %s[%p]", c_name, c); free(client->name); client->name = NULL; free(client); free(c_name); } /*! * \brief Connect to IPC and add it as a main loop source * * \param[in,out] ipc IPC connection to add * \param[in] priority Event source priority to use for connection * \param[in] userdata Data to register with callbacks * \param[in] callbacks Dispatch and destroy callbacks for connection * \param[out] source Newly allocated event source * * \return Standard Pacemaker return code * * \note On failure, the caller is still responsible for ipc. On success, the * caller should call mainloop_del_ipc_client() when source is no longer * needed, which will lead to the disconnection of the IPC later in the * main loop if it is connected. However the IPC disconnects, * mainloop_gio_destroy() will free ipc and source after calling the * destroy callback. */ int pcmk__add_mainloop_ipc(crm_ipc_t *ipc, int priority, void *userdata, const struct ipc_client_callbacks *callbacks, mainloop_io_t **source) { + int rc = pcmk_rc_ok; + int fd = -1; + const char *ipc_name = NULL; + CRM_CHECK((ipc != NULL) && (callbacks != NULL), return EINVAL); - if (!crm_ipc_connect(ipc)) { - int rc = errno; - crm_debug("Connection to %s failed: %d", crm_ipc_name(ipc), errno); + ipc_name = pcmk__s(crm_ipc_name(ipc), "Pacemaker"); + rc = pcmk__connect_generic_ipc(ipc); + if (rc != pcmk_rc_ok) { + crm_debug("Connection to %s failed: %s", ipc_name, pcmk_rc_str(rc)); + return rc; + } + + rc = pcmk__ipc_fd(ipc, &fd); + if (rc != pcmk_rc_ok) { + crm_debug("Could not obtain file descriptor for %s IPC: %s", + ipc_name, pcmk_rc_str(rc)); + crm_ipc_close(ipc); return rc; } - *source = mainloop_add_fd(crm_ipc_name(ipc), priority, crm_ipc_get_fd(ipc), - userdata, NULL); - if (*source == NULL) { - int rc = errno; + *source = mainloop_add_fd(ipc_name, priority, fd, userdata, NULL); + if (*source == NULL) { + rc = errno; crm_ipc_close(ipc); return rc; } + (*source)->ipc = ipc; (*source)->destroy_fn = callbacks->destroy; (*source)->dispatch_fn_ipc = callbacks->dispatch; return pcmk_rc_ok; } /*! * \brief Get period for mainloop timer * * \param[in] timer Timer * * \return Period in ms */ guint pcmk__mainloop_timer_get_period(const mainloop_timer_t *timer) { if (timer) { return timer->period_ms; } return 0; } mainloop_io_t * mainloop_add_ipc_client(const char *name, int priority, size_t max_size, void *userdata, struct ipc_client_callbacks *callbacks) { crm_ipc_t *ipc = crm_ipc_new(name, max_size); mainloop_io_t *source = NULL; int rc = pcmk__add_mainloop_ipc(ipc, priority, userdata, callbacks, &source); if (rc != pcmk_rc_ok) { if (crm_log_level == LOG_STDOUT) { fprintf(stderr, "Connection to %s failed: %s", name, pcmk_rc_str(rc)); } crm_ipc_destroy(ipc); if (rc > 0) { errno = rc; } else { errno = ENOTCONN; } return NULL; } return source; } void mainloop_del_ipc_client(mainloop_io_t * client) { mainloop_del_fd(client); } crm_ipc_t * mainloop_get_ipc_client(mainloop_io_t * client) { if (client) { return client->ipc; } return NULL; } mainloop_io_t * mainloop_add_fd(const char *name, int priority, int fd, void *userdata, struct mainloop_fd_callbacks * callbacks) { mainloop_io_t *client = NULL; if (fd >= 0) { client = calloc(1, sizeof(mainloop_io_t)); if (client == NULL) { return NULL; } client->name = strdup(name); client->userdata = userdata; if (callbacks) { client->destroy_fn = callbacks->destroy; client->dispatch_fn_io = callbacks->dispatch; } client->fd = fd; client->channel = g_io_channel_unix_new(fd); client->source = g_io_add_watch_full(client->channel, priority, (G_IO_IN | G_IO_HUP | G_IO_NVAL | G_IO_ERR), mainloop_gio_callback, client, mainloop_gio_destroy); /* Now that mainloop now holds a reference to channel, * thanks to g_io_add_watch_full(), drop ours from g_io_channel_unix_new(). * * This means that channel will be free'd by: * g_main_context_dispatch() or g_source_remove() * -> g_source_destroy_internal() * -> g_source_callback_unref() * shortly after mainloop_gio_destroy() completes */ g_io_channel_unref(client->channel); crm_trace("Added connection %d for %s[%p].%d", client->source, client->name, client, fd); } else { errno = EINVAL; } return client; } void mainloop_del_fd(mainloop_io_t * client) { if (client != NULL) { crm_trace("Removing client %s[%p]", client->name, client); if (client->source) { /* Results in mainloop_gio_destroy() being called just * before the source is removed from mainloop */ g_source_remove(client->source); } } } static GList *child_list = NULL; pid_t mainloop_child_pid(mainloop_child_t * child) { return child->pid; } const char * mainloop_child_name(mainloop_child_t * child) { return child->desc; } int mainloop_child_timeout(mainloop_child_t * child) { return child->timeout; } void * mainloop_child_userdata(mainloop_child_t * child) { return child->privatedata; } void mainloop_clear_child_userdata(mainloop_child_t * child) { child->privatedata = NULL; } /* good function name */ static void child_free(mainloop_child_t *child) { if (child->timerid != 0) { crm_trace("Removing timer %d", child->timerid); g_source_remove(child->timerid); child->timerid = 0; } free(child->desc); free(child); } /* terrible function name */ static int child_kill_helper(mainloop_child_t *child) { int rc; if (child->flags & mainloop_leave_pid_group) { crm_debug("Kill pid %d only. leave group intact.", child->pid); rc = kill(child->pid, SIGKILL); } else { crm_debug("Kill pid %d's group", child->pid); rc = kill(-child->pid, SIGKILL); } if (rc < 0) { if (errno != ESRCH) { crm_perror(LOG_ERR, "kill(%d, KILL) failed", child->pid); } return -errno; } return 0; } static gboolean child_timeout_callback(gpointer p) { mainloop_child_t *child = p; int rc = 0; child->timerid = 0; if (child->timeout) { crm_warn("%s process (PID %d) will not die!", child->desc, (int)child->pid); return FALSE; } rc = child_kill_helper(child); if (rc == -ESRCH) { /* Nothing left to do. pid doesn't exist */ return FALSE; } child->timeout = TRUE; crm_debug("%s process (PID %d) timed out", child->desc, (int)child->pid); child->timerid = g_timeout_add(5000, child_timeout_callback, child); return FALSE; } static bool child_waitpid(mainloop_child_t *child, int flags) { int rc = 0; int core = 0; int signo = 0; int status = 0; int exitcode = 0; bool callback_needed = true; rc = waitpid(child->pid, &status, flags); if (rc == 0) { // WNOHANG in flags, and child status is not available crm_trace("Child process %d (%s) still active", child->pid, child->desc); callback_needed = false; } else if (rc != child->pid) { /* According to POSIX, possible conditions: * - child->pid was non-positive (process group or any child), * and rc is specific child * - errno ECHILD (pid does not exist or is not child) * - errno EINVAL (invalid flags) * - errno EINTR (caller interrupted by signal) * * @TODO Handle these cases more specifically. */ signo = SIGCHLD; exitcode = 1; crm_notice("Wait for child process %d (%s) interrupted: %s", child->pid, child->desc, pcmk_rc_str(errno)); } else if (WIFEXITED(status)) { exitcode = WEXITSTATUS(status); crm_trace("Child process %d (%s) exited with status %d", child->pid, child->desc, exitcode); } else if (WIFSIGNALED(status)) { signo = WTERMSIG(status); crm_trace("Child process %d (%s) exited with signal %d (%s)", child->pid, child->desc, signo, strsignal(signo)); #ifdef WCOREDUMP // AIX, SunOS, maybe others } else if (WCOREDUMP(status)) { core = 1; crm_err("Child process %d (%s) dumped core", child->pid, child->desc); #endif } else { // flags must contain WUNTRACED and/or WCONTINUED to reach this crm_trace("Child process %d (%s) stopped or continued", child->pid, child->desc); callback_needed = false; } if (callback_needed && child->callback) { child->callback(child, child->pid, core, signo, exitcode); } return callback_needed; } static void child_death_dispatch(int signal) { for (GList *iter = child_list; iter; ) { GList *saved = iter; mainloop_child_t *child = iter->data; iter = iter->next; if (child_waitpid(child, WNOHANG)) { crm_trace("Removing completed process %d from child list", child->pid); child_list = g_list_remove_link(child_list, saved); g_list_free(saved); child_free(child); } } } static gboolean child_signal_init(gpointer p) { crm_trace("Installed SIGCHLD handler"); /* Do NOT use g_child_watch_add() and friends, they rely on pthreads */ mainloop_add_signal(SIGCHLD, child_death_dispatch); /* In case they terminated before the signal handler was installed */ child_death_dispatch(SIGCHLD); return FALSE; } gboolean mainloop_child_kill(pid_t pid) { GList *iter; mainloop_child_t *child = NULL; mainloop_child_t *match = NULL; /* It is impossible to block SIGKILL, this allows us to * call waitpid without WNOHANG flag.*/ int waitflags = 0, rc = 0; for (iter = child_list; iter != NULL && match == NULL; iter = iter->next) { child = iter->data; if (pid == child->pid) { match = child; } } if (match == NULL) { return FALSE; } rc = child_kill_helper(match); if(rc == -ESRCH) { /* It's gone, but hasn't shown up in waitpid() yet. Wait until we get * SIGCHLD and let handler clean it up as normal (so we get the correct * return code/status). The blocking alternative would be to call * child_waitpid(match, 0). */ crm_trace("Waiting for signal that child process %d completed", match->pid); return TRUE; } else if(rc != 0) { /* If KILL for some other reason set the WNOHANG flag since we * can't be certain what happened. */ waitflags = WNOHANG; } if (!child_waitpid(match, waitflags)) { /* not much we can do if this occurs */ return FALSE; } child_list = g_list_remove(child_list, match); child_free(match); return TRUE; } /* Create/Log a new tracked process * To track a process group, use -pid * * @TODO Using a non-positive pid (i.e. any child, or process group) would * likely not be useful since we will free the child after the first * completed process. */ void mainloop_child_add_with_flags(pid_t pid, int timeout, const char *desc, void *privatedata, enum mainloop_child_flags flags, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)) { static bool need_init = TRUE; mainloop_child_t *child = calloc(1, sizeof(mainloop_child_t)); child->pid = pid; child->timerid = 0; child->timeout = FALSE; child->privatedata = privatedata; child->callback = callback; child->flags = flags; pcmk__str_update(&child->desc, desc); if (timeout) { child->timerid = g_timeout_add(timeout, child_timeout_callback, child); } child_list = g_list_append(child_list, child); if(need_init) { need_init = FALSE; /* SIGCHLD processing has to be invoked from mainloop. * We do not want it to be possible to both add a child pid * to mainloop, and have the pid's exit callback invoked within * the same callstack. */ g_timeout_add(1, child_signal_init, NULL); } } void mainloop_child_add(pid_t pid, int timeout, const char *desc, void *privatedata, void (*callback) (mainloop_child_t * p, pid_t pid, int core, int signo, int exitcode)) { mainloop_child_add_with_flags(pid, timeout, desc, privatedata, 0, callback); } static gboolean mainloop_timer_cb(gpointer user_data) { int id = 0; bool repeat = FALSE; struct mainloop_timer_s *t = user_data; CRM_ASSERT(t != NULL); id = t->id; t->id = 0; /* Ensure it's unset during callbacks so that * mainloop_timer_running() works as expected */ if(t->cb) { crm_trace("Invoking callbacks for timer %s", t->name); repeat = t->repeat; if(t->cb(t->userdata) == FALSE) { crm_trace("Timer %s complete", t->name); repeat = FALSE; } } if(repeat) { /* Restore if repeating */ t->id = id; } return repeat; } bool mainloop_timer_running(mainloop_timer_t *t) { if(t && t->id != 0) { return TRUE; } return FALSE; } void mainloop_timer_start(mainloop_timer_t *t) { mainloop_timer_stop(t); if(t && t->period_ms > 0) { crm_trace("Starting timer %s", t->name); t->id = g_timeout_add(t->period_ms, mainloop_timer_cb, t); } } void mainloop_timer_stop(mainloop_timer_t *t) { if(t && t->id != 0) { crm_trace("Stopping timer %s", t->name); g_source_remove(t->id); t->id = 0; } } guint mainloop_timer_set_period(mainloop_timer_t *t, guint period_ms) { guint last = 0; if(t) { last = t->period_ms; t->period_ms = period_ms; } if(t && t->id != 0 && last != t->period_ms) { mainloop_timer_start(t); } return last; } mainloop_timer_t * mainloop_timer_add(const char *name, guint period_ms, bool repeat, GSourceFunc cb, void *userdata) { mainloop_timer_t *t = calloc(1, sizeof(mainloop_timer_t)); if(t) { if(name) { t->name = crm_strdup_printf("%s-%u-%d", name, period_ms, repeat); } else { t->name = crm_strdup_printf("%p-%u-%d", t, period_ms, repeat); } t->id = 0; t->period_ms = period_ms; t->repeat = repeat; t->cb = cb; t->userdata = userdata; crm_trace("Created timer %s with %p %p", t->name, userdata, t->userdata); } return t; } void mainloop_timer_del(mainloop_timer_t *t) { if(t) { crm_trace("Destroying timer %s", t->name); mainloop_timer_stop(t); free(t->name); free(t); } } /* * Helpers to make sure certain events aren't lost at shutdown */ static gboolean drain_timeout_cb(gpointer user_data) { bool *timeout_popped = (bool*) user_data; *timeout_popped = TRUE; return FALSE; } /*! * \brief Drain some remaining main loop events then quit it * * \param[in,out] mloop Main loop to drain and quit * \param[in] n Drain up to this many pending events */ void pcmk_quit_main_loop(GMainLoop *mloop, unsigned int n) { if ((mloop != NULL) && g_main_loop_is_running(mloop)) { GMainContext *ctx = g_main_loop_get_context(mloop); /* Drain up to n events in case some memory clean-up is pending * (helpful to reduce noise in valgrind output). */ for (int i = 0; (i < n) && g_main_context_pending(ctx); ++i) { g_main_context_dispatch(ctx); } g_main_loop_quit(mloop); } } /*! * \brief Process main loop events while a certain condition is met * * \param[in,out] mloop Main loop to process * \param[in] timer_ms Don't process longer than this amount of time * \param[in] check Function that returns true if events should be * processed * * \note This function is intended to be called at shutdown if certain important * events should not be missed. The caller would likely quit the main loop * or exit after calling this function. The check() function will be * passed the remaining timeout in milliseconds. */ void pcmk_drain_main_loop(GMainLoop *mloop, guint timer_ms, bool (*check)(guint)) { bool timeout_popped = FALSE; guint timer = 0; GMainContext *ctx = NULL; CRM_CHECK(mloop && check, return); ctx = g_main_loop_get_context(mloop); if (ctx) { time_t start_time = time(NULL); timer = g_timeout_add(timer_ms, drain_timeout_cb, &timeout_popped); while (!timeout_popped && check(timer_ms - (time(NULL) - start_time) * 1000)) { g_main_context_iteration(ctx, TRUE); } } if (!timeout_popped && (timer > 0)) { g_source_remove(timer); } } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START #include gboolean crm_signal(int sig, void (*dispatch) (int sig)) { return crm_signal_handler(sig, dispatch) != SIG_ERR; } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/fencing/st_client.c b/lib/fencing/st_client.c index e2783d59e4..338bf653d4 100644 --- a/lib/fencing/st_client.c +++ b/lib/fencing/st_client.c @@ -1,2702 +1,2709 @@ /* - * Copyright 2004-2022 the Pacemaker project contributors + * Copyright 2004-2023 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 "fencing_private.h" CRM_TRACE_INIT_DATA(stonith); // Used as stonith_t:st_private typedef struct stonith_private_s { char *token; crm_ipc_t *ipc; mainloop_io_t *source; GHashTable *stonith_op_callback_table; GList *notify_list; int notify_refcnt; bool notify_deletes; void (*op_callback) (stonith_t * st, stonith_callback_data_t * data); } stonith_private_t; // Used as stonith_event_t:opaque struct event_private { pcmk__action_result_t result; }; typedef struct stonith_notify_client_s { const char *event; const char *obj_id; /* implement one day */ const char *obj_type; /* implement one day */ void (*notify) (stonith_t * st, stonith_event_t * e); bool delete; } stonith_notify_client_t; typedef struct stonith_callback_client_s { void (*callback) (stonith_t * st, stonith_callback_data_t * data); const char *id; void *user_data; gboolean only_success; gboolean allow_timeout_updates; struct timer_rec_s *timer; } stonith_callback_client_t; struct notify_blob_s { stonith_t *stonith; xmlNode *xml; }; struct timer_rec_s { int call_id; int timeout; guint ref; stonith_t *stonith; }; typedef int (*stonith_op_t) (const char *, int, const char *, xmlNode *, xmlNode *, xmlNode *, xmlNode **, xmlNode **); bool stonith_dispatch(stonith_t * st); xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options); static int stonith_send_command(stonith_t *stonith, const char *op, xmlNode *data, xmlNode **output_data, int call_options, int timeout); static void stonith_connection_destroy(gpointer user_data); static void stonith_send_notification(gpointer data, gpointer user_data); static int stonith_api_del_notification(stonith_t *stonith, const char *event); /*! * \brief Get agent namespace by name * * \param[in] namespace_s Name of namespace as string * * \return Namespace as enum value */ enum stonith_namespace stonith_text2namespace(const char *namespace_s) { if (pcmk__str_eq(namespace_s, "any", pcmk__str_null_matches)) { return st_namespace_any; } else if (!strcmp(namespace_s, "redhat") || !strcmp(namespace_s, "stonith-ng")) { return st_namespace_rhcs; } else if (!strcmp(namespace_s, "internal")) { return st_namespace_internal; } else if (!strcmp(namespace_s, "heartbeat")) { return st_namespace_lha; } return st_namespace_invalid; } /*! * \brief Get agent namespace name * * \param[in] namespace Namespace as enum value * * \return Namespace name as string */ const char * stonith_namespace2text(enum stonith_namespace st_namespace) { switch (st_namespace) { case st_namespace_any: return "any"; case st_namespace_rhcs: return "stonith-ng"; case st_namespace_internal: return "internal"; case st_namespace_lha: return "heartbeat"; default: break; } return "unsupported"; } /*! * \brief Determine namespace of a fence agent * * \param[in] agent Fence agent type * \param[in] namespace_s Name of agent namespace as string, if known * * \return Namespace of specified agent, as enum value */ enum stonith_namespace stonith_get_namespace(const char *agent, const char *namespace_s) { if (pcmk__str_eq(namespace_s, "internal", pcmk__str_none)) { return st_namespace_internal; } if (stonith__agent_is_rhcs(agent)) { return st_namespace_rhcs; } #if HAVE_STONITH_STONITH_H if (stonith__agent_is_lha(agent)) { return st_namespace_lha; } #endif crm_err("Unknown fence agent: %s", agent); return st_namespace_invalid; } gboolean stonith__watchdog_fencing_enabled_for_node_api(stonith_t *st, const char *node) { gboolean rv = FALSE; stonith_t *stonith_api = st?st:stonith_api_new(); char *list = NULL; if(stonith_api) { if (stonith_api->state == stonith_disconnected) { int rc = stonith_api->cmds->connect(stonith_api, "stonith-api", NULL); if (rc != pcmk_ok) { crm_err("Failed connecting to Stonith-API for watchdog-fencing-query."); } } if (stonith_api->state != stonith_disconnected) { /* caveat!!! * this might fail when when stonithd is just updating the device-list * probably something we should fix as well for other api-calls */ int rc = stonith_api->cmds->list(stonith_api, st_opt_sync_call, STONITH_WATCHDOG_ID, &list, 0); if ((rc != pcmk_ok) || (list == NULL)) { /* due to the race described above it can happen that * we drop in here - so as not to make remote nodes * panic on that answer */ if (rc == -ENODEV) { crm_notice("Cluster does not have watchdog fencing device"); } else { crm_warn("Could not check for watchdog fencing device: %s", pcmk_strerror(rc)); } } else if (list[0] == '\0') { rv = TRUE; } else { GList *targets = stonith__parse_targets(list); rv = pcmk__str_in_list(node, targets, pcmk__str_casei); g_list_free_full(targets, free); } free(list); if (!st) { /* if we're provided the api we still might have done the * connection - but let's assume the caller won't bother */ stonith_api->cmds->disconnect(stonith_api); } } if (!st) { stonith_api_delete(stonith_api); } } else { crm_err("Stonith-API for watchdog-fencing-query couldn't be created."); } crm_trace("Pacemaker assumes node %s %sto do watchdog-fencing.", node, rv?"":"not "); return rv; } gboolean stonith__watchdog_fencing_enabled_for_node(const char *node) { return stonith__watchdog_fencing_enabled_for_node_api(NULL, node); } /* when cycling through the list we don't want to delete items so just mark them and when we know nobody is using the list loop over it to remove the marked items */ static void foreach_notify_entry (stonith_private_t *private, GFunc func, gpointer user_data) { private->notify_refcnt++; g_list_foreach(private->notify_list, func, user_data); private->notify_refcnt--; if ((private->notify_refcnt == 0) && private->notify_deletes) { GList *list_item = private->notify_list; private->notify_deletes = FALSE; while (list_item != NULL) { stonith_notify_client_t *list_client = list_item->data; GList *next = g_list_next(list_item); if (list_client->delete) { free(list_client); private->notify_list = g_list_delete_link(private->notify_list, list_item); } list_item = next; } } } static void stonith_connection_destroy(gpointer user_data) { stonith_t *stonith = user_data; stonith_private_t *native = NULL; struct notify_blob_s blob; crm_trace("Sending destroyed notification"); blob.stonith = stonith; blob.xml = create_xml_node(NULL, "notify"); native = stonith->st_private; native->ipc = NULL; native->source = NULL; free(native->token); native->token = NULL; stonith->state = stonith_disconnected; crm_xml_add(blob.xml, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(blob.xml, F_SUBTYPE, T_STONITH_NOTIFY_DISCONNECT); foreach_notify_entry(native, stonith_send_notification, &blob); free_xml(blob.xml); } xmlNode * create_device_registration_xml(const char *id, enum stonith_namespace namespace, const char *agent, const stonith_key_value_t *params, const char *rsc_provides) { xmlNode *data = create_xml_node(NULL, F_STONITH_DEVICE); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); #if HAVE_STONITH_STONITH_H if (namespace == st_namespace_any) { namespace = stonith_get_namespace(agent, NULL); } if (namespace == st_namespace_lha) { hash2field((gpointer) "plugin", (gpointer) agent, args); agent = "fence_legacy"; } #endif crm_xml_add(data, XML_ATTR_ID, id); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add(data, "agent", agent); if ((namespace != st_namespace_any) && (namespace != st_namespace_invalid)) { crm_xml_add(data, "namespace", stonith_namespace2text(namespace)); } if (rsc_provides) { crm_xml_add(data, "rsc_provides", rsc_provides); } for (; params; params = params->next) { hash2field((gpointer) params->key, (gpointer) params->value, args); } return data; } static int stonith_api_register_device(stonith_t *st, int call_options, const char *id, const char *namespace, const char *agent, const stonith_key_value_t *params) { int rc = 0; xmlNode *data = NULL; data = create_device_registration_xml(id, stonith_text2namespace(namespace), agent, params, NULL); rc = stonith_send_command(st, STONITH_OP_DEVICE_ADD, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_device(stonith_t * st, int call_options, const char *name) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add(data, XML_ATTR_ID, name); rc = stonith_send_command(st, STONITH_OP_DEVICE_DEL, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_level_full(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level) { int rc = 0; xmlNode *data = NULL; CRM_CHECK(node || pattern || (attr && value), return -EINVAL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __func__); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); rc = stonith_send_command(st, STONITH_OP_LEVEL_DEL, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_remove_level(stonith_t * st, int options, const char *node, int level) { return stonith_api_remove_level_full(st, options, node, NULL, NULL, NULL, level); } /*! * \internal * \brief Create XML for fence topology level registration request * * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to register * \param[in] device_list List of devices in level * * \return Newly allocated XML tree on success, NULL otherwise * * \note The caller should set only one of node, pattern or attr/value. */ xmlNode * create_level_registration_xml(const char *node, const char *pattern, const char *attr, const char *value, int level, const stonith_key_value_t *device_list) { GString *list = NULL; xmlNode *data; CRM_CHECK(node || pattern || (attr && value), return NULL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); CRM_CHECK(data, return NULL); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add_int(data, XML_ATTR_ID, level); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } for (; device_list; device_list = device_list->next) { pcmk__add_separated_word(&list, 1024, device_list->value, ","); } if (list != NULL) { crm_xml_add(data, XML_ATTR_STONITH_DEVICES, (const char *) list->str); g_string_free(list, TRUE); } return data; } static int stonith_api_register_level_full(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level, const stonith_key_value_t *device_list) { int rc = 0; xmlNode *data = create_level_registration_xml(node, pattern, attr, value, level, device_list); CRM_CHECK(data != NULL, return -EINVAL); rc = stonith_send_command(st, STONITH_OP_LEVEL_ADD, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_register_level(stonith_t * st, int options, const char *node, int level, const stonith_key_value_t * device_list) { return stonith_api_register_level_full(st, options, node, NULL, NULL, NULL, level, device_list); } static int stonith_api_device_list(stonith_t * stonith, int call_options, const char *namespace, stonith_key_value_t ** devices, int timeout) { int count = 0; enum stonith_namespace ns = stonith_text2namespace(namespace); if (devices == NULL) { crm_err("Parameter error: stonith_api_device_list"); return -EFAULT; } #if HAVE_STONITH_STONITH_H // Include Linux-HA agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_lha)) { count += stonith__list_lha_agents(devices); } #endif // Include Red Hat agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_rhcs)) { count += stonith__list_rhcs_agents(devices); } return count; } // See stonith_api_operations_t:metadata() documentation static int stonith_api_device_metadata(stonith_t *stonith, int call_options, const char *agent, const char *namespace, char **output, int timeout_sec) { /* By executing meta-data directly, we can get it from stonith_admin when * the cluster is not running, which is important for higher-level tools. */ enum stonith_namespace ns = stonith_get_namespace(agent, namespace); if (timeout_sec <= 0) { timeout_sec = CRMD_METADATA_CALL_TIMEOUT; } crm_trace("Looking up metadata for %s agent %s", stonith_namespace2text(ns), agent); switch (ns) { case st_namespace_rhcs: return stonith__rhcs_metadata(agent, timeout_sec, output); #if HAVE_STONITH_STONITH_H case st_namespace_lha: return stonith__lha_metadata(agent, timeout_sec, output); #endif default: crm_err("Can't get fence agent '%s' meta-data: No such agent", agent); break; } return -ENODEV; } static int stonith_api_query(stonith_t * stonith, int call_options, const char *target, stonith_key_value_t ** devices, int timeout) { int rc = 0, lpc = 0, max = 0; xmlNode *data = NULL; xmlNode *output = NULL; xmlXPathObjectPtr xpathObj = NULL; CRM_CHECK(devices != NULL, return -EINVAL); data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add(data, F_STONITH_TARGET, target); crm_xml_add(data, F_STONITH_ACTION, "off"); rc = stonith_send_command(stonith, STONITH_OP_QUERY, data, &output, call_options, timeout); if (rc < 0) { return rc; } xpathObj = xpath_search(output, "//@agent"); if (xpathObj) { max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { xmlChar *match_path = xmlGetNodePath(match); crm_info("%s[%d] = %s", "//@agent", lpc, match_path); free(match_path); *devices = stonith_key_value_add(*devices, NULL, crm_element_value(match, XML_ATTR_ID)); } } freeXpathObject(xpathObj); } free_xml(output); free_xml(data); return max; } /*! * \internal * \brief Make a STONITH_OP_EXEC request * * \param[in,out] stonith Fencer connection * \param[in] call_options Bitmask of \c stonith_call_options * \param[in] id Fence device ID that request is for * \param[in] action Agent action to request (list, status, monitor) * \param[in] target Name of target node for requested action * \param[in] timeout_sec Error if not completed within this many seconds * \param[out] output Where to set agent output */ static int stonith_api_call(stonith_t *stonith, int call_options, const char *id, const char *action, const char *target, int timeout_sec, xmlNode **output) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add(data, F_STONITH_DEVICE, id); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add(data, F_STONITH_TARGET, target); rc = stonith_send_command(stonith, STONITH_OP_EXEC, data, output, call_options, timeout_sec); free_xml(data); return rc; } static int stonith_api_list(stonith_t * stonith, int call_options, const char *id, char **list_info, int timeout) { int rc; xmlNode *output = NULL; rc = stonith_api_call(stonith, call_options, id, "list", NULL, timeout, &output); if (output && list_info) { const char *list_str; list_str = crm_element_value(output, F_STONITH_OUTPUT); if (list_str) { *list_info = strdup(list_str); } } if (output) { free_xml(output); } return rc; } static int stonith_api_monitor(stonith_t * stonith, int call_options, const char *id, int timeout) { return stonith_api_call(stonith, call_options, id, "monitor", NULL, timeout, NULL); } static int stonith_api_status(stonith_t * stonith, int call_options, const char *id, const char *port, int timeout) { return stonith_api_call(stonith, call_options, id, "status", port, timeout, NULL); } static int stonith_api_fence_with_delay(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance, int delay) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, __func__); crm_xml_add(data, F_STONITH_TARGET, node); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add_int(data, F_STONITH_TIMEOUT, timeout); crm_xml_add_int(data, F_STONITH_TOLERANCE, tolerance); crm_xml_add_int(data, F_STONITH_DELAY, delay); rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout); free_xml(data); return rc; } static int stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance) { return