diff --git a/daemons/controld/controld_execd.c b/daemons/controld/controld_execd.c index e0dd959004..2060db60ca 100644 --- a/daemons/controld/controld_execd.c +++ b/daemons/controld/controld_execd.c @@ -1,2446 +1,2446 @@ /* * Copyright 2004-2024 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 // lrmd_event_data_t, lrmd_rsc_info_t, etc. #include #include #include #include #include #include #include #define START_DELAY_THRESHOLD 5 * 60 * 1000 #define MAX_LRM_REG_FAILS 30 struct delete_event_s { int rc; const char *rsc; lrm_state_t *lrm_state; }; static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id); static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list); static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data); static lrmd_event_data_t *construct_op(const lrm_state_t *lrm_state, const xmlNode *rsc_op, const char *rsc_id, const char *operation); static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, xmlNode *msg, struct ra_metadata_s *md); static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level); static void lrm_connection_destroy(void) { if (pcmk_is_set(controld_globals.fsa_input_register, R_LRM_CONNECTED)) { crm_crit("Lost connection to local executor"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); controld_clear_fsa_input_flags(R_LRM_CONNECTED); } } static char * make_stop_id(const char *rsc, int call_id) { return crm_strdup_printf("%s:%d", rsc, call_id); } static void copy_instance_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") == NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } static void copy_meta_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") != NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } /*! * \internal * \brief Remove a recurring operation from a resource's history * * \param[in,out] history Resource history to modify * \param[in] op Operation to remove * * \return TRUE if the operation was found and removed, FALSE otherwise */ static gboolean history_remove_recurring_op(rsc_history_t *history, const lrmd_event_data_t *op) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_event_data_t *existing = iter->data; if ((op->interval_ms == existing->interval_ms) && pcmk__str_eq(op->rsc_id, existing->rsc_id, pcmk__str_none) && pcmk__str_eq(op->op_type, existing->op_type, pcmk__str_casei)) { history->recurring_op_list = g_list_delete_link(history->recurring_op_list, iter); lrmd_free_event(existing); return TRUE; } } return FALSE; } /*! * \internal * \brief Free all recurring operations in resource history * * \param[in,out] history Resource history to modify */ static void history_free_recurring_ops(rsc_history_t *history) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_free_event(iter->data); } g_list_free(history->recurring_op_list); history->recurring_op_list = NULL; } /*! * \internal * \brief Free resource history * * \param[in,out] history Resource history to free */ void history_free(gpointer data) { rsc_history_t *history = (rsc_history_t*)data; if (history->stop_params) { g_hash_table_destroy(history->stop_params); } /* Don't need to free history->rsc.id because it's set to history->id */ free(history->rsc.type); free(history->rsc.standard); free(history->rsc.provider); lrmd_free_event(history->failed); lrmd_free_event(history->last); free(history->id); history_free_recurring_ops(history); free(history); } static void update_history_cache(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { int target_rc = 0; rsc_history_t *entry = NULL; if (op->rsc_deleted) { crm_debug("Purged history for '%s' after %s", op->rsc_id, op->op_type); controld_delete_resource_history(op->rsc_id, lrm_state->node_name, NULL, crmd_cib_smart_opt()); return; } if (pcmk__str_eq(op->op_type, PCMK_ACTION_NOTIFY, pcmk__str_casei)) { return; } crm_debug("Updating history for '%s' with %s op", op->rsc_id, op->op_type); entry = g_hash_table_lookup(lrm_state->resource_history, op->rsc_id); if (entry == NULL && rsc) { entry = calloc(1, sizeof(rsc_history_t)); entry->id = strdup(op->rsc_id); g_hash_table_insert(lrm_state->resource_history, entry->id, entry); entry->rsc.id = entry->id; entry->rsc.type = strdup(rsc->type); entry->rsc.standard = strdup(rsc->standard); pcmk__str_update(&entry->rsc.provider, rsc->provider); } else if (entry == NULL) { crm_info("Resource %s no longer exists, not updating cache", op->rsc_id); return; } entry->last_callid = op->call_id; target_rc = rsc_op_expected_rc(op); if (op->op_status == PCMK_EXEC_CANCELLED) { if (op->interval_ms > 0) { crm_trace("Removing cancelled recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); history_remove_recurring_op(entry, op); return; } else { crm_trace("Skipping " PCMK__OP_FMT " rc=%d, status=%d", op->rsc_id, op->op_type, op->interval_ms, op->rc, op->op_status); } } else if (did_rsc_op_fail(op, target_rc)) { /* Store failed monitors here, otherwise the block below will cause them * to be forgotten when a stop happens. */ if (entry->failed) { lrmd_free_event(entry->failed); } entry->failed = lrmd_copy_event(op); } else if (op->interval_ms == 0) { if (entry->last) { lrmd_free_event(entry->last); } entry->last = lrmd_copy_event(op); if (op->params && pcmk__strcase_any_of(op->op_type, PCMK_ACTION_START, PCMK_ACTION_RELOAD, PCMK_ACTION_RELOAD_AGENT, PCMK_ACTION_MONITOR, NULL)) { if (entry->stop_params) { g_hash_table_destroy(entry->stop_params); } entry->stop_params = pcmk__strkey_table(free, free); g_hash_table_foreach(op->params, copy_instance_keys, entry->stop_params); } } if (op->interval_ms > 0) { /* Ensure there are no duplicates */ history_remove_recurring_op(entry, op); crm_trace("Adding recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); entry->recurring_op_list = g_list_prepend(entry->recurring_op_list, lrmd_copy_event(op)); } else if ((entry->recurring_op_list != NULL) && !pcmk__str_eq(op->op_type, PCMK_ACTION_MONITOR, pcmk__str_casei)) { crm_trace("Dropping %d recurring ops because of: " PCMK__OP_FMT, g_list_length(entry->recurring_op_list), op->rsc_id, op->op_type, op->interval_ms); history_free_recurring_ops(entry); } } /*! * \internal * \brief Send a direct OK ack for a resource task * * \param[in] lrm_state LRM connection * \param[in] input Input message being ack'ed * \param[in] rsc_id ID of affected resource * \param[in] rsc Affected resource (if available) * \param[in] task Operation task being ack'ed * \param[in] ack_host Name of host to send ack to * \param[in] ack_sys IPC system name to ack */ static void send_task_ok_ack(const lrm_state_t *lrm_state, const ha_msg_input_t *input, const char *rsc_id, const lrmd_rsc_info_t *rsc, const char *task, const char *ack_host, const char *ack_sys) { lrmd_event_data_t *op = construct_op(lrm_state, input->xml, rsc_id, task); lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(ack_host, ack_sys, rsc, op, rsc_id); lrmd_free_event(op); } static inline const char * op_node_name(lrmd_event_data_t *op) { return pcmk__s(op->remote_nodename, controld_globals.our_nodename); } void lrm_op_callback(lrmd_event_data_t * op) { CRM_CHECK(op != NULL, return); switch (op->type) { case lrmd_event_disconnect: if (op->remote_nodename == NULL) { /* If this is the local executor IPC connection, set the right * bits in the controller when the connection goes down. */ lrm_connection_destroy(); } break; case lrmd_event_exec_complete: { lrm_state_t *lrm_state = lrm_state_find(op_node_name(op)); CRM_ASSERT(lrm_state != NULL); process_lrm_event(lrm_state, op, NULL, NULL); } break; default: break; } } static void try_local_executor_connect(long long action, fsa_data_t *msg_data, lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; crm_debug("Connecting to the local executor"); // If we can connect, great rc = controld_connect_local_executor(lrm_state); if (rc == pcmk_rc_ok) { controld_set_fsa_input_flags(R_LRM_CONNECTED); crm_info("Connection to the local executor established"); return; } // Otherwise, if we can try again, set a timer to do so if (lrm_state->num_lrm_register_fails < MAX_LRM_REG_FAILS) { crm_warn("Failed to connect to the local executor %d time%s " "(%d max): %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), MAX_LRM_REG_FAILS, pcmk_rc_str(rc)); controld_start_wait_timer(); crmd_fsa_stall(FALSE); return; } // Otherwise give up crm_err("Failed to connect to the executor the max allowed " "%d time%s: %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), pcmk_rc_str(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } /* A_LRM_CONNECT */ void do_lrm_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) { /* This only pertains to local executor connections. Remote connections are * handled as resources within the scheduler. Connecting and disconnecting * from remote executor instances is handled differently. */ lrm_state_t *lrm_state = NULL; if (controld_globals.our_nodename == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(controld_globals.our_nodename); if (lrm_state == NULL) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (action & A_LRM_DISCONNECT) { if (lrm_state_verify_stopped(lrm_state, cur_state, LOG_INFO) == FALSE) { if (action == A_LRM_DISCONNECT) { crmd_fsa_stall(FALSE); return; } } controld_clear_fsa_input_flags(R_LRM_CONNECTED); lrm_state_disconnect(lrm_state); lrm_state_reset_tables(lrm_state, FALSE); } if (action & A_LRM_CONNECT) { try_local_executor_connect(action, msg_data, lrm_state); } if (action & ~(A_LRM_CONNECT | A_LRM_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __func__); } } static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level) { int counter = 0; gboolean rc = TRUE; const char *when = "lrm disconnect"; GHashTableIter gIter; const char *key = NULL; rsc_history_t *entry = NULL; active_op_t *pending = NULL; crm_debug("Checking for active resources before exit"); if (cur_state == S_TERMINATE) { log_level = LOG_ERR; when = "shutdown"; } else if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { when = "shutdown... waiting"; } if ((lrm_state->active_ops != NULL) && lrm_state_is_connected(lrm_state)) { guint removed = g_hash_table_foreach_remove(lrm_state->active_ops, stop_recurring_actions, lrm_state); guint nremaining = g_hash_table_size(lrm_state->active_ops); if (removed || nremaining) { crm_notice("Stopped %u recurring operation%s at %s (%u remaining)", removed, pcmk__plural_s(removed), when, nremaining); } } if (lrm_state->active_ops != NULL) { g_hash_table_iter_init(&gIter, lrm_state->active_ops); while (g_hash_table_iter_next(&gIter, NULL, (void **)&pending)) { /* Ignore recurring actions in the shutdown calculations */ if (pending->interval_ms == 0) { counter++; } } } if (counter > 0) { do_crm_log(log_level, "%d pending executor operation%s at %s", counter, pcmk__plural_s(counter), when); if ((cur_state == S_TERMINATE) || !pcmk_is_set(controld_globals.fsa_input_register, R_SENT_RSC_STOP)) { g_hash_table_iter_init(&gIter, lrm_state->active_ops); while (g_hash_table_iter_next(&gIter, (gpointer*)&key, (gpointer*)&pending)) { do_crm_log(log_level, "Pending action: %s (%s)", key, pending->op_key); } } else { rc = FALSE; } return rc; } if (lrm_state->resource_history == NULL) { return rc; } if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { /* At this point we're not waiting, we're just shutting down */ when = "shutdown"; } counter = 0; g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (gpointer*)&entry)) { if (is_rsc_active(lrm_state, entry->id) == FALSE) { continue; } counter++; if (log_level == LOG_ERR) { crm_info("Found %s active at %s", entry->id, when); } else { crm_trace("Found %s active at %s", entry->id, when); } if (lrm_state->active_ops != NULL) { GHashTableIter hIter; g_hash_table_iter_init(&hIter, lrm_state->active_ops); while (g_hash_table_iter_next(&hIter, (gpointer*)&key, (gpointer*)&pending)) { if (pcmk__str_eq(entry->id, pending->rsc_id, pcmk__str_none)) { crm_notice("%sction %s (%s) incomplete at %s", pending->interval_ms == 0 ? "A" : "Recurring a", key, pending->op_key, when); } } } } if (counter) { crm_err("%d resource%s active at %s", counter, (counter == 1)? " was" : "s were", when); } return rc; } static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id) { rsc_history_t *entry = NULL; entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->last == NULL) { return FALSE; } crm_trace("Processing %s: %s.%d=%d", rsc_id, entry->last->op_type, entry->last->interval_ms, entry->last->rc); if ((entry->last->rc == PCMK_OCF_OK) && pcmk__str_eq(entry->last->op_type, PCMK_ACTION_STOP, pcmk__str_casei)) { return FALSE; } else if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, PCMK_ACTION_MIGRATE_TO, pcmk__str_casei)) { // A stricter check is too complex ... leave that to the scheduler return FALSE; } else if (entry->last->rc == PCMK_OCF_NOT_RUNNING) { return FALSE; } else if ((entry->last->interval_ms == 0) && (entry->last->rc == PCMK_OCF_NOT_CONFIGURED)) { /* Badly configured resources can't be reliably stopped */ return FALSE; } return TRUE; } static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list) { GHashTableIter iter; rsc_history_t *entry = NULL; g_hash_table_iter_init(&iter, lrm_state->resource_history); while (g_hash_table_iter_next(&iter, NULL, (void **)&entry)) { GList *gIter = NULL; xmlNode *xml_rsc = create_xml_node(rsc_list, XML_LRM_TAG_RESOURCE); crm_xml_add(xml_rsc, PCMK_XA_ID, entry->id); crm_xml_add(xml_rsc, PCMK_XA_TYPE, entry->rsc.type); crm_xml_add(xml_rsc, PCMK_XA_CLASS, entry->rsc.standard); crm_xml_add(xml_rsc, PCMK_XA_PROVIDER, entry->rsc.provider); if (entry->last && entry->last->params) { static const char *name = CRM_META "_" PCMK__META_CONTAINER; const char *container = g_hash_table_lookup(entry->last->params, name); if (container) { crm_trace("Resource %s is a part of container resource %s", entry->id, container); crm_xml_add(xml_rsc, PCMK__META_CONTAINER, container); } } controld_add_resource_history_xml(xml_rsc, &(entry->rsc), entry->failed, lrm_state->node_name); controld_add_resource_history_xml(xml_rsc, &(entry->rsc), entry->last, lrm_state->node_name); for (gIter = entry->recurring_op_list; gIter != NULL; gIter = gIter->next) { controld_add_resource_history_xml(xml_rsc, &(entry->rsc), gIter->data, lrm_state->node_name); } } return FALSE; } xmlNode * controld_query_executor_state(void) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; lrm_state_t *lrm_state = lrm_state_find(controld_globals.our_nodename); if (!lrm_state) { crm_err("Could not find executor state for node %s", controld_globals.our_nodename); return NULL; } peer = crm_get_peer_full(0, lrm_state->node_name, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return NULL); xml_state = create_node_state_update(peer, node_update_cluster|node_update_peer, NULL, __func__); if (xml_state == NULL) { return NULL; } xml_data = create_xml_node(xml_state, XML_CIB_TAG_LRM); crm_xml_add(xml_data, PCMK_XA_ID, peer->uuid); rsc_list = create_xml_node(xml_data, XML_LRM_TAG_RESOURCES); /* Build a list of active (not always running) resources */ build_active_RAs(lrm_state, rsc_list); crm_log_xml_trace(xml_state, "Current executor state"); return xml_state; } /*! * \internal * \brief Map standard Pacemaker return code to operation status and OCF code * * \param[out] event Executor event whose status and return code should be set * \param[in] rc Standard Pacemaker return code */ void controld_rc2event(lrmd_event_data_t *event, int rc) { /* This is called for cleanup requests from controller peers/clients, not * for resource actions, so no exit reason is needed. */ switch (rc) { case pcmk_rc_ok: lrmd__set_result(event, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); break; case EACCES: lrmd__set_result(event, PCMK_OCF_INSUFFICIENT_PRIV, PCMK_EXEC_ERROR, NULL); break; default: lrmd__set_result(event, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, NULL); break; } } /*! * \internal * \brief Trigger a new transition after CIB status was deleted * * If a CIB status delete was not expected (as part of the transition graph), * trigger a new transition by updating the (arbitrary) "last-lrm-refresh" * cluster property. * * \param[in] from_sys IPC name that requested the delete * \param[in] rsc_id Resource whose status was deleted (for logging only) */ void controld_trigger_delete_refresh(const char *from_sys, const char *rsc_id) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { char *now_s = crm_strdup_printf("%lld", (long long) time(NULL)); crm_debug("Triggering a refresh after %s cleaned %s", from_sys, rsc_id); cib__update_node_attr(controld_globals.logger_out, controld_globals.cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, NULL, NULL); free(now_s); } } static void notify_deleted(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, int rc) { lrmd_event_data_t *op = NULL; const char *from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); const char *from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "localhost"), rsc_id, ((rc == pcmk_ok)? "" : " not")); op = construct_op(lrm_state, input->xml, rsc_id, PCMK_ACTION_DELETE); controld_rc2event(op, pcmk_legacy2rc(rc)); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); controld_trigger_delete_refresh(from_sys, rsc_id); } static gboolean lrm_remove_deleted_rsc(gpointer key, gpointer value, gpointer user_data) { struct delete_event_s *event = user_data; struct pending_deletion_op_s *op = value; if (pcmk__str_eq(event->rsc, op->rsc, pcmk__str_none)) { notify_deleted(event->lrm_state, op->input, event->rsc, event->rc); return TRUE; } return FALSE; } static gboolean lrm_remove_deleted_op(gpointer key, gpointer value, gpointer user_data) { const char *rsc = user_data; active_op_t *pending = value; if (pcmk__str_eq(rsc, pending->rsc_id, pcmk__str_none)) { crm_info("Removing op %s:%d for deleted resource %s", pending->op_key, pending->call_id, rsc); return TRUE; } return FALSE; } static void delete_rsc_entry(lrm_state_t *lrm_state, ha_msg_input_t *input, const char *rsc_id, GHashTableIter *rsc_iter, int rc, const char *user_name, bool from_cib) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_iter) { g_hash_table_iter_remove(rsc_iter); } else { g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); } if (from_cib) { controld_delete_resource_history(rsc_id_copy, lrm_state->node_name, user_name, crmd_cib_smart_opt()); } g_hash_table_foreach_remove(lrm_state->active_ops, lrm_remove_deleted_op, rsc_id_copy); free(rsc_id_copy); } if (input) { notify_deleted(lrm_state, input, rsc_id, rc); } event.rc = rc; event.rsc = rsc_id; event.lrm_state = lrm_state; g_hash_table_foreach_remove(lrm_state->deletion_ops, lrm_remove_deleted_rsc, &event); } static inline gboolean last_failed_matches_op(rsc_history_t *entry, const char *op, guint interval_ms) { if (entry == NULL) { return FALSE; } if (op == NULL) { return TRUE; } return (pcmk__str_eq(op, entry->failed->op_type, pcmk__str_casei) && (interval_ms == entry->failed->interval_ms)); } /*! * \internal * \brief Clear a resource's last failure * * Erase a resource's last failure on a particular node from both the * LRM resource history in the CIB, and the resource history remembered * for the LRM state. * * \param[in] rsc_id Resource name * \param[in] node_name Node name * \param[in] operation If specified, only clear if matching this operation * \param[in] interval_ms If operation is specified, it has this interval */ void lrm_clear_last_failure(const char *rsc_id, const char *node_name, const char *operation, guint interval_ms) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { return; } if (lrm_state->resource_history != NULL) { rsc_history_t *entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (last_failed_matches_op(entry, operation, interval_ms)) { lrmd_free_event(entry->failed); entry->failed = NULL; } } } /* Returns: gboolean - cancellation is in progress */ static gboolean cancel_op(lrm_state_t * lrm_state, const char *rsc_id, const char *key, int op, gboolean remove) { int rc = pcmk_ok; char *local_key = NULL; active_op_t *pending = NULL; CRM_CHECK(op != 0, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); if (key == NULL) { local_key = make_stop_id(rsc_id, op); key = local_key; } pending = g_hash_table_lookup(lrm_state->active_ops, key); if (pending) { if (remove && !pcmk_is_set(pending->flags, active_op_remove)) { controld_set_active_op_flags(pending, active_op_remove); crm_debug("Scheduling %s for removal", key); } if (pcmk_is_set(pending->flags, active_op_cancelled)) { crm_debug("Operation %s already cancelled", key); free(local_key); return FALSE; } controld_set_active_op_flags(pending, active_op_cancelled); } else { crm_info("No pending op found for %s", key); free(local_key); return FALSE; } crm_debug("Cancelling op %d for %s (%s)", op, rsc_id, key); rc = lrm_state_cancel(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms); if (rc == pcmk_ok) { crm_debug("Op %d for %s (%s): cancelled", op, rsc_id, key); free(local_key); return TRUE; } crm_debug("Op %d for %s (%s): Nothing to cancel", op, rsc_id, key); /* The caller needs to make sure the entry is * removed from the active operations list * * Usually by returning TRUE inside the worker function * supplied to g_hash_table_foreach_remove() * * Not removing the entry from active operations will block * the node from shutting down */ free(local_key); return FALSE; } struct cancel_data { gboolean done; gboolean remove; const char *key; lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean cancel_action_by_key(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct cancel_data *data = user_data; active_op_t *op = value; if (pcmk__str_eq(op->op_key, data->key, pcmk__str_none)) { data->done = TRUE; remove = !cancel_op(data->lrm_state, data->rsc->id, key, op->call_id, data->remove); } return remove; } static gboolean cancel_op_key(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *key, gboolean remove) { guint removed = 0; struct cancel_data data; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(key != NULL, return FALSE); data.key = key; data.rsc = rsc; data.done = FALSE; data.remove = remove; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->active_ops, cancel_action_by_key, &data); crm_trace("Removed %u op cache entries, new size: %u", removed, g_hash_table_size(lrm_state->active_ops)); return data.done; } /*! * \internal * \brief Retrieve resource information from LRM * * \param[in,out] lrm_state Executor connection state to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource if not already * \param[out] rsc_info Where to store information obtained from executor * * \retval pcmk_ok Success (and rsc_info holds newly allocated result) * \retval -EINVAL Required information is missing from arguments * \retval -ENOTCONN No active connection to LRM * \retval -ENODEV Resource not found * \retval -errno Error communicating with executor when registering resource * * \note Caller is responsible for freeing result on success. */ static int get_lrm_resource(lrm_state_t *lrm_state, const xmlNode *rsc_xml, gboolean do_create, lrmd_rsc_info_t **rsc_info) { const char *id = ID(rsc_xml); CRM_CHECK(lrm_state && rsc_xml && rsc_info, return -EINVAL); CRM_CHECK(id, return -EINVAL); if (lrm_state_is_connected(lrm_state) == FALSE) { return -ENOTCONN; } crm_trace("Retrieving resource information for %s from the executor", id); *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); // If resource isn't known by ID, try clone name, if provided if (!*rsc_info) { const char *long_id = crm_element_value(rsc_xml, PCMK__XA_LONG_ID); if (long_id) { *rsc_info = lrm_state_get_rsc_info(lrm_state, long_id, 0); } } if ((*rsc_info == NULL) && do_create) { const char *class = crm_element_value(rsc_xml, PCMK_XA_CLASS); const char *provider = crm_element_value(rsc_xml, PCMK_XA_PROVIDER); const char *type = crm_element_value(rsc_xml, PCMK_XA_TYPE); int rc; crm_trace("Registering resource %s with the executor", id); rc = lrm_state_register_rsc(lrm_state, id, class, provider, type, lrmd_opt_drop_recurring); if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Could not register resource %s with the executor on %s: %s " CRM_XS " rc=%d", id, lrm_state->node_name, pcmk_strerror(rc), rc); /* Register this as an internal error if this involves the local * executor. Otherwise, we're likely dealing with an unresponsive * remote node, which is not an FSA failure. */ if (lrm_state_is_local(lrm_state) == TRUE) { register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } return rc; } *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); } return *rsc_info? pcmk_ok : -ENODEV; } static void delete_resource(lrm_state_t *lrm_state, const char *id, lrmd_rsc_info_t *rsc, GHashTableIter *iter, const char *sys, const char *user, ha_msg_input_t *request, bool unregister, bool from_cib) { int rc = pcmk_ok; crm_info("Removing resource %s from executor for %s%s%s", id, sys, (user? " as " : ""), (user? user : "")); if (rsc && unregister) { rc = lrm_state_unregister_rsc(lrm_state, id, 0); } if (rc == pcmk_ok) { crm_trace("Resource %s deleted from executor", id); } else if (rc == -EINPROGRESS) { crm_info("Deletion of resource '%s' from executor is pending", id); if (request) { struct pending_deletion_op_s *op = NULL; - char *ref = crm_element_value_copy(request->msg, XML_ATTR_REFERENCE); + char *ref = crm_element_value_copy(request->msg, PCMK_XA_REFERENCE); op = calloc(1, sizeof(struct pending_deletion_op_s)); op->rsc = strdup(rsc->id); op->input = copy_ha_msg_input(request); g_hash_table_insert(lrm_state->deletion_ops, ref, op); } return; } else { crm_warn("Could not delete '%s' from executor for %s%s%s: %s " CRM_XS " rc=%d", id, sys, (user? " as " : ""), (user? user : ""), pcmk_strerror(rc), rc); } delete_rsc_entry(lrm_state, request, id, iter, rc, user, from_cib); } static int get_fake_call_id(lrm_state_t *lrm_state, const char *rsc_id) { int call_id = 999999999; rsc_history_t *entry = NULL; if(lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* Make sure the call id is greater than the last successful operation, * otherwise the failure will not result in a possible recovery of the resource * as it could appear the failure occurred before the successful start */ if (entry) { call_id = entry->last_callid + 1; } if (call_id < 0) { call_id = 1; } return call_id; } static void fake_op_status(lrm_state_t *lrm_state, lrmd_event_data_t *op, int op_status, enum ocf_exitcode op_exitcode, const char *exit_reason) { op->call_id = get_fake_call_id(lrm_state, op->rsc_id); op->t_run = time(NULL); op->t_rcchange = op->t_run; lrmd__set_result(op, op_exitcode, op_status, exit_reason); } static void force_reprobe(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node, bool reprobe_all_nodes) { GHashTableIter gIter; rsc_history_t *entry = NULL; crm_info("Clearing resource history on node %s", lrm_state->node_name); g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { /* only unregister the resource during a reprobe if it is not a remote connection * resource. otherwise unregistering the connection will terminate remote-node * membership */ bool unregister = true; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { unregister = false; if (reprobe_all_nodes) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state != NULL) { /* If reprobing all nodes, be sure to reprobe the remote * node before clearing its connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE, reprobe_all_nodes); } } } /* Don't delete from the CIB, since we'll delete the whole node's LRM * state from the CIB soon */ delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, user_name, NULL, unregister, false); } /* Now delete the copy in the CIB */ controld_delete_node_state(lrm_state->node_name, controld_section_lrm, cib_scope_local); // @COMPAT DCs < 1.1.14 need this deleted (in case it was explicitly false) update_attrd(lrm_state->node_name, CRM_OP_PROBED, NULL, user_name, is_remote_node); } /*! * \internal * \brief Fail a requested action without actually executing it * * For an action that can't be executed, process it similarly to an actual * execution result, with specified error status (except for notify actions, * which will always be treated as successful). * * \param[in,out] lrm_state Executor connection that action is for * \param[in] action Action XML from request * \param[in] rc Desired return code to use * \param[in] op_status Desired operation status to use * \param[in] exit_reason Human-friendly detail, if error */ static void synthesize_lrmd_failure(lrm_state_t *lrm_state, const xmlNode *action, int op_status, enum ocf_exitcode rc, const char *exit_reason) { lrmd_event_data_t *op = NULL; const char *operation = crm_element_value(action, XML_LRM_ATTR_TASK); const char *target_node = crm_element_value(action, XML_LRM_ATTR_TARGET); xmlNode *xml_rsc = find_xml_node(action, XML_CIB_TAG_RESOURCE, TRUE); if ((xml_rsc == NULL) || (ID(xml_rsc) == NULL)) { /* @TODO Should we do something else, like direct ack? */ crm_info("Can't fake %s failure (%d) on %s without resource configuration", crm_element_value(action, XML_LRM_ATTR_TASK_KEY), rc, target_node); return; } else if(operation == NULL) { /* This probably came from crm_resource -C, nothing to do */ crm_info("Can't fake %s failure (%d) on %s without operation", ID(xml_rsc), rc, target_node); return; } op = construct_op(lrm_state, action, ID(xml_rsc), operation); if (pcmk__str_eq(operation, PCMK_ACTION_NOTIFY, pcmk__str_casei)) { // Notifications can't fail fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_OK, NULL); } else { fake_op_status(lrm_state, op, op_status, rc, exit_reason); } crm_info("Faking " PCMK__OP_FMT " result (%d) on %s", op->rsc_id, op->op_type, op->interval_ms, op->rc, target_node); // Process the result as if it came from the LRM process_lrm_event(lrm_state, op, NULL, action); lrmd_free_event(op); } /*! * \internal * \brief Get target of an LRM operation (replacing \p NULL with local node * name) * * \param[in] xml LRM operation data XML * * \return LRM operation target node name (local node or Pacemaker Remote node) */ static const char * lrm_op_target(const xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = controld_globals.our_nodename; } return target; } static void fail_lrm_resource(xmlNode *xml, lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { lrmd_event_data_t *op = NULL; lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(xml, XML_CIB_TAG_RESOURCE, TRUE); CRM_CHECK(xml_rsc != NULL, return); /* The executor simply executes operations and reports the results, without * any concept of success or failure, so to fail a resource, we must fake * what a failure looks like. * * To do this, we create a fake executor operation event for the resource, * and pass that event to the executor client callback so it will be * processed as if it came from the executor. */ op = construct_op(lrm_state, xml, ID(xml_rsc), "asyncmon"); free((char*) op->user_data); op->user_data = NULL; op->interval_ms = 0; if (user_name && !pcmk__is_privileged(user_name)) { crm_err("%s does not have permission to fail %s", user_name, ID(xml_rsc)); fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, PCMK_OCF_INSUFFICIENT_PRIV, "Unprivileged user cannot fail resources"); controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); return; } if (get_lrm_resource(lrm_state, xml_rsc, TRUE, &rsc) == pcmk_ok) { crm_info("Failing resource %s...", rsc->id); fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_UNKNOWN_ERROR, "Simulated failure"); process_lrm_event(lrm_state, op, NULL, xml); op->rc = PCMK_OCF_OK; // The request to fail the resource succeeded lrmd_free_rsc_info(rsc); } else { crm_info("Cannot find/create resource in order to fail it..."); crm_log_xml_warn(xml, "bad input"); fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, PCMK_OCF_UNKNOWN_ERROR, "Cannot fail unknown resource"); } controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); } static void handle_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node, bool reprobe_all_nodes) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node, reprobe_all_nodes); if (!pcmk__strcase_any_of(from_sys, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, NULL, from_host, from_sys, CRM_SYSTEM_LRMD, controld_globals.our_uuid); crm_debug("ACK'ing re-probe from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } } static bool do_lrm_cancel(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_host, const char *from_sys) { char *op_key = NULL; char *meta_key = NULL; int call = 0; const char *call_id = NULL; const char *op_task = NULL; guint interval_ms = 0; gboolean in_progress = FALSE; xmlNode *params = find_xml_node(input->xml, XML_TAG_ATTRS, TRUE); CRM_CHECK(params != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_TASK); op_task = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(op_task != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); if (crm_element_value_ms(params, meta_key, &interval_ms) != pcmk_ok) { free(meta_key); return FALSE; } free(meta_key); op_key = pcmk__op_key(rsc->id, op_task, interval_ms); meta_key = crm_meta_name(XML_LRM_ATTR_CALLID); call_id = crm_element_value(params, meta_key); free(meta_key); crm_debug("Scheduler requested op %s (call=%s) be cancelled", op_key, (call_id? call_id : "NA")); pcmk__scan_min_int(call_id, &call, 0); if (call == 0) { // Normal case when the scheduler cancels a recurring op in_progress = cancel_op_key(lrm_state, rsc, op_key, TRUE); } else { // Normal case when the scheduler cancels an orphan op in_progress = cancel_op(lrm_state, rsc->id, NULL, call, TRUE); } // Acknowledge cancellation operation if for a remote connection resource if (!in_progress || is_remote_lrmd_ra(NULL, NULL, rsc->id)) { char *op_id = make_stop_id(rsc->id, call); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) == FALSE) { crm_info("Nothing known about operation %d for %s", call, op_key); } controld_delete_action_history_by_key(rsc->id, lrm_state->node_name, op_key, call); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); /* needed at least for cancellation of a remote operation */ if (lrm_state->active_ops != NULL) { g_hash_table_remove(lrm_state->active_ops, op_id); } free(op_id); } else { /* No ack is needed since abcdaa8, but peers with older versions * in a rolling upgrade need one. We didn't bump the feature set * at that commit, so we can only compare against the previous * CRM version (3.0.8). If any peers have feature set 3.0.9 but * not abcdaa8, they will time out waiting for the ack (no * released versions of Pacemaker are affected). */ const char *peer_version = crm_element_value(params, PCMK_XA_CRM_FEATURE_SET); if (compare_version(peer_version, "3.0.8") <= 0) { crm_info("Sending compatibility ack for %s cancellation to %s (CRM version %s)", op_key, from_host, peer_version); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); } } free(op_key); return TRUE; } static void do_lrm_delete(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_sys, const char *from_host, bool crm_rsc_delete, const char *user_name) { bool unregister = true; int cib_rc = controld_delete_resource_history(rsc->id, lrm_state->node_name, user_name, cib_dryrun|cib_sync_call); if (cib_rc != pcmk_rc_ok) { lrmd_event_data_t *op = NULL; op = construct_op(lrm_state, input->xml, rsc->id, PCMK_ACTION_DELETE); /* These are resource clean-ups, not actions, so no exit reason is * needed. */ lrmd__set_result(op, pcmk_rc2ocf(cib_rc), PCMK_EXEC_ERROR, NULL); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc->id); lrmd_free_event(op); return; } if (crm_rsc_delete && is_remote_lrmd_ra(NULL, NULL, rsc->id)) { unregister = false; } delete_resource(lrm_state, rsc->id, rsc, NULL, from_sys, user_name, input, unregister, true); } // User data for asynchronous metadata execution struct metadata_cb_data { lrmd_rsc_info_t *rsc; // Copy of resource information xmlNode *input_xml; // Copy of FSA input XML }; static struct metadata_cb_data * new_metadata_cb_data(lrmd_rsc_info_t *rsc, xmlNode *input_xml) { struct metadata_cb_data *data = NULL; data = calloc(1, sizeof(struct metadata_cb_data)); CRM_ASSERT(data != NULL); data->input_xml = copy_xml(input_xml); data->rsc = lrmd_copy_rsc_info(rsc); return data; } static void free_metadata_cb_data(struct metadata_cb_data *data) { lrmd_free_rsc_info(data->rsc); free_xml(data->input_xml); free(data); } /*! * \internal * \brief Execute an action after metadata has been retrieved * * \param[in] pid Ignored * \param[in] result Result of metadata action * \param[in] user_data Metadata callback data */ static void metadata_complete(int pid, const pcmk__action_result_t *result, void *user_data) { struct metadata_cb_data *data = (struct metadata_cb_data *) user_data; struct ra_metadata_s *md = NULL; lrm_state_t *lrm_state = lrm_state_find(lrm_op_target(data->input_xml)); if ((lrm_state != NULL) && pcmk__result_ok(result)) { md = controld_cache_metadata(lrm_state->metadata_cache, data->rsc, result->action_stdout); } if (!pcmk_is_set(controld_globals.fsa_input_register, R_HA_DISCONNECTED)) { do_lrm_rsc_op(lrm_state, data->rsc, data->input_xml, md); } free_metadata_cb_data(data); } /* A_LRM_INVOKE */ void do_lrm_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) { lrm_state_t *lrm_state = NULL; const char *crm_op = NULL; const char *from_sys = NULL; const char *from_host = NULL; const char *operation = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *user_name = NULL; const char *target_node = lrm_op_target(input->xml); gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; // Message routed to the local node is targeting a specific, non-local node is_remote_node = !pcmk__str_eq(target_node, controld_globals.our_nodename, pcmk__str_casei); lrm_state = lrm_state_find(target_node); if ((lrm_state == NULL) && is_remote_node) { crm_err("Failing action because local node has never had connection to remote node %s", target_node); synthesize_lrmd_failure(NULL, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Local node has no connection to remote"); return; } CRM_ASSERT(lrm_state != NULL); user_name = pcmk__update_acl_user(input->msg, F_CRM_USER, NULL); crm_op = crm_element_value(input->msg, F_CRM_TASK); from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } if (pcmk__str_eq(crm_op, PCMK_ACTION_LRM_DELETE, pcmk__str_none)) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { crm_rsc_delete = TRUE; // from crm_resource } operation = PCMK_ACTION_DELETE; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } CRM_CHECK(!pcmk__str_empty(crm_op) || !pcmk__str_empty(operation), return); crm_trace("'%s' execution request from %s as %s user", pcmk__s(crm_op, operation), pcmk__s(from_sys, "unknown subsystem"), pcmk__s(user_name, "current")); if (pcmk__str_eq(crm_op, CRM_OP_LRM_FAIL, pcmk__str_none)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_REFRESH, pcmk__str_none)) { /* @COMPAT This can only be sent by crm_resource --refresh on a * Pacemaker Remote node running Pacemaker 1.1.9, which is extremely * unlikely. It previously would cause the controller to re-write its * resource history to the CIB. Just ignore it. */ crm_notice("Ignoring refresh request from Pacemaker Remote 1.1.9 node"); // @COMPAT