stonith_api_fence_with_delay(stonith, call_options, node, action, timeout, tolerance, 0); } static int stonith_api_confirm(stonith_t * stonith, int call_options, const char *target) { stonith__set_call_options(call_options, target, st_opt_manual_ack); return stonith_api_fence(stonith, call_options, target, "off", 0, 0); } static int stonith_api_history(stonith_t * stonith, int call_options, const char *node, stonith_history_t ** history, int timeout) { int rc = 0; xmlNode *data = NULL; xmlNode *output = NULL; stonith_history_t *last = NULL; *history = NULL; if (node) { data = create_xml_node(NULL, __func__); crm_xml_add(data, F_STONITH_TARGET, node); } stonith__set_call_options(call_options, node, st_opt_sync_call); rc = stonith_send_command(stonith, STONITH_OP_FENCE_HISTORY, data, &output, call_options, timeout); free_xml(data); if (rc == 0) { xmlNode *op = NULL; xmlNode *reply = get_xpath_object("//" F_STONITH_HISTORY_LIST, output, LOG_NEVER); for (op = pcmk__xml_first_child(reply); op != NULL; op = pcmk__xml_next(op)) { stonith_history_t *kvp; long long completed; long long completed_nsec = 0L; kvp = calloc(1, sizeof(stonith_history_t)); kvp->target = crm_element_value_copy(op, F_STONITH_TARGET); kvp->action = crm_element_value_copy(op, F_STONITH_ACTION); kvp->origin = crm_element_value_copy(op, F_STONITH_ORIGIN); kvp->delegate = crm_element_value_copy(op, F_STONITH_DELEGATE); kvp->client = crm_element_value_copy(op, F_STONITH_CLIENTNAME); crm_element_value_ll(op, F_STONITH_DATE, &completed); kvp->completed = (time_t) completed; crm_element_value_ll(op, F_STONITH_DATE_NSEC, &completed_nsec); kvp->completed_nsec = completed_nsec; crm_element_value_int(op, F_STONITH_STATE, &kvp->state); kvp->exit_reason = crm_element_value_copy(op, XML_LRM_ATTR_EXIT_REASON); if (last) { last->next = kvp; } else { *history = kvp; } last = kvp; } } free_xml(output); return rc; } void stonith_history_free(stonith_history_t *history) { stonith_history_t *hp, *hp_old; for (hp = history; hp; hp_old = hp, hp = hp->next, free(hp_old)) { free(hp->target); free(hp->action); free(hp->origin); free(hp->delegate); free(hp->client); free(hp->exit_reason); } } static gint stonithlib_GCompareFunc(gconstpointer a, gconstpointer b) { int rc = 0; const stonith_notify_client_t *a_client = a; const stonith_notify_client_t *b_client = b; if (a_client->delete || b_client->delete) { /* make entries marked for deletion not findable */ return -1; } CRM_CHECK(a_client->event != NULL && b_client->event != NULL, return 0); rc = strcmp(a_client->event, b_client->event); if (rc == 0) { if (a_client->notify == NULL || b_client->notify == NULL) { return 0; } else if (a_client->notify == b_client->notify) { return 0; } else if (((long)a_client->notify) < ((long)b_client->notify)) { crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return -1; } crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return 1; } return rc; } xmlNode * stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options) { xmlNode *op_msg = create_xml_node(NULL, "stonith_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "stonith_command"); crm_xml_add(op_msg, F_TYPE, T_STONITH_NG); crm_xml_add(op_msg, F_STONITH_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_STONITH_OPERATION, op); crm_xml_add_int(op_msg, F_STONITH_CALLID, call_id); crm_trace("Sending call options: %.8lx, %d", (long)call_options, call_options); crm_xml_add_int(op_msg, F_STONITH_CALLOPTS, call_options); if (data != NULL) { add_message_xml(op_msg, F_STONITH_CALLDATA, data); } return op_msg; } static void stonith_destroy_op_callback(gpointer data) { stonith_callback_client_t *blob = data; if (blob->timer && blob->timer->ref > 0) { g_source_remove(blob->timer->ref); } free(blob->timer); free(blob); } static int stonith_api_signoff(stonith_t * stonith) { stonith_private_t *native = stonith->st_private; crm_debug("Disconnecting from the fencer"); if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } free(native->token); native->token = NULL; stonith->state = stonith_disconnected; return pcmk_ok; } static int stonith_api_del_callback(stonith_t * stonith, int call_id, bool all_callbacks) { stonith_private_t *private = stonith->st_private; if (all_callbacks) { private->op_callback = NULL; g_hash_table_destroy(private->stonith_op_callback_table); private->stonith_op_callback_table = pcmk__intkey_table(stonith_destroy_op_callback); } else if (call_id == 0) { private->op_callback = NULL; } else { pcmk__intkey_table_remove(private->stonith_op_callback_table, call_id); } return pcmk_ok; } /*! * \internal * \brief Invoke a (single) specified fence action callback * * \param[in,out] st Fencer API connection * \param[in] call_id If positive, call ID of completed fence action, * otherwise legacy return code for early failure * \param[in,out] result Full result for action * \param[in,out] userdata User data to pass to callback * \param[in] callback Fence action callback to invoke */ static void invoke_fence_action_callback(stonith_t *st, int call_id, pcmk__action_result_t *result, void *userdata, void (*callback) (stonith_t *st, stonith_callback_data_t *data)) { stonith_callback_data_t data = { 0, }; data.call_id = call_id; data.rc = pcmk_rc2legacy(stonith__result2rc(result)); data.userdata = userdata; data.opaque = (void *) result; callback(st, &data); } /*! * \internal * \brief Invoke any callbacks registered for a specified fence action result * * Given a fence action result from the fencer, invoke any callback registered * for that action, as well as any global callback registered. * * \param[in,out] stonith Fencer API connection * \param[in] msg If non-NULL, fencer reply * \param[in] call_id If \p msg is NULL, call ID of action that timed out */ static void invoke_registered_callbacks(stonith_t *stonith, const xmlNode *msg, int call_id) { stonith_private_t *private = NULL; stonith_callback_client_t *cb_info = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; CRM_CHECK(stonith != NULL, return); CRM_CHECK(stonith->st_private != NULL, return); private = stonith->st_private; if (msg == NULL) { // Fencer didn't reply in time pcmk__set_result(&result, CRM_EX_ERROR, PCMK_EXEC_TIMEOUT, "Fencer accepted request but did not reply in time"); CRM_LOG_ASSERT(call_id > 0); } else { // We have the fencer reply if ((crm_element_value_int(msg, F_STONITH_CALLID, &call_id) != 0) || (call_id <= 0)) { crm_log_xml_warn(msg, "Bad fencer reply"); } stonith__xe_get_result(msg, &result); } if (call_id > 0) { cb_info = pcmk__intkey_table_lookup(private->stonith_op_callback_table, call_id); } if ((cb_info != NULL) && (cb_info->callback != NULL) && (pcmk__result_ok(&result) || !(cb_info->only_success))) { crm_trace("Invoking callback %s for call %d", pcmk__s(cb_info->id, "without ID"), call_id); invoke_fence_action_callback(stonith, call_id, &result, cb_info->user_data, cb_info->callback); } else if ((private->op_callback == NULL) && !pcmk__result_ok(&result)) { crm_warn("Fencing action without registered callback failed: %d (%s%s%s)", result.exit_status, pcmk_exec_status_str(result.execution_status), ((result.exit_reason == NULL)? "" : ": "), ((result.exit_reason == NULL)? "" : result.exit_reason)); crm_log_xml_debug(msg, "Failed fence update"); } if (private->op_callback != NULL) { crm_trace("Invoking global callback for call %d", call_id); invoke_fence_action_callback(stonith, call_id, &result, NULL, private->op_callback); } if (cb_info != NULL) { stonith_api_del_callback(stonith, call_id, FALSE); } pcmk__reset_result(&result); } static gboolean stonith_async_timeout_handler(gpointer data) { struct timer_rec_s *timer = data; crm_err("Async call %d timed out after %dms", timer->call_id, timer->timeout); invoke_registered_callbacks(timer->stonith, NULL, timer->call_id); /* Always return TRUE, never remove the handler * We do that in stonith_del_callback() */ return TRUE; } static void set_callback_timeout(stonith_callback_client_t * callback, stonith_t * stonith, int call_id, int timeout) { struct timer_rec_s *async_timer = callback->timer; if (timeout <= 0) { return; } if (!async_timer) { async_timer = calloc(1, sizeof(struct timer_rec_s)); callback->timer = async_timer; } async_timer->stonith = stonith; async_timer->call_id = call_id; /* Allow a fair bit of grace to allow the server to tell us of a timeout * This is only a fallback */ async_timer->timeout = (timeout + 60) * 1000; if (async_timer->ref) { g_source_remove(async_timer->ref); } async_timer->ref = g_timeout_add(async_timer->timeout, stonith_async_timeout_handler, async_timer); } static void update_callback_timeout(int call_id, int timeout, stonith_t * st) { stonith_callback_client_t *callback = NULL; stonith_private_t *private = st->st_private; callback = pcmk__intkey_table_lookup(private->stonith_op_callback_table, call_id); if (!callback || !callback->allow_timeout_updates) { return; } set_callback_timeout(callback, st, call_id, timeout); } static int stonith_dispatch_internal(const char *buffer, ssize_t length, gpointer userdata) { const char *type = NULL; struct notify_blob_s blob; stonith_t *st = userdata; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; blob.stonith = st; blob.xml = string2xml(buffer); if (blob.xml == NULL) { crm_warn("Received malformed message from fencer: %s", buffer); return 0; } /* do callbacks */ type = crm_element_value(blob.xml, F_TYPE); crm_trace("Activating %s callbacks...", type); if (pcmk__str_eq(type, T_STONITH_NG, pcmk__str_none)) { invoke_registered_callbacks(st, blob.xml, 0); } else if (pcmk__str_eq(type, T_STONITH_NOTIFY, pcmk__str_none)) { foreach_notify_entry(private, stonith_send_notification, &blob); } else if (pcmk__str_eq(type, T_STONITH_TIMEOUT_VALUE, pcmk__str_none)) { int call_id = 0; int timeout = 0; crm_element_value_int(blob.xml, F_STONITH_TIMEOUT, &timeout); crm_element_value_int(blob.xml, F_STONITH_CALLID, &call_id); update_callback_timeout(call_id, timeout, st); } else { crm_err("Unknown message type: %s", type); crm_log_xml_warn(blob.xml, "BadReply"); } free_xml(blob.xml); return 1; } static int stonith_api_signon(stonith_t * stonith, const char *name, int *stonith_fd) { int rc = pcmk_ok; stonith_private_t *native = NULL; const char *display_name = name? name : "client"; struct ipc_client_callbacks st_callbacks = { .dispatch = stonith_dispatch_internal, .destroy = stonith_connection_destroy }; CRM_CHECK(stonith != NULL, return -EINVAL); native = stonith->st_private; CRM_ASSERT(native != NULL); crm_debug("Attempting fencer connection by %s with%s mainloop", display_name, (stonith_fd? "out" : "")); stonith->state = stonith_connected_command; if (stonith_fd) { /* No mainloop */ native->ipc = crm_ipc_new("stonith-ng", 0); - - if (native->ipc && crm_ipc_connect(native->ipc)) { - *stonith_fd = crm_ipc_get_fd(native->ipc); - } else if (native->ipc) { - crm_ipc_close(native->ipc); - crm_ipc_destroy(native->ipc); - native->ipc = NULL; + if (native->ipc != NULL) { + rc = pcmk__connect_generic_ipc(native->ipc); + if (rc == pcmk_rc_ok) { + rc = pcmk__ipc_fd(native->ipc, stonith_fd); + if (rc != pcmk_rc_ok) { + crm_debug("Couldn't get file descriptor for IPC: %s", + pcmk_rc_str(rc)); + } + } + if (rc != pcmk_rc_ok) { + crm_ipc_close(native->ipc); + crm_ipc_destroy(native->ipc); + native->ipc = NULL; + } } } else { /* With mainloop */ native->source = mainloop_add_ipc_client("stonith-ng", G_PRIORITY_MEDIUM, 0, stonith, &st_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { rc = -ENOTCONN; } else { xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "stonith_command"); crm_xml_add(hello, F_TYPE, T_STONITH_NG); crm_xml_add(hello, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_STONITH_CLIENTNAME, name); rc = crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply); if (rc < 0) { crm_debug("Couldn't register with the fencer: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); rc = -ECOMM; } else if (reply == NULL) { crm_debug("Couldn't register with the fencer: no reply"); rc = -EPROTO; } else { const char *msg_type = crm_element_value(reply, F_STONITH_OPERATION); native->token = crm_element_value_copy(reply, F_STONITH_CLIENTID); if (!pcmk__str_eq(msg_type, CRM_OP_REGISTER, pcmk__str_none)) { crm_debug("Couldn't register with the fencer: invalid reply type '%s'", (msg_type? msg_type : "(missing)")); crm_log_xml_debug(reply, "Invalid fencer reply"); rc = -EPROTO; } else if (native->token == NULL) { crm_debug("Couldn't register with the fencer: no token in reply"); crm_log_xml_debug(reply, "Invalid fencer reply"); rc = -EPROTO; } else { crm_debug("Connection to fencer by %s succeeded (registration token: %s)", display_name, native->token); rc = pcmk_ok; } } free_xml(reply); free_xml(hello); } if (rc != pcmk_ok) { crm_debug("Connection attempt to fencer by %s failed: %s " CRM_XS " rc=%d", display_name, pcmk_strerror(rc), rc); stonith->cmds->disconnect(stonith); } return rc; } static int stonith_set_notification(stonith_t * stonith, const char *callback, int enabled) { int rc = pcmk_ok; xmlNode *notify_msg = create_xml_node(NULL, __func__); stonith_private_t *native = stonith->st_private; if (stonith->state != stonith_disconnected) { crm_xml_add(notify_msg, F_STONITH_OPERATION, T_STONITH_NOTIFY); if (enabled) { crm_xml_add(notify_msg, F_STONITH_NOTIFY_ACTIVATE, callback); } else { crm_xml_add(notify_msg, F_STONITH_NOTIFY_DEACTIVATE, callback); } rc = crm_ipc_send(native->ipc, notify_msg, crm_ipc_client_response, -1, NULL); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't register for fencing notifications: %d", rc); rc = -ECOMM; } else { rc = pcmk_ok; } } free_xml(notify_msg); return rc; } static int stonith_api_add_notification(stonith_t * stonith, const char *event, void (*callback) (stonith_t * stonith, stonith_event_t * e)) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = NULL; private = stonith->st_private; crm_trace("Adding callback for %s events (%d)", event, g_list_length(private->notify_list)); new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = callback; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); if (list_item != NULL) { crm_warn("Callback already present"); free(new_client); return -ENOTUNIQ; } else { private->notify_list = g_list_append(private->notify_list, new_client); stonith_set_notification(stonith, event, 1); crm_trace("Callback added (%d)", g_list_length(private->notify_list)); } return pcmk_ok; } static void del_notify_entry(gpointer data, gpointer user_data) { stonith_notify_client_t *entry = data; stonith_t * stonith = user_data; if (!entry->delete) { crm_debug("Removing callback for %s events", entry->event); stonith_api_del_notification(stonith, entry->event); } } static int stonith_api_del_notification(stonith_t * stonith, const char *event) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = stonith->st_private; if (event == NULL) { foreach_notify_entry(private, del_notify_entry, stonith); crm_trace("Removed callback"); return pcmk_ok; } crm_debug("Removing callback for %s events", event); new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = NULL; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); stonith_set_notification(stonith, event, 0); if (list_item != NULL) { stonith_notify_client_t *list_client = list_item->data; if (private->notify_refcnt) { list_client->delete = TRUE; private->notify_deletes = TRUE; } else { private->notify_list = g_list_remove(private->notify_list, list_client); free(list_client); } crm_trace("Removed callback"); } else { crm_trace("Callback not present"); } free(new_client); return pcmk_ok; } static int stonith_api_add_callback(stonith_t * stonith, int call_id, int timeout, int options, void *user_data, const char *callback_name, void (*callback) (stonith_t * st, stonith_callback_data_t * data)) { stonith_callback_client_t *blob = NULL; stonith_private_t *private = NULL; CRM_CHECK(stonith != NULL, return -EINVAL); CRM_CHECK(stonith->st_private != NULL, return -EINVAL); private = stonith->st_private; if (call_id == 0) { // Add global callback private->op_callback = callback; } else if (call_id < 0) { // Call failed immediately, so call callback now if (!(options & st_opt_report_only_success)) { pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; crm_trace("Call failed, calling %s: %s", callback_name, pcmk_strerror(call_id)); pcmk__set_result(&result, CRM_EX_ERROR, stonith__legacy2status(call_id), NULL); invoke_fence_action_callback(stonith, call_id, &result, user_data, callback); } else { crm_warn("Fencer call failed: %s", pcmk_strerror(call_id)); } return FALSE; } blob = calloc(1, sizeof(stonith_callback_client_t)); blob->id = callback_name; blob->only_success = (options & st_opt_report_only_success) ? TRUE : FALSE; blob->user_data = user_data; blob->callback = callback; blob->allow_timeout_updates = (options & st_opt_timeout_updates) ? TRUE : FALSE; if (timeout > 0) { set_callback_timeout(blob, stonith, call_id, timeout); } pcmk__intkey_table_insert(private->stonith_op_callback_table, call_id, blob); crm_trace("Added callback to %s for call %d", callback_name, call_id); return TRUE; } static void stonith_dump_pending_op(gpointer key, gpointer value, gpointer user_data) { int call = GPOINTER_TO_INT(key); stonith_callback_client_t *blob = value; crm_debug("Call %d (%s): pending", call, pcmk__s(blob->id, "no ID")); } void stonith_dump_pending_callbacks(stonith_t * stonith) { stonith_private_t *private = stonith->st_private; if (private->stonith_op_callback_table == NULL) { return; } return g_hash_table_foreach(private->stonith_op_callback_table, stonith_dump_pending_op, NULL); } /*! * \internal * \brief Get the data section of a fencer notification * * \param[in] msg Notification XML * \param[in] ntype Notification type */ static xmlNode * get_event_data_xml(xmlNode *msg, const char *ntype) { char *data_addr = crm_strdup_printf("//%s", ntype); xmlNode *data = get_xpath_object(data_addr, msg, LOG_DEBUG); free(data_addr); return data; } /* */ static stonith_event_t * xml_to_event(xmlNode *msg) { stonith_event_t *event = calloc(1, sizeof(stonith_event_t)); struct event_private *event_private = NULL; CRM_ASSERT(event != NULL); event->opaque = calloc(1, sizeof(struct event_private)); CRM_ASSERT(event->opaque != NULL); event_private = (struct event_private *) event->opaque; crm_log_xml_trace(msg, "stonith_notify"); // All notification types have the operation result and notification subtype stonith__xe_get_result(msg, &event_private->result); event->operation = crm_element_value_copy(msg, F_STONITH_OPERATION); // @COMPAT The API originally provided the result as a legacy return code event->result = pcmk_rc2legacy(stonith__result2rc(&event_private->result)); // Some notification subtypes have additional information if (pcmk__str_eq(event->operation, T_STONITH_NOTIFY_FENCE, pcmk__str_none)) { xmlNode *data = get_event_data_xml(msg, event->operation); if (data == NULL) { crm_err("No data for %s event", event->operation); crm_log_xml_notice(msg, "BadEvent"); } else { event->origin = crm_element_value_copy(data, F_STONITH_ORIGIN); event->action = crm_element_value_copy(data, F_STONITH_ACTION); event->target = crm_element_value_copy(data, F_STONITH_TARGET); event->executioner = crm_element_value_copy(data, F_STONITH_DELEGATE); event->id = crm_element_value_copy(data, F_STONITH_REMOTE_OP_ID); event->client_origin = crm_element_value_copy(data, F_STONITH_CLIENTNAME); event->device = crm_element_value_copy(data, F_STONITH_DEVICE); } } else if (pcmk__str_any_of(event->operation, STONITH_OP_DEVICE_ADD, STONITH_OP_DEVICE_DEL, STONITH_OP_LEVEL_ADD, STONITH_OP_LEVEL_DEL, NULL)) { xmlNode *data = get_event_data_xml(msg, event->operation); if (data == NULL) { crm_err("No data for %s event", event->operation); crm_log_xml_notice(msg, "BadEvent"); } else { event->device = crm_element_value_copy(data, F_STONITH_DEVICE); } } return event; } static void event_free(stonith_event_t * event) { struct event_private *event_private = event->opaque; free(event->id); free(event->type); free(event->message); free(event->operation); free(event->origin); free(event->action); free(event->target); free(event->executioner); free(event->device); free(event->client_origin); pcmk__reset_result(&event_private->result); free(event->opaque); free(event); } static void stonith_send_notification(gpointer data, gpointer user_data) { struct notify_blob_s *blob = user_data; stonith_notify_client_t *entry = data; stonith_event_t *st_event = NULL; const char *event = NULL; if (blob->xml == NULL) { crm_warn("Skipping callback - NULL message"); return; } event = crm_element_value(blob->xml, F_SUBTYPE); if (entry == NULL) { crm_warn("Skipping callback - NULL callback client"); return; } else if (entry->delete) { crm_trace("Skipping callback - marked for deletion"); return; } else if (entry->notify == NULL) { crm_warn("Skipping callback - NULL callback"); return; } else if (!pcmk__str_eq(entry->event, event, pcmk__str_none)) { crm_trace("Skipping callback - event mismatch %p/%s vs. %s", entry, entry->event, event); return; } st_event = xml_to_event(blob->xml); crm_trace("Invoking callback for %p/%s event...", entry, event); entry->notify(blob->stonith, st_event); crm_trace("Callback invoked..."); event_free(st_event); } /*! * \internal * \brief Create and send an API request * * \param[in,out] stonith Stonith connection * \param[in] op API operation to request * \param[in] data Data to attach to request * \param[out] output_data If not NULL, will be set to reply if synchronous * \param[in] call_options Bitmask of stonith_call_options to use * \param[in] timeout Error if not completed within this many seconds * * \return pcmk_ok (for synchronous requests) or positive call ID * (for asynchronous requests) on success, -errno otherwise */ static int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout) { int rc = 0; int reply_id = -1; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; stonith_private_t *native = NULL; CRM_ASSERT(stonith && stonith->st_private && op); native = stonith->st_private; if (output_data != NULL) { *output_data = NULL; } if ((stonith->state == stonith_disconnected) || (native->token == NULL)) { return -ENOTCONN; } /* Increment the call ID, which must be positive to avoid conflicting with * error codes. This shouldn't be a problem unless the client mucked with * it or the counter wrapped around. */ stonith->call_id++; if (stonith->call_id < 1) { stonith->call_id = 1; } op_msg = stonith_create_op(stonith->call_id, native->token, op, data, call_options); if (op_msg == NULL) { return -EINVAL; } crm_xml_add_int(op_msg, F_STONITH_TIMEOUT, timeout); crm_trace("Sending %s message to fencer with timeout %ds", op, timeout); if (data) { const char *delay_s = crm_element_value(data, F_STONITH_DELAY); if (delay_s) { crm_xml_add(op_msg, F_STONITH_DELAY, delay_s); } } { enum crm_ipc_flags ipc_flags = crm_ipc_flags_none; if (call_options & st_opt_sync_call) { pcmk__set_ipc_flags(ipc_flags, "stonith command", crm_ipc_client_response); } rc = crm_ipc_send(native->ipc, op_msg, ipc_flags, 1000 * (timeout + 60), &op_reply); } free_xml(op_msg); if (rc < 0) { crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%ds): %d", op, timeout, rc); rc = -ECOMM; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (!(call_options & st_opt_sync_call)) { crm_trace("Async call %d, returning", stonith->call_id); free_xml(op_reply); return stonith->call_id; } crm_element_value_int(op_reply, F_STONITH_CALLID, &reply_id); if (reply_id == stonith->call_id) { pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; crm_trace("Synchronous reply %d received", reply_id); stonith__xe_get_result(op_reply, &result); rc = pcmk_rc2legacy(stonith__result2rc(&result)); pcmk__reset_result(&result); if ((call_options & st_opt_discard_reply) || output_data == NULL) { crm_trace("Discarding reply"); } else { *output_data = op_reply; op_reply = NULL; /* Prevent subsequent free */ } } else if (reply_id <= 0) { crm_err("Received bad reply: No id set"); crm_log_xml_err(op_reply, "Bad reply"); free_xml(op_reply); rc = -ENOMSG; } else { crm_err("Received bad reply: %d (wanted %d)", reply_id, stonith->call_id); crm_log_xml_err(op_reply, "Old reply"); free_xml(op_reply); rc = -ENOMSG; } done: if (!crm_ipc_connected(native->ipc)) { crm_err("Fencer disconnected"); free(native->token); native->token = NULL; stonith->state = stonith_disconnected; } free_xml(op_reply); return rc; } /* Not used with mainloop */ bool stonith_dispatch(stonith_t * st) { gboolean stay_connected = TRUE; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; while (crm_ipc_ready(private->ipc)) { if (crm_ipc_read(private->ipc) > 0) { const char *msg = crm_ipc_buffer(private->ipc); stonith_dispatch_internal(msg, strlen(msg), st); } if (!crm_ipc_connected(private->ipc)) { crm_err("Connection closed"); stay_connected = FALSE; } } return stay_connected; } static int stonith_api_free(stonith_t * stonith) { int rc = pcmk_ok; crm_trace("Destroying %p", stonith); if (stonith->state != stonith_disconnected) { crm_trace("Unregistering notifications and disconnecting %p first", stonith); stonith->cmds->remove_notification(stonith, NULL); rc = stonith->cmds->disconnect(stonith); } if (stonith->state == stonith_disconnected) { stonith_private_t *private = stonith->st_private; crm_trace("Removing %d callbacks", g_hash_table_size(private->stonith_op_callback_table)); g_hash_table_destroy(private->stonith_op_callback_table); crm_trace("Destroying %d notification clients", g_list_length(private->notify_list)); g_list_free_full(private->notify_list, free); free(stonith->st_private); free(stonith->cmds); free(stonith); } else { crm_err("Not free'ing active connection: %s (%d)", pcmk_strerror(rc), rc); } return rc; } void stonith_api_delete(stonith_t * stonith) { crm_trace("Destroying %p", stonith); if(stonith) { stonith->cmds->free(stonith); } } static int stonith_api_validate(stonith_t *st, int call_options, const char *rsc_id, const char *namespace_s, const char *agent, const stonith_key_value_t *params, int timeout_sec, char **output, char **error_output) { /* Validation should be done directly via the agent, so we can get it from * stonith_admin when the cluster is not running, which is important for * higher-level tools. */ int rc = pcmk_ok; /* Use a dummy node name in case the agent requires a target. We assume the * actual target doesn't matter for validation purposes (if in practice, * that is incorrect, we will need to allow the caller to pass the target). */ const char *target = "node1"; const char *host_arg = NULL; GHashTable *params_table = pcmk__strkey_table(free, free); // Convert parameter list to a hash table for (; params; params = params->next) { if (pcmk__str_eq(params->key, PCMK_STONITH_HOST_ARGUMENT, pcmk__str_none)) { host_arg = params->value; } if (!pcmk_stonith_param(params->key)) { g_hash_table_insert(params_table, strdup(params->key), strdup(params->value)); } } #if SUPPORT_CIBSECRETS rc = pcmk__substitute_secrets(rsc_id, params_table); if (rc != pcmk_rc_ok) { crm_warn("Could not replace secret parameters for validation of %s: %s", agent, pcmk_rc_str(rc)); // rc is standard return value, don't return it in this function } #endif if (output) { *output = NULL; } if (error_output) { *error_output = NULL; } if (timeout_sec <= 0) { timeout_sec = CRMD_METADATA_CALL_TIMEOUT; // Questionable } switch (stonith_get_namespace(agent, namespace_s)) { case st_namespace_rhcs: rc = stonith__rhcs_validate(st, call_options, target, agent, params_table, host_arg, timeout_sec, output, error_output); break; #if HAVE_STONITH_STONITH_H case st_namespace_lha: rc = stonith__lha_validate(st, call_options, target, agent, params_table, timeout_sec, output, error_output); break; #endif case st_namespace_invalid: errno = ENOENT; rc = -errno; if (error_output) { *error_output = crm_strdup_printf("Agent %s not found", agent); } else { crm_err("Agent %s not found", agent); } break; default: errno = EOPNOTSUPP; rc = -errno; if (error_output) { *error_output = crm_strdup_printf("Agent %s does not support validation", agent); } else { crm_err("Agent %s does not support validation", agent); } break; } g_hash_table_destroy(params_table); return rc; } stonith_t * stonith_api_new(void) { stonith_t *new_stonith = NULL; stonith_private_t *private = NULL; new_stonith = calloc(1, sizeof(stonith_t)); if (new_stonith == NULL) { return NULL; } private = calloc(1, sizeof(stonith_private_t)); if (private == NULL) { free(new_stonith); return NULL; } new_stonith->st_private = private; private->stonith_op_callback_table = pcmk__intkey_table(stonith_destroy_op_callback); private->notify_list = NULL; private->notify_refcnt = 0; private->notify_deletes = FALSE; new_stonith->call_id = 1; new_stonith->state = stonith_disconnected; new_stonith->cmds = calloc(1, sizeof(stonith_api_operations_t)); if (new_stonith->cmds == NULL) { free(new_stonith->st_private); free(new_stonith); return NULL; } /* *INDENT-OFF* */ new_stonith->cmds->free = stonith_api_free; new_stonith->cmds->connect = stonith_api_signon; new_stonith->cmds->disconnect = stonith_api_signoff; new_stonith->cmds->list = stonith_api_list; new_stonith->cmds->monitor = stonith_api_monitor; new_stonith->cmds->status = stonith_api_status; new_stonith->cmds->fence = stonith_api_fence; new_stonith->cmds->fence_with_delay = stonith_api_fence_with_delay; new_stonith->cmds->confirm = stonith_api_confirm; new_stonith->cmds->history = stonith_api_history; new_stonith->cmds->list_agents = stonith_api_device_list; new_stonith->cmds->metadata = stonith_api_device_metadata; new_stonith->cmds->query = stonith_api_query; new_stonith->cmds->remove_device = stonith_api_remove_device; new_stonith->cmds->register_device = stonith_api_register_device; new_stonith->cmds->remove_level = stonith_api_remove_level; new_stonith->cmds->remove_level_full = stonith_api_remove_level_full; new_stonith->cmds->register_level = stonith_api_register_level; new_stonith->cmds->register_level_full = stonith_api_register_level_full; new_stonith->cmds->remove_callback = stonith_api_del_callback; new_stonith->cmds->register_callback = stonith_api_add_callback; new_stonith->cmds->remove_notification = stonith_api_del_notification; new_stonith->cmds->register_notification = stonith_api_add_notification; new_stonith->cmds->validate = stonith_api_validate; /* *INDENT-ON* */ return new_stonith; } /*! * \brief Make a blocking connection attempt to the fencer * * \param[in,out] st Fencer API object * \param[in] name Client name to use with fencer * \param[in] max_attempts Return error if this many attempts fail * * \return pcmk_ok on success, result of last attempt otherwise */ int stonith_api_connect_retry(stonith_t *st, const char *name, int max_attempts) { int rc = -EINVAL; // if max_attempts is not positive for (int attempt = 1; attempt <= max_attempts; attempt++) { rc = st->cmds->connect(st, name, NULL); if (rc == pcmk_ok) { return pcmk_ok; } else if (attempt < max_attempts) { crm_notice("Fencer connection attempt %d of %d failed (retrying in 2s): %s " CRM_XS " rc=%d", attempt, max_attempts, pcmk_strerror(rc), rc); sleep(2); } } crm_notice("Could not connect to fencer: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); return rc; } stonith_key_value_t * stonith_key_value_add(stonith_key_value_t * head, const char *key, const char *value) { stonith_key_value_t *p, *end; p = calloc(1, sizeof(stonith_key_value_t)); pcmk__str_update(&p->key, key); pcmk__str_update(&p->value, value); end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void stonith_key_value_freeall(stonith_key_value_t * head, int keys, int values) { stonith_key_value_t *p; while (head) { p = head->next; if (keys) { free(head->key); } if (values) { free(head->value); } free(head); head = p; } } #define api_log_open() openlog("stonith-api", LOG_CONS | LOG_NDELAY | LOG_PID, LOG_DAEMON) #define api_log(level, fmt, args...) syslog(level, "%s: "fmt, __func__, args) int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off) { int rc = pcmk_ok; stonith_t *st = stonith_api_new(); const char *action = off? "off" : "reboot"; api_log_open(); if (st == NULL) { api_log(LOG_ERR, "API initialization failed, could not kick (%s) node %u/%s", action, nodeid, uname); return -EPROTO; } rc = st->cmds->connect(st, "stonith-api", NULL); if (rc != pcmk_ok) { api_log(LOG_ERR, "Connection failed, could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); } else { char *name = (uname == NULL)? pcmk__itoa(nodeid) : strdup(uname); int opts = 0; stonith__set_call_options(opts, name, st_opt_sync_call|st_opt_allow_suicide); if ((uname == NULL) && (nodeid > 0)) { stonith__set_call_options(opts, name, st_opt_cs_nodeid); } rc = st->cmds->fence(st, opts, name, action, timeout, 0); free(name); if (rc != pcmk_ok) { api_log(LOG_ERR, "Could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); } else { api_log(LOG_NOTICE, "Node %u/%s kicked: %s", nodeid, uname, action); } } stonith_api_delete(st); return rc; } time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress) { int rc = pcmk_ok; time_t when = 0; stonith_t *st = stonith_api_new(); stonith_history_t *history = NULL, *hp = NULL; if (st == NULL) { api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: " "API initialization failed", nodeid, uname); return when; } rc = st->cmds->connect(st, "stonith-api", NULL); if (rc != pcmk_ok) { api_log(LOG_NOTICE, "Connection failed: %s (%d)", pcmk_strerror(rc), rc); } else { int entries = 0; int progress = 0; int completed = 0; int opts = 0; char *name = (uname == NULL)? pcmk__itoa(nodeid) : strdup(uname); stonith__set_call_options(opts, name, st_opt_sync_call); if ((uname == NULL) && (nodeid > 0)) { stonith__set_call_options(opts, name, st_opt_cs_nodeid); } rc = st->cmds->history(st, opts, name, &history, 120); free(name); for (hp = history; hp; hp = hp->next) { entries++; if (in_progress) { progress++; if (hp->state != st_done && hp->state != st_failed) { when = time(NULL); } } else if (hp->state == st_done) { completed++; if (hp->completed > when) { when = hp->completed; } } } stonith_history_free(history); if(rc == pcmk_ok) { api_log(LOG_INFO, "Found %d entries for %u/%s: %d in progress, %d completed", entries, nodeid, uname, progress, completed); } else { api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: %s (%d)", nodeid, uname, pcmk_strerror(rc), rc); } } stonith_api_delete(st); if(when) { api_log(LOG_INFO, "Node %u/%s last kicked at: %ld", nodeid, uname, (long int)when); } return when; } bool stonith_agent_exists(const char *agent, int timeout) { stonith_t *st = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; bool rc = FALSE; if (agent == NULL) { return rc; } st = stonith_api_new(); if (st == NULL) { crm_err("Could not list fence agents: API memory allocation failed"); return FALSE; } st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout == 0 ? 