DCs <1.1.14 in a rolling upgrade might schedule this op } else if (pcmk__str_eq(operation, CRM_OP_PROBED, pcmk__str_none)) { update_attrd(lrm_state->node_name, CRM_OP_PROBED, XML_BOOLEAN_TRUE, user_name, is_remote_node); } else if (pcmk__str_eq(crm_op, CRM_OP_REPROBE, pcmk__str_none) || pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) { const char *raw_target = NULL; if (input->xml != NULL) { // For CRM_OP_REPROBE, a NULL target means we're targeting all nodes raw_target = crm_element_value(input->xml, XML_LRM_ATTR_TARGET); } handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node, (raw_target == NULL)); } else if (operation != NULL) { lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(input->xml, XML_CIB_TAG_RESOURCE, TRUE); gboolean create_rsc = !pcmk__str_eq(operation, PCMK_ACTION_DELETE, pcmk__str_none); int rc; // We can't return anything meaningful without a resource ID CRM_CHECK(xml_rsc && ID(xml_rsc), return); rc = get_lrm_resource(lrm_state, xml_rsc, create_rsc, &rsc); if (rc == -ENOTCONN) { synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Not connected to remote executor"); return; } else if ((rc < 0) && !create_rsc) { /* Delete of malformed or nonexistent resource * (deleting something that does not exist is a success) */ crm_notice("Not registering resource '%s' for a %s event " CRM_XS " get-rc=%d (%s) transition-key=%s", ID(xml_rsc), operation, rc, pcmk_strerror(rc), ID(input->xml)); delete_rsc_entry(lrm_state, input, ID(xml_rsc), NULL, pcmk_ok, user_name, true); return; } else if (rc == -EINVAL) { // Resource operation on malformed resource crm_err("Invalid resource definition for %s", ID(xml_rsc)); crm_log_xml_warn(input->msg, "invalid resource"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, PCMK_OCF_NOT_CONFIGURED, // fatal error "Invalid resource definition"); return; } else if (rc < 0) { // Error communicating with the executor crm_err("Could not register resource '%s' with executor: %s " CRM_XS " rc=%d", ID(xml_rsc), pcmk_strerror(rc), rc); crm_log_xml_warn(input->msg, "failed registration"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, PCMK_OCF_INVALID_PARAM, // hard error "Could not register resource with executor"); return; } if (pcmk__str_eq(operation, PCMK_ACTION_CANCEL, pcmk__str_none)) { if (!do_lrm_cancel(input, lrm_state, rsc, from_host, from_sys)) { crm_log_xml_warn(input->xml, "Bad command"); } } else if (pcmk__str_eq(operation, PCMK_ACTION_DELETE, pcmk__str_none)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else { struct ra_metadata_s *md = NULL; /* Getting metadata from cache is OK except for start actions -- * always refresh from the agent for those, in case the resource * agent was updated. * * @TODO Only refresh metadata for starts if the agent actually * changed (using something like inotify, or a hash or modification * time of the agent executable). */ if (strcmp(operation, PCMK_ACTION_START) != 0) { md = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); } if ((md == NULL) && crm_op_needs_metadata(rsc->standard, operation)) { /* Most likely, we'll need the agent metadata to record the * pending operation and the operation result. Get it now rather * than wait until then, so the metadata action doesn't eat into * the real action's timeout. * * @TODO Metadata is retrieved via direct execution of the * agent, which has a couple of related issues: the executor * should execute agents, not the controller; and metadata for * Pacemaker Remote nodes should be collected on those nodes, * not locally. */ struct metadata_cb_data *data = NULL; data = new_metadata_cb_data(rsc, input->xml); crm_info("Retrieving metadata for %s (%s%s%s:%s) asynchronously", rsc->id, rsc->standard, ((rsc->provider == NULL)? "" : ":"), ((rsc->provider == NULL)? "" : rsc->provider), rsc->type); (void) lrmd__metadata_async(rsc, metadata_complete, (void *) data); } else { do_lrm_rsc_op(lrm_state, rsc, input->xml, md); } } lrmd_free_rsc_info(rsc); } else { crm_err("Invalid execution request: unknown command '%s' (bug?)", crm_op); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } static lrmd_event_data_t * construct_op(const lrm_state_t *lrm_state, const xmlNode *rsc_op, const char *rsc_id, const char *operation) { lrmd_event_data_t *op = NULL; const char *op_delay = NULL; const char *op_timeout = NULL; GHashTable *params = NULL; xmlNode *primitive = NULL; const char *class = NULL; const char *transition = NULL; CRM_ASSERT(rsc_id && operation); op = lrmd_new_event(rsc_id, operation, 0); op->type = lrmd_event_exec_complete; op->timeout = 0; op->start_delay = 0; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); if (rsc_op == NULL) { CRM_LOG_ASSERT(pcmk__str_eq(operation, PCMK_ACTION_STOP, pcmk__str_casei)); op->user_data = NULL; /* the stop_all_resources() case * by definition there is no DC (or they'd be shutting * us down). * So we should put our version here. */ op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(PCMK_XA_CRM_FEATURE_SET), strdup(CRM_FEATURE_SET)); crm_trace("Constructed %s op for %s", operation, rsc_id); return op; } params = xml2list(rsc_op); g_hash_table_remove(params, CRM_META "_op_target_rc"); op_delay = crm_meta_value(params, XML_OP_ATTR_START_DELAY); pcmk__scan_min_int(op_delay, &op->start_delay, 0); op_timeout = crm_meta_value(params, XML_ATTR_TIMEOUT); pcmk__scan_min_int(op_timeout, &op->timeout, 0); if (pcmk__guint_from_hash(params, CRM_META "_" XML_LRM_ATTR_INTERVAL_MS, 0, &(op->interval_ms)) != pcmk_rc_ok) { op->interval_ms = 0; } /* Use pcmk_monitor_timeout instead of meta timeout for stonith recurring monitor, if set */ primitive = find_xml_node(rsc_op, XML_CIB_TAG_RESOURCE, FALSE); class = crm_element_value(primitive, PCMK_XA_CLASS); if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_fence_params) && pcmk__str_eq(operation, PCMK_ACTION_MONITOR, pcmk__str_casei) && (op->interval_ms > 0)) { op_timeout = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (op_timeout != NULL) { op->timeout = crm_get_msec(op_timeout); } } if (!pcmk__str_eq(operation, PCMK_ACTION_STOP, pcmk__str_casei)) { op->params = params; } else { rsc_history_t *entry = NULL; if (lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* If we do not have stop parameters cached, use * whatever we are given */ if (!entry || !entry->stop_params) { op->params = params; } else { /* Copy the cached parameter list so that we stop the resource * with the old attributes, not the new ones */ op->params = pcmk__strkey_table(free, free); g_hash_table_foreach(params, copy_meta_keys, op->params); g_hash_table_foreach(entry->stop_params, copy_instance_keys, op->params); g_hash_table_destroy(params); params = NULL; } } /* sanity */ if (op->timeout <= 0) { op->timeout = op->interval_ms; } if (op->start_delay < 0) { op->start_delay = 0; } transition = crm_element_value(rsc_op, XML_ATTR_TRANSITION_KEY); CRM_CHECK(transition != NULL, return op); op->user_data = strdup(transition); if (op->interval_ms != 0) { if (pcmk__strcase_any_of(operation, PCMK_ACTION_START, PCMK_ACTION_STOP, NULL)) { crm_err("Start and Stop actions cannot have an interval: %u", op->interval_ms); op->interval_ms = 0; } } crm_trace("Constructed %s op for %s: interval=%u", operation, rsc_id, op->interval_ms); return op; } /*! * \internal * \brief Send a (synthesized) event result * * Reply with a synthesized event result directly, as opposed to going through * the executor. * * \param[in] to_host Host to send result to * \param[in] to_sys IPC name to send result (NULL for transition engine) * \param[in] rsc Type information about resource the result is for * \param[in,out] op Event with result to send * \param[in] rsc_id ID of resource the result is for */ void controld_ack_event_directly(const char *to_host, const char *to_sys, const lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, const char *rsc_id) { xmlNode *reply = NULL; xmlNode *update, *iter; crm_node_t *peer = NULL; CRM_CHECK(op != NULL, return); if (op->rsc_id == NULL) { CRM_ASSERT(rsc_id != NULL); op->rsc_id = strdup(rsc_id); } if (to_sys == NULL) { to_sys = CRM_SYSTEM_TENGINE; } peer = crm_get_peer(0, controld_globals.our_nodename); update = create_node_state_update(peer, node_update_none, NULL, __func__); iter = create_xml_node(update, XML_CIB_TAG_LRM); crm_xml_add(iter, PCMK_XA_ID, controld_globals.our_uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, PCMK_XA_ID, op->rsc_id); controld_add_resource_history_xml(iter, rsc, op, controld_globals.our_nodename); reply = create_request(CRM_OP_INVOKE_LRM, update, to_host, to_sys, CRM_SYSTEM_LRMD, NULL); crm_log_xml_trace(update, "[direct ACK]"); crm_debug("ACK'ing resource op " PCMK__OP_FMT " from %s: %s", op->rsc_id, op->op_type, op->interval_ms, op->user_data, - crm_element_value(reply, XML_ATTR_REFERENCE)); + crm_element_value(reply, PCMK_XA_REFERENCE)); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(update); free_xml(reply); } gboolean verify_stopped(enum crmd_fsa_state cur_state, int log_level) { gboolean res = TRUE; GList *lrm_state_list = lrm_state_get_list(); GList *state_entry; for (state_entry = lrm_state_list; state_entry != NULL; state_entry = state_entry->next) { lrm_state_t *lrm_state = state_entry->data; if (!lrm_state_verify_stopped(lrm_state, cur_state, log_level)) { /* keep iterating through all even when false is returned */ res = FALSE; } } controld_set_fsa_input_flags(R_SENT_RSC_STOP); g_list_free(lrm_state_list); lrm_state_list = NULL; return res; } struct stop_recurring_action_s { lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean stop_recurring_action_by_rsc(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct stop_recurring_action_s *event = user_data; active_op_t *op = value; if ((op->interval_ms != 0) && pcmk__str_eq(op->rsc_id, event->rsc->id, pcmk__str_none)) { crm_debug("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (char*)key); remove = !cancel_op(event->lrm_state, event->rsc->id, key, op->call_id, FALSE); } return remove; } static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; lrm_state_t *lrm_state = user_data; active_op_t *op = value; if (op->interval_ms != 0) { crm_info("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (const char *) key); remove = !cancel_op(lrm_state, op->rsc_id, key, op->call_id, FALSE); } return remove; } /*! * \internal * \brief Check whether recurring actions should be cancelled before an action * * \param[in] rsc_id Resource that action is for * \param[in] action Action being performed * \param[in] interval_ms Operation interval of \p action (in milliseconds) * * \return true if recurring actions should be cancelled, otherwise false */ static bool should_cancel_recurring(const char *rsc_id, const char *action, guint interval_ms) { if (is_remote_lrmd_ra(NULL, NULL, rsc_id) && (interval_ms == 0) && (strcmp(action, PCMK_ACTION_MIGRATE_TO) == 0)) { /* Don't stop monitoring a migrating Pacemaker Remote connection * resource until the entire migration has completed. We must detect if * the connection is unexpectedly severed, even during a migration. */ return false; } // Cancel recurring actions before changing resource state return (interval_ms == 0) && !pcmk__str_any_of(action, PCMK_ACTION_MONITOR, PCMK_ACTION_NOTIFY, NULL); } /*! * \internal * \brief Check whether an action should not be performed at this time * * \param[in] operation Action to be performed * * \return Readable description of why action should not be performed, * or NULL if it should be performed */ static const char * should_nack_action(const char *action) { if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN) && pcmk__str_eq(action, PCMK_ACTION_START, pcmk__str_none)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); return "Not attempting start due to shutdown in progress"; } switch (controld_globals.fsa_state) { case S_NOT_DC: case S_POLICY_ENGINE: // Recalculating case S_TRANSITION_ENGINE: break; default: if (!pcmk__str_eq(action, PCMK_ACTION_STOP, pcmk__str_none)) { return "Controller cannot attempt actions at this time"; } break; } return NULL; } static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, xmlNode *msg, struct ra_metadata_s *md) { int rc; int call_id = 0; char *op_id = NULL; lrmd_event_data_t *op = NULL; fsa_data_t *msg_data = NULL; const char *transition = NULL; const char *operation = NULL; const char *nack_reason = NULL; CRM_CHECK((rsc != NULL) && (msg != NULL), return); operation = crm_element_value(msg, XML_LRM_ATTR_TASK); CRM_CHECK(!pcmk__str_empty(operation), return); transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (pcmk__str_empty(transition)) { crm_log_xml_err(msg, "Missing transition number"); } if (lrm_state == NULL) { // This shouldn't be possible, but provide a failsafe just in case crm_err("Cannot execute %s of %s: No executor connection " CRM_XS " transition_key=%s", operation, rsc->id, pcmk__s(transition, "")); synthesize_lrmd_failure(NULL, msg, PCMK_EXEC_INVALID, PCMK_OCF_UNKNOWN_ERROR, "No executor connection"); return; } if (pcmk__str_any_of(operation, PCMK_ACTION_RELOAD, PCMK_ACTION_RELOAD_AGENT, NULL)) { /* Pre-2.1.0 DCs will schedule reload actions only, and 2.1.0+ DCs * will schedule reload-agent actions only. In either case, we need * to map that to whatever the resource agent actually supports. * Default to the OCF 1.1 name. */ if ((md != NULL) && pcmk_is_set(md->ra_flags, ra_supports_legacy_reload)) { operation = PCMK_ACTION_RELOAD; } else { operation = PCMK_ACTION_RELOAD_AGENT; } } op = construct_op(lrm_state, msg, rsc->id, operation); CRM_CHECK(op != NULL, return); if (should_cancel_recurring(rsc->id, operation, op->interval_ms)) { guint removed = 0; struct stop_recurring_action_s data; data.rsc = rsc; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->active_ops, stop_recurring_action_by_rsc, &data); if (removed) { crm_debug("Stopped %u recurring operation%s in preparation for " PCMK__OP_FMT, removed, pcmk__plural_s(removed), rsc->id, operation, op->interval_ms); } } /* now do the op */ crm_notice("Requesting local execution of %s operation for %s on %s " CRM_XS " transition_key=%s op_key=" PCMK__OP_FMT, pcmk__readable_action(op->op_type, op->interval_ms), rsc->id, lrm_state->node_name, pcmk__s(transition, ""), rsc->id, operation, op->interval_ms); nack_reason = should_nack_action(operation); if (nack_reason != NULL) { crm_notice("Discarding attempt to perform action %s on %s in state %s " "(shutdown=%s)", operation, rsc->id, fsa_state2string(controld_globals.fsa_state), pcmk__btoa(pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN))); lrmd__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_INVALID, nack_reason); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); lrmd_free_event(op); free(op_id); return; } controld_record_pending_op(lrm_state->node_name, rsc, op); op_id = pcmk__op_key(rsc->id, op->op_type, op->interval_ms); if (op->interval_ms > 0) { /* cancel it so we can then restart it without conflict */ cancel_op_key(lrm_state, rsc, op_id, FALSE); } rc = controld_execute_resource_agent(lrm_state, rsc->id, op->op_type, op->user_data, op->interval_ms, op->timeout, op->start_delay, op->params, &call_id); if (rc == pcmk_rc_ok) { /* record all operations so we can wait * for them to complete during shutdown */ char *call_id_s = make_stop_id(rsc->id, call_id); active_op_t *pending = NULL; pending = calloc(1, sizeof(active_op_t)); crm_trace("Recording pending op: %d - %s %s", call_id, op_id, call_id_s); pending->call_id = call_id; pending->interval_ms = op->interval_ms; pending->op_type = strdup(operation); pending->op_key = strdup(op_id); pending->rsc_id = strdup(rsc->id); pending->start_time = time(NULL); pcmk__str_update(&pending->user_data, op->user_data); if (crm_element_value_epoch(msg, PCMK_OPT_SHUTDOWN_LOCK, &(pending->lock_time)) != pcmk_ok) { pending->lock_time = 0; } g_hash_table_replace(lrm_state->active_ops, call_id_s, pending); if ((op->interval_ms > 0) && (op->start_delay > START_DELAY_THRESHOLD)) { int target_rc = PCMK_OCF_OK; crm_info("Faking confirmation of %s: execution postponed for over 5 minutes", op_id); decode_transition_key(op->user_data, NULL, NULL, NULL, &target_rc); lrmd__set_result(op, target_rc, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); } pending->params = op->params; op->params = NULL; } else if (lrm_state_is_local(lrm_state)) { crm_err("Could not initiate %s action for resource %s locally: %s " CRM_XS " rc=%d", operation, rsc->id, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } else { crm_err("Could not initiate %s action for resource %s remotely on %s: " "%s " CRM_XS " rc=%d", operation, rsc->id, lrm_state->node_name, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); } free(op_id); lrmd_free_event(op); } void do_lrm_event(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input cur_input, fsa_data_t * msg_data) { CRM_CHECK(FALSE, return); } static char * unescape_newlines(const char *string) { char *pch = NULL; char *ret = NULL; static const char *escaped_newline = "\\n"; if (!string) { return NULL; } ret = strdup(string); pch = strstr(ret, escaped_newline); while (pch != NULL) { /* Replace newline escape pattern with actual newline (and a space so we * don't have to shuffle the rest of the buffer) */ pch[0] = '\n'; pch[1] = ' '; pch = strstr(pch, escaped_newline); } return ret; } static bool did_lrm_rsc_op_fail(lrm_state_t *lrm_state, const char * rsc_id, const char * op_type, guint interval_ms) { rsc_history_t *entry = NULL; CRM_CHECK(lrm_state != NULL, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); CRM_CHECK(op_type != NULL, return FALSE); entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->failed == NULL) { return FALSE; } if (pcmk__str_eq(entry->failed->rsc_id, rsc_id, pcmk__str_none) && pcmk__str_eq(entry->failed->op_type, op_type, pcmk__str_casei) && entry->failed->interval_ms == interval_ms) { return TRUE; } return FALSE; } /*! * \internal * \brief Log the result of an executor action (actual or synthesized) * * \param[in] op Executor action to log result for * \param[in] op_key Operation key for action * \param[in] node_name Name of node action was performed on, if known * \param[in] confirmed Whether to log that graph action was confirmed */ static void log_executor_event(const lrmd_event_data_t *op, const char *op_key, const char *node_name, gboolean confirmed) { int log_level = LOG_ERR; GString *str = g_string_sized_new(100); // reasonable starting size pcmk__g_strcat(str, "Result of ", pcmk__readable_action(op->op_type, op->interval_ms), " operation for ", op->rsc_id, NULL); if (node_name != NULL) { pcmk__g_strcat(str, " on ", node_name, NULL); } switch (op->op_status) { case PCMK_EXEC_DONE: log_level = LOG_NOTICE; pcmk__g_strcat(str, ": ", services_ocf_exitcode_str(op->rc), NULL); break; case PCMK_EXEC_TIMEOUT: pcmk__g_strcat(str, ": ", pcmk_exec_status_str(op->op_status), " after ", pcmk__readable_interval(op->timeout), NULL); break; case PCMK_EXEC_CANCELLED: log_level = LOG_INFO; /* order of __attribute__ and Fall through comment is IMPORTANT! * do not change it without proper testing with both clang and gcc * in multiple versions. * the clang check allows to build with all versions of clang. * the has_c_attribute check is to workaround a bug in clang version * in rhel7. has_attribute would happily return "YES SIR WE GOT IT" * and fail the build the next line. */ #ifdef __clang__ #ifdef __has_c_attribute #if __has_attribute(fallthrough) __attribute__((fallthrough)); #endif #endif #endif // Fall through default: pcmk__g_strcat(str, ": ", pcmk_exec_status_str(op->op_status), NULL); } if ((op->exit_reason != NULL) && ((op->op_status != PCMK_EXEC_DONE) || (op->rc != PCMK_OCF_OK))) { pcmk__g_strcat(str, " (", op->exit_reason, ")", NULL); } g_string_append(str, " " CRM_XS); g_string_append_printf(str, " graph action %sconfirmed; call=%d key=%s", (confirmed? "" : "un"), op->call_id, op_key); if (op->op_status == PCMK_EXEC_DONE) { g_string_append_printf(str, " rc=%d", op->rc); } do_crm_log(log_level, "%s", str->str); g_string_free(str, TRUE); /* The services library has already logged the output at info or debug * level, so just raise to notice if it looks like a failure. */ if ((op->output != NULL) && (op->rc != PCMK_OCF_OK)) { char *prefix = crm_strdup_printf(PCMK__OP_FMT "@%s output", op->rsc_id, op->op_type, op->interval_ms, node_name); crm_log_output(LOG_NOTICE, prefix, op->output); free(prefix); } } void process_lrm_event(lrm_state_t *lrm_state, lrmd_event_data_t *op, active_op_t *pending, const xmlNode *action_xml) { char *op_id = NULL; char *op_key = NULL; gboolean remove = FALSE; gboolean removed = FALSE; bool need_direct_ack = FALSE; lrmd_rsc_info_t *rsc = NULL; const char *node_name = NULL; CRM_CHECK(op != NULL, return); CRM_CHECK(op->rsc_id != NULL, return); // Remap new status codes for older DCs if (compare_version(controld_globals.dc_version, "3.2.0") < 0) { switch (op->op_status) { case PCMK_EXEC_NOT_CONNECTED: lrmd__set_result(op, PCMK_OCF_CONNECTION_DIED, PCMK_EXEC_ERROR, op->exit_reason); break; case PCMK_EXEC_INVALID: lrmd__set_result(op, CRM_DIRECT_NACK_RC, PCMK_EXEC_ERROR, op->exit_reason); break; default: break; } } op_id = make_stop_id(op->rsc_id, op->call_id); op_key = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); // Get resource info if available (from executor state or action XML) if (lrm_state) { rsc = lrm_state_get_rsc_info(lrm_state, op->rsc_id, 0); } if ((rsc == NULL) && action_xml) { xmlNode *xml = find_xml_node(action_xml, XML_CIB_TAG_RESOURCE, TRUE); const char *standard = crm_element_value(xml, PCMK_XA_CLASS); const char *provider = crm_element_value(xml, PCMK_XA_PROVIDER); const char *type = crm_element_value(xml, PCMK_XA_TYPE); if (standard && type) { crm_info("%s agent information not cached, using %s%s%s:%s from action XML", op->rsc_id, standard, (provider? ":" : ""), (provider? provider : ""), type); rsc = lrmd_new_rsc_info(op->rsc_id, standard, provider, type); } else { crm_err("Can't process %s result because %s agent information not cached or in XML", op_key, op->rsc_id); } } // Get node name if available (from executor state or action XML) if (lrm_state) { node_name = lrm_state->node_name; } else if (action_xml) { node_name = crm_element_value(action_xml, XML_LRM_ATTR_TARGET); } if(pending == NULL) { remove = TRUE; if (lrm_state) { pending = g_hash_table_lookup(lrm_state->active_ops, op_id); } } if (op->op_status == PCMK_EXEC_ERROR) { switch(op->rc) { case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Leave it to the TE/scheduler to decide if this is an error op->op_status = PCMK_EXEC_DONE; break; default: /* Nothing to do */ break; } } if (op->op_status != PCMK_EXEC_CANCELLED) { /* We might not record the result, so directly acknowledge it to the * originator instead, so it doesn't time out waiting for the result * (especially important if part of a transition). */ need_direct_ack = TRUE; if (controld_action_is_recordable(op->op_type)) { if (node_name && rsc) { // We should record the result, and happily, we can time_t lock_time = (pending == NULL)? 0 : pending->lock_time; controld_update_resource_history(node_name, rsc, op, lock_time); need_direct_ack = FALSE; } else if (op->rsc_deleted) { /* We shouldn't record the result (likely the resource was * refreshed, cleaned, or removed while this operation was * in flight). */ crm_notice("Not recording %s result in CIB because " "resource information was removed since it was initiated", op_key); } else { /* This shouldn't be possible; the executor didn't consider the * resource deleted, but we couldn't find resource or node * information. */ crm_err("Unable to record %s result in CIB: %s", op_key, (node_name? "No resource information" : "No node name")); } } } else if (op->interval_ms == 0) { /* A non-recurring operation was cancelled. Most likely, the * never-initiated action was removed from the executor's pending * operations list upon resource removal. */ need_direct_ack = TRUE; } else if (pending == NULL) { /* This recurring operation was cancelled, but was not pending. No * transition actions are waiting on it, nothing needs to be done. */ } else if (op->user_data == NULL) { /* This recurring operation was cancelled and pending, but we don't * have a transition key. This should never happen. */ crm_err("Recurring operation %s was cancelled without transition information", op_key); } else if (pcmk_is_set(pending->flags, active_op_remove)) { /* This recurring operation was cancelled (by us) and pending, and we * have been waiting for it to finish. */ if (lrm_state) { controld_delete_action_history(op); } /* Directly acknowledge failed recurring actions here. The above call to * controld_delete_action_history() will not erase any corresponding * last_failure entry, which means that the DC won't confirm the * cancellation via process_op_deletion(), and the transition would * otherwise wait for the action timer to pop. */ if (did_lrm_rsc_op_fail(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms)) { need_direct_ack = TRUE; } } else if (op->rsc_deleted) { /* This recurring operation was cancelled (but not by us, and the * executor does not have resource information, likely due to resource * cleanup, refresh, or removal) and pending. */ crm_debug("Recurring op %s was cancelled due to resource deletion", op_key); need_direct_ack = TRUE; } else { /* This recurring operation was cancelled (but not by us, likely by the * executor before stopping the resource) and pending. We don't need to * do anything special. */ } if (need_direct_ack) { controld_ack_event_directly(NULL, NULL, NULL, op, op->rsc_id); } if(remove == FALSE) { /* The caller will do this afterwards, but keep the logging consistent */ removed = TRUE; } else if (lrm_state && ((op->interval_ms == 0) || (op->op_status == PCMK_EXEC_CANCELLED))) { gboolean found = g_hash_table_remove(lrm_state->active_ops, op_id); if (op->interval_ms != 0) { removed = TRUE; } else if (found) { removed = TRUE; crm_trace("Op %s (call=%d, stop-id=%s, remaining=%u): Confirmed", op_key, op->call_id, op_id, g_hash_table_size(lrm_state->active_ops)); } } log_executor_event(op, op_key, node_name, removed); if (lrm_state) { if (!pcmk__str_eq(op->op_type, PCMK_ACTION_META_DATA, pcmk__str_casei)) { crmd_alert_resource_op(lrm_state->node_name, op); } else if (rsc && (op->rc == PCMK_OCF_OK)) { char *metadata = unescape_newlines(op->output); controld_cache_metadata(lrm_state->metadata_cache, rsc, metadata); free(metadata); } } if (op->rsc_deleted) { crm_info("Deletion of resource '%s' complete after %s", op->rsc_id, op_key); if (lrm_state) { delete_rsc_entry(lrm_state, NULL, op->rsc_id, NULL, pcmk_ok, NULL, true); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ controld_trigger_fsa(); if (lrm_state && rsc) { update_history_cache(lrm_state, rsc, op); } lrmd_free_rsc_info(rsc); free(op_key); free(op_id); } diff --git a/daemons/controld/controld_join_dc.c b/daemons/controld/controld_join_dc.c index 5c876bd5e5..9916845a19 100644 --- a/daemons/controld/controld_join_dc.c +++ b/daemons/controld/controld_join_dc.c @@ -1,1034 +1,1034 @@ /* * Copyright 2004-2024 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 static char *max_generation_from = NULL; static xmlNodePtr max_generation_xml = NULL; /*! * \internal * \brief Nodes from which a CIB sync has failed since the peer joined * * This table is of the form (node_name -> join_id). \p node_name is * the name of a client node from which a CIB \p sync_from() call has failed in * \p do_dc_join_finalize() since the client joined the cluster as a peer. * \p join_id is the ID of the join round in which the \p sync_from() failed, * and is intended for use in nack log messages. */ static GHashTable *failed_sync_nodes = NULL; void finalize_join_for(gpointer key, gpointer value, gpointer user_data); void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data); gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); /* Numeric counter used to identify join rounds (an unsigned int would be * appropriate, except we get and set it in XML as int) */ static int current_join_id = 0; /*! * \internal * \brief Destroy the hash table containing failed sync nodes */ void controld_destroy_failed_sync_table(void) { if (failed_sync_nodes != NULL) { g_hash_table_destroy(failed_sync_nodes); failed_sync_nodes = NULL; } } /*! * \internal * \brief Remove a node from the failed sync nodes table if present * * \param[in] node_name Node name to remove */ void controld_remove_failed_sync_node(const char *node_name) { if (failed_sync_nodes != NULL) { g_hash_table_remove(failed_sync_nodes, (gchar *) node_name); } } /*! * \internal * \brief Add to a hash table a node whose CIB failed to sync * * \param[in] node_name Name of node whose CIB failed to sync * \param[in] join_id Join round when the failure occurred */ static void record_failed_sync_node(const char *node_name, gint join_id) { if (failed_sync_nodes == NULL) { failed_sync_nodes = pcmk__strikey_table(g_free, NULL); } /* If the node is already in the table then we failed to nack it during the * filter offer step */ CRM_LOG_ASSERT(g_hash_table_insert(failed_sync_nodes, g_strdup(node_name), GINT_TO_POINTER(join_id))); } /*! * \internal * \brief Look up a node name in the failed sync table * * \param[in] node_name Name of node to look up * \param[out] join_id Where to store the join ID of when the sync failed * * \return Standard Pacemaker return code. Specifically, \p pcmk_rc_ok if the * node name was found, or \p pcmk_rc_node_unknown otherwise. * \note \p *join_id is set to -1 if the node is not found. */ static int lookup_failed_sync_node(const char *node_name, gint *join_id) { *join_id = -1; if (failed_sync_nodes != NULL) { gpointer result = g_hash_table_lookup(failed_sync_nodes, (gchar *) node_name); if (result != NULL) { *join_id = GPOINTER_TO_INT(result); return pcmk_rc_ok; } } return pcmk_rc_node_unknown; } void crm_update_peer_join(const char *source, crm_node_t * node, enum crm_join_phase phase) { enum crm_join_phase last = 0; CRM_CHECK(node != NULL, return); /* Remote nodes do not participate in joins */ if (pcmk_is_set(node->flags, crm_remote_node)) { return; } last = node->join; if(phase == last) { crm_trace("Node %s join-%d phase is still %s " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(last), node->id, source); } else if ((phase <= crm_join_none) || (phase == (last + 1))) { node->join = phase; crm_trace("Node %s join-%d phase is now %s (was %s) " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(phase), crm_join_phase_str(last), node->id, source); } else { crm_warn("Rejecting join-%d phase update for node %s because " "can't go from %s to %s " CRM_XS " nodeid=%u source=%s", current_join_id, node->uname, crm_join_phase_str(last), crm_join_phase_str(phase), node->id, source); } } static void start_join_round(void) { GHashTableIter iter; crm_node_t *peer = NULL; crm_debug("Starting new join round join-%d", current_join_id); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { crm_update_peer_join(__func__, peer, crm_join_none); } if (max_generation_from != NULL) { free(max_generation_from); max_generation_from = NULL; } if (max_generation_xml != NULL) { free_xml(max_generation_xml); max_generation_xml = NULL; } controld_clear_fsa_input_flags(R_HAVE_CIB); } /*! * \internal * \brief Create a join message from the DC * * \param[in] join_op Join operation name * \param[in] host_to Recipient of message */ static xmlNode * create_dc_message(const char *join_op, const char *host_to) { xmlNode *msg = create_request(join_op, NULL, host_to, CRM_SYSTEM_CRMD, CRM_SYSTEM_DC, NULL); /* Identify which election this is a part of */ crm_xml_add_int(msg, F_CRM_JOIN_ID, current_join_id); /* Add a field specifying whether the DC is shutting down. This keeps the * joining node from fencing the old DC if it becomes the new DC. */ pcmk__xe_set_bool_attr(msg, F_CRM_DC_LEAVING, pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)); return msg; } static void join_make_offer(gpointer key, gpointer value, gpointer user_data) { xmlNode *offer = NULL; crm_node_t *member = (crm_node_t *)value; CRM_ASSERT(member != NULL); if (crm_is_peer_active(member) == FALSE) { crm_info("Not making join-%d offer to inactive node %s", current_join_id, (member->uname? member->uname : "with unknown name")); if(member->expected == NULL && pcmk__str_eq(member->state, CRM_NODE_LOST, pcmk__str_casei)) { /* You would think this unsafe, but in fact this plus an * active resource is what causes it to be fenced. * * Yes, this does mean that any node that dies at the same * time as the old DC and is not running resource (still) * won't be fenced. * * I'm not happy about this either. */ pcmk__update_peer_expected(__func__, member, CRMD_JOINSTATE_DOWN); } return; } if (member->uname == NULL) { crm_info("Not making join-%d offer to node uuid %s with unknown name", current_join_id, member->uuid); return; } if (controld_globals.membership_id != crm_peer_seq) { controld_globals.membership_id = crm_peer_seq; crm_info("Making join-%d offers based on membership event %llu", current_join_id, crm_peer_seq); } if(user_data && member->join > crm_join_none) { crm_info("Not making join-%d offer to already known node %s (%s)", current_join_id, member->uname, crm_join_phase_str(member->join)); return; } crm_update_peer_join(__func__, (crm_node_t*)member, crm_join_none); offer = create_dc_message(CRM_OP_JOIN_OFFER, member->uname); // Advertise our feature set so the joining node can bail if not compatible crm_xml_add(offer, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET); crm_info("Sending join-%d offer to %s", current_join_id, member->uname); send_cluster_message(member, crm_msg_crmd, offer, TRUE); free_xml(offer); crm_update_peer_join(__func__, member, crm_join_welcomed); } /* A_DC_JOIN_OFFER_ALL */ void do_dc_join_offer_all(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) { int count; /* Reset everyone's status back to down or in_ccm in the CIB. * Any nodes that are active in the CIB but not in the cluster membership * will be seen as offline by the scheduler anyway. */ current_join_id++; start_join_round(); update_dc(NULL); if (cause == C_HA_MESSAGE && current_input == I_NODE_JOIN) { crm_info("A new node joined the cluster"); } g_hash_table_foreach(crm_peer_cache, join_make_offer, NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } /* A_DC_JOIN_OFFER_ONE */ void do_dc_join_offer_one(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) { crm_node_t *member; ha_msg_input_t *welcome = NULL; int count; const char *join_to = NULL; if (msg_data->data == NULL) { crm_info("Making join-%d offers to any unconfirmed nodes " "because an unknown node joined", current_join_id); g_hash_table_foreach(crm_peer_cache, join_make_offer, &member); check_join_state(cur_state, __func__); return; } welcome = fsa_typed_data(fsa_dt_ha_msg); if (welcome == NULL) { // fsa_typed_data() already logged an error return; } join_to = crm_element_value(welcome->msg, F_CRM_HOST_FROM); if (join_to == NULL) { crm_err("Can't make join-%d offer to unknown node", current_join_id); return; } member = crm_get_peer(0, join_to); /* It is possible that a node will have been sick or starting up when the * original offer was made. However, it will either re-announce itself in * due course, or we can re-store the original offer on the client. */ crm_update_peer_join(__func__, member, crm_join_none); join_make_offer(NULL, member, NULL); /* If the offer isn't to the local node, make an offer to the local node as * well, to ensure the correct value for max_generation_from. */ if (strcasecmp(join_to, controld_globals.our_nodename) != 0) { member = crm_get_peer(0, controld_globals.our_nodename); join_make_offer(NULL, member, NULL); } /* This was a genuine join request; cancel any existing transition and * invoke the scheduler. */ abort_transition(INFINITY, pcmk__graph_restart, "Node join", NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } static int compare_int_fields(xmlNode * left, xmlNode * right, const char *field) { const char *elem_l = crm_element_value(left, field); const char *elem_r = crm_element_value(right, field); long long int_elem_l; long long int_elem_r; pcmk__scan_ll(elem_l, &int_elem_l, -1LL); pcmk__scan_ll(elem_r, &int_elem_r, -1LL); if (int_elem_l < int_elem_r) { return -1; } else if (int_elem_l > int_elem_r) { return 1; } return 0; } /* A_DC_JOIN_PROCESS_REQ */ void do_dc_join_filter_offer(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) { xmlNode *generation = NULL; int cmp = 0; int join_id = -1; int count = 0; gint value = 0; gboolean ack_nack_bool = TRUE; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *join_from = crm_element_value(join_ack->msg, F_CRM_HOST_FROM); - const char *ref = crm_element_value(join_ack->msg, F_CRM_REFERENCE); + const char *ref = crm_element_value(join_ack->msg, PCMK_XA_REFERENCE); const char *join_version = crm_element_value(join_ack->msg, PCMK_XA_CRM_FEATURE_SET); crm_node_t *join_node = NULL; if (join_from == NULL) { crm_err("Ignoring invalid join request without node name"); return; } join_node = crm_get_peer(0, join_from); crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id); if (join_id != current_join_id) { crm_debug("Ignoring join-%d request from %s because we are on join-%d", join_id, join_from, current_join_id); check_join_state(cur_state, __func__); return; } generation = join_ack->xml; if (max_generation_xml != NULL && generation != NULL) { int lpc = 0; const char *attributes[] = { PCMK_XA_ADMIN_EPOCH, PCMK_XA_EPOCH, PCMK_XA_NUM_UPDATES, }; for (lpc = 0; cmp == 0 && lpc < PCMK__NELEM(attributes); lpc++) { cmp = compare_int_fields(max_generation_xml, generation, attributes[lpc]); } } if (ref == NULL) { ref = "none"; // for logging only } if (lookup_failed_sync_node(join_from, &value) == pcmk_rc_ok) { crm_err("Rejecting join-%d request from node %s because we failed to " "sync its CIB in join-%d " CRM_XS " ref=%s", join_id, join_from, value, ref); ack_nack_bool = FALSE; } else if (!crm_is_peer_active(join_node)) { if (match_down_event(join_from) != NULL) { /* The join request was received after the node was fenced or * otherwise shutdown in a way that we're aware of. No need to log * an error in this rare occurrence; we know the client was recently * shut down, and receiving a lingering in-flight request is not * cause for alarm. */ crm_debug("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } else { crm_err("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } ack_nack_bool = FALSE; } else if (generation == NULL) { crm_err("Rejecting invalid join-%d request from node %s " "missing CIB generation " CRM_XS " ref=%s", join_id, join_from, ref); ack_nack_bool = FALSE; } else if ((join_version == NULL) || !feature_set_compatible(CRM_FEATURE_SET, join_version)) { crm_err("Rejecting join-%d request from node %s because feature set %s" " is incompatible with ours (%s) " CRM_XS " ref=%s", join_id, join_from, (join_version? join_version : "pre-3.1.0"), CRM_FEATURE_SET, ref); ack_nack_bool = FALSE; } else if (max_generation_xml == NULL) { const char *validation = crm_element_value(generation, PCMK_XA_VALIDATE_WITH); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with