120 : timeout); for (dIter = devices; dIter != NULL; dIter = dIter->next) { if (pcmk__str_eq(dIter->value, agent, pcmk__str_none)) { rc = TRUE; break; } } stonith_key_value_freeall(devices, 1, 1); stonith_api_delete(st); return rc; } const char * stonith_action_str(const char *action) { if (action == NULL) { return "fencing"; } else if (!strcmp(action, "on")) { return "unfencing"; } else if (!strcmp(action, "off")) { return "turning off"; } else { return action; } } /*! * \internal * \brief Parse a target name from one line of a target list string * * \param[in] line One line of a target list string * \param[in] len String length of line * \param[in,out] output List to add newly allocated target name to */ static void parse_list_line(const char *line, int len, GList **output) { size_t i = 0; size_t entry_start = 0; /* Skip complaints about additional parameters device doesn't understand * * @TODO Document or eliminate the implied restriction of target names */ if (strstr(line, "invalid") || strstr(line, "variable")) { crm_debug("Skipping list output line: %s", line); return; } // Process line content, character by character for (i = 0; i <= len; i++) { if (isspace(line[i]) || (line[i] == ',') || (line[i] == ';') || (line[i] == '\0')) { // We've found a separator (i.e. the end of an entry) int rc = 0; char *entry = NULL; if (i == entry_start) { // Skip leading and sequential separators entry_start = i + 1; continue; } entry = calloc(i - entry_start + 1, sizeof(char)); CRM_ASSERT(entry != NULL); /* Read entry, stopping at first separator * * @TODO Document or eliminate these character restrictions */ rc = sscanf(line + entry_start, "%[a-zA-Z0-9_-.]", entry); if (rc != 1) { crm_warn("Could not parse list output entry: %s " CRM_XS " entry_start=%d position=%d", line + entry_start, entry_start, i); free(entry); } else if (pcmk__strcase_any_of(entry, "on", "off", NULL)) { /* Some agents print the target status in the list output, * though none are known now (the separate list-status command * is used for this, but it can also print "UNKNOWN"). To handle * this possibility, skip such entries. * * @TODO Document or eliminate the implied restriction of target * names. */ free(entry); } else { // We have a valid entry *output = g_list_append(*output, entry); } entry_start = i + 1; } } } /*! * \internal * \brief Parse a list of targets from a string * * \param[in] list_output Target list as a string * * \return List of target names * \note The target list string format is flexible, to allow for user-specified * lists such pcmk_host_list and the output of an agent's list action * (whether direct or via the API, which escapes newlines). There may be * multiple lines, separated by either a newline or an escaped newline * (backslash n). Each line may have one or more target names, separated * by any combination of whitespace, commas, and semi-colons. Lines * containing "invalid" or "variable" will be ignored entirely. Target * names "on" or "off" (case-insensitive) will be ignored. Target names * may contain only alphanumeric characters, underbars (_), dashes (-), * and dots (.) (if any other character occurs in the name, it and all * subsequent characters in the name will be ignored). * \note The caller is responsible for freeing the result with * g_list_free_full(result, free). */ GList * stonith__parse_targets(const char *target_spec) { GList *targets = NULL; if (target_spec != NULL) { size_t out_len = strlen(target_spec); size_t line_start = 0; // Starting index of line being processed for (size_t i = 0; i <= out_len; ++i) { if ((target_spec[i] == '\n') || (target_spec[i] == '\0') || ((target_spec[i] == '\\') && (target_spec[i + 1] == 'n'))) { // We've reached the end of one line of output int len = i - line_start; if (len > 0) { char *line = strndup(target_spec + line_start, len); line[len] = '\0'; // Because it might be a newline parse_list_line(line, len, &targets); free(line); } if (target_spec[i] == '\\') { ++i; // backslash-n takes up two positions } line_start = i + 1; } } } return targets; } /*! * \internal * \brief Check whether a fencing failure was followed by an equivalent success * * \param[in] event Fencing failure * \param[in] top_history Complete fencing history (must be sorted by * stonith__sort_history() beforehand) * * \return The name of the node that executed the fencing if a later successful * event exists, or NULL if no such event exists */ const char * stonith__later_succeeded(const stonith_history_t *event, const stonith_history_t *top_history) { const char *other = NULL; for (const stonith_history_t *prev_hp = top_history; prev_hp != NULL; prev_hp = prev_hp->next) { if (prev_hp == event) { break; } if ((prev_hp->state == st_done) && pcmk__str_eq(event->target, prev_hp->target, pcmk__str_casei) && pcmk__str_eq(event->action, prev_hp->action, pcmk__str_none) && ((event->completed < prev_hp->completed) || ((event->completed == prev_hp->completed) && (event->completed_nsec < prev_hp->completed_nsec)))) { if ((event->delegate == NULL) || pcmk__str_eq(event->delegate, prev_hp->delegate, pcmk__str_casei)) { // Prefer equivalent fencing by same executioner return prev_hp->delegate; } else if (other == NULL) { // Otherwise remember first successful executioner other = (prev_hp->delegate == NULL)? "some node" : prev_hp->delegate; } } } return other; } /*! * \internal * \brief Sort fencing history, pending first then by most recently completed * * \param[in,out] history List of stonith actions * * \return New head of sorted \p history */ stonith_history_t * stonith__sort_history(stonith_history_t *history) { stonith_history_t *new = NULL, *pending = NULL, *hp, *np, *tmp; for (hp = history; hp; ) { tmp = hp->next; if ((hp->state == st_done) || (hp->state == st_failed)) { /* sort into new */ if ((!new) || (hp->completed > new->completed) || ((hp->completed == new->completed) && (hp->completed_nsec > new->completed_nsec))) { hp->next = new; new = hp; } else { np = new; do { if ((!np->next) || (hp->completed > np->next->completed) || ((hp->completed == np->next->completed) && (hp->completed_nsec > np->next->completed_nsec))) { hp->next = np->next; np->next = hp; break; } np = np->next; } while (1); } } else { /* put into pending */ hp->next = pending; pending = hp; } hp = tmp; } /* pending actions don't have a completed-stamp so make them go front */ if (pending) { stonith_history_t *last_pending = pending; while (last_pending->next) { last_pending = last_pending->next; } last_pending->next = new; new = pending; } return new; } /*! * \brief Return string equivalent of an operation state value * * \param[in] state Fencing operation state value * * \return Human-friendly string equivalent of state */ const char * stonith_op_state_str(enum op_state state) { switch (state) { case st_query: return "querying"; case st_exec: return "executing"; case st_done: return "completed"; case st_duplicate: return "duplicate"; case st_failed: return "failed"; } return "unknown"; } stonith_history_t * stonith__first_matching_event(stonith_history_t *history, bool (*matching_fn)(stonith_history_t *, void *), void *user_data) { for (stonith_history_t *hp = history; hp; hp = hp->next) { if (matching_fn(hp, user_data)) { return hp; } } return NULL; } bool stonith__event_state_pending(stonith_history_t *history, void *user_data) { return history->state != st_failed && history->state != st_done; } bool stonith__event_state_eq(stonith_history_t *history, void *user_data) { return history->state == GPOINTER_TO_INT(user_data); } bool stonith__event_state_neq(stonith_history_t *history, void *user_data) { return history->state != GPOINTER_TO_INT(user_data); } void stonith__device_parameter_flags(uint32_t *device_flags, const char *device_name, xmlNode *metadata) { xmlXPathObjectPtr xpath = NULL; int max = 0; int lpc = 0; CRM_CHECK((device_flags != NULL) && (metadata != NULL), return); xpath = xpath_search(metadata, "//parameter"); max = numXpathResults(xpath); if (max <= 0) { freeXpathObject(xpath); return; } for (lpc = 0; lpc < max; lpc++) { const char *parameter = NULL; xmlNode *match = getXpathResult(xpath, lpc); CRM_LOG_ASSERT(match != NULL); if (match == NULL) { continue; } parameter = crm_element_value(match, "name"); if (pcmk__str_eq(parameter, "plug", pcmk__str_casei)) { stonith__set_device_flags(*device_flags, device_name, st_device_supports_parameter_plug); } else if (pcmk__str_eq(parameter, "port", pcmk__str_casei)) { stonith__set_device_flags(*device_flags, device_name, st_device_supports_parameter_port); } } freeXpathObject(xpath); } /*! * \internal * \brief Retrieve fence agent meta-data asynchronously * * \param[in] agent Agent to execute * \param[in] timeout_sec Error if not complete within this time * \param[in] callback Function to call with result (this will always be * called, whether by this function directly or * later via the main loop, and on success the * metadata will be in its result argument's * action_stdout) * \param[in,out] user_data User data to pass to callback * * \return Standard Pacemaker return code * \note The caller must use a main loop. This function is not a * stonith_api_operations_t method because it does not need a stonith_t * object and does not go through the fencer, but executes the agent * directly. */ int stonith__metadata_async(const char *agent, int timeout_sec, void (*callback)(int pid, const pcmk__action_result_t *result, void *user_data), void *user_data) { switch (stonith_get_namespace(agent, NULL)) { case st_namespace_rhcs: { stonith_action_t *action = NULL; int rc = pcmk_ok; action = stonith__action_create(agent, "metadata", NULL, 0, timeout_sec, NULL, NULL, NULL); rc = stonith__execute_async(action, user_data, callback, NULL); if (rc != pcmk_ok) { callback(0, stonith__action_result(action), user_data); stonith__destroy_action(action); } return pcmk_legacy2rc(rc); } #if HAVE_STONITH_STONITH_H case st_namespace_lha: // LHA metadata is simply synthesized, so simulate async { pcmk__action_result_t result = { .exit_status = CRM_EX_OK, .execution_status = PCMK_EXEC_DONE, .exit_reason = NULL, .action_stdout = NULL, .action_stderr = NULL, }; stonith__lha_metadata(agent, timeout_sec, &result.action_stdout); callback(0, &result, user_data); pcmk__reset_result(&result); return pcmk_rc_ok; } #endif default: { pcmk__action_result_t result = { .exit_status = CRM_EX_NOSUCH, .execution_status = PCMK_EXEC_ERROR_HARD, .exit_reason = crm_strdup_printf("No such agent '%s'", agent), .action_stdout = NULL, .action_stderr = NULL, }; callback(0, &result, user_data); pcmk__reset_result(&result); return ENOENT; } } } /*! * \internal * \brief Return the exit status from an async action callback * * \param[in] data Callback data * * \return Exit status from callback data */ int stonith__exit_status(const stonith_callback_data_t *data) { if ((data == NULL) || (data->opaque == NULL)) { return CRM_EX_ERROR; } return ((pcmk__action_result_t *) data->opaque)->exit_status; } /*! * \internal * \brief Return the execution status from an async action callback * * \param[in] data Callback data * * \return Execution status from callback data */ int stonith__execution_status(const stonith_callback_data_t *data) { if ((data == NULL) || (data->opaque == NULL)) { return PCMK_EXEC_UNKNOWN; } return ((pcmk__action_result_t *) data->opaque)->execution_status; } /*! * \internal * \brief Return the exit reason from an async action callback * * \param[in] data Callback data * * \return Exit reason from callback data */ const char * stonith__exit_reason(const stonith_callback_data_t *data) { if ((data == NULL) || (data->opaque == NULL)) { return NULL; } return ((pcmk__action_result_t *) data->opaque)->exit_reason; } /*! * \internal * \brief Return the exit status from an event notification * * \param[in] event Event * * \return Exit status from event */ int stonith__event_exit_status(const stonith_event_t *event) { if ((event == NULL) || (event->opaque == NULL)) { return CRM_EX_ERROR; } else { struct event_private *event_private = event->opaque; return event_private->result.exit_status; } } /*! * \internal * \brief Return the execution status from an event notification * * \param[in] event Event * * \return Execution status from event */ int stonith__event_execution_status(const stonith_event_t *event) { if ((event == NULL) || (event->opaque == NULL)) { return PCMK_EXEC_UNKNOWN; } else { struct event_private *event_private = event->opaque; return event_private->result.execution_status; } } /*! * \internal * \brief Return the exit reason from an event notification * * \param[in] event Event * * \return Exit reason from event */ const char * stonith__event_exit_reason(const stonith_event_t *event) { if ((event == NULL) || (event->opaque == NULL)) { return NULL; } else { struct event_private *event_private = event->opaque; return event_private->result.exit_reason; } } /*! * \internal * \brief Return a human-friendly description of a fencing event * * \param[in] event Event to describe * * \return Newly allocated string with description of \p event * \note The caller is responsible for freeing the return value. * This function asserts on memory errors and never returns NULL. */ char * stonith__event_description(const stonith_event_t *event) { // Use somewhat readable defaults const char *origin = pcmk__s(event->client_origin, "a client"); const char *origin_node = pcmk__s(event->origin, "a node"); const char *executioner = pcmk__s(event->executioner, "the cluster"); const char *device = pcmk__s(event->device, "unknown"); const char *action = pcmk__s(event->action, event->operation); const char *target = pcmk__s(event->target, "no node"); const char *reason = stonith__event_exit_reason(event); const char *status; if (action == NULL) { action = "(unknown)"; } if (stonith__event_execution_status(event) != PCMK_EXEC_DONE) { status = pcmk_exec_status_str(stonith__event_execution_status(event)); } else if (stonith__event_exit_status(event) != CRM_EX_OK) { status = pcmk_exec_status_str(PCMK_EXEC_ERROR); } else { status = crm_exit_str(CRM_EX_OK); } if (pcmk__str_eq(event->operation, T_STONITH_NOTIFY_HISTORY, pcmk__str_none)) { return crm_strdup_printf("Fencing history may have changed"); } else if (pcmk__str_eq(event->operation, STONITH_OP_DEVICE_ADD, pcmk__str_none)) { return crm_strdup_printf("A fencing device (%s) was added", device); } else if (pcmk__str_eq(event->operation, STONITH_OP_DEVICE_DEL, pcmk__str_none)) { return crm_strdup_printf("A fencing device (%s) was removed", device); } else if (pcmk__str_eq(event->operation, STONITH_OP_LEVEL_ADD, pcmk__str_none)) { return crm_strdup_printf("A fencing topology level (%s) was added", device); } else if (pcmk__str_eq(event->operation, STONITH_OP_LEVEL_DEL, pcmk__str_none)) { return crm_strdup_printf("A fencing topology level (%s) was removed", device); } // event->operation should be T_STONITH_NOTIFY_FENCE at this point return crm_strdup_printf("Operation %s of %s by %s for %s@%s: %s%s%s%s (ref=%s)", action, target, executioner, origin, origin_node, status, ((reason == NULL)? "" : " ("), pcmk__s(reason, ""), ((reason == NULL)? "" : ")"), pcmk__s(event->id, "(none)")); } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START const char *get_stonith_provider(const char *agent, const char *provider); const char * get_stonith_provider(const char *agent, const char *provider) { return stonith_namespace2text(stonith_get_namespace(agent, provider)); } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/lrmd/lrmd_client.c b/lib/lrmd/lrmd_client.c index 82434b94a5..291882a5ee 100644 --- a/lib/lrmd/lrmd_client.c +++ b/lib/lrmd/lrmd_client.c @@ -1,2552 +1,2556 @@ /* * Copyright 2012-2023 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 // uint32_t, uint64_t #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_GNUTLS_GNUTLS_H # include #endif #include #include #include #include #include #define MAX_TLS_RECV_WAIT 10000 CRM_TRACE_INIT_DATA(lrmd); static int lrmd_api_disconnect(lrmd_t * lrmd); static int lrmd_api_is_connected(lrmd_t * lrmd); /* IPC proxy functions */ int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg); static void lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg); void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)); #ifdef HAVE_GNUTLS_GNUTLS_H # define LRMD_CLIENT_HANDSHAKE_TIMEOUT 5000 /* 5 seconds */ gnutls_psk_client_credentials_t psk_cred_s; static void lrmd_tls_disconnect(lrmd_t * lrmd); static int global_remote_msg_id = 0; static void lrmd_tls_connection_destroy(gpointer userdata); #endif typedef struct lrmd_private_s { uint64_t type; char *token; mainloop_io_t *source; /* IPC parameters */ crm_ipc_t *ipc; pcmk__remote_t *remote; /* Extra TLS parameters */ char *remote_nodename; #ifdef HAVE_GNUTLS_GNUTLS_H char *server; int port; gnutls_psk_client_credentials_t psk_cred_c; /* while the async connection is occurring, this is the id * of the connection timeout timer. */ int async_timer; int sock; /* since tls requires a round trip across the network for a * request/reply, there are times where we just want to be able * to send a request from the client and not wait around (or even care * about) what the reply is. */ int expected_late_replies; GList *pending_notify; crm_trigger_t *process_notify; #endif lrmd_event_callback callback; /* Internal IPC proxy msg passing for remote guests */ void (*proxy_callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg); void *proxy_callback_userdata; char *peer_version; } lrmd_private_t; static lrmd_list_t * lrmd_list_add(lrmd_list_t * head, const char *value) { lrmd_list_t *p, *end; p = calloc(1, sizeof(lrmd_list_t)); p->val = strdup(value); end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void lrmd_list_freeall(lrmd_list_t * head) { lrmd_list_t *p; while (head) { char *val = (char *)head->val; p = head->next; free(val); free(head); head = p; } } lrmd_key_value_t * lrmd_key_value_add(lrmd_key_value_t * head, const char *key, const char *value) { lrmd_key_value_t *p, *end; p = calloc(1, sizeof(lrmd_key_value_t)); p->key = strdup(key); p->value = strdup(value); end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void lrmd_key_value_freeall(lrmd_key_value_t * head) { lrmd_key_value_t *p; while (head) { p = head->next; free(head->key); free(head->value); free(head); head = p; } } /*! * \brief Create a new lrmd_event_data_t object * * \param[in] rsc_id ID of resource involved in event * \param[in] task Action name * \param[in] interval_ms Action interval * * \return Newly allocated and initialized lrmd_event_data_t * \note This functions asserts on memory errors, so the return value is * guaranteed to be non-NULL. The caller is responsible for freeing the * result with lrmd_free_event(). */ lrmd_event_data_t * lrmd_new_event(const char *rsc_id, const char *task, guint interval_ms) { lrmd_event_data_t *event = calloc(1, sizeof(lrmd_event_data_t)); CRM_ASSERT(event != NULL); pcmk__str_update((char **) &event->rsc_id, rsc_id); pcmk__str_update((char **) &event->op_type, task); event->interval_ms = interval_ms; return event; } lrmd_event_data_t * lrmd_copy_event(lrmd_event_data_t * event) { lrmd_event_data_t *copy = NULL; copy = calloc(1, sizeof(lrmd_event_data_t)); copy->type = event->type; pcmk__str_update((char **) ©->rsc_id, event->rsc_id); pcmk__str_update((char **) ©->op_type, event->op_type); pcmk__str_update((char **) ©->user_data, event->user_data); copy->call_id = event->call_id; copy->timeout = event->timeout; copy->interval_ms = event->interval_ms; copy->start_delay = event->start_delay; copy->rsc_deleted = event->rsc_deleted; copy->rc = event->rc; copy->op_status = event->op_status; pcmk__str_update((char **) ©->output, event->output); copy->t_run = event->t_run; copy->t_rcchange = event->t_rcchange; copy->exec_time = event->exec_time; copy->queue_time = event->queue_time; copy->connection_rc = event->connection_rc; copy->params = pcmk__str_table_dup(event->params); pcmk__str_update((char **) ©->remote_nodename, event->remote_nodename); pcmk__str_update((char **) ©->exit_reason, event->exit_reason); return copy; } /*! * \brief Free an executor event * * \param[in,out] Executor event object to free */ void lrmd_free_event(lrmd_event_data_t *event) { if (event == NULL) { return; } // @TODO Why are these const char *? free((void *) event->rsc_id); free((void *) event->op_type); free((void *) event->user_data); free((void *) event->remote_nodename); lrmd__reset_result(event); if (event->params != NULL) { g_hash_table_destroy(event->params); } free(event); } static void lrmd_dispatch_internal(lrmd_t * lrmd, xmlNode * msg) { const char *type; const char *proxy_session = crm_element_value(msg, F_LRMD_IPC_SESSION); lrmd_private_t *native = lrmd->lrmd_private; lrmd_event_data_t event = { 0, }; if (proxy_session != NULL) { /* this is proxy business */ lrmd_internal_proxy_dispatch(lrmd, msg); return; } else if (!native->callback) { /* no callback set */ crm_trace("notify event received but client has not set callback"); return; } event.remote_nodename = native->remote_nodename; type = crm_element_value(msg, F_LRMD_OPERATION); crm_element_value_int(msg, F_LRMD_CALLID, &event.call_id); event.rsc_id = crm_element_value(msg, F_LRMD_RSC_ID); if (pcmk__str_eq(type, LRMD_OP_RSC_REG, pcmk__str_none)) { event.type = lrmd_event_register; } else if (pcmk__str_eq(type, LRMD_OP_RSC_UNREG, pcmk__str_none)) { event.type = lrmd_event_unregister; } else if (pcmk__str_eq(type, LRMD_OP_RSC_EXEC, pcmk__str_none)) { time_t epoch = 0; crm_element_value_int(msg, F_LRMD_TIMEOUT, &event.timeout); crm_element_value_ms(msg, F_LRMD_RSC_INTERVAL, &event.interval_ms); crm_element_value_int(msg, F_LRMD_RSC_START_DELAY, &event.start_delay); crm_element_value_int(msg, F_LRMD_EXEC_RC, (int *)&event.rc); crm_element_value_int(msg, F_LRMD_OP_STATUS, &event.op_status); crm_element_value_int(msg, F_LRMD_RSC_DELETED, &event.rsc_deleted); crm_element_value_epoch(msg, F_LRMD_RSC_RUN_TIME, &epoch); event.t_run = (unsigned int) epoch; crm_element_value_epoch(msg, F_LRMD_RSC_RCCHANGE_TIME, &epoch); event.t_rcchange = (unsigned int) epoch; crm_element_value_int(msg, F_LRMD_RSC_EXEC_TIME, (int *)&event.exec_time); crm_element_value_int(msg, F_LRMD_RSC_QUEUE_TIME, (int *)&event.queue_time); event.op_type = crm_element_value(msg, F_LRMD_RSC_ACTION); event.user_data = crm_element_value(msg, F_LRMD_RSC_USERDATA_STR); event.type = lrmd_event_exec_complete; /* output and exit_reason may be freed by a callback */ event.output = crm_element_value_copy(msg, F_LRMD_RSC_OUTPUT); lrmd__set_result(&event, event.rc, event.op_status, crm_element_value(msg, F_LRMD_RSC_EXIT_REASON)); event.params = xml2list(msg); } else if (pcmk__str_eq(type, LRMD_OP_NEW_CLIENT, pcmk__str_none)) { event.type = lrmd_event_new_client; } else if (pcmk__str_eq(type, LRMD_OP_POKE, pcmk__str_none)) { event.type = lrmd_event_poke; } else { return; } crm_trace("op %s notify event received", type); native->callback(&event); if (event.params) { g_hash_table_destroy(event.params); } lrmd__reset_result(&event); } // \return Always 0, to indicate that IPC mainloop source should be kept static int lrmd_ipc_dispatch(const char *buffer, ssize_t length, gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; if (native->callback != NULL) { xmlNode *msg = string2xml(buffer); lrmd_dispatch_internal(lrmd, msg); free_xml(msg); } return 0; } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_free_xml(gpointer userdata) { free_xml((xmlNode *) userdata); } static bool remote_executor_connected(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; return (native->remote->tls_session != NULL); } /*! * \internal * \brief TLS dispatch function (for both trigger and file descriptor sources) * * \param[in,out] userdata API connection * * \return Always return a nonnegative value, which as a file descriptor * dispatch function means keep the mainloop source, and as a * trigger dispatch function, 0 means remove the trigger from the * mainloop while 1 means keep it (and job completed) */ static int lrmd_tls_dispatch(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *xml = NULL; int rc = pcmk_rc_ok; if (!remote_executor_connected(lrmd)) { crm_trace("TLS dispatch triggered after disconnect"); return 0; } crm_trace("TLS dispatch triggered"); /* First check if there are any pending notifies to process that came * while we were waiting for replies earlier. */ if (native->pending_notify) { GList *iter = NULL; crm_trace("Processing pending notifies"); for (iter = native->pending_notify; iter; iter = iter->next) { lrmd_dispatch_internal(lrmd, iter->data); } g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } /* Next read the current buffer and see if there are any messages to handle. */ switch (pcmk__remote_ready(native->remote, 0)) { case pcmk_rc_ok: rc = pcmk__read_remote_message(native->remote, -1); xml = pcmk__remote_message_xml(native->remote); break; case ETIME: // Nothing to read, check if a full message is already in buffer xml = pcmk__remote_message_xml(native->remote); break; default: rc = ENOTCONN; break; } while (xml) { const char *msg_type = crm_element_value(xml, F_LRMD_REMOTE_MSG_TYPE); if (pcmk__str_eq(msg_type, "notify", pcmk__str_casei)) { lrmd_dispatch_internal(lrmd, xml); } else if (pcmk__str_eq(msg_type, "reply", pcmk__str_casei)) { if (native->expected_late_replies > 0) { native->expected_late_replies--; } else { int reply_id = 0; crm_element_value_int(xml, F_LRMD_CALLID, &reply_id); /* if this happens, we want to know about it */ crm_err("Got outdated Pacemaker Remote reply %d", reply_id); } } free_xml(xml); xml = pcmk__remote_message_xml(native->remote); } if (rc == ENOTCONN) { crm_info("Lost %s executor connection while reading data", (native->remote_nodename? native->remote_nodename : "local")); lrmd_tls_disconnect(lrmd); return 0; } return 1; } #endif /* Not used with mainloop */ int lrmd_poll(lrmd_t * lrmd, int timeout) { lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: return crm_ipc_ready(native->ipc); #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: if (native->pending_notify) { return 1; } else { int rc = pcmk__remote_ready(native->remote, 