first CIB " "generation) due to unknown schema version %s " CRM_XS " ref=%s", join_id, join_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with first CIB " "generation) " CRM_XS " ref=%s", join_id, join_from, ref); max_generation_xml = copy_xml(generation); pcmk__str_update(&max_generation_from, join_from); } } else if ((cmp < 0) || ((cmp == 0) && pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei))) { const char *validation = crm_element_value(generation, PCMK_XA_VALIDATE_WITH); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with better CIB " "generation than current best from %s) due to unknown " "schema version %s " CRM_XS " ref=%s", join_id, join_from, max_generation_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with better CIB " "generation than current best from %s) " CRM_XS " ref=%s", join_id, join_from, max_generation_from, ref); crm_log_xml_debug(max_generation_xml, "Old max generation"); crm_log_xml_debug(generation, "New max generation"); free_xml(max_generation_xml); max_generation_xml = copy_xml(join_ack->xml); pcmk__str_update(&max_generation_from, join_from); } } else { crm_debug("Accepting join-%d request from %s " CRM_XS " ref=%s", join_id, join_from, ref); } if (!ack_nack_bool) { if (compare_version(join_version, "3.17.0") < 0) { /* Clients with CRM_FEATURE_SET < 3.17.0 may respawn infinitely * after a nack message, don't send one */ crm_update_peer_join(__func__, join_node, crm_join_nack_quiet); } else { crm_update_peer_join(__func__, join_node, crm_join_nack); } pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_NACK); } else { crm_update_peer_join(__func__, join_node, crm_join_integrated); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); } count = crmd_join_phase_count(crm_join_integrated); crm_debug("%d node%s currently integrated in join-%d", count, pcmk__plural_s(count), join_id); if (check_join_state(cur_state, __func__) == FALSE) { // Don't waste time by invoking the scheduler yet count = crmd_join_phase_count(crm_join_welcomed); crm_debug("Waiting on join-%d requests from %d outstanding node%s", join_id, count, pcmk__plural_s(count)); } } /* A_DC_JOIN_FINALIZE */ void do_dc_join_finalize(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) { char *sync_from = NULL; int rc = pcmk_ok; int count_welcomed = crmd_join_phase_count(crm_join_welcomed); int count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); /* This we can do straight away and avoid clients timing us out * while we compute the latest CIB */ if (count_welcomed != 0) { crm_debug("Waiting on join-%d requests from %d outstanding node%s " "before finalizing join", current_join_id, count_welcomed, pcmk__plural_s(count_welcomed)); crmd_join_phase_log(LOG_DEBUG); /* crmd_fsa_stall(FALSE); Needed? */ return; } else if (count_finalizable == 0) { crm_debug("Finalization not needed for join-%d at the current time", current_join_id); crmd_join_phase_log(LOG_DEBUG); check_join_state(controld_globals.fsa_state, __func__); return; } controld_clear_fsa_input_flags(R_HAVE_CIB); if (pcmk__str_eq(max_generation_from, controld_globals.our_nodename, pcmk__str_null_matches|pcmk__str_casei)) { controld_set_fsa_input_flags(R_HAVE_CIB); } if (!controld_globals.transition_graph->complete) { crm_warn("Delaying join-%d finalization while transition in progress", current_join_id); crmd_join_phase_log(LOG_DEBUG); crmd_fsa_stall(FALSE); return; } if (pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { // Send our CIB out to everyone pcmk__str_update(&sync_from, controld_globals.our_nodename); crm_debug("Finalizing join-%d for %d node%s (sync'ing from local CIB)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable)); crm_log_xml_debug(max_generation_xml, "Requested CIB version"); } else { // Ask for the agreed best CIB pcmk__str_update(&sync_from, max_generation_from); crm_notice("Finalizing join-%d for %d node%s (sync'ing CIB from %s)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable), sync_from); crm_log_xml_notice(max_generation_xml, "Requested CIB version"); } crmd_join_phase_log(LOG_DEBUG); rc = controld_globals.cib_conn->cmds->sync_from(controld_globals.cib_conn, sync_from, NULL, cib_none); fsa_register_cib_callback(rc, sync_from, finalize_sync_callback); } void free_max_generation(void) { free(max_generation_from); max_generation_from = NULL; free_xml(max_generation_xml); max_generation_xml = NULL; } void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { CRM_LOG_ASSERT(-EPERM != rc); if (rc != pcmk_ok) { const char *sync_from = (const char *) user_data; do_crm_log(((rc == -pcmk_err_old_data)? LOG_WARNING : LOG_ERR), "Could not sync CIB from %s in join-%d: %s", sync_from, current_join_id, pcmk_strerror(rc)); if (rc != -pcmk_err_old_data) { record_failed_sync_node(sync_from, current_join_id); } /* restart the whole join process */ register_fsa_error_adv(C_FSA_INTERNAL, I_ELECTION_DC, NULL, NULL, __func__); } else if (!AM_I_DC) { crm_debug("Sync'ed CIB for join-%d but no longer DC", current_join_id); } else if (controld_globals.fsa_state != S_FINALIZE_JOIN) { crm_debug("Sync'ed CIB for join-%d but no longer in S_FINALIZE_JOIN " "(%s)", current_join_id, fsa_state2string(controld_globals.fsa_state)); } else { controld_set_fsa_input_flags(R_HAVE_CIB); /* make sure dc_uuid is re-set to us */ if (!check_join_state(controld_globals.fsa_state, __func__)) { int count_finalizable = 0; count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); crm_debug("Notifying %d node%s of join-%d results", count_finalizable, pcmk__plural_s(count_finalizable), current_join_id); g_hash_table_foreach(crm_peer_cache, finalize_join_for, NULL); } } } static void join_node_state_commit_callback(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { const char *node = user_data; if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; // for register_fsa_error() macro crm_crit("join-%d node history update (via CIB call %d) for node %s " "failed: %s", current_join_id, call_id, node, pcmk_strerror(rc)); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } crm_debug("join-%d node history update (via CIB call %d) for node %s " "complete", current_join_id, call_id, node); check_join_state(controld_globals.fsa_state, __func__); } /* A_DC_JOIN_PROCESS_ACK */ void do_dc_join_ack(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) { int join_id = -1; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *op = crm_element_value(join_ack->msg, F_CRM_TASK); char *join_from = crm_element_value_copy(join_ack->msg, F_CRM_HOST_FROM); crm_node_t *peer = NULL; enum controld_section_e section = controld_section_lrm; char *xpath = NULL; xmlNode *state = join_ack->xml; xmlNode *execd_state = NULL; cib_t *cib = controld_globals.cib_conn; int rc = pcmk_ok; // Sanity checks if (join_from == NULL) { crm_warn("Ignoring message received without node identification"); goto done; } if (op == NULL) { crm_warn("Ignoring message received from %s without task", join_from); goto done; } if (strcmp(op, CRM_OP_JOIN_CONFIRM)) { crm_debug("Ignoring '%s' message from %s while waiting for '%s'", op, join_from, CRM_OP_JOIN_CONFIRM); goto done; } if (crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id) != 0) { crm_warn("Ignoring join confirmation from %s without valid join ID", join_from); goto done; } peer = crm_get_peer(0, join_from); if (peer->join != crm_join_finalized) { crm_info("Ignoring out-of-sequence join-%d confirmation from %s " "(currently %s not %s)", join_id, join_from, crm_join_phase_str(peer->join), crm_join_phase_str(crm_join_finalized)); goto done; } if (join_id != current_join_id) { crm_err("Rejecting join-%d confirmation from %s " "because currently on join-%d", join_id, join_from, current_join_id); crm_update_peer_join(__func__, peer, crm_join_nack); goto done; } crm_update_peer_join(__func__, peer, crm_join_confirmed); /* Update CIB with node's current executor state. A new transition will be * triggered later, when the CIB manager notifies us of the change. * * The delete and modify requests are part of an atomic transaction. */ rc = cib->cmds->init_transaction(cib); if (rc != pcmk_ok) { goto done; } // Delete relevant parts of node's current executor state from CIB if (pcmk_is_set(controld_globals.flags, controld_shutdown_lock_enabled)) { section = controld_section_lrm_unlocked; } controld_node_state_deletion_strings(join_from, section, &xpath, NULL); rc = cib->cmds->remove(cib, xpath, NULL, cib_scope_local |cib_xpath |cib_multiple |cib_transaction); if (rc != pcmk_ok) { goto done; } // Update CIB with node's latest known executor state if (pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei)) { // Use the latest possible state if processing our own join ack execd_state = controld_query_executor_state(); if (execd_state != NULL) { crm_debug("Updating local node history for join-%d from query " "result", current_join_id); state = execd_state; } else { crm_warn("Updating local node history from join-%d confirmation " "because query failed", current_join_id); } } else { crm_debug("Updating node history for %s from join-%d confirmation", join_from, current_join_id); } rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, state, cib_scope_local|cib_can_create|cib_transaction); free_xml(execd_state); if (rc != pcmk_ok) { goto done; } // Commit the transaction rc = cib->cmds->end_transaction(cib, true, cib_scope_local); fsa_register_cib_callback(rc, join_from, join_node_state_commit_callback); if (rc > 0) { // join_from will be freed after callback join_from = NULL; rc = pcmk_ok; } done: if (rc != pcmk_ok) { crm_crit("join-%d node history update for node %s failed: %s", current_join_id, join_from, pcmk_strerror(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } free(join_from); free(xpath); } void finalize_join_for(gpointer key, gpointer value, gpointer user_data) { xmlNode *acknak = NULL; xmlNode *tmp1 = NULL; crm_node_t *join_node = value; const char *join_to = join_node->uname; bool integrated = false; switch (join_node->join) { case crm_join_integrated: integrated = true; break; case crm_join_nack: case crm_join_nack_quiet: break; default: crm_trace("Not updating non-integrated and non-nacked node %s (%s) " "for join-%d", join_to, crm_join_phase_str(join_node->join), current_join_id); return; } /* Update the element with the node's name and UUID, in case they * weren't known before */ crm_trace("Updating node name and UUID in CIB for %s", join_to); tmp1 = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(tmp1, PCMK_XA_ID, crm_peer_uuid(join_node)); crm_xml_add(tmp1, PCMK_XA_UNAME, join_to); fsa_cib_anon_update(XML_CIB_TAG_NODES, tmp1); free_xml(tmp1); if (join_node->join == crm_join_nack_quiet) { crm_trace("Not sending nack message to node %s with feature set older " "than 3.17.0", join_to); return; } join_node = crm_get_peer(0, join_to); if (!crm_is_peer_active(join_node)) { /* * NACK'ing nodes that the membership layer doesn't know about yet * simply creates more churn * * Better to leave them waiting and let the join restart when * the new membership event comes in * * All other NACKs (due to versions etc) should still be processed */ pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_PENDING); return; } // Acknowledge or nack node's join request crm_debug("%sing join-%d request from %s", integrated? "Acknowledg" : "Nack", current_join_id, join_to); acknak = create_dc_message(CRM_OP_JOIN_ACKNAK, join_to); pcmk__xe_set_bool_attr(acknak, CRM_OP_JOIN_ACKNAK, integrated); if (integrated) { // No change needed for a nacked node crm_update_peer_join(__func__, join_node, crm_join_finalized); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); /* Iterate through the remote peer cache and add information on which * node hosts each to the ACK message. This keeps new controllers in * sync with what has already happened. */ if (crm_remote_peer_cache_size() != 0) { GHashTableIter iter; crm_node_t *node = NULL; xmlNode *remotes = create_xml_node(acknak, XML_CIB_TAG_NODES); g_hash_table_iter_init(&iter, crm_remote_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *remote = NULL; if (!node->conn_host) { continue; } remote = create_xml_node(remotes, XML_CIB_TAG_NODE); pcmk__xe_set_props(remote, PCMK_XA_ID, node->uname, XML_CIB_TAG_STATE, node->state, PCMK__XA_CONN_HOST, node->conn_host, NULL); } } } send_cluster_message(join_node, crm_msg_crmd, acknak, TRUE); free_xml(acknak); return; } gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source) { static unsigned long long highest_seq = 0; if (controld_globals.membership_id != crm_peer_seq) { crm_debug("join-%d: Membership changed from %llu to %llu " CRM_XS " highest=%llu state=%s for=%s", current_join_id, controld_globals.membership_id, crm_peer_seq, highest_seq, fsa_state2string(cur_state), source); if(highest_seq < crm_peer_seq) { /* Don't spam the FSA with duplicates */ highest_seq = crm_peer_seq; register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } } else if (cur_state == S_INTEGRATION) { if (crmd_join_phase_count(crm_join_welcomed) == 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Integration of %d peer%s complete " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); register_fsa_input_before(C_FSA_INTERNAL, I_INTEGRATED, NULL); return TRUE; } } else if (cur_state == S_FINALIZE_JOIN) { if (!pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { crm_debug("join-%d: Delaying finalization until we have CIB " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); return TRUE; } else if (crmd_join_phase_count(crm_join_welcomed) != 0) { int count = crmd_join_phase_count(crm_join_welcomed); crm_debug("join-%d: Still waiting on %d welcomed node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_integrated) != 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Still waiting on %d integrated node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_finalized) != 0) { int count = crmd_join_phase_count(crm_join_finalized); crm_debug("join-%d: Still waiting on %d finalized node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else { crm_debug("join-%d: Complete " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); register_fsa_input_later(C_FSA_INTERNAL, I_FINALIZED, NULL); return TRUE; } } return FALSE; } void do_dc_join_final(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) { crm_debug("Ensuring DC, quorum and node attributes are up-to-date"); crm_update_quorum(crm_have_quorum, TRUE); } int crmd_join_phase_count(enum crm_join_phase phase) { int count = 0; crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { if(peer->join == phase) { count++; } } return count; } void crmd_join_phase_log(int level) { crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { do_crm_log(level, "join-%d: %s=%s", current_join_id, peer->uname, crm_join_phase_str(peer->join)); } } diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c index 1928aa125f..f55afe318b 100644 --- a/daemons/controld/controld_messages.c +++ b/daemons/controld/controld_messages.c @@ -1,1337 +1,1337 @@ /* * Copyright 2004-2024 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 extern void crm_shutdown(int nsig); static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause); static void handle_response(xmlNode *stored_msg); static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause); static enum crmd_fsa_input handle_shutdown_request(xmlNode *stored_msg); static void send_msg_via_ipc(xmlNode * msg, const char *sys); /* debug only, can wrap all it likes */ static int last_data_id = 0; void register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, fsa_data_t * cur_data, void *new_data, const char *raised_from) { /* save the current actions if any */ if (controld_globals.fsa_actions != A_NOTHING) { register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL, I_NULL, cur_data ? cur_data->data : NULL, controld_globals.fsa_actions, TRUE, __func__); } /* reset the action list */ crm_info("Resetting the current action list"); fsa_dump_actions(controld_globals.fsa_actions, "Drop"); controld_globals.fsa_actions = A_NOTHING; /* register the error */ register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from); } void register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, void *data, uint64_t with_actions, gboolean prepend, const char *raised_from) { unsigned old_len = g_list_length(controld_globals.fsa_message_queue); fsa_data_t *fsa_data = NULL; if (raised_from == NULL) { raised_from = ""; } if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) { /* no point doing anything */ crm_err("Cannot add entry to queue: no input and no action"); return; } if (input == I_WAIT_FOR_EVENT) { controld_set_global_flags(controld_fsa_is_stalled); crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d", raised_from, fsa_cause2string(cause), data, old_len); if (old_len > 0) { fsa_dump_queue(LOG_TRACE); prepend = FALSE; } if (data == NULL) { controld_set_fsa_action_flags(with_actions); fsa_dump_actions(with_actions, "Restored"); return; } /* Store everything in the new event and reset * controld_globals.fsa_actions */ with_actions |= controld_globals.fsa_actions; controld_globals.fsa_actions = A_NOTHING; } last_data_id++; crm_trace("%s %s FSA input %d (%s) due to %s, %s data", raised_from, (prepend? "prepended" : "appended"), last_data_id, fsa_input2string(input), fsa_cause2string(cause), (data? "with" : "without")); fsa_data = calloc(1, sizeof(fsa_data_t)); fsa_data->id = last_data_id; fsa_data->fsa_input = input; fsa_data->fsa_cause = cause; fsa_data->origin = raised_from; fsa_data->data = NULL; fsa_data->data_type = fsa_dt_none; fsa_data->actions = with_actions; if (with_actions != A_NOTHING) { crm_trace("Adding actions %.16llx to input", (unsigned long long) with_actions); } if (data != NULL) { switch (cause) { case C_FSA_INTERNAL: case C_CRMD_STATUS_CALLBACK: case C_IPC_MESSAGE: case C_HA_MESSAGE: CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL, crm_err("Bogus data from %s", raised_from)); crm_trace("Copying %s data from %s as cluster message data", fsa_cause2string(cause), raised_from); fsa_data->data = copy_ha_msg_input(data); fsa_data->data_type = fsa_dt_ha_msg; break; case C_LRM_OP_CALLBACK: crm_trace("Copying %s data from %s as lrmd_event_data_t", fsa_cause2string(cause), raised_from); fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data); fsa_data->data_type = fsa_dt_lrm; break; case C_TIMER_POPPED: case C_SHUTDOWN: case C_UNKNOWN: case C_STARTUP: crm_crit("Copying %s data (from %s) is not yet implemented", fsa_cause2string(cause), raised_from); crmd_exit(CRM_EX_SOFTWARE); break; } } /* make sure to free it properly later */ if (prepend) { controld_globals.fsa_message_queue = g_list_prepend(controld_globals.fsa_message_queue, fsa_data); } else { controld_globals.fsa_message_queue = g_list_append(controld_globals.fsa_message_queue, fsa_data); } crm_trace("FSA message queue length is %d", g_list_length(controld_globals.fsa_message_queue)); /* fsa_dump_queue(LOG_TRACE); */ if (old_len == g_list_length(controld_globals.fsa_message_queue)) { crm_err("Couldn't add message to the queue"); } if (input != I_WAIT_FOR_EVENT) { controld_trigger_fsa(); } } void fsa_dump_queue(int log_level) { int offset = 0; for (GList *iter = controld_globals.fsa_message_queue; iter != NULL; iter = iter->next) { fsa_data_t *data = (fsa_data_t *) iter->data; do_crm_log_unlikely(log_level, "queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)", offset++, data->id, fsa_input2string(data->fsa_input), data->origin, data->data, data->data_type, fsa_cause2string(data->fsa_cause)); } } ha_msg_input_t * copy_ha_msg_input(ha_msg_input_t * orig) { ha_msg_input_t *copy = calloc(1, sizeof(ha_msg_input_t)); CRM_ASSERT(copy != NULL); copy->msg = (orig && orig->msg)? copy_xml(orig->msg) : NULL; copy->xml = get_message_xml(copy->msg, F_CRM_DATA); return copy; } void delete_fsa_input(fsa_data_t * fsa_data) { lrmd_event_data_t *op = NULL; xmlNode *foo = NULL; if (fsa_data == NULL) { return; } crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause)); if (fsa_data->data != NULL) { switch (fsa_data->data_type) { case fsa_dt_ha_msg: delete_ha_msg_input(fsa_data->data); break; case fsa_dt_xml: foo = fsa_data->data; free_xml(foo); break; case fsa_dt_lrm: op = (lrmd_event_data_t *) fsa_data->data; lrmd_free_event(op); break; case fsa_dt_none: if (fsa_data->data != NULL) { crm_err("Don't know how to free %s data from %s", fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin); crmd_exit(CRM_EX_SOFTWARE); } break; } crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause)); } free(fsa_data); } /* returns the next message */ fsa_data_t * get_message(void) { fsa_data_t *message = (fsa_data_t *) controld_globals.fsa_message_queue->data; controld_globals.fsa_message_queue = g_list_remove(controld_globals.fsa_message_queue, message); crm_trace("Processing input %d", message->id); return message; } void * fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller) { void *ret_val = NULL; if (fsa_data == NULL) { crm_err("%s: No FSA data available", caller); } else if (fsa_data->data == NULL) { crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin); } else if (fsa_data->data_type != a_type) { crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s", caller, fsa_data->data_type, a_type, fsa_data->origin); CRM_ASSERT(fsa_data->data_type == a_type); } else { ret_val = fsa_data->data; } return ret_val; } /* A_MSG_ROUTE */ void do_msg_route(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) { ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); route_message(msg_data->fsa_cause, input->msg); } void route_message(enum crmd_fsa_cause cause, xmlNode * input) { ha_msg_input_t fsa_input; enum crmd_fsa_input result = I_NULL; fsa_input.msg = input; CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return); /* try passing the buck first */ if (relay_message(input, cause == C_IPC_MESSAGE)) { return; } /* handle locally */ result = handle_message(input, cause); /* done or process later? */ switch (result) { case I_NULL: case I_CIB_OP: case I_ROUTER: case I_NODE_JOIN: case I_JOIN_REQUEST: case I_JOIN_RESULT: break; default: /* Defering local processing of message */ register_fsa_input_later(cause, result, &fsa_input); return; } if (result != I_NULL) { /* add to the front of the queue */ register_fsa_input(cause, result, &fsa_input); } } gboolean relay_message(xmlNode * msg, gboolean originated_locally) { enum crm_ais_msg_types dest = crm_msg_ais; bool is_for_dc = false; bool is_for_dcib = false; bool is_for_te = false; bool is_for_crm = false; bool is_for_cib = false; bool is_local = false; bool broadcast = false; const char *host_to = NULL; const char *sys_to = NULL; const char *sys_from = NULL; const char *type = NULL; const char *task = NULL; const char *ref = NULL; crm_node_t *node_to = NULL; CRM_CHECK(msg != NULL, return TRUE); host_to = crm_element_value(msg, F_CRM_HOST_TO); sys_to = crm_element_value(msg, F_CRM_SYS_TO); sys_from = crm_element_value(msg, F_CRM_SYS_FROM); type = crm_element_value(msg, F_TYPE); task = crm_element_value(msg, F_CRM_TASK); - ref = crm_element_value(msg, XML_ATTR_REFERENCE); + ref = crm_element_value(msg, PCMK_XA_REFERENCE); broadcast = pcmk__str_empty(host_to); if (ref == NULL) { ref = "without reference ID"; } if (pcmk__str_eq(task, CRM_OP_HELLO, pcmk__str_casei)) { crm_trace("Received hello %s from %s (no processing needed)", ref, pcmk__s(sys_from, "unidentified source")); crm_log_xml_trace(msg, "hello"); return TRUE; } // Require message type (set by create_request()) if (!pcmk__str_eq(type, T_CRM, pcmk__str_casei)) { crm_warn("Ignoring invalid message %s with type '%s' (not '" T_CRM "')", ref, pcmk__s(type, "")); crm_log_xml_trace(msg, "ignored"); return TRUE; } // Require a destination subsystem (also set by create_request()) if (sys_to == NULL) { crm_warn("Ignoring invalid message %s with no " F_CRM_SYS_TO, ref); crm_log_xml_trace(msg, "ignored"); return TRUE; } // Get the message type appropriate to the destination subsystem if (is_corosync_cluster()) { dest = text2msg_type(sys_to); if ((dest < crm_msg_ais) || (dest > crm_msg_stonith_ng)) { /* Unrecognized value, use a sane default * * @TODO Maybe we should bail instead */ dest = crm_msg_crmd; } } is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0); is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0); is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0); is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0); is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0); // Check whether message should be processed locally is_local = false; if (broadcast) { if (is_for_dc || is_for_te) { is_local = false; } else if (is_for_crm) { if (pcmk__strcase_any_of(task, CRM_OP_NODE_INFO, PCMK__CONTROLD_CMD_NODES, NULL)) { /* Node info requests do not specify a host, which is normally * treated as "all hosts", because the whole point is that the * client may not know the local node name. Always handle these * requests locally. */ is_local = true; } else { is_local = !originated_locally; } } else { is_local = true; } } else if (pcmk__str_eq(controld_globals.our_nodename, host_to, pcmk__str_casei)) { is_local = true; } else if (is_for_crm && pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { xmlNode *msg_data = get_message_xml(msg, F_CRM_DATA); const char *mode = crm_element_value(msg_data, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_casei)) { // Local delete of an offline node's resource history is_local = true; } } // Check whether message should be relayed if (is_for_dc || is_for_dcib || is_for_te) { if (AM_I_DC) { if (is_for_te) { crm_trace("Route message %s locally as transition request", ref); crm_log_xml_trace(msg, sys_to); send_msg_via_ipc(msg, sys_to); return TRUE; // No further processing of message is needed } crm_trace("Route message %s locally as DC request", ref); return FALSE; // More to be done by caller } if (originated_locally && !pcmk__strcase_any_of(sys_from, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { crm_trace("Relay message %s to DC (via %s)", ref, pcmk__s(host_to, "broadcast")); crm_log_xml_trace(msg, "relayed"); if (!broadcast) { node_to = crm_get_peer(0, host_to); } send_cluster_message(node_to, dest, msg, TRUE); return TRUE; } /* Transition engine and scheduler messages are sent only to the DC on * the same node. If we are no longer the DC, discard this message. */ crm_trace("Ignoring message %s because we are no longer DC", ref); crm_log_xml_trace(msg, "ignored"); return TRUE; // No further processing of message is needed } if (is_local) { if (is_for_crm || is_for_cib) { crm_trace("Route message %s locally as controller request", ref); return FALSE; // More to be done by caller } crm_trace("Relay message %s locally to %s", ref, sys_to); crm_log_xml_trace(msg, "IPC-relay"); send_msg_via_ipc(msg, sys_to); return TRUE; } if (!broadcast) { node_to = pcmk__search_cluster_node_cache(0, host_to, NULL); if (node_to == NULL) { crm_warn("Ignoring message %s because node %s is unknown", ref, host_to); crm_log_xml_trace(msg, "ignored"); return TRUE; } } crm_trace("Relay message %s to %s", ref, pcmk__s(host_to, "all peers")); crm_log_xml_trace(msg, "relayed"); send_cluster_message(node_to, dest, msg, TRUE); return TRUE; } // Return true if field contains a positive integer static bool authorize_version(xmlNode *message_data, const char *field, const char *client_name, const char *ref, const char *uuid) { const char *version = crm_element_value(message_data, field); long long version_num; if ((pcmk__scan_ll(version, &version_num, -1LL) != pcmk_rc_ok) || (version_num < 0LL)) { crm_warn("Rejected IPC hello from %s: '%s' is not a valid protocol %s " CRM_XS " ref=%s uuid=%s", client_name, ((version == NULL)? "" : version), field, (ref? ref : "none"), uuid); return false; } return true; } /*! * \internal * \brief Check whether a client IPC message is acceptable * * If a given client IPC message is a hello, "authorize" it by ensuring it has * valid information such as a protocol version, and return false indicating * that nothing further needs to be done with the message. If the message is not * a hello, just return true to indicate it needs further processing. * * \param[in] client_msg XML of IPC message * \param[in,out] curr_client If IPC is not proxied, client that sent message * \param[in] proxy_session If IPC is proxied, the session ID * * \return true if message needs further processing, false if it doesn't */ bool controld_authorize_ipc_message(const xmlNode *client_msg, pcmk__client_t *curr_client, const char *proxy_session) { xmlNode *message_data = NULL; const char *client_name = NULL; const char *op = crm_element_value(client_msg, F_CRM_TASK); - const char *ref = crm_element_value(client_msg, XML_ATTR_REFERENCE); + const char *ref = crm_element_value(client_msg, PCMK_XA_REFERENCE); const char *uuid = (curr_client? curr_client->id : proxy_session); if (uuid == NULL) { crm_warn("IPC message from client rejected: No client identifier " CRM_XS " ref=%s", (ref? ref : "none")); goto rejected; } if (!pcmk__str_eq(CRM_OP_HELLO, op, pcmk__str_casei)) { // Only hello messages need to be authorized return true; } message_data = get_message_xml(client_msg, F_CRM_DATA); client_name = crm_element_value(message_data, "client_name"); if (pcmk__str_empty(client_name)) { crm_warn("IPC hello from client rejected: No client name", CRM_XS " ref=%s uuid=%s", (ref? ref : "none"), uuid); goto rejected; } if (!authorize_version(message_data, "major_version", client_name, ref, uuid)) { goto rejected; } if (!authorize_version(message_data, "minor_version", client_name, ref, uuid)) { goto rejected; } crm_trace("Validated IPC hello from client %s", client_name); crm_log_xml_trace(client_msg, "hello"); if (curr_client) { curr_client->userdata = strdup(client_name); } controld_trigger_fsa(); return false; rejected: crm_log_xml_trace(client_msg, "rejected"); if (curr_client) { qb_ipcs_disconnect(curr_client->ipcs); } return false; } static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause) { const char *type = NULL; CRM_CHECK(msg != NULL, return I_NULL); type = crm_element_value(msg, F_CRM_MSG_TYPE); if (pcmk__str_eq(type, PCMK__VALUE_REQUEST, pcmk__str_none)) { return handle_request(msg, cause); } if (pcmk__str_eq(type, PCMK__VALUE_RESPONSE, pcmk__str_none)) { handle_response(msg); return I_NULL; } crm_warn("Ignoring message with unknown " F_CRM_MSG_TYPE " '%s'", pcmk__s(type, "")); crm_log_xml_trace(msg, "bad"); return I_NULL; } static enum crmd_fsa_input handle_failcount_op(xmlNode * stored_msg) { const char *rsc = NULL; const char *uname = NULL; const char *op = NULL; char *interval_spec = NULL; guint interval_ms = 0; gboolean is_remote_node = FALSE; xmlNode *xml_op = get_message_xml(stored_msg, F_CRM_DATA); if (xml_op) { xmlNode *xml_rsc = first_named_child(xml_op, XML_CIB_TAG_RESOURCE); xmlNode *xml_attrs = first_named_child(xml_op, XML_TAG_ATTRS); if (xml_rsc) { rsc = ID(xml_rsc); } if (xml_attrs) { op = crm_element_value(xml_attrs, CRM_META "_" PCMK__META_CLEAR_FAILURE_OP); crm_element_value_ms(xml_attrs, CRM_META "_" PCMK__META_CLEAR_FAILURE_INTERVAL, &interval_ms); } } uname = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if ((rsc == NULL) || (uname == NULL)) { crm_log_xml_warn(stored_msg, "invalid failcount op"); return I_NULL; } if (crm_element_value(xml_op, XML_LRM_ATTR_ROUTER_NODE)) { is_remote_node = TRUE; } crm_debug("Clearing failures for %s-interval %s on %s " "from attribute manager, CIB, and executor state", pcmk__readable_interval(interval_ms), rsc, uname); if (interval_ms) { interval_spec = crm_strdup_printf("%ums", interval_ms); } update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node); free(interval_spec); controld_cib_delete_last_failure(rsc, uname, op, interval_ms); lrm_clear_last_failure(rsc, uname, op, interval_ms); return I_NULL; } static enum crmd_fsa_input handle_lrm_delete(xmlNode *stored_msg) { const char *mode = NULL; xmlNode *msg_data = get_message_xml(stored_msg, F_CRM_DATA); CRM_CHECK(msg_data != NULL, return I_NULL); /* CRM_OP_LRM_DELETE has two distinct modes. The default behavior is to * relay the operation to the affected node, which will unregister the * resource from the local executor, clear the resource's history from the * CIB, and do some bookkeeping in the controller. * * However, if the affected node is offline, the client will specify * mode="cib" which means the controller receiving the operation should * clear the resource's history from the CIB and nothing else. This is used * to clear shutdown locks. */ mode = crm_element_value(msg_data, PCMK__XA_MODE); if ((mode == NULL) || strcmp(mode, XML_TAG_CIB)) { // Relay to affected node crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else { // Delete CIB history locally (compare with do_lrm_delete()) const char *from_sys = NULL; const char *user_name = NULL; const char *rsc_id = NULL; const char *node = NULL; xmlNode *rsc_xml = NULL; int rc = pcmk_rc_ok; rsc_xml = first_named_child(msg_data, XML_CIB_TAG_RESOURCE); CRM_CHECK(rsc_xml != NULL, return I_NULL); rsc_id = ID(rsc_xml); from_sys = crm_element_value(stored_msg, F_CRM_SYS_FROM); node = crm_element_value(msg_data, XML_LRM_ATTR_TARGET); user_name = pcmk__update_acl_user(stored_msg, F_CRM_USER, NULL); crm_debug("Handling " CRM_OP_LRM_DELETE " for %s on %s locally%s%s " "(clearing CIB resource history only)", rsc_id, node, (user_name? " for user " : ""), (user_name? user_name : "")); rc = controld_delete_resource_history(rsc_id, node, user_name, cib_dryrun|cib_sync_call); if (rc == pcmk_rc_ok) { rc = controld_delete_resource_history(rsc_id, node, user_name, crmd_cib_smart_opt()); } //Notify client and tengine.(Only notify tengine if mode = "cib" and CRM_OP_LRM_DELETE.) if (from_sys) { lrmd_event_data_t *op = NULL; const char *from_host = crm_element_value(stored_msg, F_CRM_HOST_FROM); const char *transition; if (strcmp(from_sys, CRM_SYSTEM_TENGINE)) { transition = crm_element_value(msg_data, XML_ATTR_TRANSITION_KEY); } else { transition = crm_element_value(stored_msg, XML_ATTR_TRANSITION_KEY); } crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "local node"), rsc_id, ((rc == pcmk_rc_ok)? "" : " not")); op = lrmd_new_event(rsc_id, PCMK_ACTION_DELETE, 0); op->type = lrmd_event_exec_complete; op->user_data = strdup(transition? transition : FAKE_TE_ID); op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(PCMK_XA_CRM_FEATURE_SET), strdup(CRM_FEATURE_SET)); controld_rc2event(op, rc); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); controld_trigger_delete_refresh(from_sys, rsc_id); } return I_NULL; } } /*! * \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_remote_state(const xmlNode *msg) { const char *conn_host = NULL; const char *remote_uname = ID(msg); crm_node_t *remote_peer; bool remote_is_up = false; int rc = pcmk_rc_ok; rc = pcmk__xe_get_bool_attr(msg, PCMK__XA_IN_CCM, &remote_is_up); CRM_CHECK(remote_uname && rc == pcmk_rc_ok, return I_NULL); remote_peer = crm_remote_peer_get(remote_uname); CRM_CHECK(remote_peer, return I_NULL); pcmk__update_peer_state(__func__, remote_peer, remote_is_up ? CRM_NODE_MEMBER : CRM_NODE_LOST, 0); conn_host = crm_element_value(msg, PCMK__XA_CONN_HOST); if (conn_host) { pcmk__str_update(&remote_peer->conn_host, conn_host); } else if (remote_peer->conn_host) { free(remote_peer->conn_host); remote_peer->conn_host = NULL; } return I_NULL; } /*! * \brief Handle a CRM_OP_PING message * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_ping(const xmlNode *msg) { const char *value = NULL; xmlNode *ping = NULL; xmlNode *reply = NULL; // Build reply ping = create_xml_node(NULL, XML_CRM_TAG_PING); value = crm_element_value(msg, F_CRM_SYS_TO); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value); // Add controller state value = fsa_state2string(controld_globals.fsa_state); crm_xml_add(ping, XML_PING_ATTR_CRMDSTATE, value); crm_notice("Current ping state: %s", value); // CTS needs this // Add controller health // @TODO maybe do some checks to determine meaningful status crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); // Send reply reply = create_reply(msg, ping); free_xml(ping); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a PCMK__CONTROLD_CMD_NODES message * * \param[in] request Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_list(const xmlNode *request) { GHashTableIter iter; crm_node_t *node = NULL; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Create message data for reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODES); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { xmlNode *xml = create_xml_node(reply_data, XML_CIB_TAG_NODE); crm_xml_add_ll(xml, PCMK_XA_ID, (long long) node->id); // uint32_t crm_xml_add(xml, PCMK_XA_UNAME, node->uname); crm_xml_add(xml, PCMK__XA_IN_CCM, node->state); } // Create and send reply reply = create_reply(request, reply_data); free_xml(reply_data); if (reply) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a CRM_OP_NODE_INFO request * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_info_request(const xmlNode *msg) { const char *value = NULL; crm_node_t *node = NULL; int node_id = 0; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Build reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(reply_data, XML_PING_ATTR_SYSFROM, CRM_SYSTEM_CRMD); // Add whether current partition has quorum pcmk__xe_set_bool_attr(reply_data, PCMK_XA_HAVE_QUORUM, pcmk_is_set(controld_globals.flags, controld_has_quorum)); // Check whether client requested node info by ID and/or name crm_element_value_int(msg, PCMK_XA_ID, &node_id); if (node_id < 0) { node_id = 0; } value = crm_element_value(msg, PCMK_XA_UNAME); // Default to local node if none given if ((node_id == 0) && (value == NULL)) { value = controld_globals.our_nodename; } node = pcmk__search_node_caches(node_id, value, CRM_GET_PEER_ANY); if (node) { crm_xml_add(reply_data, PCMK_XA_ID, node->uuid); crm_xml_add(reply_data, PCMK_XA_UNAME, node->uname); crm_xml_add(reply_data, PCMK__XA_CRMD, node->state); pcmk__xe_set_bool_attr(reply_data, XML_NODE_IS_REMOTE, pcmk_is_set(node->flags, crm_remote_node)); } // Send reply reply = create_reply(msg, reply_data); free_xml(reply_data); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } static void verify_feature_set(xmlNode *msg) { const char *dc_version = crm_element_value(msg, PCMK_XA_CRM_FEATURE_SET); if (dc_version == NULL) { /* All we really know is that the DC feature set is older than 3.1.0, * but that's also all that really matters. */ dc_version = "3.0.14"; } if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) { crm_trace("Local feature set (%s) is compatible with DC's (%s)", CRM_FEATURE_SET, dc_version); } else { crm_err("Local feature set (%s) is incompatible with DC's (%s)", CRM_FEATURE_SET, dc_version); // Nothing is likely to improve without administrator involvement controld_set_fsa_input_flags(R_STAYDOWN); crmd_exit(CRM_EX_FATAL); } } // DC gets own shutdown all-clear static enum crmd_fsa_input handle_shutdown_self_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { // The expected case -- we