0); switch (rc) { case pcmk_rc_ok: return 1; case ETIME: return 0; default: return pcmk_rc2legacy(rc); } } #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); return -EPROTONOSUPPORT; } } /* Not used with mainloop */ bool lrmd_dispatch(lrmd_t * lrmd) { lrmd_private_t *private = NULL; CRM_ASSERT(lrmd != NULL); private = lrmd->lrmd_private; switch (private->type) { case pcmk__client_ipc: while (crm_ipc_ready(private->ipc)) { if (crm_ipc_read(private->ipc) > 0) { const char *msg = crm_ipc_buffer(private->ipc); lrmd_ipc_dispatch(msg, strlen(msg), lrmd); } } break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: lrmd_tls_dispatch(lrmd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", private->type); } if (lrmd_api_is_connected(lrmd) == FALSE) { crm_err("Connection closed"); return FALSE; } return TRUE; } static xmlNode * lrmd_create_op(const char *token, const char *op, xmlNode *data, int timeout, enum lrmd_call_options options) { xmlNode *op_msg = create_xml_node(NULL, "lrmd_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "lrmd_command"); crm_xml_add(op_msg, F_TYPE, T_LRMD); crm_xml_add(op_msg, F_LRMD_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_LRMD_OPERATION, op); crm_xml_add_int(op_msg, F_LRMD_TIMEOUT, timeout); crm_xml_add_int(op_msg, F_LRMD_CALLOPTS, options); if (data != NULL) { add_message_xml(op_msg, F_LRMD_CALLDATA, data); } crm_trace("Created executor %s command with call options %.8lx (%d)", op, (long)options, options); return op_msg; } static void lrmd_ipc_connection_destroy(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; crm_info("IPC connection destroyed"); /* Prevent these from being cleaned up in lrmd_api_disconnect() */ native->ipc = NULL; native->source = NULL; if (native->callback) { lrmd_event_data_t event = { 0, }; event.type = lrmd_event_disconnect; event.remote_nodename = native->remote_nodename; native->callback(&event); } } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_tls_connection_destroy(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; crm_info("TLS connection destroyed"); if (native->remote->tls_session) { gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); } if (native->psk_cred_c) { gnutls_psk_free_client_credentials(native->psk_cred_c); } if (native->sock) { close(native->sock); } if (native->process_notify) { mainloop_destroy_trigger(native->process_notify); native->process_notify = NULL; } if (native->pending_notify) { g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } free(native->remote->buffer); free(native->remote->start_state); native->remote->buffer = NULL; native->remote->start_state = NULL; native->source = 0; native->sock = 0; native->psk_cred_c = NULL; native->remote->tls_session = NULL; native->sock = 0; if (native->callback) { lrmd_event_data_t event = { 0, }; event.remote_nodename = native->remote_nodename; event.type = lrmd_event_disconnect; native->callback(&event); } return; } // \return Standard Pacemaker return code int lrmd__remote_send_xml(pcmk__remote_t *session, xmlNode *msg, uint32_t id, const char *msg_type) { crm_xml_add_int(msg, F_LRMD_REMOTE_MSG_ID, id); crm_xml_add(msg, F_LRMD_REMOTE_MSG_TYPE, msg_type); return pcmk__remote_send_xml(session, msg); } // \return Standard Pacemaker return code static int read_remote_reply(lrmd_t *lrmd, int total_timeout, int expected_reply_id, xmlNode **reply) { lrmd_private_t *native = lrmd->lrmd_private; time_t start = time(NULL); const char *msg_type = NULL; int reply_id = 0; int remaining_timeout = 0; int rc = pcmk_rc_ok; /* A timeout of 0 here makes no sense. We have to wait a period of time * for the response to come back. If -1 or 0, default to 10 seconds. */ if (total_timeout <= 0 || total_timeout > MAX_TLS_RECV_WAIT) { total_timeout = MAX_TLS_RECV_WAIT; } for (*reply = NULL; *reply == NULL; ) { *reply = pcmk__remote_message_xml(native->remote); if (*reply == NULL) { /* read some more off the tls buffer if we still have time left. */ if (remaining_timeout) { remaining_timeout = total_timeout - ((time(NULL) - start) * 1000); } else { remaining_timeout = total_timeout; } if (remaining_timeout <= 0) { return ETIME; } rc = pcmk__read_remote_message(native->remote, remaining_timeout); if (rc != pcmk_rc_ok) { return rc; } *reply = pcmk__remote_message_xml(native->remote); if (*reply == NULL) { return ENOMSG; } } crm_element_value_int(*reply, F_LRMD_REMOTE_MSG_ID, &reply_id); msg_type = crm_element_value(*reply, F_LRMD_REMOTE_MSG_TYPE); if (!msg_type) { crm_err("Empty msg type received while waiting for reply"); free_xml(*reply); *reply = NULL; } else if (pcmk__str_eq(msg_type, "notify", pcmk__str_casei)) { /* got a notify while waiting for reply, trigger the notify to be processed later */ crm_info("queueing notify"); native->pending_notify = g_list_append(native->pending_notify, *reply); if (native->process_notify) { crm_info("notify trigger set."); mainloop_set_trigger(native->process_notify); } *reply = NULL; } else if (!pcmk__str_eq(msg_type, "reply", pcmk__str_casei)) { /* msg isn't a reply, make some noise */ crm_err("Expected a reply, got %s", msg_type); free_xml(*reply); *reply = NULL; } else if (reply_id != expected_reply_id) { if (native->expected_late_replies > 0) { native->expected_late_replies--; } else { crm_err("Got outdated reply, expected id %d got id %d", expected_reply_id, reply_id); } free_xml(*reply); *reply = NULL; } } if (native->remote->buffer && native->process_notify) { mainloop_set_trigger(native->process_notify); } return rc; } // \return Standard Pacemaker return code static int send_remote_message(lrmd_t *lrmd, xmlNode *msg) { int rc = pcmk_rc_ok; lrmd_private_t *native = lrmd->lrmd_private; global_remote_msg_id++; if (global_remote_msg_id <= 0) { global_remote_msg_id = 1; } rc = lrmd__remote_send_xml(native->remote, msg, global_remote_msg_id, "request"); if (rc != pcmk_rc_ok) { crm_err("Disconnecting because TLS message could not be sent to " "Pacemaker Remote: %s", pcmk_rc_str(rc)); lrmd_tls_disconnect(lrmd); } return rc; } static int lrmd_tls_send_recv(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply) { int rc = 0; xmlNode *xml = NULL; if (!remote_executor_connected(lrmd)) { return -ENOTCONN; } rc = send_remote_message(lrmd, msg); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } rc = read_remote_reply(lrmd, timeout, global_remote_msg_id, &xml); if (rc != pcmk_rc_ok) { crm_err("Disconnecting remote after request %d reply not received: %s " CRM_XS " rc=%d timeout=%dms", global_remote_msg_id, pcmk_rc_str(rc), rc, timeout); lrmd_tls_disconnect(lrmd); } if (reply) { *reply = xml; } else { free_xml(xml); } return pcmk_rc2legacy(rc); } #endif static int lrmd_send_xml(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = crm_ipc_send(native->ipc, msg, crm_ipc_client_response, timeout, reply); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_send_recv(lrmd, msg, timeout, reply); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static int lrmd_send_xml_no_reply(lrmd_t * lrmd, xmlNode * msg) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = crm_ipc_send(native->ipc, msg, crm_ipc_flags_none, 0, NULL); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = send_remote_message(lrmd, msg); if (rc == pcmk_rc_ok) { /* we don't want to wait around for the reply, but * since the request/reply protocol needs to behave the same * as libqb, a reply will eventually come later anyway. */ native->expected_late_replies++; } rc = pcmk_rc2legacy(rc); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static int lrmd_api_is_connected(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: return crm_ipc_connected(native->ipc); #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: return remote_executor_connected(lrmd); #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); return 0; } } /*! * \internal * \brief Send a prepared API command to the executor * * \param[in,out] lrmd Existing connection to the executor * \param[in] op Name of API command to send * \param[in] data Command data XML to add to the sent command * \param[out] output_data If expecting a reply, it will be stored here * \param[in] timeout Timeout in milliseconds (if 0, defaults to * a sensible value per the type of connection, * standard vs. pacemaker remote); * also propagated to the command XML * \param[in] call_options Call options to pass to server when sending * \param[in] expect_reply If TRUE, wait for a reply from the server; * must be TRUE for IPC (as opposed to TLS) clients * * \return pcmk_ok on success, -errno on error */ static int lrmd_send_command(lrmd_t *lrmd, const char *op, xmlNode *data, xmlNode **output_data, int timeout, enum lrmd_call_options options, gboolean expect_reply) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; if (!lrmd_api_is_connected(lrmd)) { return -ENOTCONN; } if (op == NULL) { crm_err("No operation specified"); return -EINVAL; } CRM_CHECK(native->token != NULL,; ); crm_trace("Sending %s op to executor", op); op_msg = lrmd_create_op(native->token, op, data, timeout, options); if (op_msg == NULL) { return -EINVAL; } if (expect_reply) { rc = lrmd_send_xml(lrmd, op_msg, timeout, &op_reply); } else { rc = lrmd_send_xml_no_reply(lrmd, op_msg); goto done; } if (rc < 0) { crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%d): %d", op, timeout, rc); goto done; } else if(op_reply == NULL) { rc = -ENOMSG; goto done; } rc = pcmk_ok; crm_trace("%s op reply received", op); if (crm_element_value_int(op_reply, F_LRMD_RC, &rc) != 0) { rc = -ENOMSG; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (output_data) { *output_data = op_reply; op_reply = NULL; /* Prevent subsequent free */ } done: if (lrmd_api_is_connected(lrmd) == FALSE) { crm_err("Executor disconnected"); } free_xml(op_msg); free_xml(op_reply); return rc; } static int lrmd_api_poke_connection(lrmd_t * lrmd) { int rc; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); rc = lrmd_send_command(lrmd, LRMD_OP_POKE, data, NULL, 0, 0, (native->type == pcmk__client_ipc)); free_xml(data); return rc < 0 ? rc : pcmk_ok; } // \return Standard Pacemaker return code int lrmd__validate_remote_settings(lrmd_t *lrmd, GHashTable *hash) { int rc = pcmk_rc_ok; const char *value; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *data = create_xml_node(NULL, F_LRMD_OPERATION); crm_xml_add(data, F_LRMD_ORIGIN, __func__); value = g_hash_table_lookup(hash, "stonith-watchdog-timeout"); if ((value) && (stonith__watchdog_fencing_enabled_for_node(native->remote_nodename))) { crm_xml_add(data, F_LRMD_WATCHDOG, value); } rc = lrmd_send_command(lrmd, LRMD_OP_CHECK, data, NULL, 0, 0, (native->type == pcmk__client_ipc)); free_xml(data); return (rc < 0)? pcmk_legacy2rc(rc) : pcmk_rc_ok; } static int lrmd_handshake(lrmd_t * lrmd, const char *name) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "lrmd_command"); crm_xml_add(hello, F_TYPE, T_LRMD); crm_xml_add(hello, F_LRMD_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_LRMD_CLIENTNAME, name); crm_xml_add(hello, F_LRMD_PROTOCOL_VERSION, LRMD_PROTOCOL_VERSION); /* advertise that we are a proxy provider */ if (native->proxy_callback) { pcmk__xe_set_bool_attr(hello, F_LRMD_IS_IPC_PROVIDER, true); } rc = lrmd_send_xml(lrmd, hello, -1, &reply); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't complete registration with the executor API: %d", rc); rc = -ECOMM; } else if (reply == NULL) { crm_err("Did not receive registration reply"); rc = -EPROTO; } else { const char *version = crm_element_value(reply, F_LRMD_PROTOCOL_VERSION); const char *msg_type = crm_element_value(reply, F_LRMD_OPERATION); const char *tmp_ticket = crm_element_value(reply, F_LRMD_CLIENTID); const char *start_state = crm_element_value(reply, PCMK__XA_NODE_START_STATE); long long uptime = -1; crm_element_value_int(reply, F_LRMD_RC, &rc); /* The remote executor may add its uptime to the XML reply, which is * useful in handling transient attributes when the connection to the * remote node unexpectedly drops. If no parameter is given, just * default to -1. */ crm_element_value_ll(reply, PCMK__XA_UPTIME, &uptime); native->remote->uptime = uptime; if (start_state) { native->remote->start_state = strdup(start_state); } if (rc == -EPROTO) { crm_err("Executor protocol version mismatch between client (%s) and server (%s)", LRMD_PROTOCOL_VERSION, version); crm_log_xml_err(reply, "Protocol Error"); } else if (!pcmk__str_eq(msg_type, CRM_OP_REGISTER, pcmk__str_casei)) { crm_err("Invalid registration message: %s", msg_type); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else if (tmp_ticket == NULL) { crm_err("No registration token provided"); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else { crm_trace("Obtained registration token: %s", tmp_ticket); native->token = strdup(tmp_ticket); native->peer_version = strdup(version?version:"1.0"); /* Included since 1.1 */ rc = pcmk_ok; } } free_xml(reply); free_xml(hello); if (rc != pcmk_ok) { lrmd_api_disconnect(lrmd); } return rc; } static int lrmd_ipc_connect(lrmd_t * lrmd, int *fd) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; struct ipc_client_callbacks lrmd_callbacks = { .dispatch = lrmd_ipc_dispatch, .destroy = lrmd_ipc_connection_destroy }; crm_info("Connecting to executor"); if (fd) { /* No mainloop */ native->ipc = crm_ipc_new(CRM_SYSTEM_LRMD, 0); - if (native->ipc && crm_ipc_connect(native->ipc)) { - *fd = crm_ipc_get_fd(native->ipc); - } else if (native->ipc) { - crm_perror(LOG_ERR, "Connection to executor failed"); - rc = -ENOTCONN; + if (native->ipc != NULL) { + rc = pcmk__connect_generic_ipc(native->ipc); + if (rc == pcmk_rc_ok) { + rc = pcmk__ipc_fd(native->ipc, fd); + } + if (rc != pcmk_rc_ok) { + crm_err("Connection to executor failed: %s", pcmk_rc_str(rc)); + rc = -ENOTCONN; + } } } else { native->source = mainloop_add_ipc_client(CRM_SYSTEM_LRMD, G_PRIORITY_HIGH, 0, lrmd, &lrmd_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { crm_debug("Could not connect to the executor API"); rc = -ENOTCONN; } return rc; } #ifdef HAVE_GNUTLS_GNUTLS_H static void copy_gnutls_datum(gnutls_datum_t *dest, gnutls_datum_t *source) { CRM_ASSERT((dest != NULL) && (source != NULL) && (source->data != NULL)); dest->data = gnutls_malloc(source->size); CRM_ASSERT(dest->data); memcpy(dest->data, source->data, source->size); dest->size = source->size; } static void clear_gnutls_datum(gnutls_datum_t *datum) { gnutls_free(datum->data); datum->data = NULL; datum->size = 0; } #define KEY_READ_LEN 256 // Chunk size for reading key from file // \return Standard Pacemaker return code static int read_gnutls_key(const char *location, gnutls_datum_t *key) { FILE *stream = NULL; size_t buf_len = KEY_READ_LEN; if ((location == NULL) || (key == NULL)) { return EINVAL; } stream = fopen(location, "r"); if (stream == NULL) { return errno; } key->data = gnutls_malloc(buf_len); key->size = 0; while (!feof(stream)) { int next = fgetc(stream); if (next == EOF) { if (!feof(stream)) { crm_warn("Pacemaker Remote key read was partially successful " "(copy in memory may be corrupted)"); } break; } if (key->size == buf_len) { buf_len = key->size + KEY_READ_LEN; key->data = gnutls_realloc(key->data, buf_len); CRM_ASSERT(key->data); } key->data[key->size++] = (unsigned char) next; } fclose(stream); if (key->size == 0) { clear_gnutls_datum(key); return ENOKEY; } return pcmk_rc_ok; } // Cache the most recently used Pacemaker Remote authentication key struct key_cache_s { time_t updated; // When cached key was read (valid for 1 minute) const char *location; // Where cached key was read from gnutls_datum_t key; // Cached key }; static bool key_is_cached(struct key_cache_s *key_cache) { return key_cache->updated != 0; } static bool key_cache_expired(struct key_cache_s *key_cache) { return (time(NULL) - key_cache->updated) >= 60; } static void clear_key_cache(struct key_cache_s *key_cache) { clear_gnutls_datum(&(key_cache->key)); if ((key_cache->updated != 0) || (key_cache->location != NULL)) { key_cache->updated = 0; key_cache->location = NULL; crm_debug("Cleared Pacemaker Remote key cache"); } } static void get_cached_key(struct key_cache_s *key_cache, gnutls_datum_t *key) { copy_gnutls_datum(key, &(key_cache->key)); crm_debug("Using cached Pacemaker Remote key from %s", pcmk__s(key_cache->location, "unknown location")); } static void cache_key(struct key_cache_s *key_cache, gnutls_datum_t *key, const char *location) { key_cache->updated = time(NULL); key_cache->location = location; copy_gnutls_datum(&(key_cache->key), key); crm_debug("Using (and cacheing) Pacemaker Remote key from %s", pcmk__s(location, "unknown location")); } /*! * \internal * \brief Get Pacemaker Remote authentication key from file or cache * * \param[in] location Path to key file to try (this memory must * persist across all calls of this function) * \param[out] key Key from location or cache * * \return Standard Pacemaker return code */ static int get_remote_key(const char *location, gnutls_datum_t *key) { static struct key_cache_s key_cache = { 0, }; int rc = pcmk_rc_ok; if ((location == NULL) || (key == NULL)) { return EINVAL; } if (key_is_cached(&key_cache)) { if (key_cache_expired(&key_cache)) { clear_key_cache(&key_cache); } else { get_cached_key(&key_cache, key); return pcmk_rc_ok; } } rc = read_gnutls_key(location, key); if (rc != pcmk_rc_ok) { return rc; } cache_key(&key_cache, key, location); return pcmk_rc_ok; } /*! * \internal * \brief Initialize the Pacemaker Remote authentication key * * Try loading the Pacemaker Remote authentication key from cache if available, * otherwise from these locations, in order of preference: the value of the * PCMK_authkey_location environment variable, if set; the Pacemaker default key * file location; or (for historical reasons) /etc/corosync/authkey. * * \param[out] key Where to store key * * \return Standard Pacemaker return code */ int lrmd__init_remote_key(gnutls_datum_t *key) { static const char *env_location = NULL; static bool need_env = true; int env_rc = pcmk_rc_ok; int default_rc = pcmk_rc_ok; int alt_rc = pcmk_rc_ok; bool env_is_default = false; bool env_is_fallback = false; if (need_env) { env_location = getenv("PCMK_authkey_location"); need_env = false; } // Try location in environment variable, if set if (env_location != NULL) { env_rc = get_remote_key(env_location, key); if (env_rc == pcmk_rc_ok) { return pcmk_rc_ok; } env_is_default = !strcmp(env_location, DEFAULT_REMOTE_KEY_LOCATION); env_is_fallback = !strcmp(env_location, ALT_REMOTE_KEY_LOCATION); /* @TODO It would be more secure to fail, rather than fall back to the * default, if an explicitly set key location is not readable, and it * would be better to never use the Corosync location as a fallback. * However, that would break any deployments currently working with the * fallbacks. */ } // Try default location, if environment wasn't explicitly set to it if (env_is_default) { default_rc = env_rc; } else { default_rc = get_remote_key(DEFAULT_REMOTE_KEY_LOCATION, key); } // Try fallback location, if environment wasn't set to it and default failed if (env_is_fallback) { alt_rc = env_rc; } else if (default_rc != pcmk_rc_ok) { alt_rc = get_remote_key(ALT_REMOTE_KEY_LOCATION, key); } // We have all results, so log and return if ((env_rc != pcmk_rc_ok) && (default_rc != pcmk_rc_ok) && (alt_rc != pcmk_rc_ok)) { // Environment set, everything failed crm_warn("Could not read Pacemaker Remote key from %s (%s%s%s%s%s): %s", env_location, env_is_default? "" : "or default location ", env_is_default? "" : DEFAULT_REMOTE_KEY_LOCATION, !env_is_default && !env_is_fallback? " " : "", env_is_fallback? "" : "or fallback location ", env_is_fallback? "" : ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(env_rc)); return ENOKEY; } if (env_rc != pcmk_rc_ok) { // Environment set but failed, using a default crm_warn("Could not read Pacemaker Remote key from %s " "(using %s location %s instead): %s", env_location, (default_rc == pcmk_rc_ok)? "default" : "fallback", (default_rc == pcmk_rc_ok)? DEFAULT_REMOTE_KEY_LOCATION : ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(env_rc)); return pcmk_rc_ok; } if ((default_rc != pcmk_rc_ok) && (alt_rc != pcmk_rc_ok)) { // Environment unset, defaults failed crm_warn("Could not read Pacemaker Remote key from default location %s" " (or fallback location %s): %s", DEFAULT_REMOTE_KEY_LOCATION, ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(default_rc)); return ENOKEY; } return pcmk_rc_ok; // Environment variable unset, a default worked } static void lrmd_gnutls_global_init(void) { static int gnutls_init = 0; if (!gnutls_init) { crm_gnutls_global_init(); } gnutls_init = 1; } #endif static void report_async_connection_result(lrmd_t * lrmd, int rc) { lrmd_private_t *native = lrmd->lrmd_private; if (native->callback) { lrmd_event_data_t event = { 0, }; event.type = lrmd_event_connect; event.remote_nodename = native->remote_nodename; event.connection_rc = rc; native->callback(&event); } } #ifdef HAVE_GNUTLS_GNUTLS_H static inline int lrmd__tls_client_handshake(pcmk__remote_t *remote) { return pcmk__tls_client_handshake(remote, LRMD_CLIENT_HANDSHAKE_TIMEOUT); } /*! * \internal * \brief Add trigger and file descriptor mainloop sources for TLS * * \param[in,out] lrmd API connection with established TLS session * \param[in] do_handshake Whether to perform executor handshake * * \return Standard Pacemaker return code */ static int add_tls_to_mainloop(lrmd_t *lrmd, bool do_handshake) { lrmd_private_t *native = lrmd->lrmd_private; int rc = pcmk_rc_ok; char *name = crm_strdup_printf("pacemaker-remote-%s:%d", native->server, native->port); struct mainloop_fd_callbacks tls_fd_callbacks = { .dispatch = lrmd_tls_dispatch, .destroy = lrmd_tls_connection_destroy, }; native->process_notify = mainloop_add_trigger(G_PRIORITY_HIGH, lrmd_tls_dispatch, lrmd); native->source = mainloop_add_fd(name, G_PRIORITY_HIGH, native->sock, lrmd, &tls_fd_callbacks); /* Async connections lose the client name provided by the API caller, so we * have to use our generated name here to perform the executor handshake. * * @TODO Keep track of the caller-provided name. Perhaps we should be using * that name in this function instead of generating one anyway. */ if (do_handshake) { rc = lrmd_handshake(lrmd, name); rc = pcmk_legacy2rc(rc); } free(name); return rc; } static void lrmd_tcp_connect_cb(void *userdata, int rc, int sock) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; gnutls_datum_t psk_key = { NULL, 0 }; native->async_timer = 0; if (rc != pcmk_rc_ok) { lrmd_tls_connection_destroy(lrmd); crm_info("Could not connect to Pacemaker Remote at %s:%d: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); return; } /* The TCP connection was successful, so establish the TLS connection. * @TODO make this async to avoid blocking code in client */ native->sock = sock; rc = lrmd__init_remote_key(&psk_key); if (rc != pcmk_rc_ok) { crm_info("Could not connect to Pacemaker Remote at %s:%d: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); return; } gnutls_psk_allocate_client_credentials(&native->psk_cred_c); gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW); gnutls_free(psk_key.data); native->remote->tls_session = pcmk__new_tls_session(sock, GNUTLS_CLIENT, GNUTLS_CRD_PSK, native->psk_cred_c); if (native->remote->tls_session == NULL) { lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, -EPROTO); return; } if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) { crm_warn("Disconnecting after TLS handshake with Pacemaker Remote server %s:%d failed", native->server, native->port); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = NULL; lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, -EKEYREJECTED); return; } crm_info("TLS connection to Pacemaker Remote server %s:%d succeeded", native->server, native->port); rc = add_tls_to_mainloop(lrmd, true); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); } static int lrmd_tls_connect_async(lrmd_t * lrmd, int timeout /*ms */ ) { int rc; int timer_id = 0; lrmd_private_t *native = lrmd->lrmd_private; lrmd_gnutls_global_init(); native->sock = -1; rc = pcmk__connect_remote(native->server, native->port, timeout, &timer_id, &(native->sock), lrmd, lrmd_tcp_connect_cb); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%d failed: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); return pcmk_rc2legacy(rc); } native->async_timer = timer_id; return pcmk_ok; } static int lrmd_tls_connect(lrmd_t * lrmd, int *fd) { int rc; lrmd_private_t *native = lrmd->lrmd_private; gnutls_datum_t psk_key = { NULL, 0 }; lrmd_gnutls_global_init(); native->sock = -1; rc = pcmk__connect_remote(native->server, native->port, 0, NULL, &(native->sock), NULL, NULL); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%d failed: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); lrmd_tls_connection_destroy(lrmd); return -ENOTCONN; } rc = lrmd__init_remote_key(&psk_key); if (rc != pcmk_rc_ok) { lrmd_tls_connection_destroy(lrmd); return pcmk_rc2legacy(rc); } gnutls_psk_allocate_client_credentials(&native->psk_cred_c); gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW); gnutls_free(psk_key.data); native->remote->tls_session = pcmk__new_tls_session(native->sock, GNUTLS_CLIENT, GNUTLS_CRD_PSK, native->psk_cred_c); if (native->remote->tls_session == NULL) { lrmd_tls_connection_destroy(lrmd); return -EPROTO; } if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) { crm_err("Session creation for %s:%d failed", native->server, native->port); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = NULL; lrmd_tls_connection_destroy(lrmd); return -EKEYREJECTED; } crm_info("Client TLS connection established with Pacemaker Remote server %s:%d", native->server, native->port); if (fd) { *fd = native->sock; } else { add_tls_to_mainloop(lrmd, false); } return pcmk_ok; } #endif static int lrmd_api_connect(lrmd_t * lrmd, const char *name, int *fd) { int rc = -ENOTCONN; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = lrmd_ipc_connect(lrmd, fd); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_connect(lrmd, fd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } if (rc == pcmk_ok) { rc = lrmd_handshake(lrmd, name); } return rc; } static int lrmd_api_connect_async(lrmd_t * lrmd, const char *name, int timeout) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; CRM_CHECK(native && native->callback, return -EINVAL); switch (native->type) { case pcmk__client_ipc: /* fake async connection with ipc. it should be fast * enough that we gain very little from async */ rc = lrmd_api_connect(lrmd, name, NULL); if (!rc) { report_async_connection_result(lrmd, rc); } break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_connect_async(lrmd, timeout); if (rc) { /* connection failed, report rc now */ report_async_connection_result(lrmd, rc); } break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static void lrmd_ipc_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_tls_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->remote->tls_session) { gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = 0; } if (native->async_timer) { g_source_remove(native->async_timer); native->async_timer = 0; } if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; } else if (native->sock) { close(native->sock); native->sock = 0; } if (native->pending_notify) { g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } } #endif static int lrmd_api_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; int rc = pcmk_ok; crm_info("Disconnecting %s %s executor connection", pcmk__client_type_str(native->type), (native->remote_nodename? native->remote_nodename : "local")); switch (native->type) { case pcmk__client_ipc: lrmd_ipc_disconnect(lrmd); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: lrmd_tls_disconnect(lrmd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } free(native->token); native->token = NULL; free(native->peer_version); native->peer_version = NULL; return rc; } static int lrmd_api_register_rsc(lrmd_t * lrmd, const char *rsc_id, const char *class, const char *provider, const char *type, enum lrmd_call_options options) { int rc = pcmk_ok; xmlNode *data = NULL; if (!class || !type || !rsc_id) { return -EINVAL; } if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) && (provider == NULL)) { return -EINVAL; } data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add(data, F_LRMD_CLASS, class); crm_xml_add(data, F_LRMD_PROVIDER, provider); crm_xml_add(data, F_LRMD_TYPE, type); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_REG, data, NULL, 0, options, TRUE); free_xml(data); return rc; } static int lrmd_api_unregister_rsc(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_UNREG, data, NULL, 0, options, TRUE); free_xml(data); return rc; } lrmd_rsc_info_t * lrmd_new_rsc_info(const char *rsc_id, const char *standard, const char *provider, const char *type) { lrmd_rsc_info_t *rsc_info = calloc(1, sizeof(lrmd_rsc_info_t)); CRM_ASSERT(rsc_info); pcmk__str_update(&rsc_info->id, rsc_id); pcmk__str_update(&rsc_info->standard, standard); pcmk__str_update(&rsc_info->provider, provider); pcmk__str_update(&rsc_info->type, type); return rsc_info; } lrmd_rsc_info_t * lrmd_copy_rsc_info(lrmd_rsc_info_t * rsc_info) { return lrmd_new_rsc_info(rsc_info->id, rsc_info->standard, rsc_info->provider, rsc_info->type); } void lrmd_free_rsc_info(lrmd_rsc_info_t * rsc_info) { if (!rsc_info) { return; } free(rsc_info->id); free(rsc_info->type); free(rsc_info->standard); free(rsc_info->provider); free(rsc_info); } static lrmd_rsc_info_t * lrmd_api_get_rsc_info(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options) { lrmd_rsc_info_t *rsc_info = NULL; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); xmlNode *output = NULL; const char *class = NULL; const char *provider = NULL; const char *type = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); lrmd_send_command(lrmd, LRMD_OP_RSC_INFO, data, &output, 0, options, TRUE); free_xml(data); if (!output) { return NULL; } class = crm_element_value(output, F_LRMD_CLASS); provider = crm_element_value(output, F_LRMD_PROVIDER); type = crm_element_value(output, F_LRMD_TYPE); if (!class || !type) { free_xml(output); return NULL; } else if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) && !provider) { free_xml(output); return NULL; } rsc_info = lrmd_new_rsc_info(rsc_id, class, provider, type); free_xml(output); return rsc_info; } void lrmd_free_op_info(lrmd_op_info_t *op_info) { if (op_info) { free(op_info->rsc_id); free(op_info->action); free(op_info->interval_ms_s); free(op_info->timeout_ms_s); free(op_info); } } static int lrmd_api_get_recurring_ops(lrmd_t *lrmd, const char *rsc_id, int timeout_ms, enum lrmd_call_options options, GList **output) { xmlNode *data = NULL; xmlNode *output_xml = NULL; int rc = pcmk_ok; if (output == NULL) { return -EINVAL; } *output = NULL; // Send request if (rsc_id) { data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); } rc = lrmd_send_command(lrmd, LRMD_OP_GET_RECURRING, data, &output_xml, timeout_ms, options, TRUE); if (data) { free_xml(data); } // Process reply if ((rc != pcmk_ok) || (output_xml == NULL)) { return rc; } for (xmlNode *rsc_xml = first_named_child(output_xml, F_LRMD_RSC); (rsc_xml != NULL) && (rc == pcmk_ok); rsc_xml = crm_next_same_xml(rsc_xml)) { rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); if (rsc_id == NULL) { crm_err("Could not parse recurring operation information from executor"); continue; } for (xmlNode *op_xml = first_named_child(rsc_xml, T_LRMD_RSC_OP); op_xml != NULL; op_xml = crm_next_same_xml(op_xml)) { lrmd_op_info_t *op_info = calloc(1, sizeof(lrmd_op_info_t)); if (op_info == NULL) { rc = -ENOMEM; break; } op_info->rsc_id = strdup(rsc_id); op_info->action = crm_element_value_copy(op_xml, F_LRMD_RSC_ACTION); op_info->interval_ms_s = crm_element_value_copy(op_xml, F_LRMD_RSC_INTERVAL); op_info->timeout_ms_s = crm_element_value_copy(op_xml, F_LRMD_TIMEOUT); *output = g_list_prepend(*output, op_info); } } free_xml(output_xml); return rc; } static void lrmd_api_set_callback(lrmd_t * lrmd, lrmd_event_callback callback) { lrmd_private_t *native = lrmd->lrmd_private; native->callback = callback; } void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)) { lrmd_private_t *native = lrmd->lrmd_private; native->proxy_callback = callback; native->proxy_callback_userdata = userdata; } void lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg) { lrmd_private_t *native = lrmd->lrmd_private; if (native->proxy_callback) { crm_log_xml_trace(msg, "PROXY_INBOUND"); native->proxy_callback(lrmd, native->proxy_callback_userdata, msg); } } int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg) { if (lrmd == NULL) { return -ENOTCONN; } crm_xml_add(msg, F_LRMD_OPERATION, CRM_OP_IPC_FWD); crm_log_xml_trace(msg, "PROXY_OUTBOUND"); return lrmd_send_xml_no_reply(lrmd, msg); } static int stonith_get_metadata(const char *provider, const char *type, char **output) { int rc = pcmk_ok; stonith_t *stonith_api = stonith_api_new(); if (stonith_api == NULL) { crm_err("Could not get fence agent meta-data: API memory allocation failed"); return -ENOMEM; } rc = stonith_api->cmds->metadata(stonith_api, st_opt_sync_call, type, provider, output, 0); if ((rc == pcmk_ok) && (*output == NULL)) { rc = -EIO; } stonith_api->cmds->free(stonith_api); return rc; } static int lrmd_api_get_metadata(lrmd_t *lrmd, const char *standard, const char *provider, const char *type, char **output, enum lrmd_call_options options) { return lrmd->cmds->get_metadata_params(lrmd, standard, provider, type, output, options, NULL); } static int lrmd_api_get_metadata_params(lrmd_t *lrmd, const char *standard, const char *provider, const char *type, char **output, enum lrmd_call_options options, lrmd_key_value_t *params) { svc_action_t *action = NULL; GHashTable *params_table = NULL; if (!standard || !type) { lrmd_key_value_freeall(params); return -EINVAL; } if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { lrmd_key_value_freeall(params); return stonith_get_metadata(provider, type, output); } params_table = pcmk__strkey_table(free, free); for (const lrmd_key_value_t *param = params; param; param = param->next) { g_hash_table_insert(params_table, strdup(param->key), strdup(param->value)); } action = services__create_resource_action(type, standard, provider, type, CRMD_ACTION_METADATA, 0, CRMD_METADATA_CALL_TIMEOUT, params_table, 0); lrmd_key_value_freeall(params); if (action == NULL) { return -ENOMEM; } if (action->rc != PCMK_OCF_UNKNOWN) { services_action_free(action); return -EINVAL; } if (!services_action_sync(action)) { crm_err("Failed to retrieve meta-data for %s:%s:%s", standard, provider, type); services_action_free(action); return -EIO; } if (!action->stdout_data) { crm_err("Failed to receive meta-data for %s:%s:%s", standard, provider, type); services_action_free(action); return -EIO; } *output = strdup(action->stdout_data); services_action_free(action); return pcmk_ok; } static int lrmd_api_exec(lrmd_t *lrmd, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout, /* ms */ int start_delay, /* ms */ enum lrmd_call_options options, lrmd_key_value_t * params) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); lrmd_key_value_t *tmp = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add(data, F_LRMD_RSC_ACTION, action); crm_xml_add(data, F_LRMD_RSC_USERDATA_STR, userdata); crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms); crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout); crm_xml_add_int(data, F_LRMD_RSC_START_DELAY, start_delay); for (tmp = params; tmp; tmp = tmp->next) { hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args); } rc = lrmd_send_command(lrmd, LRMD_OP_RSC_EXEC, data, NULL, timeout, options, TRUE); free_xml(data); lrmd_key_value_freeall(params); return rc; } /* timeout is in ms */ static int lrmd_api_exec_alert(lrmd_t *lrmd, const char *alert_id, const char *alert_path, int timeout, lrmd_key_value_t *params) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_ALERT); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); lrmd_key_value_t *tmp = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_ALERT_ID, alert_id); crm_xml_add(data, F_LRMD_ALERT_PATH, alert_path); crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout); for (tmp = params; tmp; tmp = tmp->next) { hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args); } rc = lrmd_send_command(lrmd, LRMD_OP_ALERT_EXEC, data, NULL, timeout, lrmd_opt_notify_orig_only, TRUE); free_xml(data); lrmd_key_value_freeall(params); return rc; } static int lrmd_api_cancel(lrmd_t *lrmd, const char *rsc_id, const char *action, guint interval_ms) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ACTION, action); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_CANCEL, data, NULL, 0, 0, TRUE); free_xml(data); return rc; } static int list_stonith_agents(lrmd_list_t ** resources) { int rc = 0; stonith_t *stonith_api = stonith_api_new(); stonith_key_value_t *stonith_resources = NULL; stonith_key_value_t *dIter = NULL; if (stonith_api == NULL) { crm_err("Could not list fence agents: API memory allocation failed"); return -ENOMEM; } stonith_api->cmds->list_agents(stonith_api, st_opt_sync_call, NULL, &stonith_resources, 0); stonith_api->cmds->free(stonith_api); for (dIter = stonith_resources; dIter; dIter = dIter->next) { rc++; if (resources) { *resources = lrmd_list_add(*resources, dIter->value); } } stonith_key_value_freeall(stonith_resources, 1, 0); return rc; } static int lrmd_api_list_agents(lrmd_t * lrmd, lrmd_list_t ** resources, const char *class, const char *provider) { int rc = 0; int stonith_count = 0; // Initially, whether to include stonith devices if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { stonith_count = 1; } else { GList *gIter = NULL; GList *agents = resources_list_agents(class, provider); for (gIter = agents; gIter != NULL; gIter = gIter->next) { *resources = lrmd_list_add(*resources, (const char *)gIter->data); rc++; } g_list_free_full(agents, free); if (!class) { stonith_count = 1; } } if (stonith_count) { // Now, if stonith devices are included, how many there are stonith_count = list_stonith_agents(resources); if (stonith_count > 0) { rc += stonith_count; } } if (rc == 0) { crm_notice("No agents found for class %s", class); rc = -EPROTONOSUPPORT; } return rc; } static bool does_provider_have_agent(const char *agent, const char *provider, const char *class) { bool found = false; GList *agents = NULL; GList *gIter2 = NULL; agents = resources_list_agents(class, provider); for (gIter2 = agents; gIter2 != NULL; gIter2 = gIter2->next) { if (pcmk__str_eq(agent, gIter2->data, pcmk__str_casei)) { found = true; } } g_list_free_full(agents, free); return found; } static int lrmd_api_list_ocf_providers(lrmd_t * lrmd, const char *agent, lrmd_list_t ** providers) { int rc = pcmk_ok; char *provider = NULL; GList *ocf_providers = NULL; GList *gIter = NULL; ocf_providers = resources_list_providers(PCMK_RESOURCE_CLASS_OCF); for (gIter = ocf_providers; gIter != NULL; gIter = gIter->next) { provider = gIter->data; if (!agent || does_provider_have_agent(agent, provider, PCMK_RESOURCE_CLASS_OCF)) { *providers = lrmd_list_add(*providers, (const char *)gIter->data); rc++; } } g_list_free_full(ocf_providers, free); return rc; } static int lrmd_api_list_standards(lrmd_t * lrmd, lrmd_list_t ** supported) { int rc = 0; GList *standards = NULL; GList *gIter = NULL; standards = resources_list_standards(); for (gIter = standards; gIter != NULL; gIter = gIter->next) { *supported = lrmd_list_add(*supported, (const char *)gIter->data); rc++; } if (list_stonith_agents(NULL) > 0) { *supported = lrmd_list_add(*supported, PCMK_RESOURCE_CLASS_STONITH); rc++; } g_list_free_full(standards, free); return rc; } /*! * \internal * \brief Create an executor API object * * \param[out] api Will be set to newly created API object (it is the * caller's responsibility to free this value with * lrmd_api_delete() if this function succeeds) * \param[in] nodename If the object will be used for a remote connection, * the node name to use in cluster for remote executor * \param[in] server If the object will be used for a remote connection, * the resolvable host name to connect to * \param[in] port If the object will be used for a remote connection, * port number on \p server to connect to * * \return Standard Pacemaker return code * \note If the caller leaves one of \p nodename or \p server NULL, the other's * value will be used for both. If the caller leaves both NULL, an API * object will be created for a local executor connection. */ int lrmd__new(lrmd_t **api, const char *nodename, const char *server, int port) { lrmd_private_t *pvt = NULL; if (api == NULL) { return EINVAL; } *api = NULL; // Allocate all memory needed *api = calloc(1, sizeof(lrmd_t)); if (*api == NULL) { return ENOMEM; } pvt = calloc(1, sizeof(lrmd_private_t)); if (pvt == NULL) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } (*api)->lrmd_private = pvt; // @TODO Do we need to do this for local connections? pvt->remote = calloc(1, sizeof(pcmk__remote_t)); (*api)->cmds = calloc(1, sizeof(lrmd_api_operations_t)); if ((pvt->remote == NULL) || ((*api)->cmds == NULL)) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } // Set methods (*api)->cmds->connect = lrmd_api_connect; (*api)->cmds->connect_async = lrmd_api_connect_async; (*api)->cmds->is_connected = lrmd_api_is_connected; (*api)->cmds->poke_connection = lrmd_api_poke_connection; (*api)->cmds->disconnect = lrmd_api_disconnect; (*api)->cmds->register_rsc = lrmd_api_register_rsc; (*api)->cmds->unregister_rsc = lrmd_api_unregister_rsc; (*api)->cmds->get_rsc_info = lrmd_api_get_rsc_info; (*api)->cmds->get_recurring_ops = lrmd_api_get_recurring_ops; (*api)->cmds->set_callback = lrmd_api_set_callback; (*api)->cmds->get_metadata = lrmd_api_get_metadata; (*api)->cmds->exec = lrmd_api_exec; (*api)->cmds->cancel = lrmd_api_cancel; (*api)->cmds->list_agents = lrmd_api_list_agents; (*api)->cmds->list_ocf_providers = lrmd_api_list_ocf_providers; (*api)->cmds->list_standards = lrmd_api_list_standards; (*api)->cmds->exec_alert = lrmd_api_exec_alert; (*api)->cmds->get_metadata_params = lrmd_api_get_metadata_params; if ((nodename == NULL) && (server == NULL)) { pvt->type = pcmk__client_ipc; } else { #ifdef HAVE_GNUTLS_GNUTLS_H if (nodename == NULL) { nodename = server; } else if (server == NULL) { server = nodename; } pvt->type = pcmk__client_tls; pvt->remote_nodename = strdup(nodename); pvt->server = strdup(server); if ((pvt->remote_nodename == NULL) || (pvt->server == NULL)) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } pvt->port = port; if (pvt->port == 0) { pvt->port = crm_default_remote_port(); } #else crm_err("Cannot communicate with Pacemaker Remote " "because GnuTLS is not enabled for this build"); lrmd_api_delete(*api); *api = NULL; return EOPNOTSUPP; #endif } return pcmk_rc_ok; } lrmd_t * lrmd_api_new(void) { lrmd_t *api = NULL; CRM_ASSERT(lrmd__new(&api, NULL, NULL, 0) == pcmk_rc_ok); return api; } lrmd_t * lrmd_remote_api_new(const char *nodename, const char *server, int port) { lrmd_t *api = NULL; CRM_ASSERT(lrmd__new(&api, nodename, server, port) == pcmk_rc_ok); return api; } void lrmd_api_delete(lrmd_t * lrmd) { if (lrmd == NULL) { return; } if (lrmd->cmds != NULL) { // Never NULL, but make static analysis happy if (lrmd->cmds->disconnect != NULL) { // Also never really NULL lrmd->cmds->disconnect(lrmd); // No-op if already disconnected } free(lrmd->cmds); } if (lrmd->lrmd_private != NULL) { lrmd_private_t *native = lrmd->lrmd_private; #ifdef HAVE_GNUTLS_GNUTLS_H free(native->server); #endif free(native->remote_nodename); free(native->remote); free(native->token); free(native->peer_version); free(lrmd->lrmd_private); } free(lrmd); } struct metadata_cb { void (*callback)(int pid, const pcmk__action_result_t *result, void *user_data); void *user_data; }; /*! * \internal * \brief Process asynchronous metadata completion * * \param[in,out] action Metadata action that completed */ static void metadata_complete(svc_action_t *action) { struct metadata_cb *metadata_cb = (struct metadata_cb *) action->cb_data; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; pcmk__set_result(&result, action->rc, action->status, services__exit_reason(action)); pcmk__set_result_output(&result, action->stdout_data, action->stderr_data); metadata_cb->callback(0, &result, metadata_cb->user_data); result.action_stdout = NULL; // Prevent free, because action owns it result.action_stderr = NULL; // Prevent free, because action owns it pcmk__reset_result(&result); free(metadata_cb); } /*! * \internal * \brief Retrieve agent metadata asynchronously * * \param[in] rsc Resource agent specification * \param[in] callback Function to call with result (this will always be * called, whether by this function directly or later * via the main loop, and on success the metadata will * be in its result argument's action_stdout) * \param[in,out] user_data User data to pass to callback * * \return Standard Pacemaker return code * \note This function is not a lrmd_api_operations_t method because it does not * need an lrmd_t object and does not go through the executor, but * executes the agent directly. */ int lrmd__metadata_async(const lrmd_rsc_info_t *rsc, void (*callback)(int pid, const pcmk__action_result_t *result, void *user_data), void *user_data) { svc_action_t *action = NULL; struct metadata_cb *metadata_cb = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; CRM_CHECK(callback != NULL, return EINVAL); if ((rsc == NULL) || (rsc->standard == NULL) || (rsc->type == NULL)) { pcmk__set_result(&result, PCMK_OCF_NOT_CONFIGURED, PCMK_EXEC_ERROR_FATAL, "Invalid resource specification"); callback(0, &result, user_data); pcmk__reset_result(&result); return EINVAL; } if (strcmp(rsc->standard, PCMK_RESOURCE_CLASS_STONITH) == 0) { return stonith__metadata_async(rsc->type, CRMD_METADATA_CALL_TIMEOUT / 1000, callback, user_data); } action = services__create_resource_action(pcmk__s(rsc->id, rsc->type), rsc->standard, rsc->provider, rsc->type, CRMD_ACTION_METADATA, 0, CRMD_METADATA_CALL_TIMEOUT, NULL, 0); if (action == NULL) { pcmk__set_result(&result, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Out of memory"); callback(0, &result, user_data); pcmk__reset_result(&result); return ENOMEM; } if (action->rc != PCMK_OCF_UNKNOWN) { pcmk__set_result(&result, action->rc, action->status, services__exit_reason(action)); callback(0, &result, user_data); pcmk__reset_result(&result); services_action_free(action); return EINVAL; } action->cb_data = calloc(1, sizeof(struct metadata_cb)); if (action->cb_data == NULL) { services_action_free(action); pcmk__set_result(&result, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Out of memory"); callback(0, &result, user_data); pcmk__reset_result(&result); return ENOMEM; } metadata_cb = (struct metadata_cb *) action->cb_data; metadata_cb->callback = callback; metadata_cb->user_data = user_data; if (!services_action_async(action, metadata_complete)) { services_action_free(action); return pcmk_rc_error; // @TODO Derive from action->rc and ->status } // The services library has taken responsibility for action return pcmk_rc_ok; } /*! * \internal * \brief Set the result of an executor event * * \param[in,out] event Executor event to set * \param[in] rc OCF exit status of event * \param[in] op_status Executor status of event * \param[in] exit_reason Human-friendly description of event */ void lrmd__set_result(lrmd_event_data_t *event, enum ocf_exitcode rc, int op_status, const char *exit_reason) { if (event == NULL) { return; } event->rc = rc; event->op_status = op_status; pcmk__str_update((char **) &event->exit_reason, exit_reason); } /*! * \internal * \brief Clear an executor event's exit reason, output, and error output * * \param[in,out] event Executor event to reset */ void lrmd__reset_result(lrmd_event_data_t *event) { if (event == NULL) { return; } free((void *) event->exit_reason); event->exit_reason = NULL; free((void *) event->output); event->output = NULL; } /*! * \internal * \brief Get the uptime of a remote resource connection * * When the cluster connects to a remote resource, part of that resource's * handshake includes the uptime of the remote resource's connection. This * uptime is stored in the lrmd_t object. * * \return The connection's uptime, or -1 if unknown */ time_t lrmd__uptime(lrmd_t *lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->remote == NULL) { return -1; } else { return native->remote->uptime; } } const char * lrmd__node_start_state(lrmd_t *lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->remote == NULL) { return NULL; } else { return native->remote->start_state; } } diff --git a/lib/pacemaker/pcmk_cluster_queries.c b/lib/pacemaker/pcmk_cluster_queries.c index 062adc6809..c99039a922 100644 --- a/lib/pacemaker/pcmk_cluster_queries.c +++ b/lib/pacemaker/pcmk_cluster_queries.c @@ -1,897 +1,899 @@ /* * Copyright 2020-2023 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 // xmlNode #include #include #include #include #include #include #include #include #include #include #include #include //! Object to store node info from the controller API typedef struct { /* Adapted from pcmk_controld_api_reply_t:data:node_info. * (char **) are convenient here for use within callbacks: we can skip * copying strings unless the caller passes a non-NULL value. */ uint32_t id; char **node_name; char **uuid; char **state; bool have_quorum; bool is_remote; } node_info_t; //! Object to store API results, a timeout, and an output object typedef struct { pcmk__output_t *out; bool show_output; int rc; unsigned int message_timeout_ms; enum pcmk_pacemakerd_state pcmkd_state; node_info_t node_info; } data_t; /*! * \internal * \brief Validate that an IPC API event is a good reply * * \param[in,out] data API results and options * \param[in] api IPC API connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * * \return Standard Pacemaker return code */ static int validate_reply_event(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status) { pcmk__output_t *out = data->out; switch (event_type) { case pcmk_ipc_event_reply: break; case pcmk_ipc_event_disconnect: if (data->rc == ECONNRESET) { // Unexpected out->err(out, "error: Lost connection to %s", pcmk_ipc_name(api, true)); } // Nothing bad but not the reply we're looking for return ENOTSUP; default: // Ditto return ENOTSUP; } if (status != CRM_EX_OK) { out->err(out, "error: Bad reply from %s: %s", pcmk_ipc_name(api, true), crm_exit_str(status)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a controller API event is a good reply of expected type * * \param[in,out] data API results and options * \param[in] api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_controld_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_controld_api_reply expected_type) { pcmk__output_t *out = data->out; int rc = pcmk_rc_ok; const pcmk_controld_api_reply_t *reply = NULL; rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from controller", pcmk__controld_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a \p pacemakerd API event is a good reply of expected * type * * \param[in,out] data API results and options * \param[in] api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_pcmkd_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_pacemakerd_api_reply expected_type) { pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_pacemakerd_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from pacemakerd", pcmk__pcmkd_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Process a controller status IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void controller_status_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "health", reply->data.ping.sys_from, reply->host_from, reply->data.ping.fsa_state, reply->data.ping.result); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a designated controller IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void designated_controller_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "dc", reply->host_from); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a node info IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void node_info_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_info); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->data.node_info.uname == NULL) { out->err(out, "Node is not known to cluster"); data->rc = pcmk_rc_node_unknown; return; } data->node_info.have_quorum = reply->data.node_info.have_quorum; data->node_info.is_remote = reply->data.node_info.is_remote; data->node_info.id = (uint32_t) reply->data.node_info.id; pcmk__str_update(data->node_info.node_name, reply->data.node_info.uname); pcmk__str_update(data->node_info.uuid, reply->data.node_info.uuid); pcmk__str_update(data->node_info.state, reply->data.node_info.state); if (data->show_output) { out->message(out, "node-info", reply->data.node_info.id, reply->data.node_info.uname, reply->data.node_info.uuid, reply->data.node_info.state, reply->data.node_info.have_quorum, reply->data.node_info.is_remote); } data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a \p pacemakerd status IPC event * * \param[in,out] pacemakerd_api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_pacemakerd_api_reply_t object * containing event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void pacemakerd_event_cb(pcmk_ipc_api_t *pacemakerd_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = user_data; pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_pcmkd_reply(data, pacemakerd_api, event_type, status, event_data, pcmk_pacemakerd_reply_ping); if (rc != pcmk_rc_ok) { return; } // Parse desired information from reply reply = (const pcmk_pacemakerd_api_reply_t *) event_data; data->pcmkd_state = reply->data.ping.state; data->rc = pcmk_rc_ok; if (!data->show_output) { return; } if (reply->data.ping.status == pcmk_rc_ok) { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, NULL, reply->data.ping.last_good); } else { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, "query failed", time(NULL)); } } static pcmk_ipc_api_t * ipc_connect(data_t *data, enum pcmk_ipc_server server, pcmk_ipc_callback_t cb, enum pcmk_ipc_dispatch dispatch_type, bool eremoteio_ok) { int rc; pcmk__output_t *out = data->out; pcmk_ipc_api_t *api = NULL; rc = pcmk_new_ipc_api(&api, server); if (api == NULL) { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); data->rc = rc; return NULL; } if (cb != NULL) { pcmk_register_ipc_callback(api, cb, data); } - rc = pcmk_connect_ipc(api, dispatch_type); - + rc = pcmk__connect_ipc(api, dispatch_type, 5); if (rc != pcmk_rc_ok) { if (rc == EREMOTEIO) { data->pcmkd_state = pcmk_pacemakerd_state_remote; if (eremoteio_ok) { /* EREMOTEIO may be expected and acceptable for some callers * on a Pacemaker Remote node */ + crm_debug("Ignoring %s connection failure: No " + "Pacemaker Remote connection", + pcmk_ipc_name(api, true)); rc = pcmk_rc_ok; } else { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); } } data->rc = rc; pcmk_free_ipc_api(api); return NULL; } return api; } /*! * \internal * \brief Poll an IPC API connection until timeout or a reply is received * * \param[in,out] data API results and options * \param[in,out] api IPC API connection * \param[in] on_node If not \p NULL, name of the node to poll (used only * for logging) * * \note Sets the \p rc member of \p data on error */ static void poll_until_reply(data_t *data, pcmk_ipc_api_t *api, const char *on_node) { pcmk__output_t *out = data->out; uint64_t start_nsec = qb_util_nano_current_get(); uint64_t end_nsec = start_nsec; uint64_t elapsed_ms = 0; uint64_t remaining_ms = data->message_timeout_ms; while (remaining_ms > 0) { int rc = pcmk_poll_ipc(api, remaining_ms); if (rc == EAGAIN) { // Poll timed out break; } if (rc != pcmk_rc_ok) { out->err(out, "error: Failed to poll %s API%s%s: %s", pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, ""), pcmk_rc_str(rc)); data->rc = rc; return; } pcmk_dispatch_ipc(api); if (data->rc != EAGAIN) { // Received a reply return; } end_nsec = qb_util_nano_current_get(); elapsed_ms = (end_nsec - start_nsec) / QB_TIME_NS_IN_MSEC; remaining_ms = data->message_timeout_ms - elapsed_ms; } out->err(out, "error: Timed out after %ums waiting for reply from %s API%s%s", data->message_timeout_ms, pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, "")); data->rc = EAGAIN; } /*! * \internal * \brief Get and output controller status * * \param[in,out] out Output object * \param[in] node_name Name of node whose status is desired * (\p NULL for DC) * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__controller_status(pcmk__output_t *out, const char *node_name, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, controller_status_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, node_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on %s: %s", pcmk__s(node_name, "DC"), pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, pcmk__s(node_name, "DC")); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_controller_status(xmlNodePtr *xml, const char *node_name, unsigned int message_timeout_ms) { 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__controller_status(out, node_name, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and output designated controller node name * * \param[in,out] out Output object * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__designated_controller(pcmk__output_t *out, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, designated_controller_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, NULL); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on DC: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "DC"); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_designated_controller(xmlNodePtr *xml, unsigned int message_timeout_ms) { 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__designated_controller(out, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and optionally output node info corresponding to a node ID from * the controller * * \param[in,out] out Output object * \param[in,out] node_id ID of node whose name to get. If \p NULL * or 0, get the local node name. If not * \p NULL, store the true node ID here on * success. * \param[out] node_name If not \p NULL, where to store the node * name * \param[out] uuid If not \p NULL, where to store the node * UUID * \param[out] state If not \p NULL, where to store the * membership state * \param[out] is_remote If not \p NULL, where to store whether the * node is a Pacemaker Remote node * \param[out] have_quorum If not \p NULL, where to store whether the * node has quorum * \param[in] show_output Whether to show the node info * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code * * \note The caller is responsible for freeing \p *node_name, \p *uuid, and * \p *state using \p free(). */ int pcmk__query_node_info(pcmk__output_t *out, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .node_info = { .id = (node_id == NULL)? 0 : *node_id, .node_name = node_name, .uuid = uuid, .state = state, }, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (node_name != NULL) { *node_name = NULL; } if (uuid != NULL) { *uuid = NULL; } if (state != NULL) { *state = NULL; } if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, node_info_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_node_info(controld_api, (node_id != NULL)? *node_id : 0); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not send request to controller API on local " "node: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "local node"); } pcmk_free_ipc_api(controld_api); } if (data.rc != pcmk_rc_ok) { return data.rc; } // String outputs are set in callback if (node_id != NULL) { *node_id = data.node_info.id; } if (have_quorum != NULL) { *have_quorum = data.node_info.have_quorum; } if (is_remote != NULL) { *is_remote = data.node_info.is_remote; } return data.rc; } // Documented in header int pcmk_query_node_info(xmlNodePtr *xml, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; CRM_ASSERT(node_name != NULL); rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__query_node_info(out, node_id, node_name, uuid, state, have_quorum, is_remote, show_output, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and optionally output \p pacemakerd status * * \param[in,out] out Output object * \param[in] ipc_name IPC name for request * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemakerd API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * \param[in] show_output Whether to output the \p pacemakerd state * \param[out] state Where to store the \p pacemakerd state, if * not \p NULL * * \return Standard Pacemaker return code * * \note This function sets \p state to \p pcmk_pacemakerd_state_remote and * returns \p pcmk_rc_ok if the IPC connection attempt returns * \p EREMOTEIO. That code indicates that this is a Pacemaker Remote node * with \p pacemaker-remoted running. The node may be connected to the * cluster. */ int pcmk__pacemakerd_status(pcmk__output_t *out, const char *ipc_name, unsigned int message_timeout_ms, bool show_output, enum pcmk_pacemakerd_state *state) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .pcmkd_state = pcmk_pacemakerd_state_invalid, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *pacemakerd_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } pacemakerd_api = ipc_connect(&data, pcmk_ipc_pacemakerd, pacemakerd_event_cb, dispatch_type, true); if (pacemakerd_api != NULL) { int rc = pcmk_pacemakerd_api_ping(pacemakerd_api, ipc_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping launcher API: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, pacemakerd_api, NULL); } pcmk_free_ipc_api(pacemakerd_api); } else if ((data.pcmkd_state == pcmk_pacemakerd_state_remote) && show_output) { // No API connection so the callback wasn't run out->message(out, "pacemakerd-health", NULL, data.pcmkd_state, NULL, time(NULL)); } if (state != NULL) { *state = data.pcmkd_state; } return data.rc; } // Documented in header int pcmk_pacemakerd_status(xmlNodePtr *xml, const char *ipc_name, unsigned int message_timeout_ms) { 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__pacemakerd_status(out, ipc_name, message_timeout_ms, true, NULL); pcmk__xml_output_finish(out, xml); return rc; } /* user data for looping through remote node xpath searches */ struct node_data { pcmk__output_t *out; int found; const char *field; /* XML attribute to check for node name */ const char *type; bool bash_export; }; static void remote_node_print_helper(xmlNode *result, void *user_data) { struct node_data *data = user_data; pcmk__output_t *out = data->out; const char *name = crm_element_value(result, XML_ATTR_UNAME); const char *id = crm_element_value(result, data->field); // node name and node id are the same for remote/guest nodes out->message(out, "crmadmin-node", data->type, pcmk__s(name, id), id, data->bash_export); data->found++; } // \return Standard Pacemaker return code int pcmk__list_nodes(pcmk__output_t *out, const char *node_types, bool bash_export) { xmlNode *xml_node = NULL; int rc; rc = cib__signon_query(out, NULL, &xml_node); if (rc == pcmk_rc_ok) { struct node_data data = { .out = out, .found = 0, .bash_export = bash_export }; out->begin_list(out, NULL, NULL, "nodes"); if (!pcmk__str_empty(node_types) && strstr(node_types, "all")) { node_types = NULL; } if (pcmk__str_empty(node_types) || strstr(node_types, "cluster")) { data.field = "id"; data.type = "cluster"; crm_foreach_xpath_result(xml_node, PCMK__XP_MEMBER_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || strstr(node_types, "guest")) { data.field = "value"; data.type = "guest"; crm_foreach_xpath_result(xml_node, PCMK__XP_GUEST_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || pcmk__str_eq(node_types, ",|^remote", pcmk__str_regex)) { data.field = "id"; data.type = "remote"; crm_foreach_xpath_result(xml_node, PCMK__XP_REMOTE_NODE_CONFIG, remote_node_print_helper, &data); } out->end_list(out); if (data.found == 0) { out->info(out, "No nodes configured"); } free_xml(xml_node); } return rc; } int pcmk_list_nodes(xmlNodePtr *xml, const char *node_types) { 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__list_nodes(out, node_types, FALSE); pcmk__xml_output_finish(out, xml); return rc; } diff --git a/tools/attrd_updater.c b/tools/attrd_updater.c index 60e4cc7c52..b615a3575d 100644 --- a/tools/attrd_updater.c +++ b/tools/attrd_updater.c @@ -1,520 +1,521 @@ /* * Copyright 2004-2023 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 #define SUMMARY "query and update Pacemaker node attributes" static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; GError *error = NULL; bool printed_values = false; struct { char command; gchar *attr_dampen; gchar *attr_name; gchar *attr_pattern; gchar *attr_node; gchar *attr_set; char *attr_value; uint32_t attr_options; gboolean query_all; gboolean quiet; } options = { .attr_options = pcmk__node_attr_none, .command = 'Q', }; static gboolean command_cb (const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&options.attr_value, optarg); if (pcmk__str_any_of(option_name, "--update-both", "-B", NULL)) { options.command = 'B'; } else if (pcmk__str_any_of(option_name, "--delete", "-D", NULL)) { options.command = 'D'; } else if (pcmk__str_any_of(option_name, "--query", "-Q", NULL)) { options.command = 'Q'; } else if (pcmk__str_any_of(option_name, "--refresh", "-R", NULL)) { options.command = 'R'; } else if (pcmk__str_any_of(option_name, "--update", "-U", "-v", NULL)) { options.command = 'U'; } else if (pcmk__str_any_of(option_name, "--update-delay", "-Y", NULL)) { options.command = 'Y'; } return TRUE; } static gboolean private_cb (const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__set_node_attr_flags(options.attr_options, pcmk__node_attr_private); return TRUE; } static gboolean section_cb (const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (pcmk__str_any_of(optarg, "nodes", "forever", NULL)) { pcmk__set_node_attr_flags(options.attr_options, pcmk__node_attr_perm); } else if (pcmk__str_any_of(optarg, "status", "reboot", NULL)) { pcmk__clear_node_attr_flags(options.attr_options, pcmk__node_attr_perm); } else { g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_USAGE, "Unknown value for --lifetime: %s", optarg); return FALSE; } return TRUE; } static gboolean attr_set_type_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-z", "--utilization", NULL)) { pcmk__set_node_attr_flags(options.attr_options, pcmk__node_attr_utilization); } return TRUE; } static gboolean wait_cb (const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (pcmk__str_eq(optarg, "no", pcmk__str_none)) { pcmk__clear_node_attr_flags(options.attr_options, pcmk__node_attr_sync_local | pcmk__node_attr_sync_cluster); return TRUE; } else if (pcmk__str_eq(optarg, PCMK__VALUE_LOCAL, pcmk__str_none)) { pcmk__clear_node_attr_flags(options.attr_options, pcmk__node_attr_sync_local | pcmk__node_attr_sync_cluster); pcmk__set_node_attr_flags(options.attr_options, pcmk__node_attr_sync_local); return TRUE; } else if (pcmk__str_eq(optarg, PCMK__VALUE_CLUSTER, pcmk__str_none)) { pcmk__clear_node_attr_flags(options.attr_options, pcmk__node_attr_sync_local | pcmk__node_attr_sync_cluster); pcmk__set_node_attr_flags(options.attr_options, pcmk__node_attr_sync_cluster); return TRUE; } else { g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--wait= must be one of 'no', 'local', 'cluster'"); return FALSE; } } #define INDENT " " static GOptionEntry required_entries[] = { { "name", 'n', 0, G_OPTION_ARG_STRING, &options.attr_name, "The attribute's name", "NAME" }, { "pattern", 'P', 0, G_OPTION_ARG_STRING, &options.attr_pattern, "Operate on all attributes matching this pattern\n" INDENT "(with -B, -D, -U, or -Y)", "PATTERN" }, { NULL } }; static GOptionEntry command_entries[] = { { "update", 'U', 0, G_OPTION_ARG_CALLBACK, command_cb, "Update attribute's value in pacemaker-attrd. If this causes the value\n" INDENT "to change, it will also be updated in the cluster configuration.", "VALUE" }, { "update-both", 'B', 0, G_OPTION_ARG_CALLBACK, command_cb, "Update attribute's value and time to wait (dampening) in\n" INDENT "pacemaker-attrd. If this causes the value or dampening to change,\n" INDENT "the attribute will also be written to the cluster configuration,\n" INDENT "so be aware that repeatedly changing the dampening reduces its\n" INDENT "effectiveness.\n" INDENT "Requires -d/--delay", "VALUE" }, { "update-delay", 'Y', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Update attribute's dampening in pacemaker-attrd. If this causes\n" INDENT "the dampening to change, the attribute will also be written\n" INDENT "to the cluster configuration, so be aware that repeatedly\n" INDENT "changing the dampening reduces its effectiveness.\n" INDENT "Requires -d/--delay", NULL }, { "query", 'Q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Query the attribute's value from pacemaker-attrd", NULL }, { "delete", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Unset attribute from pacemaker-attrd. At the moment, there is no way\n" INDENT "to remove an attribute. This option will instead set its value\n" INDENT "to the empty string.", NULL }, { "refresh", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "(Advanced) Force the pacemaker-attrd daemon to resend all current\n" INDENT "values to the CIB", NULL }, { NULL } }; static GOptionEntry addl_entries[] = { { "delay", 'd', 0, G_OPTION_ARG_STRING, &options.attr_dampen, "The time to wait (dampening) in seconds for further changes\n" INDENT "before sending to the CIB", "SECONDS" }, { "set", 's', 0, G_OPTION_ARG_STRING, &options.attr_set, "(Advanced) The attribute set in which to place the value", "SET" }, { "node", 'N', 0, G_OPTION_ARG_STRING, &options.attr_node, "Set the attribute for the named node (instead of the local one)", "NODE" }, { "all", 'A', 0, G_OPTION_ARG_NONE, &options.query_all, "Show values of the attribute for all nodes (query only)", NULL }, { "lifetime", 'l', 0, G_OPTION_ARG_CALLBACK, section_cb, "(Not yet implemented) Lifetime of the node attribute (silently\n" INDENT "ignored by cluster)", "SECTION" }, { "private", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, private_cb, "If this creates a new attribute, never write the attribute to CIB", NULL }, { "wait", 'W', 0, G_OPTION_ARG_CALLBACK, wait_cb, "Wait for some event to occur before returning. Values are 'no' (wait\n" INDENT "only for the attribute daemon to acknowledge the request),\n" INDENT "'local' (wait until the change has propagated to where a local\n" INDENT "query will return the request value, or the value set by a\n" INDENT "later request), or 'cluster' (wait until the change has propagated\n" INDENT "to where a query anywhere on the cluster will return the requested\n" INDENT "value, or the value set by a later request). Default is 'no'.", "UNTIL" }, { "utilization", 'z', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb, "When creating a new attribute, create it as a node utilization attribute\n" INDENT "instead of an instance attribute. If the attribute already exists,\n" INDENT "its existing type (utilization vs. instance) will be used regardless.\n" INDENT "(with -B, -U, -Y)", NULL }, { NULL } }; static GOptionEntry deprecated_entries[] = { { "quiet", 'q', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.quiet, NULL, NULL }, { "update", 'v', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, command_cb, NULL, NULL }, { "section", 'S', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, section_cb, NULL, NULL }, { NULL } }; static int send_attrd_query(pcmk__output_t *out, const char *attr_name, const char *attr_node, gboolean query_all); static int send_attrd_update(char command, const char *attr_node, const char *attr_name, const char *attr_value, const char *attr_set, const char *attr_dampen, uint32_t attr_options); static bool pattern_used_correctly(void) { /* --pattern can only be used with: * -B (update-both), -D (delete), -U (update), or -Y (update-delay) */ return options.command == 'B' || options.command == 'D' || options.command == 'U' || options.command == 'Y'; } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); pcmk__add_arg_group(context, "required", "Required Arguments:", "Show required arguments", required_entries); pcmk__add_arg_group(context, "command", "Command:", "Show command options (mutually exclusive)", command_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; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_exit_t exit_code = CRM_EX_OK; pcmk__output_t *out = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); GOptionContext *context = build_arg_context(args, &output_group); gchar **processed_args = pcmk__cmdline_preproc(argv, "dlnsvBNUS"); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("attrd_updater", args->verbosity); 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 (args->version) { out->version(out, false); goto done; } if (options.attr_pattern) { if (options.attr_name) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error: --name and --pattern cannot be used at the same time"); goto done; } if (!pattern_used_correctly()) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error: pattern can only be used with delete or update"); goto done; } g_free(options.attr_name); options.attr_name = options.attr_pattern; options.attr_options |= pcmk__node_attr_pattern; } if (options.command != 'R' && options.attr_name == NULL) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Command requires --name or --pattern argument"); goto done; } else if ((options.command == 'B'|| options.command == 'Y') && options.attr_dampen == NULL) { out->info(out, "Warning: '%c' command given without required --delay", options.command); } pcmk__register_lib_messages(out); if (options.command == 'Q') { int rc = send_attrd_query(out, options.attr_name, options.attr_node, options.query_all); exit_code = pcmk_rc2exitc(rc); } else { /* @TODO We don't know whether the specified node is a Pacemaker Remote * node or not, so we can't set pcmk__node_attr_remote when appropriate. * However, it's not a big problem, because pacemaker-attrd will learn * and remember a node's "remoteness". */ int rc = send_attrd_update(options.command, options.attr_node, options.attr_name, options.attr_value, options.attr_set, options.attr_dampen, options.attr_options); exit_code = pcmk_rc2exitc(rc); } done: g_strfreev(processed_args); pcmk__free_arg_context(context); g_free(options.attr_dampen); g_free(options.attr_name); g_free(options.attr_node); g_free(options.attr_set); free(options.attr_value); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } /*! * \brief Print the attribute values in a pacemaker-attrd XML query reply * * \param[in,out] out Output object * \param[in] reply List of attribute name/value pairs * * \return true if any values were printed */ static void print_attrd_values(pcmk__output_t *out, const GList *reply) { for (const GList *iter = reply; iter != NULL; iter = iter->next) { const pcmk__attrd_query_pair_t *pair = iter->data; out->message(out, "attribute", NULL, NULL, pair->name, pair->value, pair->node); printed_values = true; } } static void attrd_event_cb(pcmk_ipc_api_t *attrd_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk__output_t *out = (pcmk__output_t *) user_data; pcmk__attrd_api_reply_t *reply = event_data; if (event_type != pcmk_ipc_event_reply || status != CRM_EX_OK) { return; } /* Print the values from the reply. */ if (reply->reply_type == pcmk__attrd_reply_query) { print_attrd_values(out, reply->data.pairs); } } /*! * \brief Submit a query to pacemaker-attrd and print reply * * \param[in,out] out Output object * \param[in] attr_name Name of attribute to be affected by request * \param[in] attr_node Name of host to query for (or NULL for localhost) * \param[in] query_all If TRUE, ignore attr_node and query all nodes * * \return Standard Pacemaker return code */ static int send_attrd_query(pcmk__output_t *out, const char *attr_name, const char *attr_node, gboolean query_all) { uint32_t options = pcmk__node_attr_none; pcmk_ipc_api_t *attrd_api = NULL; int rc = pcmk_rc_ok; // Create attrd IPC object rc = pcmk_new_ipc_api(&attrd_api, pcmk_ipc_attrd); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not connect to attrd: %s", pcmk_rc_str(rc)); return ENOTCONN; } pcmk_register_ipc_callback(attrd_api, attrd_event_cb, out); // Connect to attrd (without main loop) - rc = pcmk_connect_ipc(attrd_api, pcmk_ipc_dispatch_sync); + rc = pcmk__connect_ipc(attrd_api, pcmk_ipc_dispatch_sync, 5); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, - "Could not connect to attrd: %s", pcmk_rc_str(rc)); + "Could not connect to %s: %s", + pcmk_ipc_name(attrd_api, true), pcmk_rc_str(rc)); pcmk_free_ipc_api(attrd_api); return rc; } /* Decide which node(s) to query */ if (query_all == TRUE) { options |= pcmk__node_attr_query_all; } rc = pcmk__attrd_api_query(attrd_api, attr_node, attr_name, options); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not query value of %s: %s (%d)", attr_name, pcmk_strerror(rc), rc); } else if (!printed_values) { rc = pcmk_rc_schema_validation; g_set_error(&error, PCMK__RC_ERROR, rc, "Could not query value of %s: attribute does not exist", attr_name); } pcmk_disconnect_ipc(attrd_api); pcmk_free_ipc_api(attrd_api); return rc; } static int send_attrd_update(char command, const char *attr_node, const char *attr_name, const char *attr_value, const char *attr_set, const char *attr_dampen, uint32_t attr_options) { int rc = pcmk_rc_ok; switch (command) { case 'B': rc = pcmk__attrd_api_update(NULL, attr_node, attr_name, attr_value, attr_dampen, attr_set, NULL, attr_options | pcmk__node_attr_value | pcmk__node_attr_delay); break; case 'D': rc = pcmk__attrd_api_delete(NULL, attr_node, attr_name, attr_options); break; case 'R': rc = pcmk__attrd_api_refresh(NULL, attr_node); break; case 'U': rc = pcmk__attrd_api_update(NULL, attr_node, attr_name, attr_value, NULL, attr_set, NULL, attr_options | pcmk__node_attr_value); break; case 'Y': rc = pcmk__attrd_api_update(NULL, attr_node, attr_name, NULL, attr_dampen, attr_set, NULL, attr_options | pcmk__node_attr_delay); break; } if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not update %s=%s: %s (%d)", attr_name, attr_value, pcmk_rc_str(rc), rc); } return rc; } diff --git a/tools/crm_node.c b/tools/crm_node.c index ac2a19022b..b90bae2a62 100644 --- a/tools/crm_node.c +++ b/tools/crm_node.c @@ -1,601 +1,603 @@ /* * Copyright 2004-2023 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 #define SUMMARY "crm_node - Tool for displaying low-level node information" struct { gboolean corosync; gboolean dangerous_cmd; gboolean force_flag; char command; int nodeid; char *target_uname; } options = { .command = '\0', .force_flag = FALSE }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); static GMainLoop *mainloop = NULL; static crm_exit_t exit_code = CRM_EX_OK; #define INDENT " " static GOptionEntry command_entries[] = { { "cluster-id", 'i', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display this node's cluster id", NULL }, { "list", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display all known members (past and present) of this cluster", NULL }, { "name", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the name used by the cluster for this node", NULL }, { "partition", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the members of this partition", NULL }, { "quorum", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display a 1 if our partition has quorum, 0 if not", NULL }, { "name-for-id", 'N', 0, G_OPTION_ARG_CALLBACK, name_cb, "Display the name used by the cluster for the node with the specified ID", "ID" }, { "remove", 'R', 0, G_OPTION_ARG_CALLBACK, remove_cb, "(Advanced) Remove the (stopped) node with the specified name from Pacemaker's\n" INDENT "configuration and caches (the node must already have been removed from\n" INDENT "the underlying cluster stack configuration", "NAME" }, { NULL } }; static GOptionEntry addl_entries[] = { { "force", 'f', 0, G_OPTION_ARG_NONE, &options.force_flag, NULL, NULL }, #if SUPPORT_COROSYNC /* Unused and deprecated */ { "corosync", 'C', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.corosync, NULL, NULL }, #endif // @TODO add timeout option for when IPC replies are needed { NULL } }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_eq("-i", option_name, pcmk__str_casei) || pcmk__str_eq("--cluster-id", option_name, pcmk__str_casei)) { options.command = 'i'; } else if (pcmk__str_eq("-l", option_name, pcmk__str_casei) || pcmk__str_eq("--list", option_name, pcmk__str_casei)) { options.command = 'l'; } else if (pcmk__str_eq("-n", option_name, pcmk__str_casei) || pcmk__str_eq("--name", option_name, pcmk__str_casei)) { options.command = 'n'; } else if (pcmk__str_eq("-p", option_name, pcmk__str_casei) || pcmk__str_eq("--partition", option_name, pcmk__str_casei)) { options.command = 'p'; } else if (pcmk__str_eq("-q", option_name, pcmk__str_casei) || pcmk__str_eq("--quorum", option_name, pcmk__str_casei)) { options.command = 'q'; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Unknown param passed to command_cb: %s\n", option_name); return FALSE; } return TRUE; } gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.command = 'N'; pcmk__scan_min_int(optarg, &(options.nodeid), 0); return TRUE; } gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (optarg == NULL) { crm_err("-R option requires an argument"); g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "-R option requires an argument"); return FALSE; } options.command = 'R'; options.dangerous_cmd = TRUE; pcmk__str_update(&options.target_uname, optarg); return TRUE; } static gint sort_node(gconstpointer a, gconstpointer b) { const pcmk_controld_api_node_t *node_a = a; const pcmk_controld_api_node_t *node_b = b; return pcmk__numeric_strcasecmp((node_a->uname? node_a->uname : ""), (node_b->uname? node_b->uname : "")); } static void controller_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk_controld_api_reply_t *reply = event_data; switch (event_type) { case pcmk_ipc_event_disconnect: if (exit_code == CRM_EX_DISCONNECT) { // Unexpected fprintf(stderr, "error: Lost connection to controller\n"); } goto done; break; case pcmk_ipc_event_reply: break; default: return; } if (status != CRM_EX_OK) { fprintf(stderr, "error: Bad reply from controller: %s\n", crm_exit_str(status)); goto done; } // Parse desired info from reply and display to user switch (options.command) { case 'i': if (reply->reply_type != pcmk_controld_reply_info) { fprintf(stderr, "error: Unknown reply type %d from controller\n", reply->reply_type); goto done; } if (reply->data.node_info.id == 0) { fprintf(stderr, "error: Controller reply did not contain node ID\n"); exit_code = CRM_EX_PROTOCOL; goto done; } printf("%d\n", reply->data.node_info.id); break; case 'n': case 'N': if (reply->reply_type != pcmk_controld_reply_info) { fprintf(stderr, "error: Unknown reply type %d from controller\n", reply->reply_type); goto done; } if (reply->data.node_info.uname == NULL) { fprintf(stderr, "Node is not known to cluster\n"); exit_code = CRM_EX_NOHOST; goto done; } printf("%s\n", reply->data.node_info.uname); break; case 'q': if (reply->reply_type != pcmk_controld_reply_info) { fprintf(stderr, "error: Unknown reply type %d from controller\n", reply->reply_type); goto done; } printf("%d\n", reply->data.node_info.have_quorum); if (!(reply->data.node_info.have_quorum)) { exit_code = CRM_EX_QUORUM; goto done; } break; case 'l': case 'p': if (reply->reply_type != pcmk_controld_reply_nodes) { fprintf(stderr, "error: Unknown reply type %d from controller\n", reply->reply_type); goto done; } reply->data.nodes = g_list_sort(reply->data.nodes, sort_node); for (GList *node_iter = reply->data.nodes; node_iter != NULL; node_iter = node_iter->next) { pcmk_controld_api_node_t *node = node_iter->data; const char *uname = (node->uname? node->uname : ""); const char *state = (node->state? node->state : ""); if (options.command == 'l') { printf("%lu %s %s\n", (unsigned long) node->id, uname, state); // i.e. CRM_NODE_MEMBER, but we don't want to include cluster.h } else if (!strcmp(state, "member")) { printf("%s ", uname); } } if (options.command == 'p') { printf("\n"); } break; default: fprintf(stderr, "internal error: Controller reply not expected\n"); exit_code = CRM_EX_SOFTWARE; goto done; } // Success exit_code = CRM_EX_OK; done: pcmk_disconnect_ipc(controld_api); pcmk_quit_main_loop(mainloop, 10); } static void run_controller_mainloop(uint32_t nodeid, bool list_nodes) { pcmk_ipc_api_t *controld_api = NULL; int rc; // Set disconnect exit code to handle unexpected disconnects exit_code = CRM_EX_DISCONNECT; // Create controller IPC object rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld); if (rc != pcmk_rc_ok) { fprintf(stderr, "error: Could not connect to controller: %s\n", pcmk_rc_str(rc)); return; } pcmk_register_ipc_callback(controld_api, controller_event_cb, NULL); // Connect to controller - rc = pcmk_connect_ipc(controld_api, pcmk_ipc_dispatch_main); + rc = pcmk__connect_ipc(controld_api, pcmk_ipc_dispatch_main, 5); if (rc != pcmk_rc_ok) { - fprintf(stderr, "error: Could not connect to controller: %s\n", - pcmk_rc_str(rc)); + fprintf(stderr, "error: Could not connect to %s: %s\n", + pcmk_ipc_name(controld_api, true), pcmk_rc_str(rc)); exit_code = pcmk_rc2exitc(rc); return; } if (list_nodes) { rc = pcmk_controld_api_list_nodes(controld_api); } else { rc = pcmk_controld_api_node_info(controld_api, nodeid); } if (rc != pcmk_rc_ok) { fprintf(stderr, "error: Could not ping controller: %s\n", pcmk_rc_str(rc)); pcmk_disconnect_ipc(controld_api); exit_code = pcmk_rc2exitc(rc); return; } // Run main loop to get controller reply via controller_event_cb() mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); mainloop = NULL; pcmk_free_ipc_api(controld_api); } static void print_node_name(void) { // Check environment first (i.e. when called by resource agent) const char *name = getenv("OCF_RESKEY_" CRM_META "_" XML_LRM_ATTR_TARGET); if (name != NULL) { printf("%s\n", name); exit_code = CRM_EX_OK; return; } else { /* Otherwise ask the controller. * FIXME: Use pcmk__query_node_name() after conversion to formatted * output. */ run_controller_mainloop(0, false); } } static int cib_remove_node(long id, const char *name) { int rc; cib_t *cib = NULL; xmlNode *node = NULL; xmlNode *node_state = NULL; crm_trace("Removing %s from the CIB", name); if(name == NULL && id == 0) { return -ENOTUNIQ; } node = create_xml_node(NULL, XML_CIB_TAG_NODE); node_state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(node, XML_ATTR_UNAME, name); crm_xml_add(node_state, XML_ATTR_UNAME, name); if (id > 0) { crm_xml_set_id(node, "%ld", id); crm_xml_add(node_state, XML_ATTR_ID, ID(node)); } cib = cib_new(); cib->cmds->signon(cib, crm_system_name, cib_command); rc = cib->cmds->remove(cib, XML_CIB_TAG_NODES, node, cib_sync_call); if (rc != pcmk_ok) { printf("Could not remove %s[%ld] from " XML_CIB_TAG_NODES ": %s", name, id, pcmk_strerror(rc)); } rc = cib->cmds->remove(cib, XML_CIB_TAG_STATUS, node_state, cib_sync_call); if (rc != pcmk_ok) { printf("Could not remove %s[%ld] from " XML_CIB_TAG_STATUS ": %s", name, id, pcmk_strerror(rc)); } cib__clean_up_connection(&cib); return rc; } static int controller_remove_node(const char *node_name, long nodeid) { pcmk_ipc_api_t *controld_api = NULL; int rc; // Create controller IPC object rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld); if (rc != pcmk_rc_ok) { fprintf(stderr, "error: Could not connect to controller: %s\n", pcmk_rc_str(rc)); return ENOTCONN; } // Connect to controller (without main loop) - rc = pcmk_connect_ipc(controld_api, pcmk_ipc_dispatch_sync); + rc = pcmk__connect_ipc(controld_api, pcmk_ipc_dispatch_sync, 5); if (rc != pcmk_rc_ok) { - fprintf(stderr, "error: Could not connect to controller: %s\n", - pcmk_rc_str(rc)); + fprintf(stderr, "error: Could not connect to %s: %s\n", + pcmk_ipc_name(controld_api, true), pcmk_rc_str(rc)); pcmk_free_ipc_api(controld_api); return rc; } rc = pcmk_ipc_purge_node(controld_api, node_name, nodeid); if (rc != pcmk_rc_ok) { fprintf(stderr, "error: Could not clear node from controller's cache: %s\n", pcmk_rc_str(rc)); } pcmk_free_ipc_api(controld_api); return pcmk_rc_ok; } static int tools_remove_node_cache(const char *node_name, long nodeid, const char *target) { int rc = -1; crm_ipc_t *conn = NULL; xmlNode *cmd = NULL; conn = crm_ipc_new(target, 0); if (!conn) { return -ENOTCONN; } - if (!crm_ipc_connect(conn)) { + rc = pcmk__connect_generic_ipc(conn); + if (rc != pcmk_rc_ok) { + errno = (rc > 0)? rc : ENOTCONN; crm_perror(LOG_ERR, "Connection to %s failed", target); crm_ipc_destroy(conn); return -ENOTCONN; } crm_trace("Removing %s[%ld] from the %s membership cache", node_name, nodeid, target); if(pcmk__str_eq(target, T_ATTRD, pcmk__str_casei)) { cmd = create_xml_node(NULL, __func__); crm_xml_add(cmd, F_TYPE, T_ATTRD); crm_xml_add(cmd, F_ORIG, crm_system_name); crm_xml_add(cmd, PCMK__XA_TASK, PCMK__ATTRD_CMD_PEER_REMOVE); pcmk__xe_add_node(cmd, node_name, nodeid); } else { // Fencer or pacemakerd cmd = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, target, crm_system_name, NULL); if (nodeid > 0) { crm_xml_set_id(cmd, "%ld", nodeid); } crm_xml_add(cmd, XML_ATTR_UNAME, node_name); } rc = crm_ipc_send(conn, cmd, 0, 0, NULL); crm_debug("%s peer cache cleanup for %s (%ld): %d", target, node_name, nodeid, rc); if (rc > 0) { // @TODO Should this be done just once after all the rest? rc = cib_remove_node(nodeid, node_name); } if (conn) { crm_ipc_close(conn); crm_ipc_destroy(conn); } free_xml(cmd); return rc > 0 ? 0 : rc; } static void remove_node(const char *target_uname) { int rc; int d = 0; long nodeid = 0; const char *node_name = NULL; char *endptr = NULL; const char *daemons[] = { "stonith-ng", T_ATTRD, CRM_SYSTEM_MCP, }; // Check whether node was specified by name or numeric ID errno = 0; nodeid = strtol(target_uname, &endptr, 10); if ((errno != 0) || (endptr == target_uname) || (*endptr != '\0') || (nodeid <= 0)) { // It's not a positive integer, so assume it's a node name nodeid = 0; node_name = target_uname; } rc = controller_remove_node(node_name, nodeid); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); return; } for (d = 0; d < PCMK__NELEM(daemons); d++) { if (tools_remove_node_cache(node_name, nodeid, daemons[d])) { crm_err("Failed to connect to %s to remove node '%s'", daemons[d], target_uname); exit_code = CRM_EX_ERROR; return; } } exit_code = CRM_EX_OK; } 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 } }; context = pcmk__build_arg_context(args, NULL, &group, NULL); /* Add the -q option, which cannot be part of the globally supported options * because some tools use that flag for something else. */ pcmk__add_main_args(context, extra_prog_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) { GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "NR"); GOptionContext *context = build_arg_context(args, output_group); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("crm_node", args->verbosity); if (args->version) { g_strfreev(processed_args); pcmk__free_arg_context(context); /* FIXME: When crm_node is converted to use formatted output, this can go. */ pcmk__cli_help('v'); } if (options.command == 0) { char *help = g_option_context_get_help(context, TRUE, NULL); fprintf(stderr, "%s", help); g_free(help); exit_code = CRM_EX_USAGE; goto done; } if (options.dangerous_cmd && options.force_flag == FALSE) { fprintf(stderr, "The supplied command is considered dangerous." " To prevent accidental destruction of the cluster," " the --force flag is required in order to proceed.\n"); exit_code = CRM_EX_USAGE; goto done; } switch (options.command) { case 'n': print_node_name(); break; case 'R': remove_node(options.target_uname); break; case 'i': case 'q': case 'N': /* FIXME: Use pcmk__query_node_name() after conversion to formatted * output */ run_controller_mainloop(options.nodeid, false); break; case 'l': case 'p': run_controller_mainloop(0, true); break; default: break; } done: g_strfreev(processed_args); pcmk__free_arg_context(context); pcmk__output_and_clear_error(&error, NULL); return crm_exit(exit_code); } diff --git a/tools/crm_resource.c b/tools/crm_resource.c index e54939bf15..a0ccef94d8 100644 --- a/tools/crm_resource.c +++ b/tools/crm_resource.c @@ -1,2182 +1,2183 @@ /* * Copyright 2004-2023 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 #define SUMMARY "crm_resource - perform tasks related to Pacemaker cluster resources" enum rsc_command { cmd_none = 0, // No command option given (yet) cmd_ban, cmd_cleanup, cmd_clear, cmd_colocations, cmd_cts, cmd_delete, cmd_delete_param, cmd_digests, cmd_execute_agent, cmd_fail, cmd_get_param, cmd_get_property, cmd_list_active_ops, cmd_list_agents, cmd_list_all_ops, cmd_list_alternatives, cmd_list_instances, cmd_list_providers, cmd_list_resources, cmd_list_standards, cmd_locate, cmd_metadata, cmd_move, cmd_query_raw_xml, cmd_query_xml, cmd_refresh, cmd_restart, cmd_set_param, cmd_set_property, cmd_wait, cmd_why, }; struct { enum rsc_command rsc_cmd; // crm_resource command to perform // Infrastructure that given command needs to work gboolean require_cib; // Whether command requires CIB IPC int cib_options; // Options to use with CIB IPC calls gboolean require_crmd; // Whether command requires controller IPC gboolean require_dataset; // Whether command requires populated data set gboolean require_resource; // Whether command requires resource specified gboolean require_node; // Whether command requires node specified int find_flags; // Flags to use when searching for resource // Command-line option values gchar *rsc_id; // Value of --resource gchar *rsc_type; // Value of --resource-type gboolean force; // --force was given gboolean clear_expired; // --expired was given gboolean recursive; // --recursive was given gboolean promoted_role_only; // --promoted was given gchar *host_uname; // Value of --node gchar *interval_spec; // Value of --interval gchar *move_lifetime; // Value of --lifetime gchar *operation; // Value of --operation const char *attr_set_type; // Instance, meta, utilization, or element attribute gchar *prop_id; // --nvpair (attribute XML ID) char *prop_name; // Attribute name gchar *prop_set; // --set-name (attribute block XML ID) gchar *prop_value; // --parameter-value (attribute value) int timeout_ms; // Parsed from --timeout value char *agent_spec; // Standard and/or provider and/or agent gchar *xml_file; // Value of (deprecated) --xml-file int check_level; // Optional value of --validate or --force-check // Resource configuration specified via command-line arguments gboolean cmdline_config; // Resource configuration was via arguments char *v_agent; // Value of --agent char *v_class; // Value of --class char *v_provider; // Value of --provider GHashTable *cmdline_params; // Resource parameters specified // Positional command-line arguments gchar **remainder; // Positional arguments as given GHashTable *override_params; // Resource parameter values that override config } options = { .attr_set_type = XML_TAG_ATTR_SETS, .check_level = -1, .cib_options = cib_sync_call, .require_cib = TRUE, .require_dataset = TRUE, .require_resource = TRUE, }; #if 0 // @COMPAT @TODO enable this at next backward compatibility break #define SET_COMMAND(cmd) do { \ if (options.rsc_cmd != cmd_none) { \ g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, \ "Only one command option may be specified"); \ return FALSE; \ } \ options.rsc_cmd = (cmd); \ } while (0) #else #define SET_COMMAND(cmd) do { \ if (options.rsc_cmd != cmd_none) { \ reset_options(); \ } \ options.rsc_cmd = (cmd); \ } while (0) #endif gboolean agent_provider_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean attr_set_type_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean class_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean cleanup_refresh_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean delete_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean expired_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean list_agents_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean list_providers_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean list_standards_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean list_alternatives_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean metadata_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean option_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean flag_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean get_param_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean list_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean set_delete_param_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean set_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean timeout_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean validate_or_force_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean restart_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean digests_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean wait_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean why_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); static crm_exit_t exit_code = CRM_EX_OK; static pcmk__output_t *out = NULL; static pcmk__common_args_t *args = NULL; // Things that should be cleaned up on exit static GError *error = NULL; static GMainLoop *mainloop = NULL; static cib_t *cib_conn = NULL; static pcmk_ipc_api_t *controld_api = NULL; static pe_working_set_t *data_set = NULL; #define MESSAGE_TIMEOUT_S 60 #define INDENT " " static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; // Clean up and exit static crm_exit_t bye(crm_exit_t ec) { pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, ec, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); if (cib_conn != NULL) { cib_t *save_cib_conn = cib_conn; cib_conn = NULL; // Ensure we can't free this twice cib__clean_up_connection(&save_cib_conn); } if (controld_api != NULL) { pcmk_ipc_api_t *save_controld_api = controld_api; controld_api = NULL; // Ensure we can't free this twice pcmk_free_ipc_api(save_controld_api); } if (mainloop != NULL) { g_main_loop_unref(mainloop); mainloop = NULL; } pe_free_working_set(data_set); data_set = NULL; crm_exit(ec); return ec; } static void quit_main_loop(crm_exit_t ec) { exit_code = ec; if (mainloop != NULL) { GMainLoop *mloop = mainloop; mainloop = NULL; // Don't re-enter this block pcmk_quit_main_loop(mloop, 10); g_main_loop_unref(mloop); } } static gboolean resource_ipc_timeout(gpointer data) { // Start with newline because "Waiting for ..." message doesn't have one if (error != NULL) { g_clear_error(&error); } g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_TIMEOUT, _("Aborting because no messages received in %d seconds"), MESSAGE_TIMEOUT_S); quit_main_loop(CRM_EX_TIMEOUT); return FALSE; } static void controller_event_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { switch (event_type) { case pcmk_ipc_event_disconnect: if (exit_code == CRM_EX_DISCONNECT) { // Unexpected crm_info("Connection to controller was terminated"); } quit_main_loop(exit_code); break; case pcmk_ipc_event_reply: if (status != CRM_EX_OK) { out->err(out, "Error: bad reply from controller: %s", crm_exit_str(status)); pcmk_disconnect_ipc(api); quit_main_loop(status); } else { if ((pcmk_controld_api_replies_expected(api) == 0) && mainloop && g_main_loop_is_running(mainloop)) { out->info(out, "... got reply (done)"); crm_debug("Got all the replies we expected"); pcmk_disconnect_ipc(api); quit_main_loop(CRM_EX_OK); } else { out->info(out, "... got reply"); } } break; default: break; } } static void start_mainloop(pcmk_ipc_api_t *capi) { unsigned int count = pcmk_controld_api_replies_expected(capi); if (count > 0) { out->info(out, "Waiting for %u %s from the controller", count, pcmk__plural_alt(count, "reply", "replies")); exit_code = CRM_EX_DISCONNECT; // For unexpected disconnects mainloop = g_main_loop_new(NULL, FALSE); g_timeout_add(MESSAGE_TIMEOUT_S * 1000, resource_ipc_timeout, NULL); g_main_loop_run(mainloop); } } static int compare_id(gconstpointer a, gconstpointer b) { return strcmp((const char *)a, (const char *)b); } static GList * build_constraint_list(xmlNode *root) { GList *retval = NULL; xmlNode *cib_constraints = NULL; xmlXPathObjectPtr xpathObj = NULL; int ndx = 0; cib_constraints = pcmk_find_cib_element(root, XML_CIB_TAG_CONSTRAINTS); xpathObj = xpath_search(cib_constraints, "//" XML_CONS_TAG_RSC_LOCATION); for (ndx = 0; ndx < numXpathResults(xpathObj); ndx++) { xmlNode *match = getXpathResult(xpathObj, ndx); retval = g_list_insert_sorted(retval, (gpointer) ID(match), compare_id); } freeXpathObject(xpathObj); return retval; } /* short option letters still available: eEJkKXyYZ */ static GOptionEntry query_entries[] = { { "list", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb, "List all cluster resources with status", NULL }, { "list-raw", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb, "List IDs of all instantiated resources (individual members\n" INDENT "rather than groups etc.)", NULL }, { "list-cts", 'c', G_OPTION_FLAG_HIDDEN|G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb, NULL, NULL }, { "list-operations", 'O', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb, "List active resource operations, optionally filtered by\n" INDENT "--resource and/or --node", NULL }, { "list-all-operations", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_cb, "List all resource operations, optionally filtered by\n" INDENT "--resource and/or --node", NULL }, { "list-standards", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_standards_cb, "List supported standards", NULL }, { "list-ocf-providers", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, list_providers_cb, "List all available OCF providers", NULL }, { "list-agents", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, list_agents_cb, "List all agents available for the named standard and/or provider", "STD:PROV" }, { "list-ocf-alternatives", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, list_alternatives_cb, "List all available providers for the named OCF agent", "AGENT" }, { "show-metadata", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, metadata_cb, "Show the metadata for the named class:provider:agent", "SPEC" }, { "query-xml", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Show XML configuration of resource (after any template expansion)", NULL }, { "query-xml-raw", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Show XML configuration of resource (before any template expansion)", NULL }, { "get-parameter", 'g', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, get_param_prop_cb, "Display named parameter for resource (use instance attribute\n" INDENT "unless --element, --meta, or --utilization is specified)", "PARAM" }, { "get-property", 'G', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, get_param_prop_cb, "Display named property of resource ('class', 'type', or 'provider') " "(requires --resource)", "PROPERTY" }, { "locate", 'W', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Show node(s) currently running resource", NULL }, { "constraints", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Display the location and colocation constraints that apply to a\n" INDENT "resource, and if --recursive is specified, to the resources\n" INDENT "directly or indirectly involved in those colocations.\n" INDENT "If the named resource is part of a group, or a clone or\n" INDENT "bundle instance, constraints for the collective resource\n" INDENT "will be shown unless --force is given.", NULL }, { "stack", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Equivalent to --constraints --recursive", NULL }, { "why", 'Y', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, why_cb, "Show why resources are not running, optionally filtered by\n" INDENT "--resource and/or --node", NULL }, { NULL } }; static GOptionEntry command_entries[] = { { "validate", 0, G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "Validate resource configuration by calling agent's validate-all\n" INDENT "action. The configuration may be specified either by giving an\n" INDENT "existing resource name with -r, or by specifying --class,\n" INDENT "--agent, and --provider arguments, along with any number of\n" INDENT "--option arguments. An optional LEVEL argument can be given\n" INDENT "to control the level of checking performed.", "LEVEL" }, { "cleanup", 'C', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, cleanup_refresh_cb, "If resource has any past failures, clear its history and fail\n" INDENT "count. Optionally filtered by --resource, --node, --operation\n" INDENT "and --interval (otherwise all). --operation and --interval\n" INDENT "apply to fail counts, but entire history is always clear, to\n" INDENT "allow current state to be rechecked. If the named resource is\n" INDENT "part of a group, or one numbered instance of a clone or bundled\n" INDENT "resource, the clean-up applies to the whole collective resource\n" INDENT "unless --force is given.", NULL }, { "refresh", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, cleanup_refresh_cb, "Delete resource's history (including failures) so its current state\n" INDENT "is rechecked. Optionally filtered by --resource and --node\n" INDENT "(otherwise all). If the named resource is part of a group, or one\n" INDENT "numbered instance of a clone or bundled resource, the refresh\n" INDENT "applies to the whole collective resource unless --force is given.", NULL }, { "set-parameter", 'p', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, set_delete_param_cb, "Set named parameter for resource (requires -v). Use instance\n" INDENT "attribute unless --element, --meta, or --utilization is " "specified.", "PARAM" }, { "delete-parameter", 'd', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, set_delete_param_cb, "Delete named parameter for resource. Use instance attribute\n" INDENT "unless --element, --meta or, --utilization is specified.", "PARAM" }, { "set-property", 'S', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, set_prop_cb, "Set named property of resource ('class', 'type', or 'provider') " "(requires -r, -t, -v)", "PROPERTY" }, { NULL } }; static GOptionEntry location_entries[] = { { "move", 'M', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Create a constraint to move resource. If --node is specified,\n" INDENT "the constraint will be to move to that node, otherwise it\n" INDENT "will be to ban the current node. Unless --force is specified\n" INDENT "this will return an error if the resource is already running\n" INDENT "on the specified node. If --force is specified, this will\n" INDENT "always ban the current node.\n" INDENT "Optional: --lifetime, --promoted. NOTE: This may prevent the\n" INDENT "resource from running on its previous location until the\n" INDENT "implicit constraint expires or is removed with --clear.", NULL }, { "ban", 'B', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Create a constraint to keep resource off a node.\n" INDENT "Optional: --node, --lifetime, --promoted.\n" INDENT "NOTE: This will prevent the resource from running on the\n" INDENT "affected node until the implicit constraint expires or is\n" INDENT "removed with --clear. If --node is not specified, it defaults\n" INDENT "to the node currently running the resource for primitives\n" INDENT "and groups, or the promoted instance of promotable clones with\n" INDENT "promoted-max=1 (all other situations result in an error as\n" INDENT "there is no sane default).", NULL }, { "clear", 'U', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, flag_cb, "Remove all constraints created by the --ban and/or --move\n" INDENT "commands. Requires: --resource. Optional: --node, --promoted,\n" INDENT "--expired. If --node is not specified, all constraints created\n" INDENT "by --ban and --move will be removed for the named resource. If\n" INDENT "--node and --force are specified, any constraint created by\n" INDENT "--move will be cleared, even if it is not for the specified\n" INDENT "node. If --expired is specified, only those constraints whose\n" INDENT "lifetimes have expired will be removed.", NULL }, { "expired", 'e', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, expired_cb, "Modifies the --clear argument to remove constraints with\n" INDENT "expired lifetimes.", NULL }, { "lifetime", 'u', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.move_lifetime, "Lifespan (as ISO 8601 duration) of created constraints (with\n" INDENT "-B, -M) see https://en.wikipedia.org/wiki/ISO_8601#Durations)", "TIMESPEC" }, { "promoted", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.promoted_role_only, "Limit scope of command to promoted role (with -B, -M, -U). For\n" INDENT "-B and -M, previously promoted instances may remain\n" INDENT "active in the unpromoted role.", NULL }, // Deprecated since 2.1.0 { "master", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.promoted_role_only, "Deprecated: Use --promoted instead", NULL }, { NULL } }; static GOptionEntry advanced_entries[] = { { "delete", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, delete_cb, "(Advanced) Delete a resource from the CIB. Required: -t", NULL }, { "fail", 'F', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, fail_cb, "(Advanced) Tell the cluster this resource has failed", NULL }, { "restart", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, restart_cb, "(Advanced) Tell the cluster to restart this resource and\n" INDENT "anything that depends on it", NULL }, { "wait", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, wait_cb, "(Advanced) Wait until the cluster settles into a stable state", NULL }, { "digests", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, digests_cb, "(Advanced) Show parameter hashes that Pacemaker uses to detect\n" INDENT "configuration changes (only accurate if there is resource\n" INDENT "history on the specified node). Required: --resource, --node.\n" INDENT "Optional: any NAME=VALUE parameters will be used to override\n" INDENT "the configuration (to see what the hash would be with those\n" INDENT "changes).", NULL }, { "force-demote", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "(Advanced) Bypass the cluster and demote a resource on the local\n" INDENT "node. Unless --force is specified, this will refuse to do so if\n" INDENT "the cluster believes the resource is a clone instance already\n" INDENT "running on the local node.", NULL }, { "force-stop", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "(Advanced) Bypass the cluster and stop a resource on the local node", NULL }, { "force-start", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "(Advanced) Bypass the cluster and start a resource on the local\n" INDENT "node. Unless --force is specified, this will refuse to do so if\n" INDENT "the cluster believes the resource is a clone instance already\n" INDENT "running on the local node.", NULL }, { "force-promote", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "(Advanced) Bypass the cluster and promote a resource on the local\n" INDENT "node. Unless --force is specified, this will refuse to do so if\n" INDENT "the cluster believes the resource is a clone instance already\n" INDENT "running on the local node.", NULL }, { "force-check", 0, G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, validate_or_force_cb, "(Advanced) Bypass the cluster and check the state of a resource on\n" INDENT "the local node. An optional LEVEL argument can be given\n" INDENT "to control the level of checking performed.", "LEVEL" }, { NULL } }; static GOptionEntry addl_entries[] = { { "node", 'N', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.host_uname, "Node name", "NAME" }, { "recursive", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.recursive, "Follow colocation chains when using --set-parameter or --constraints", NULL }, { "resource-type", 't', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.rsc_type, "Resource XML element (primitive, group, etc.) (with -D)", "ELEMENT" }, { "parameter-value", 'v', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_value, "Value to use with -p", "PARAM" }, { "meta", 'm', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb, "Use resource meta-attribute instead of instance attribute\n" INDENT "(with -p, -g, -d)", NULL }, { "utilization", 'z', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb, "Use resource utilization attribute instead of instance attribute\n" INDENT "(with -p, -g, -d)", NULL }, { "element", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attr_set_type_cb, "Use resource element attribute instead of instance attribute\n" INDENT "(with -p, -g, -d)", NULL }, { "operation", 'n', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.operation, "Operation to clear instead of all (with -C -r)", "OPERATION" }, { "interval", 'I', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.interval_spec, "Interval of operation to clear (default 0) (with -C -r -n)", "N" }, { "class", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, class_cb, "The standard the resource agent conforms to (for example, ocf).\n" INDENT "Use with --agent, --provider, --option, and --validate.", "CLASS" }, { "agent", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, agent_provider_cb, "The agent to use (for example, IPaddr). Use with --class,\n" INDENT "--provider, --option, and --validate.", "AGENT" }, { "provider", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, agent_provider_cb, "The vendor that supplies the resource agent (for example,\n" INDENT "heartbeat). Use with --class, --agent, --option, and --validate.", "PROVIDER" }, { "option", 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, option_cb, "Specify a device configuration parameter as NAME=VALUE (may be\n" INDENT "specified multiple times). Use with --validate and without the\n" INDENT "-r option.", "PARAM" }, { "set-name", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_set, "(Advanced) XML ID of attributes element to use (with -p, -d)", "ID" }, { "nvpair", 'i', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.prop_id, "(Advanced) XML ID of nvpair element to use (with -p, -d)", "ID" }, { "timeout", 'T', G_OPTION_FLAG_NONE, G_OPTION_ARG_CALLBACK, timeout_cb, "(Advanced) Abort if command does not finish in this time (with\n" INDENT "--restart, --wait, --force-*)", "N" }, { "force", 'f', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &options.force, "Force the action to be performed. See help for individual commands for\n" INDENT "additional behavior.", NULL }, { "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_FILENAME, &options.xml_file, NULL, "FILE" }, { "host-uname", 'H', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_STRING, &options.host_uname, NULL, "HOST" }, { NULL } }; static void reset_options(void) { options.require_crmd = FALSE; options.require_node = FALSE; options.require_cib = TRUE; options.require_dataset = TRUE; options.require_resource = TRUE; options.find_flags = 0; } gboolean agent_provider_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.cmdline_config = TRUE; options.require_resource = FALSE; if (pcmk__str_eq(option_name, "--provider", pcmk__str_casei)) { pcmk__str_update(&options.v_provider, optarg); } else { pcmk__str_update(&options.v_agent, optarg); } return TRUE; } gboolean attr_set_type_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-m", "--meta", NULL)) { options.attr_set_type = XML_TAG_META_SETS; } else if (pcmk__str_any_of(option_name, "-z", "--utilization", NULL)) { options.attr_set_type = XML_TAG_UTILIZATION; } else if (pcmk__str_eq(option_name, "--element", pcmk__str_casei)) { options.attr_set_type = ATTR_SET_ELEMENT; } return TRUE; } gboolean class_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__str_update(&options.v_class, optarg); options.cmdline_config = TRUE; options.require_resource = FALSE; return TRUE; } gboolean cleanup_refresh_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-C", "--cleanup", NULL)) { SET_COMMAND(cmd_cleanup); } else { SET_COMMAND(cmd_refresh); } options.require_resource = FALSE; if (getenv("CIB_file") == NULL) { options.require_crmd = TRUE; } options.find_flags = pe_find_renamed|pe_find_anon; return TRUE; } gboolean delete_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_delete); options.require_dataset = FALSE; options.find_flags = pe_find_renamed|pe_find_any; return TRUE; } gboolean expired_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.clear_expired = TRUE; options.require_resource = FALSE; return TRUE; } static void get_agent_spec(const gchar *optarg) { options.require_cib = FALSE; options.require_dataset = FALSE; options.require_resource = FALSE; pcmk__str_update(&options.agent_spec, optarg); } gboolean list_agents_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_list_agents); get_agent_spec(optarg); return TRUE; } gboolean list_providers_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_list_providers); get_agent_spec(optarg); return TRUE; } gboolean list_standards_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_list_standards); options.require_cib = FALSE; options.require_dataset = FALSE; options.require_resource = FALSE; return TRUE; } gboolean list_alternatives_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_list_alternatives); get_agent_spec(optarg); return TRUE; } gboolean metadata_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_metadata); get_agent_spec(optarg); return TRUE; } gboolean option_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { char *name = NULL; char *value = NULL; if (pcmk__scan_nvpair(optarg, &name, &value) != 2) { return FALSE; } if (options.cmdline_params == NULL) { options.cmdline_params = pcmk__strkey_table(free, free); } g_hash_table_replace(options.cmdline_params, name, value); return TRUE; } gboolean fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_fail); options.require_crmd = TRUE; options.require_node = TRUE; return TRUE; } gboolean flag_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-U", "--clear", NULL)) { SET_COMMAND(cmd_clear); options.find_flags = pe_find_renamed|pe_find_anon; } else if (pcmk__str_any_of(option_name, "-B", "--ban", NULL)) { SET_COMMAND(cmd_ban); options.find_flags = pe_find_renamed|pe_find_anon; } else if (pcmk__str_any_of(option_name, "-M", "--move", NULL)) { SET_COMMAND(cmd_move); options.find_flags = pe_find_renamed|pe_find_anon; } else if (pcmk__str_any_of(option_name, "-q", "--query-xml", NULL)) { SET_COMMAND(cmd_query_xml); options.find_flags = pe_find_renamed|pe_find_any; } else if (pcmk__str_any_of(option_name, "-w", "--query-xml-raw", NULL)) { SET_COMMAND(cmd_query_raw_xml); options.find_flags = pe_find_renamed|pe_find_any; } else if (pcmk__str_any_of(option_name, "-W", "--locate", NULL)) { SET_COMMAND(cmd_locate); options.find_flags = pe_find_renamed|pe_find_anon; } else if (pcmk__str_any_of(option_name, "-a", "--constraints", NULL)) { SET_COMMAND(cmd_colocations); options.find_flags = pe_find_renamed|pe_find_anon; } else if (pcmk__str_any_of(option_name, "-A", "--stack", NULL)) { SET_COMMAND(cmd_colocations); options.find_flags = pe_find_renamed|pe_find_anon; options.recursive = TRUE; } return TRUE; } gboolean get_param_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-g", "--get-parameter", NULL)) { SET_COMMAND(cmd_get_param); } else { SET_COMMAND(cmd_get_property); } pcmk__str_update(&options.prop_name, optarg); options.find_flags = pe_find_renamed|pe_find_any; return TRUE; } gboolean list_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-c", "--list-cts", NULL)) { SET_COMMAND(cmd_cts); } else if (pcmk__str_any_of(option_name, "-L", "--list", NULL)) { SET_COMMAND(cmd_list_resources); } else if (pcmk__str_any_of(option_name, "-l", "--list-raw", NULL)) { SET_COMMAND(cmd_list_instances); } else if (pcmk__str_any_of(option_name, "-O", "--list-operations", NULL)) { SET_COMMAND(cmd_list_active_ops); } else { SET_COMMAND(cmd_list_all_ops); } options.require_resource = FALSE; return TRUE; } gboolean set_delete_param_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_any_of(option_name, "-p", "--set-parameter", NULL)) { SET_COMMAND(cmd_set_param); } else { SET_COMMAND(cmd_delete_param); } pcmk__str_update(&options.prop_name, optarg); options.find_flags = pe_find_renamed|pe_find_any; return TRUE; } gboolean set_prop_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_set_property); options.require_dataset = FALSE; pcmk__str_update(&options.prop_name, optarg); options.find_flags = pe_find_renamed|pe_find_any; return TRUE; } gboolean timeout_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.timeout_ms = crm_get_msec(optarg); return TRUE; } gboolean validate_or_force_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_execute_agent); if (options.operation) { g_free(options.operation); } options.operation = g_strdup(option_name + 2); // skip "--" options.find_flags = pe_find_renamed|pe_find_anon; if (options.override_params == NULL) { options.override_params = pcmk__strkey_table(free, free); } if (optarg != NULL) { if (pcmk__scan_min_int(optarg, &options.check_level, 0) != pcmk_rc_ok) { g_set_error(error, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, _("Invalid check level setting: %s"), optarg); return FALSE; } } return TRUE; } gboolean restart_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_restart); options.find_flags = pe_find_renamed|pe_find_anon; return TRUE; } gboolean digests_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_digests); options.find_flags = pe_find_renamed|pe_find_anon; if (options.override_params == NULL) { options.override_params = pcmk__strkey_table(free, free); } options.require_node = TRUE; options.require_dataset = TRUE; return TRUE; } gboolean wait_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_wait); options.require_resource = FALSE; options.require_dataset = FALSE; return TRUE; } gboolean why_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { SET_COMMAND(cmd_why); options.require_resource = FALSE; options.find_flags = pe_find_renamed|pe_find_anon; return TRUE; } static int ban_or_move(pcmk__output_t *out, pe_resource_t *rsc, const char *move_lifetime) { int rc = pcmk_rc_ok; pe_node_t *current = NULL; unsigned int nactive = 0; CRM_CHECK(rsc != NULL, return EINVAL); current = pe__find_active_requires(rsc, &nactive); if (nactive == 1) { rc = cli_resource_ban(out, options.rsc_id, current->details->uname, move_lifetime, NULL, cib_conn, options.cib_options, options.promoted_role_only); } else if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { int count = 0; GList *iter = NULL; current = NULL; for(iter = rsc->children; iter; iter = iter->next) { pe_resource_t *child = (pe_resource_t *)iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if (child_role == RSC_ROLE_PROMOTED) { count++; current = pe__current_node(child); } } if(count == 1 && current) { rc = cli_resource_ban(out, options.rsc_id, current->details->uname, move_lifetime, NULL, cib_conn, options.cib_options, options.promoted_role_only); } else { rc = EINVAL; g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("Resource '%s' not moved: active in %d locations (promoted in %d).\n" "To prevent '%s' from running on a specific location, " "specify a node." "To prevent '%s' from being promoted at a specific " "location, specify a node and the --promoted option."), options.rsc_id, nactive, count, options.rsc_id, options.rsc_id); } } else { rc = EINVAL; g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("Resource '%s' not moved: active in %d locations.\n" "To prevent '%s' from running on a specific location, " "specify a node."), options.rsc_id, nactive, options.rsc_id); } return rc; } static void cleanup(pcmk__output_t *out, pe_resource_t *rsc, pe_node_t *node) { int rc = pcmk_rc_ok; if (options.force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Erasing failures of %s (%s requested) on %s", rsc->id, options.rsc_id, (options.host_uname? options.host_uname: "all nodes")); rc = cli_resource_delete(controld_api, options.host_uname, rsc, options.operation, options.interval_spec, TRUE, data_set, options.force); if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) { // Show any reasons why resource might stay stopped cli_resource_check(out, rsc, node); } if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } } static int clear_constraints(pcmk__output_t *out, xmlNodePtr *cib_xml_copy) { GList *before = NULL; GList *after = NULL; GList *remaining = NULL; GList *ele = NULL; pe_node_t *dest = NULL; int rc = pcmk_rc_ok; if (!out->is_quiet(out)) { before = build_constraint_list(data_set->input); } if (options.clear_expired) { rc = cli_resource_clear_all_expired(data_set->input, cib_conn, options.cib_options, options.rsc_id, options.host_uname, options.promoted_role_only); } else if (options.host_uname) { dest = pe_find_node(data_set->nodes, options.host_uname); if (dest == NULL) { rc = pcmk_rc_node_unknown; if (!out->is_quiet(out)) { g_list_free(before); } return rc; } rc = cli_resource_clear(options.rsc_id, dest->details->uname, NULL, cib_conn, options.cib_options, TRUE, options.force); } else { rc = cli_resource_clear(options.rsc_id, NULL, data_set->nodes, cib_conn, options.cib_options, TRUE, options.force); } if (!out->is_quiet(out)) { rc = cib_conn->cmds->query(cib_conn, NULL, cib_xml_copy, cib_scope_local | cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, _("Could not get modified CIB: %s\n"), pcmk_strerror(rc)); g_list_free(before); free_xml(*cib_xml_copy); *cib_xml_copy = NULL; return rc; } data_set->input = *cib_xml_copy; cluster_status(data_set); after = build_constraint_list(data_set->input); remaining = pcmk__subtract_lists(before, after, (GCompareFunc) strcmp); for (ele = remaining; ele != NULL; ele = ele->next) { out->info(out, "Removing constraint: %s", (char *) ele->data); } g_list_free(before); g_list_free(after); g_list_free(remaining); } return rc; } static int delete(void) { int rc = pcmk_rc_ok; xmlNode *msg_data = NULL; if (options.rsc_type == NULL) { rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("You need to specify a resource type with -t")); return rc; } msg_data = create_xml_node(NULL, options.rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, options.rsc_id); rc = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, options.cib_options); rc = pcmk_legacy2rc(rc); free_xml(msg_data); return rc; } static int list_agents(pcmk__output_t *out, const char *agent_spec) { int rc = pcmk_rc_ok; char *provider = strchr(agent_spec, ':'); lrmd_t *lrmd_conn = NULL; lrmd_list_t *list = NULL; rc = lrmd__new(&lrmd_conn, NULL, NULL, 0); if (rc != pcmk_rc_ok) { goto error; } if (provider) { *provider++ = 0; } rc = lrmd_conn->cmds->list_agents(lrmd_conn, &list, agent_spec, provider); if (rc > 0) { rc = out->message(out, "agents-list", list, agent_spec, provider); } else { rc = pcmk_rc_error; } error: if (rc != pcmk_rc_ok) { if (provider == NULL) { g_set_error(&error, PCMK__RC_ERROR, rc, _("No agents found for standard '%s'"), agent_spec); } else { g_set_error(&error, PCMK__RC_ERROR, rc, _("No agents found for standard '%s' and provider '%s'"), agent_spec, provider); } } lrmd_api_delete(lrmd_conn); return rc; } static int list_providers(pcmk__output_t *out, const char *agent_spec) { int rc; const char *text = NULL; lrmd_t *lrmd_conn = NULL; lrmd_list_t *list = NULL; rc = lrmd__new(&lrmd_conn, NULL, NULL, 0); if (rc != pcmk_rc_ok) { goto error; } switch (options.rsc_cmd) { case cmd_list_alternatives: rc = lrmd_conn->cmds->list_ocf_providers(lrmd_conn, agent_spec, &list); if (rc > 0) { rc = out->message(out, "alternatives-list", list, agent_spec); } else { rc = pcmk_rc_error; } text = "OCF providers"; break; case cmd_list_standards: rc = lrmd_conn->cmds->list_standards(lrmd_conn, &list); if (rc > 0) { rc = out->message(out, "standards-list", list); } else { rc = pcmk_rc_error; } text = "standards"; break; case cmd_list_providers: rc = lrmd_conn->cmds->list_ocf_providers(lrmd_conn, agent_spec, &list); if (rc > 0) { rc = out->message(out, "providers-list", list, agent_spec); } else { rc = pcmk_rc_error; } text = "OCF providers"; break; default: g_set_error(&error, PCMK__RC_ERROR, pcmk_rc_error, "Bug"); lrmd_api_delete(lrmd_conn); return pcmk_rc_error; } error: if (rc != pcmk_rc_ok) { if (agent_spec != NULL) { rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("No %s found for %s"), text, agent_spec); } else { rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("No %s found"), text); } } lrmd_api_delete(lrmd_conn); return rc; } static int populate_working_set(xmlNodePtr *cib_xml_copy) { int rc = pcmk_rc_ok; if (options.xml_file != NULL) { *cib_xml_copy = filename2xml(options.xml_file); if (*cib_xml_copy == NULL) { rc = pcmk_rc_cib_corrupt; } } else { rc = cib_conn->cmds->query(cib_conn, NULL, cib_xml_copy, cib_scope_local | cib_sync_call); rc = pcmk_legacy2rc(rc); } if (rc == pcmk_rc_ok) { data_set = pe_new_working_set(); if (data_set == NULL) { rc = ENOMEM; } else { pe__set_working_set_flags(data_set, pe_flag_no_counts|pe_flag_no_compat); data_set->priv = out; rc = update_working_set_xml(data_set, cib_xml_copy); } } if (rc != pcmk_rc_ok) { free_xml(*cib_xml_copy); *cib_xml_copy = NULL; return rc; } cluster_status(data_set); return pcmk_rc_ok; } static int refresh(pcmk__output_t *out) { int rc = pcmk_rc_ok; const char *router_node = options.host_uname; int attr_options = pcmk__node_attr_none; if (options.host_uname) { pe_node_t *node = pe_find_node(data_set->nodes, options.host_uname); if (pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("No cluster connection to Pacemaker Remote node %s detected"), options.host_uname); return rc; } router_node = node->details->uname; attr_options |= pcmk__node_attr_remote; } } if (controld_api == NULL) { out->info(out, "Dry run: skipping clean-up of %s due to CIB_file", options.host_uname? options.host_uname : "all nodes"); rc = pcmk_rc_ok; return rc; } crm_debug("Re-checking the state of all resources on %s", options.host_uname?options.host_uname:"all nodes"); rc = pcmk__attrd_api_clear_failures(NULL, options.host_uname, NULL, NULL, NULL, NULL, attr_options); if (pcmk_controld_api_reprobe(controld_api, options.host_uname, router_node) == pcmk_rc_ok) { start_mainloop(controld_api); } return rc; } static void refresh_resource(pcmk__output_t *out, pe_resource_t *rsc, pe_node_t *node) { int rc = pcmk_rc_ok; if (options.force == FALSE) { rsc = uber_parent(rsc); } crm_debug("Re-checking the state of %s (%s requested) on %s", rsc->id, options.rsc_id, (options.host_uname? options.host_uname: "all nodes")); rc = cli_resource_delete(controld_api, options.host_uname, rsc, NULL, 0, FALSE, data_set, options.force); if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) { // Show any reasons why resource might stay stopped cli_resource_check(out, rsc, node); } if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } } static int set_property(void) { int rc = pcmk_rc_ok; xmlNode *msg_data = NULL; if (pcmk__str_empty(options.rsc_type)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("Must specify -t with resource type")); rc = ENXIO; return rc; } else if (pcmk__str_empty(options.prop_value)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("Must supply -v with new value")); rc = ENXIO; return rc; } CRM_LOG_ASSERT(options.prop_name != NULL); msg_data = create_xml_node(NULL, options.rsc_type); crm_xml_add(msg_data, XML_ATTR_ID, options.rsc_id); crm_xml_add(msg_data, options.prop_name, options.prop_value); rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_RESOURCES, msg_data, options.cib_options); rc = pcmk_legacy2rc(rc); free_xml(msg_data); return rc; } static int show_metadata(pcmk__output_t *out, const char *agent_spec) { int rc = pcmk_rc_ok; char *standard = NULL; char *provider = NULL; char *type = NULL; char *metadata = NULL; lrmd_t *lrmd_conn = NULL; rc = lrmd__new(&lrmd_conn, NULL, NULL, 0); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, _("Could not create executor connection")); lrmd_api_delete(lrmd_conn); return rc; } rc = crm_parse_agent_spec(agent_spec, &standard, &provider, &type); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { rc = lrmd_conn->cmds->get_metadata(lrmd_conn, standard, provider, type, &metadata, 0); rc = pcmk_legacy2rc(rc); if (metadata) { out->output_xml(out, "metadata", metadata); free(metadata); } else { /* We were given a validly formatted spec, but it doesn't necessarily * match up with anything that exists. Use ENXIO as the return code * here because that maps to an exit code of CRM_EX_NOSUCH, which * probably is the most common reason to get here. */ rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("Metadata query for %s failed: %s"), agent_spec, pcmk_rc_str(rc)); } } else { rc = ENXIO; g_set_error(&error, PCMK__RC_ERROR, rc, _("'%s' is not a valid agent specification"), agent_spec); } lrmd_api_delete(lrmd_conn); return rc; } static void validate_cmdline_config(void) { // Cannot use both --resource and command-line resource configuration if (options.rsc_id != NULL) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("--resource cannot be used with --class, --agent, and --provider")); // Not all commands support command-line resource configuration } else if (options.rsc_cmd != cmd_execute_agent) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("--class, --agent, and --provider can only be used with " "--validate and --force-*")); // Not all of --class, --agent, and --provider need to be given. Not all // classes support the concept of a provider. Check that what we were given // is valid. } else if (pcmk__str_eq(options.v_class, "stonith", pcmk__str_none)) { if (options.v_provider != NULL) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("stonith does not support providers")); } else if (stonith_agent_exists(options.v_agent, 0) == FALSE) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("%s is not a known stonith agent"), options.v_agent ? options.v_agent : ""); } } else if (resources_agent_exists(options.v_class, options.v_provider, options.v_agent) == FALSE) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, _("%s:%s:%s is not a known resource"), options.v_class ? options.v_class : "", options.v_provider ? options.v_provider : "", options.v_agent ? options.v_agent : ""); } if (error != NULL) { return; } if (options.cmdline_params == NULL) { options.cmdline_params = pcmk__strkey_table(free, free); } options.require_resource = FALSE; options.require_dataset = FALSE; options.require_cib = FALSE; } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; GOptionEntry extra_prog_entries[] = { { "quiet", 'Q', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &(args->quiet), "Be less descriptive in output.", NULL }, { "resource", 'r', G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING, &options.rsc_id, "Resource ID", "ID" }, { G_OPTION_REMAINING, 0, G_OPTION_FLAG_NONE, G_OPTION_ARG_STRING_ARRAY, &options.remainder, NULL, NULL }, { NULL } }; const char *description = "Examples:\n\n" "List the available OCF agents:\n\n" "\t# crm_resource --list-agents ocf\n\n" "List the available OCF agents from the linux-ha project:\n\n" "\t# crm_resource --list-agents ocf:heartbeat\n\n" "Move 'myResource' to a specific node:\n\n" "\t# crm_resource --resource myResource --move --node altNode\n\n" "Allow (but not force) 'myResource' to move back to its original " "location:\n\n" "\t# crm_resource --resource myResource --clear\n\n" "Stop 'myResource' (and anything that depends on it):\n\n" "\t# crm_resource --resource myResource --set-parameter target-role " "--meta --parameter-value Stopped\n\n" "Tell the cluster not to manage 'myResource' (the cluster will not " "attempt to start or stop the\n" "resource under any circumstances; useful when performing maintenance " "tasks on a resource):\n\n" "\t# crm_resource --resource myResource --set-parameter is-managed " "--meta --parameter-value false\n\n" "Erase the operation history of 'myResource' on 'aNode' (the cluster " "will 'forget' the existing\n" "resource state, including any errors, and attempt to recover the" "resource; useful when a resource\n" "had failed permanently and has been repaired by an administrator):\n\n" "\t# crm_resource --resource myResource --cleanup --node aNode\n\n"; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); g_option_context_set_description(context, description); /* Add the -Q option, which cannot be part of the globally supported options * because some tools use that flag for something else. */ pcmk__add_main_args(context, extra_prog_entries); 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, "locations", "Locations:", "Show location help", location_entries); pcmk__add_arg_group(context, "advanced", "Advanced:", "Show advanced option help", advanced_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); return context; } int main(int argc, char **argv) { xmlNode *cib_xml_copy = NULL; pe_resource_t *rsc = NULL; pe_node_t *node = NULL; int rc = pcmk_rc_ok; GOptionGroup *output_group = NULL; gchar **processed_args = NULL; GOptionContext *context = NULL; /* * Parse command line arguments */ args = pcmk__new_common_args(SUMMARY); processed_args = pcmk__cmdline_preproc(argv, "GHINSTdginpstuvx"); context = build_arg_context(args, &output_group); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("crm_resource", args->verbosity); 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; } pe__register_messages(out); crm_resource_register_messages(out); lrmd__register_messages(out); pcmk__register_lib_messages(out); out->quiet = args->quiet; crm_log_args(argc, argv); /* * Validate option combinations */ // If the user didn't explicitly specify a command, list resources if (options.rsc_cmd == cmd_none) { options.rsc_cmd = cmd_list_resources; options.require_resource = FALSE; } // --expired without --clear/-U doesn't make sense if (options.clear_expired && (options.rsc_cmd != cmd_clear)) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("--expired requires --clear or -U")); goto done; } if ((options.remainder != NULL) && (options.override_params != NULL)) { // Commands that use positional arguments will create override_params for (gchar **s = options.remainder; *s; s++) { char *name = calloc(1, strlen(*s)); char *value = calloc(1, strlen(*s)); int rc = sscanf(*s, "%[^=]=%s", name, value); if (rc == 2) { g_hash_table_replace(options.override_params, name, value); } else { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Error parsing '%s' as a name=value pair"), argv[optind]); free(value); free(name); goto done; } } } else if (options.remainder != NULL) { gchar **strv = NULL; gchar *msg = NULL; int i = 1; int len = 0; for (gchar **s = options.remainder; *s; s++) { len++; } CRM_ASSERT(len > 0); /* Add 1 for the strv[0] string below, and add another 1 for the NULL * at the end of the array so g_strjoinv knows when to stop. */ strv = calloc(len+2, sizeof(char *)); strv[0] = strdup("non-option ARGV-elements:\n"); for (gchar **s = options.remainder; *s; s++) { strv[i] = crm_strdup_printf("[%d of %d] %s\n", i, len, *s); i++; } strv[i] = NULL; exit_code = CRM_EX_USAGE; msg = g_strjoinv("", strv); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "%s", msg); g_free(msg); /* Don't try to free the last element, which is just NULL. */ for(i = 0; i < len+1; i++) { free(strv[i]); } free(strv); goto done; } if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) { /* Kind of a hack to display XML lists using a real tag instead of . This just * saves from having to write custom messages to build the lists around all these things */ switch (options.rsc_cmd) { case cmd_execute_agent: case cmd_list_resources: case cmd_query_xml: case cmd_query_raw_xml: case cmd_list_active_ops: case cmd_list_all_ops: case cmd_colocations: pcmk__force_args(context, &error, "%s --xml-simple-list --xml-substitute", g_get_prgname()); break; default: pcmk__force_args(context, &error, "%s --xml-substitute", g_get_prgname()); break; } } else if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches)) { if ((options.rsc_cmd == cmd_colocations) || options.rsc_cmd == cmd_list_resources) { pcmk__force_args(context, &error, "%s --text-fancy", g_get_prgname()); } } if (args->version) { out->version(out, false); goto done; } if (options.cmdline_config) { /* A resource configuration was given on the command line. Sanity-check * the values and set error if they don't make sense. */ validate_cmdline_config(); if (error != NULL) { exit_code = CRM_EX_USAGE; goto done; } } else if (options.cmdline_params != NULL) { // @COMPAT @TODO error out here when we can break backward compatibility g_hash_table_destroy(options.cmdline_params); options.cmdline_params = NULL; } if (options.require_resource && (options.rsc_id == NULL)) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Must supply a resource id with -r")); goto done; } if (options.require_node && (options.host_uname == NULL)) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Must supply a node name with -N")); goto done; } /* * Set up necessary connections */ if (options.find_flags && options.rsc_id) { options.require_dataset = TRUE; } // Establish a connection to the CIB if needed if (options.require_cib) { cib_conn = cib_new(); if ((cib_conn == NULL) || (cib_conn->cmds == NULL)) { exit_code = CRM_EX_DISCONNECT; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Could not create CIB connection")); goto done; } rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); 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 connect to the CIB: %s"), pcmk_rc_str(rc)); goto done; } } /* Populate working set from XML file if specified or CIB query otherwise */ if (options.require_dataset) { rc = populate_working_set(&cib_xml_copy); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); goto done; } } // If command requires that resource exist if specified, find it if (options.find_flags && options.rsc_id) { rsc = pe_find_resource_with_flags(data_set->resources, options.rsc_id, options.find_flags); if (rsc == NULL) { exit_code = CRM_EX_NOSUCH; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Resource '%s' not found"), options.rsc_id); goto done; } /* The --ban, --clear, --move, and --restart commands do not work with * instances of clone resourcs. */ if (strchr(options.rsc_id, ':') != NULL && pe_rsc_is_clone(rsc->parent) && (options.rsc_cmd == cmd_ban || options.rsc_cmd == cmd_clear || options.rsc_cmd == cmd_move || options.rsc_cmd == cmd_restart)) { exit_code = CRM_EX_INVALID_PARAM; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Cannot operate on clone resource instance '%s'"), options.rsc_id); goto done; } } // If user supplied a node name, check whether it exists if ((options.host_uname != NULL) && (data_set != NULL)) { node = pe_find_node(data_set->nodes, options.host_uname); if (node == NULL) { exit_code = CRM_EX_NOSUCH; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Node '%s' not found"), options.host_uname); goto done; } } // Establish a connection to the controller if needed if (options.require_crmd) { rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Error connecting to the controller: %s"), pcmk_rc_str(rc)); goto done; } pcmk_register_ipc_callback(controld_api, controller_event_callback, NULL); - rc = pcmk_connect_ipc(controld_api, pcmk_ipc_dispatch_main); + rc = pcmk__connect_ipc(controld_api, pcmk_ipc_dispatch_main, 5); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, - _("Error connecting to the controller: %s"), pcmk_rc_str(rc)); + _("Error connecting to %s: %s"), + pcmk_ipc_name(controld_api, true), pcmk_rc_str(rc)); goto done; } } /* * Handle requested command */ switch (options.rsc_cmd) { case cmd_list_resources: { GList *all = NULL; all = g_list_prepend(all, (gpointer) "*"); rc = out->message(out, "resource-list", data_set, pcmk_show_inactive_rscs | pcmk_show_rsc_only | pcmk_show_pending, true, all, all, false); g_list_free(all); if (rc == pcmk_rc_no_output) { rc = ENXIO; } break; } case cmd_list_instances: rc = out->message(out, "resource-names-list", data_set->resources); if (rc != pcmk_rc_ok) { rc = ENXIO; } break; case cmd_list_standards: case cmd_list_providers: case cmd_list_alternatives: rc = list_providers(out, options.agent_spec); break; case cmd_list_agents: rc = list_agents(out, options.agent_spec); break; case cmd_metadata: rc = show_metadata(out, options.agent_spec); break; case cmd_restart: /* We don't pass data_set because rsc needs to stay valid for the * entire lifetime of cli_resource_restart(), but it will reset and * update the working set multiple times, so it needs to use its own * copy. */ rc = cli_resource_restart(out, rsc, node, options.move_lifetime, options.timeout_ms, cib_conn, options.cib_options, options.promoted_role_only, options.force); break; case cmd_wait: rc = wait_till_stable(out, options.timeout_ms, cib_conn); break; case cmd_execute_agent: if (options.cmdline_config) { exit_code = cli_resource_execute_from_params(out, NULL, options.v_class, options.v_provider, options.v_agent, options.operation, options.cmdline_params, options.override_params, options.timeout_ms, args->verbosity, options.force, options.check_level); } else { exit_code = cli_resource_execute(rsc, options.rsc_id, options.operation, options.override_params, options.timeout_ms, cib_conn, data_set, args->verbosity, options.force, options.check_level); } goto done; case cmd_digests: node = pe_find_node(data_set->nodes, options.host_uname); if (node == NULL) { rc = pcmk_rc_node_unknown; } else { rc = pcmk__resource_digests(out, rsc, node, options.override_params); } break; case cmd_colocations: rc = out->message(out, "locations-and-colocations", rsc, options.recursive, (bool) options.force); break; case cmd_cts: rc = pcmk_rc_ok; g_list_foreach(data_set->resources, (GFunc) cli_resource_print_cts, out); cli_resource_print_cts_constraints(data_set); break; case cmd_fail: rc = cli_resource_fail(controld_api, options.host_uname, options.rsc_id, data_set); if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } break; case cmd_list_active_ops: rc = cli_resource_print_operations(options.rsc_id, options.host_uname, TRUE, data_set); break; case cmd_list_all_ops: rc = cli_resource_print_operations(options.rsc_id, options.host_uname, FALSE, data_set); break; case cmd_locate: { GList *nodes = cli_resource_search(rsc, options.rsc_id, data_set); rc = out->message(out, "resource-search-list", nodes, options.rsc_id); g_list_free_full(nodes, free); break; } case cmd_query_xml: rc = cli_resource_print(rsc, data_set, true); break; case cmd_query_raw_xml: rc = cli_resource_print(rsc, data_set, false); break; case cmd_why: if ((options.host_uname != NULL) && (node == NULL)) { rc = pcmk_rc_node_unknown; } else { rc = out->message(out, "resource-reasons-list", data_set->resources, rsc, node); } break; case cmd_clear: rc = clear_constraints(out, &cib_xml_copy); break; case cmd_move: if (options.host_uname == NULL) { rc = ban_or_move(out, rsc, options.move_lifetime); } else { rc = cli_resource_move(rsc, options.rsc_id, options.host_uname, options.move_lifetime, cib_conn, options.cib_options, data_set, options.promoted_role_only, options.force); } if (rc == EINVAL) { exit_code = CRM_EX_USAGE; goto done; } break; case cmd_ban: if (options.host_uname == NULL) { rc = ban_or_move(out, rsc, options.move_lifetime); } else if (node == NULL) { rc = pcmk_rc_node_unknown; } else { rc = cli_resource_ban(out, options.rsc_id, node->details->uname, options.move_lifetime, NULL, cib_conn, options.cib_options, options.promoted_role_only); } if (rc == EINVAL) { exit_code = CRM_EX_USAGE; goto done; } break; case cmd_get_property: rc = out->message(out, "property-list", rsc, options.prop_name); if (rc == pcmk_rc_no_output) { rc = ENXIO; } break; case cmd_set_property: rc = set_property(); break; case cmd_get_param: { unsigned int count = 0; GHashTable *params = NULL; pe_node_t *current = rsc->fns->active_node(rsc, &count, NULL); bool free_params = true; const char* value = NULL; if (count > 1) { out->err(out, "%s is active on more than one node," " returning the default value for %s", rsc->id, pcmk__s(options.prop_name, "unspecified property")); current = NULL; } crm_debug("Looking up %s in %s", options.prop_name, rsc->id); if (pcmk__str_eq(options.attr_set_type, XML_TAG_ATTR_SETS, pcmk__str_none)) { params = pe_rsc_params(rsc, current, data_set); free_params = false; value = g_hash_table_lookup(params, options.prop_name); } else if (pcmk__str_eq(options.attr_set_type, XML_TAG_META_SETS, pcmk__str_none)) { params = pcmk__strkey_table(free, free); get_meta_attributes(params, rsc, current, data_set); value = g_hash_table_lookup(params, options.prop_name); } else if (pcmk__str_eq(options.attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) { value = crm_element_value(rsc->xml, options.prop_name); free_params = false; } else { params = pcmk__strkey_table(free, free); pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_UTILIZATION, NULL, params, NULL, FALSE, data_set); value = g_hash_table_lookup(params, options.prop_name); } rc = out->message(out, "attribute-list", rsc, options.prop_name, value); if (free_params) { g_hash_table_destroy(params); } break; } case cmd_set_param: if (pcmk__str_empty(options.prop_value)) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("You need to supply a value with the -v option")); goto done; } /* coverity[var_deref_model] False positive */ rc = cli_resource_update_attribute(rsc, options.rsc_id, options.prop_set, options.attr_set_type, options.prop_id, options.prop_name, options.prop_value, options.recursive, cib_conn, options.cib_options, options.force); break; case cmd_delete_param: /* coverity[var_deref_model] False positive */ rc = cli_resource_delete_attribute(rsc, options.rsc_id, options.prop_set, options.attr_set_type, options.prop_id, options.prop_name, cib_conn, options.cib_options, options.force); break; case cmd_cleanup: if (rsc == NULL) { rc = cli_cleanup_all(controld_api, options.host_uname, options.operation, options.interval_spec, data_set); if (rc == pcmk_rc_ok) { start_mainloop(controld_api); } } else { cleanup(out, rsc, node); } break; case cmd_refresh: if (rsc == NULL) { rc = refresh(out); } else { refresh_resource(out, rsc, node); } break; case cmd_delete: rc = delete(); break; default: exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Unimplemented command: %d"), (int) options.rsc_cmd); goto done; } /* Convert rc into an exit code. */ if (rc != pcmk_rc_ok && rc != pcmk_rc_no_output) { exit_code = pcmk_rc2exitc(rc); } /* * Clean up and exit */ done: /* When we get here, exit_code has been set one of two ways - either at one of * the spots where there's a "goto done" (which itself could have happened either * directly or by calling pcmk_rc2exitc), or just up above after any of the break * statements. * * Thus, we can use just exit_code here to decide what to do. */ if (exit_code != CRM_EX_OK && exit_code != CRM_EX_USAGE) { if (error != NULL) { char *msg = crm_strdup_printf("%s\nError performing operation: %s", error->message, crm_exit_str(exit_code)); g_clear_error(&error); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "%s", msg); free(msg); } else { g_set_error(&error, PCMK__EXITC_ERROR, exit_code, _("Error performing operation: %s"), crm_exit_str(exit_code)); } } g_free(options.host_uname); g_free(options.interval_spec); g_free(options.move_lifetime); g_free(options.operation); g_free(options.prop_id); free(options.prop_name); g_free(options.prop_set); g_free(options.prop_value); g_free(options.rsc_id); g_free(options.rsc_type); free(options.agent_spec); free(options.v_agent); free(options.v_class); free(options.v_provider); g_free(options.xml_file); g_strfreev(options.remainder); if (options.override_params != NULL) { g_hash_table_destroy(options.override_params); } /* options.cmdline_params does not need to be destroyed here. See the * comments in cli_resource_execute_from_params. */ g_strfreev(processed_args); g_option_context_free(context); return bye(exit_code); }