initiated own shutdown sequence crm_info("Shutting down controller"); return I_STOP; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_casei)) { // Must be logic error -- DC confirming its own unrequested shutdown crm_err("Shutting down controller immediately due to " "unexpected shutdown confirmation"); return I_TERMINATE; } if (controld_globals.fsa_state != S_STOPPING) { // Shouldn't happen -- non-DC confirming unrequested shutdown crm_err("Starting new DC election because %s is " "confirming shutdown we did not request", (host_from? host_from : "another node")); return I_ELECTION; } // Shouldn't happen, but we are already stopping anyway crm_debug("Ignoring unexpected shutdown confirmation from %s", (host_from? host_from : "another node")); return I_NULL; } // Non-DC gets shutdown all-clear from DC static enum crmd_fsa_input handle_shutdown_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { crm_warn("Ignoring shutdown request without origin specified"); return I_NULL; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_null_matches|pcmk__str_casei)) { if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { crm_info("Shutting down controller after confirmation from %s", host_from); } else { crm_err("Shutting down controller after unexpected " "shutdown request from %s", host_from); controld_set_fsa_input_flags(R_STAYDOWN); } return I_STOP; } crm_warn("Ignoring shutdown request from %s because DC is %s", host_from, controld_globals.dc_name); return I_NULL; } static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause) { xmlNode *msg = NULL; const char *op = crm_element_value(stored_msg, F_CRM_TASK); /* Optimize this for the DC - it has the most to do */ crm_log_xml_trace(stored_msg, "request"); if (op == NULL) { crm_warn("Ignoring request without " F_CRM_TASK); return I_NULL; } if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { const char *from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_node_t *node = pcmk__search_cluster_node_cache(0, from, NULL); pcmk__update_peer_expected(__func__, node, CRMD_JOINSTATE_DOWN); if(AM_I_DC == FALSE) { return I_NULL; /* Done */ } } /*========== DC-Only Actions ==========*/ if (AM_I_DC) { if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) { return I_NODE_JOIN; } else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) { return I_JOIN_REQUEST; } else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) { return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_self_ack(stored_msg); } else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { // Another controller wants to shut down its node return handle_shutdown_request(stored_msg); } } /*========== common actions ==========*/ if (strcmp(op, CRM_OP_NOVOTE) == 0) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); } else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) { /* a remote connection host is letting us know the node state */ return handle_remote_state(stored_msg); } else if (strcmp(op, CRM_OP_THROTTLE) == 0) { throttle_update(stored_msg); if (AM_I_DC && (controld_globals.transition_graph != NULL) && !controld_globals.transition_graph->complete) { crm_debug("The throttle changed. Trigger a graph."); trigger_graph(); } return I_NULL; } else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) { return handle_failcount_op(stored_msg); } else if (strcmp(op, CRM_OP_VOTE) == 0) { /* count the vote and decide what to do after that */ ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); /* Sometimes we _must_ go into S_ELECTION */ if (controld_globals.fsa_state == S_HALT) { crm_debug("Forcing an election from S_HALT"); return I_ELECTION; } } else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) { verify_feature_set(stored_msg); crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_OFFER; } else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) { crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_LRM_DELETE) == 0) { return handle_lrm_delete(stored_msg); } else if ((strcmp(op, CRM_OP_LRM_FAIL) == 0) || (strcmp(op, CRM_OP_LRM_REFRESH) == 0) // @COMPAT || (strcmp(op, CRM_OP_REPROBE) == 0)) { crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else if (strcmp(op, CRM_OP_NOOP) == 0) { return I_NULL; } else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) { crm_shutdown(SIGTERM); /*return I_SHUTDOWN; */ return I_NULL; } else if (strcmp(op, CRM_OP_PING) == 0) { return handle_ping(stored_msg); } else if (strcmp(op, CRM_OP_NODE_INFO) == 0) { return handle_node_info_request(stored_msg); } else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) { int id = 0; const char *name = NULL; crm_element_value_int(stored_msg, PCMK_XA_ID, &id); name = crm_element_value(stored_msg, PCMK_XA_UNAME); if(cause == C_IPC_MESSAGE) { msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) { crm_err("Could not instruct peers to remove references to node %s/%u", name, id); } else { crm_notice("Instructing peers to remove references to node %s/%u", name, id); } free_xml(msg); } else { reap_crm_member(id, name); /* If we're forgetting this node, also forget any failures to fence * it, so we don't carry that over to any node added later with the * same name. */ st_fail_count_reset(name); } } else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) { xmlNode *xml = get_message_xml(stored_msg, F_CRM_DATA); remote_ra_process_maintenance_nodes(xml); } else if (strcmp(op, PCMK__CONTROLD_CMD_NODES) == 0) { return handle_node_list(stored_msg); /*========== (NOT_DC)-Only Actions ==========*/ } else if (!AM_I_DC) { if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_ack(stored_msg); } } else { crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node"); crm_log_xml_err(stored_msg, "Unexpected"); } return I_NULL; } static void handle_response(xmlNode *stored_msg) { const char *op = crm_element_value(stored_msg, F_CRM_TASK); crm_log_xml_trace(stored_msg, "reply"); if (op == NULL) { crm_warn("Ignoring reply without " F_CRM_TASK); } else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) { // Check whether scheduler answer been superseded by subsequent request - const char *msg_ref = crm_element_value(stored_msg, XML_ATTR_REFERENCE); + const char *msg_ref = crm_element_value(stored_msg, PCMK_XA_REFERENCE); if (msg_ref == NULL) { crm_err("%s - Ignoring calculation with no reference", op); } else if (pcmk__str_eq(msg_ref, controld_globals.fsa_pe_ref, pcmk__str_none)) { ha_msg_input_t fsa_input; controld_stop_sched_timer(); fsa_input.msg = stored_msg; register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input); } else { crm_info("%s calculation %s is obsolete", op, msg_ref); } } else if (strcmp(op, CRM_OP_VOTE) == 0 || strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) { } else { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_err("Unexpected response (op=%s, src=%s) sent to the %s", op, host_from, AM_I_DC ? "DC" : "controller"); } } static enum crmd_fsa_input handle_shutdown_request(xmlNode * stored_msg) { /* handle here to avoid potential version issues * where the shutdown message/procedure may have * been changed in later versions. * * This way the DC is always in control of the shutdown */ char *now_s = NULL; const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { /* we're shutting down and the DC */ host_from = controld_globals.our_nodename; } crm_info("Creating shutdown request for %s (state=%s)", host_from, fsa_state2string(controld_globals.fsa_state)); crm_log_xml_trace(stored_msg, "message"); now_s = pcmk__ttoa(time(NULL)); update_attrd(host_from, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, FALSE); free(now_s); /* will be picked up by the TE as long as its running */ return I_NULL; } static void send_msg_via_ipc(xmlNode * msg, const char *sys) { pcmk__client_t *client_channel = NULL; CRM_CHECK(sys != NULL, return); client_channel = pcmk__find_client_by_id(sys); if (crm_element_value(msg, F_CRM_HOST_FROM) == NULL) { crm_xml_add(msg, F_CRM_HOST_FROM, controld_globals.our_nodename); } if (client_channel != NULL) { /* Transient clients such as crmadmin */ pcmk__ipc_send_xml(client_channel, 0, msg, crm_ipc_server_event); } else if (pcmk__str_eq(sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { xmlNode *data = get_message_xml(msg, F_CRM_DATA); process_te_message(msg, data); } else if (pcmk__str_eq(sys, CRM_SYSTEM_LRMD, pcmk__str_none)) { fsa_data_t fsa_data; ha_msg_input_t fsa_input; fsa_input.msg = msg; fsa_input.xml = get_message_xml(msg, F_CRM_DATA); fsa_data.id = 0; fsa_data.actions = 0; fsa_data.data = &fsa_input; fsa_data.fsa_input = I_MESSAGE; fsa_data.fsa_cause = C_IPC_MESSAGE; fsa_data.origin = __func__; fsa_data.data_type = fsa_dt_ha_msg; do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, controld_globals.fsa_state, I_MESSAGE, &fsa_data); } else if (crmd_is_proxy_session(sys)) { crmd_proxy_send(sys, msg); } else { crm_info("Received invalid request: unknown subsystem '%s'", sys); } } void delete_ha_msg_input(ha_msg_input_t * orig) { if (orig == NULL) { return; } free_xml(orig->msg); free(orig); } /*! * \internal * \brief Notify the cluster of a remote node state change * * \param[in] node_name Node's name * \param[in] node_up true if node is up, false if down */ void broadcast_remote_state_message(const char *node_name, bool node_up) { xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); crm_info("Notifying cluster of Pacemaker Remote node %s %s", node_name, node_up? "coming up" : "going down"); crm_xml_add(msg, PCMK_XA_ID, node_name); pcmk__xe_set_bool_attr(msg, PCMK__XA_IN_CCM, node_up); if (node_up) { crm_xml_add(msg, PCMK__XA_CONN_HOST, controld_globals.our_nodename); } send_cluster_message(NULL, crm_msg_crmd, msg, TRUE); free_xml(msg); } diff --git a/daemons/controld/controld_schedulerd.c b/daemons/controld/controld_schedulerd.c index 4501b7ef46..d3abde1123 100644 --- a/daemons/controld/controld_schedulerd.c +++ b/daemons/controld/controld_schedulerd.c @@ -1,505 +1,505 @@ /* * Copyright 2004-2024 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] 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("Lost connection to the scheduler " CRM_XS " CIB will be saved to " PE_STATE_DIR "/pe-core-%s.bz2", 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); } 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, PCMK_XA_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, 3); if (rc != pcmk_rc_ok) { 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, PCMK_XA_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, PCMK_XA_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, PCMK_XA_NAME, attr_name); crm_xml_add(xml, PCMK_XA_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, PCMK_XA_DC_UUID, controld_globals.our_uuid); pcmk__xe_set_bool_attr(output, PCMK_XA_HAVE_QUORUM, pcmk_is_set(controld_globals.flags, controld_has_quorum)); force_local_option(output, PCMK_OPT_HAVE_WATCHDOG, pcmk__btoa(watchdog)); if (pcmk_is_set(controld_globals.flags, controld_ever_had_quorum) && !crm_have_quorum) { crm_xml_add_int(output, PCMK_XA_NO_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/controld_te_callbacks.c b/daemons/controld/controld_te_callbacks.c index e4056517af..eb4ff44ba7 100644 --- a/daemons/controld/controld_te_callbacks.c +++ b/daemons/controld/controld_te_callbacks.c @@ -1,689 +1,689 @@ /* * Copyright 2004-2024 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 /* For ONLINESTATUS etc */ #include void te_update_confirm(const char *event, xmlNode * msg); #define RSC_OP_PREFIX "//" XML_TAG_DIFF_ADDED "//" XML_TAG_CIB \ "//" XML_LRM_TAG_RSC_OP "[@" PCMK_XA_ID "='" // An explicit PCMK_OPT_SHUTDOWN_LOCK of 0 means the lock has been cleared static bool shutdown_lock_cleared(xmlNode *lrm_resource) { time_t shutdown_lock = 0; return (crm_element_value_epoch(lrm_resource, PCMK_OPT_SHUTDOWN_LOCK, &shutdown_lock) == pcmk_ok) && (shutdown_lock == 0); } static void te_update_diff_v1(const char *event, xmlNode *diff) { int lpc, max; xmlXPathObject *xpathObj = NULL; GString *rsc_op_xpath = NULL; CRM_CHECK(diff != NULL, return); pcmk__output_set_log_level(controld_globals.logger_out, LOG_TRACE); controld_globals.logger_out->message(controld_globals.logger_out, "xml-patchset", diff); if (cib__config_changed_v1(NULL, NULL, &diff)) { abort_transition(INFINITY, pcmk__graph_restart, "Non-status change", diff); goto bail; /* configuration changed */ } /* Tickets Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, pcmk__graph_restart, "Ticket attribute: update", aborted); goto bail; } freeXpathObject(xpathObj); /* Tickets Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, pcmk__graph_restart, "Ticket attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* Transient Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_TRANSIENT_NODEATTRS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, pcmk__graph_restart, "Transient attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); // Check for lrm_resource entries xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RESOURCE); max = numXpathResults(xpathObj); /* * Updates by, or in response to, graph actions will never affect more than * one resource at a time, so such updates indicate an LRM refresh. In that * case, start a new transition rather than check each result individually, * which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if ((controld_globals.transition_graph->pending == 0) && (max > 1)) { crm_debug("Ignoring resource operation updates due to history refresh of %d resources", max); crm_log_xml_trace(diff, "lrm-refresh"); abort_transition(INFINITY, pcmk__graph_restart, "History refresh", NULL); goto bail; } if (max == 1) { xmlNode *lrm_resource = getXpathResult(xpathObj, 0); if (shutdown_lock_cleared(lrm_resource)) { // @TODO would be more efficient to abort once after transition done abort_transition(INFINITY, pcmk__graph_restart, "Shutdown lock cleared", lrm_resource); // Still process results, so we stop timers and update failcounts } } freeXpathObject(xpathObj); /* Process operation updates */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); if (max > 0) { int lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } } freeXpathObject(xpathObj); /* Detect deleted (as opposed to replaced or added) actions - eg. crm_resource -C */ xpathObj = xpath_search(diff, "//" XML_TAG_DIFF_REMOVED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { const char *op_id = NULL; xmlXPathObject *op_match = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; op_id = ID(match); if (rsc_op_xpath == NULL) { rsc_op_xpath = g_string_new(RSC_OP_PREFIX); } else { g_string_truncate(rsc_op_xpath, sizeof(RSC_OP_PREFIX) - 1); } pcmk__g_strcat(rsc_op_xpath, op_id, "']", NULL); op_match = xpath_search(diff, (const char *) rsc_op_xpath->str); if (numXpathResults(op_match) == 0) { /* Prevent false positives by matching cancelations too */ const char *node = get_node_id(match); pcmk__graph_action_t *cancelled = get_cancel_action(op_id, node); if (cancelled == NULL) { crm_debug("No match for deleted action %s (%s on %s)", (const char *) rsc_op_xpath->str, op_id, node); abort_transition(INFINITY, pcmk__graph_restart, "Resource op removal", match); freeXpathObject(op_match); goto bail; } else { crm_debug("Deleted lrm_rsc_op %s on %s was for graph event %d", op_id, node, cancelled->id); } } freeXpathObject(op_match); } bail: freeXpathObject(xpathObj); if (rsc_op_xpath != NULL) { g_string_free(rsc_op_xpath, TRUE); } } static void process_lrm_resource_diff(xmlNode *lrm_resource, const char *node) { for (xmlNode *rsc_op = pcmk__xml_first_child(lrm_resource); rsc_op != NULL; rsc_op = pcmk__xml_next(rsc_op)) { process_graph_event(rsc_op, node); } if (shutdown_lock_cleared(lrm_resource)) { // @TODO would be more efficient to abort once after transition done abort_transition(INFINITY, pcmk__graph_restart, "Shutdown lock cleared", lrm_resource); } } static void process_resource_updates(const char *node, xmlNode *xml, xmlNode *change, const char *op, const char *xpath) { xmlNode *rsc = NULL; if (xml == NULL) { return; } if (pcmk__xe_is(xml, XML_CIB_TAG_LRM)) { xml = first_named_child(xml, XML_LRM_TAG_RESOURCES); CRM_CHECK(xml != NULL, return); } CRM_CHECK(pcmk__xe_is(xml, XML_LRM_TAG_RESOURCES), return); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if ((controld_globals.transition_graph->pending == 0) && (xml->children != NULL) && (xml->children->next != NULL)) { crm_log_xml_trace(change, "lrm-refresh"); abort_transition(INFINITY, pcmk__graph_restart, "History refresh", NULL); return; } for (rsc = pcmk__xml_first_child(xml); rsc != NULL; rsc = pcmk__xml_next(rsc)) { crm_trace("Processing %s", ID(rsc)); process_lrm_resource_diff(rsc, node); } } static char *extract_node_uuid(const char *xpath) { char *mutable_path = strdup(xpath); char *node_uuid = NULL; char *search = NULL; char *match = NULL; match = strstr(mutable_path, "node_state[@" PCMK_XA_ID "=\'"); if (match == NULL) { free(mutable_path); return NULL; } match += strlen("node_state[@" PCMK_XA_ID "=\'"); search = strchr(match, '\''); if (search == NULL) { free(mutable_path); return NULL; } search[0] = 0; node_uuid = strdup(match); free(mutable_path); return node_uuid; } static void abort_unless_down(const char *xpath, const char *op, xmlNode *change, const char *reason) { char *node_uuid = NULL; pcmk__graph_action_t *down = NULL; if(!pcmk__str_eq(op, "delete", pcmk__str_casei)) { abort_transition(INFINITY, pcmk__graph_restart, reason, change); return; } node_uuid = extract_node_uuid(xpath); if(node_uuid == NULL) { crm_err("Could not extract node ID from %s", xpath); abort_transition(INFINITY, pcmk__graph_restart, reason, change); return; } down = match_down_event(node_uuid); if (down == NULL) { crm_trace("Not expecting %s to be down (%s)", node_uuid, xpath); abort_transition(INFINITY, pcmk__graph_restart, reason, change); } else { crm_trace("Expecting changes to %s (%s)", node_uuid, xpath); } free(node_uuid); } static void process_op_deletion(const char *xpath, xmlNode *change) { char *mutable_key = strdup(xpath); char *key; char *node_uuid; // Extract the part of xpath between last pair of single quotes key = strrchr(mutable_key, '\''); if (key != NULL) { *key = '\0'; key = strrchr(mutable_key, '\''); } if (key == NULL) { crm_warn("Ignoring malformed CIB update (resource deletion of %s)", xpath); free(mutable_key); return; } ++key; node_uuid = extract_node_uuid(xpath); if (confirm_cancel_action(key, node_uuid) == FALSE) { abort_transition(INFINITY, pcmk__graph_restart, "Resource operation removal", change); } free(mutable_key); free(node_uuid); } static void process_delete_diff(const char *xpath, const char *op, xmlNode *change) { if (strstr(xpath, "/" XML_LRM_TAG_RSC_OP "[")) { process_op_deletion(xpath, change); } else if (strstr(xpath, "/" XML_CIB_TAG_LRM "[")) { abort_unless_down(xpath, op, change, "Resource state removal"); } else if (strstr(xpath, "/" XML_CIB_TAG_STATE "[")) { abort_unless_down(xpath, op, change, "Node state removal"); } else { crm_trace("Ignoring delete of %s", xpath); } } static void process_node_state_diff(xmlNode *state, xmlNode *change, const char *op, const char *xpath) { xmlNode *lrm = first_named_child(state, XML_CIB_TAG_LRM); process_resource_updates(ID(state), lrm, change, op, xpath); } static void process_status_diff(xmlNode *status, xmlNode *change, const char *op, const char *xpath) { for (xmlNode *state = pcmk__xml_first_child(status); state != NULL; state = pcmk__xml_next(state)) { process_node_state_diff(state, change, op, xpath); } } static void process_cib_diff(xmlNode *cib, xmlNode *change, const char *op, const char *xpath) { xmlNode *status = first_named_child(cib, XML_CIB_TAG_STATUS); xmlNode *config = first_named_child(cib, XML_CIB_TAG_CONFIGURATION); if (status) { process_status_diff(status, change, op, xpath); } if (config) { abort_transition(INFINITY, pcmk__graph_restart, "Non-status-only change", change); } } static void te_update_diff_v2(xmlNode *diff) { crm_log_xml_trace(diff, "Patch:Raw"); for (xmlNode *change = pcmk__xml_first_child(diff); change != NULL; change = pcmk__xml_next(change)) { xmlNode *match = NULL; const char *name = NULL; const char *xpath = crm_element_value(change, XML_DIFF_PATH); // Possible ops: create, modify, delete, move const char *op = crm_element_value(change, XML_DIFF_OP); // Ignore uninteresting updates if (op == NULL) { continue; } else if (xpath == NULL) { crm_trace("Ignoring %s change for version field", op); continue; } else if ((strcmp(op, "move") == 0) && (strstr(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES) == NULL)) { /* We still need to consider moves within the resources section, * since they affect placement order. */ crm_trace("Ignoring move change at %s", xpath); continue; } // Find the result of create/modify ops if (strcmp(op, "create") == 0) { match = change->children; } else if (strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } else if (!pcmk__str_any_of(op, "delete", "move", NULL)) { crm_warn("Ignoring malformed CIB update (%s operation on %s is unrecognized)", op, xpath); continue; } if (match) { if (match->type == XML_COMMENT_NODE) { crm_trace("Ignoring %s operation for comment at %s", op, xpath); continue; } name = (const char *)match->name; } crm_trace("Handling %s operation for %s%s%s", op, (xpath? xpath : "CIB"), (name? " matched by " : ""), (name? name : "")); if (strstr(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION)) { abort_transition(INFINITY, pcmk__graph_restart, "Configuration change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_CIB_TAG_TICKETS) || pcmk__str_eq(name, XML_CIB_TAG_TICKETS, pcmk__str_none)) { abort_transition(INFINITY, pcmk__graph_restart, "Ticket attribute change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_TAG_TRANSIENT_NODEATTRS "[") || pcmk__str_eq(name, XML_TAG_TRANSIENT_NODEATTRS, pcmk__str_none)) { abort_unless_down(xpath, op, change, "Transient attribute change"); break; // Won't be packaged with operation results we may be waiting for } else if (strcmp(op, "delete") == 0) { process_delete_diff(xpath, op, change); } else if (name == NULL) { crm_warn("Ignoring malformed CIB update (%s at %s has no result)", op, xpath); } else if (strcmp(name, XML_TAG_CIB) == 0) { process_cib_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATUS) == 0) { process_status_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATE) == 0) { process_node_state_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_LRM) == 0) { process_resource_updates(ID(match), match, change, op, xpath); } else if (strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = pcmk__xpath_node_id(xpath, "lrm"); process_resource_updates(local_node, match, change, op, xpath); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = pcmk__xpath_node_id(xpath, "lrm"); process_lrm_resource_diff(match, local_node); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = pcmk__xpath_node_id(xpath, "lrm"); process_graph_event(match, local_node); free(local_node); } else { crm_warn("Ignoring malformed CIB update (%s at %s has unrecognized result %s)", op, xpath, name); } } } void te_update_diff(const char *event, xmlNode * msg) { xmlNode *diff = NULL; const char *op = NULL; int rc = -EINVAL; int format = 1; int p_add[] = { 0, 0, 0 }; int p_del[] = { 0, 0, 0 }; CRM_CHECK(msg != NULL, return); crm_element_value_int(msg, F_CIB_RC, &rc); if (controld_globals.transition_graph == NULL) { crm_trace("No graph"); return; } else if (rc < pcmk_ok) { crm_trace("Filter rc=%d (%s)", rc, pcmk_strerror(rc)); return; } else if (controld_globals.transition_graph->complete && (controld_globals.fsa_state != S_IDLE) && (controld_globals.fsa_state != S_TRANSITION_ENGINE) && (controld_globals.fsa_state != S_POLICY_ENGINE)) { crm_trace("Filter state=%s (complete)", fsa_state2string(controld_globals.fsa_state)); return; } op = crm_element_value(msg, F_CIB_OPERATION); diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); xml_patch_versions(diff, p_add, p_del); crm_debug("Processing (%s) diff: %d.%d.%d -> %d.%d.%d (%s)", op, p_del[0], p_del[1], p_del[2], p_add[0], p_add[1], p_add[2], fsa_state2string(controld_globals.fsa_state)); crm_element_value_int(diff, PCMK_XA_FORMAT, &format); switch (format) { case 1: te_update_diff_v1(event, diff); break; case 2: te_update_diff_v2(diff); break; default: crm_warn("Ignoring malformed CIB update (unknown patch format %d)", format); } controld_remove_all_outside_events(); } void process_te_message(xmlNode * msg, xmlNode * xml_data) { const char *value = NULL; xmlXPathObject *xpathObj = NULL; int nmatches = 0; CRM_CHECK(msg != NULL, return); // Transition requests must specify transition engine as subsystem value = crm_element_value(msg, F_CRM_SYS_TO); if (pcmk__str_empty(value) || !pcmk__str_eq(value, CRM_SYSTEM_TENGINE, pcmk__str_none)) { crm_info("Received invalid transition request: subsystem '%s' not '" CRM_SYSTEM_TENGINE "'", pcmk__s(value, "")); return; } // Only the lrm_invoke command is supported as a transition request value = crm_element_value(msg, F_CRM_TASK); if (!pcmk__str_eq(value, CRM_OP_INVOKE_LRM, pcmk__str_none)) { crm_info("Received invalid transition request: command '%s' not '" CRM_OP_INVOKE_LRM "'", pcmk__s(value, "")); return; } // Transition requests must be marked as coming from the executor value = crm_element_value(msg, F_CRM_SYS_FROM); if (!pcmk__str_eq(value, CRM_SYSTEM_LRMD, pcmk__str_none)) { crm_info("Received invalid transition request: from '%s' not '" CRM_SYSTEM_LRMD "'", pcmk__s(value, "")); return; } crm_debug("Processing transition request with ref='%s' origin='%s'", - pcmk__s(crm_element_value(msg, F_CRM_REFERENCE), ""), + pcmk__s(crm_element_value(msg, PCMK_XA_REFERENCE), ""), pcmk__s(crm_element_value(msg, F_ORIG), "")); xpathObj = xpath_search(xml_data, "//" XML_LRM_TAG_RSC_OP); nmatches = numXpathResults(xpathObj); if (nmatches == 0) { crm_err("Received transition request with no results (bug?)"); } else { for (int lpc = 0; lpc < nmatches; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } } freeXpathObject(xpathObj); } void cib_action_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc < pcmk_ok) { crm_err("Update %d FAILED: %s", call_id, pcmk_strerror(rc)); } } /*! * \brief Handle a timeout in node-to-node communication * * \param[in,out] data Pointer to graph action * * \return FALSE (indicating that source should be not be re-added) */ gboolean action_timer_callback(gpointer data) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) data; const char *task = NULL; const char *on_node = NULL; const char *via_node = NULL; CRM_CHECK(data != NULL, return FALSE); stop_te_timer(action); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); via_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (controld_globals.transition_graph->complete) { crm_notice("Node %s did not send %s result (via %s) within %dms " "(ignoring because transition not in progress)", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), action->timeout); } else { /* fail the action */ crm_err("Node %s did not send %s result (via %s) within %dms " "(action timeout plus " PCMK_OPT_CLUSTER_DELAY ")", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), (action->timeout + controld_globals.transition_graph->network_delay)); pcmk__log_graph_action(LOG_ERR, action); pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); te_action_confirmed(action, controld_globals.transition_graph); abort_transition(INFINITY, pcmk__graph_restart, "Action lost", NULL); // Record timeout in the CIB if appropriate if ((action->type == pcmk__rsc_graph_action) && controld_action_is_recordable(task)) { controld_record_action_timeout(action); } } return FALSE; } diff --git a/daemons/controld/controld_transition.c b/daemons/controld/controld_transition.c index 897c6d30ea..b2485332db 100644 --- a/daemons/controld/controld_transition.c +++ b/daemons/controld/controld_transition.c @@ -1,186 +1,186 @@ /* - * Copyright 2004-2023 the Pacemaker project contributors + * Copyright 2004-2024 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 static pcmk__graph_t * create_blank_graph(void) { pcmk__graph_t *a_graph = pcmk__unpack_graph(NULL, NULL); a_graph->complete = true; a_graph->abort_reason = "DC Takeover"; a_graph->completion_action = pcmk__graph_restart; return a_graph; } /* A_TE_START, A_TE_STOP, O_TE_RESTART */ void do_te_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) { cib_t *cib_conn = controld_globals.cib_conn; gboolean init_ok = TRUE; if (pcmk_is_set(action, A_TE_STOP)) { pcmk__free_graph(controld_globals.transition_graph); controld_globals.transition_graph = NULL; if (cib_conn != NULL) { cib_conn->cmds->del_notify_callback(cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff); } controld_clear_fsa_input_flags(R_TE_CONNECTED); crm_info("Transitioner is now inactive"); } if ((action & A_TE_START) == 0) { return; } else if (pcmk_is_set(controld_globals.fsa_input_register, R_TE_CONNECTED)) { crm_debug("The transitioner is already active"); return; } else if ((action & A_TE_START) && cur_state == S_STOPPING) { crm_info("Ignoring request to start the transitioner while shutting down"); return; } if (controld_globals.te_uuid == NULL) { controld_globals.te_uuid = crm_generate_uuid(); crm_info("Registering TE UUID: %s", controld_globals.te_uuid); } if (cib_conn == NULL) { crm_err("Could not set CIB callbacks"); init_ok = FALSE; } else { if (cib_conn->cmds->add_notify_callback(cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff) != pcmk_ok) { crm_err("Could not set CIB notification callback"); init_ok = FALSE; } } if (init_ok) { controld_register_graph_functions(); pcmk__free_graph(controld_globals.transition_graph); /* create a blank one */ crm_debug("Transitioner is now active"); controld_globals.transition_graph = create_blank_graph(); controld_set_fsa_input_flags(R_TE_CONNECTED); } } /* A_TE_INVOKE, A_TE_CANCEL */ void do_te_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) { if (!AM_I_DC || ((controld_globals.fsa_state != S_TRANSITION_ENGINE) && pcmk_is_set(action, A_TE_INVOKE))) { crm_notice("No need to invoke the TE (%s) in state %s", fsa_action2string(action), fsa_state2string(controld_globals.fsa_state)); return; } if (action & A_TE_CANCEL) { crm_debug("Cancelling the transition: %sactive", controld_globals.transition_graph->complete? "in" : ""); abort_transition(INFINITY, pcmk__graph_restart, "Peer Cancelled", NULL); if (!controld_globals.transition_graph->complete) { crmd_fsa_stall(FALSE); } } else if (action & A_TE_HALT) { abort_transition(INFINITY, pcmk__graph_wait, "Peer Halt", NULL); if (!controld_globals.transition_graph->complete) { crmd_fsa_stall(FALSE); } } else if (action & A_TE_INVOKE) { ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); xmlNode *graph_data = input->xml; - const char *ref = crm_element_value(input->msg, XML_ATTR_REFERENCE); + const char *ref = crm_element_value(input->msg, PCMK_XA_REFERENCE); const char *graph_file = crm_element_value(input->msg, F_CRM_TGRAPH); const char *graph_input = crm_element_value(input->msg, F_CRM_TGRAPH_INPUT); if (graph_file == NULL && graph_data == NULL) { crm_log_xml_err(input->msg, "Bad command"); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); return; } if (!controld_globals.transition_graph->complete) { crm_info("Another transition is already active"); abort_transition(INFINITY, pcmk__graph_restart, "Transition Active", NULL); return; } if ((controld_globals.fsa_pe_ref == NULL) || !pcmk__str_eq(controld_globals.fsa_pe_ref, ref, pcmk__str_none)) { crm_info("Transition is redundant: %s expected but %s received", pcmk__s(controld_globals.fsa_pe_ref, "no reference"), pcmk__s(ref, "no reference")); abort_transition(INFINITY, pcmk__graph_restart, "Transition Redundant", NULL); } if (graph_data == NULL && graph_file != NULL) { graph_data = filename2xml(graph_file); } if (controld_is_started_transition_timer()) { crm_debug("The transitioner wait for a transition timer"); return; } CRM_CHECK(graph_data != NULL, crm_err("Input raised by %s is invalid", msg_data->origin); crm_log_xml_err(input->msg, "Bad command"); return); pcmk__free_graph(controld_globals.transition_graph); controld_globals.transition_graph = pcmk__unpack_graph(graph_data, graph_input); CRM_CHECK(controld_globals.transition_graph != NULL, controld_globals.transition_graph = create_blank_graph(); return); crm_info("Processing graph %d (ref=%s) derived from %s", controld_globals.transition_graph->id, ref, graph_input); te_reset_job_counts(); trigger_graph(); pcmk__log_graph(LOG_TRACE, controld_globals.transition_graph); if (graph_data != input->xml) { free_xml(graph_data); } } } diff --git a/daemons/pacemakerd/pcmkd_messages.c b/daemons/pacemakerd/pcmkd_messages.c index 35c339d276..a972ee0f5d 100644 --- a/daemons/pacemakerd/pcmkd_messages.c +++ b/daemons/pacemakerd/pcmkd_messages.c @@ -1,278 +1,278 @@ /* * Copyright 2010-2024 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 static GHashTable *pcmkd_handlers = NULL; static xmlNode * handle_node_cache_request(pcmk__request_t *request) { crm_trace("Ignoring request from client %s to purge node " "because peer cache is not used", pcmk__client_name(request->ipc_client)); pcmk__ipc_send_ack(request->ipc_client, request->ipc_id, request->ipc_flags, "ack", NULL, CRM_EX_OK); return NULL; } static xmlNode * handle_ping_request(pcmk__request_t *request) { xmlNode *msg = request->xml; const char *value = NULL; xmlNode *ping = NULL; xmlNode *reply = NULL; const char *from = crm_element_value(msg, F_CRM_SYS_FROM); /* Pinged for status */ crm_trace("Pinged from " F_CRM_SYS_FROM "='%s' " F_CRM_ORIGIN "='%s'", pcmk__s(from, ""), pcmk__s(crm_element_value(msg, F_CRM_ORIGIN), "")); pcmk__ipc_send_ack(request->ipc_client, request->ipc_id, request->ipc_flags, "ack", NULL, CRM_EX_INDETERMINATE); ping = create_xml_node(NULL, XML_CRM_TAG_PING); value = crm_element_value(msg, F_CRM_SYS_TO); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value); crm_xml_add(ping, XML_PING_ATTR_PACEMAKERDSTATE, pacemakerd_state); crm_xml_add_ll(ping, PCMK_XA_CRM_TIMESTAMP, (long long) subdaemon_check_progress); crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); reply = create_reply(msg, ping); free_xml(ping); if (reply == NULL) { pcmk__format_result(&request->result, CRM_EX_ERROR, PCMK_EXEC_ERROR, "Failed building ping reply for client %s", pcmk__client_name(request->ipc_client)); } else { pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); } /* just proceed state on sbd pinging us */ if (from && strstr(from, "sbd")) { if (pcmk__str_eq(pacemakerd_state, XML_PING_ATTR_PACEMAKERDSTATE_SHUTDOWNCOMPLETE, pcmk__str_none)) { if (pcmk__get_sbd_sync_resource_startup()) { crm_notice("Shutdown-complete-state passed to SBD."); } shutdown_complete_state_reported_to = request->ipc_client->pid; } else if (pcmk__str_eq(pacemakerd_state, XML_PING_ATTR_PACEMAKERDSTATE_WAITPING, pcmk__str_none)) { crm_notice("Received startup-trigger from SBD."); pacemakerd_state = XML_PING_ATTR_PACEMAKERDSTATE_STARTINGDAEMONS; mainloop_set_trigger(startup_trigger); } } return reply; } static xmlNode * handle_shutdown_request(pcmk__request_t *request) { xmlNode *msg = request->xml; xmlNode *shutdown = NULL; xmlNode *reply = NULL; /* Only allow privileged users (i.e. root or hacluster) to shut down * Pacemaker from the command line (or direct IPC), so that other users * are forced to go through the CIB and have ACLs applied. */ bool allowed = pcmk_is_set(request->ipc_client->flags, pcmk__client_privileged); pcmk__ipc_send_ack(request->ipc_client, request->ipc_id, request->ipc_flags, "ack", NULL, CRM_EX_INDETERMINATE); shutdown = create_xml_node(NULL, XML_CIB_ATTR_SHUTDOWN); if (allowed) { crm_notice("Shutting down in response to IPC request %s from %s", - crm_element_value(msg, F_CRM_REFERENCE), + crm_element_value(msg, PCMK_XA_REFERENCE), crm_element_value(msg, F_CRM_ORIGIN)); crm_xml_add_int(shutdown, XML_LRM_ATTR_OPSTATUS, CRM_EX_OK); } else { crm_warn("Ignoring shutdown request from unprivileged client %s", pcmk__client_name(request->ipc_client)); crm_xml_add_int(shutdown, XML_LRM_ATTR_OPSTATUS, CRM_EX_INSUFFICIENT_PRIV); } reply = create_reply(msg, shutdown); free_xml(shutdown); if (reply == NULL) { pcmk__format_result(&request->result, CRM_EX_ERROR, PCMK_EXEC_ERROR, "Failed building shutdown reply for client %s", pcmk__client_name(request->ipc_client)); } else { pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); } if (allowed) { pcmk_shutdown(15); } return reply; } static xmlNode * handle_unknown_request(pcmk__request_t *request) { pcmk__ipc_send_ack(request->ipc_client, request->ipc_id, request->ipc_flags, "ack", NULL, CRM_EX_INVALID_PARAM); pcmk__format_result(&request->result, CRM_EX_PROTOCOL, PCMK_EXEC_INVALID, "Unknown IPC request type '%s' (bug?)", pcmk__client_name(request->ipc_client)); return NULL; } static void pcmkd_register_handlers(void) { pcmk__server_command_t handlers[] = { { CRM_OP_RM_NODE_CACHE, handle_node_cache_request }, { CRM_OP_PING, handle_ping_request }, { CRM_OP_QUIT, handle_shutdown_request }, { NULL, handle_unknown_request }, }; pcmkd_handlers = pcmk__register_handlers(handlers); } static int32_t pcmk_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { crm_trace("Connection %p", c); if (pcmk__new_client(c, uid, gid) == NULL) { return -EIO; } return 0; } /* Error code means? */ static int32_t pcmk_ipc_closed(qb_ipcs_connection_t * c) { pcmk__client_t *client = pcmk__find_client(c); if (client == NULL) { return 0; } crm_trace("Connection %p", c); if (shutdown_complete_state_reported_to == client->pid) { shutdown_complete_state_reported_client_closed = TRUE; if (shutdown_trigger) { mainloop_set_trigger(shutdown_trigger); } } pcmk__free_client(client); return 0; } static void pcmk_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); pcmk_ipc_closed(c); } /* Exit code means? */ static int32_t pcmk_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; xmlNode *msg = NULL; pcmk__client_t *c = pcmk__find_client(qbc); CRM_CHECK(c != NULL, return 0); if (pcmkd_handlers == NULL) { pcmkd_register_handlers(); } msg = pcmk__client_data2xml(c, data, &id, &flags); if (msg == NULL) { pcmk__ipc_send_ack(c, id, flags, "ack", NULL, CRM_EX_PROTOCOL); return 0; } else { char *log_msg = NULL; const char *reason = NULL; xmlNode *reply = NULL; pcmk__request_t request = { .ipc_client = c, .ipc_id = id, .ipc_flags = flags, .peer = NULL, .xml = msg, .call_options = 0, .result = PCMK__UNKNOWN_RESULT, }; request.op = crm_element_value_copy(request.xml, F_CRM_TASK); CRM_CHECK(request.op != NULL, return 0); reply = pcmk__process_request(&request, pcmkd_handlers); if (reply != NULL) { pcmk__ipc_send_xml(c, id, reply, crm_ipc_server_event); free_xml(reply); } reason = request.result.exit_reason; log_msg = crm_strdup_printf("Processed %s request from %s %s: %s%s%s%s", request.op, pcmk__request_origin_type(&request), pcmk__request_origin(&request), pcmk_exec_status_str(request.result.execution_status), (reason == NULL)? "" : " (", (reason == NULL)? "" : reason, (reason == NULL)? "" : ")"); if (!pcmk__result_ok(&request.result)) { crm_warn("%s", log_msg); } else { crm_debug("%s", log_msg); } free(log_msg); pcmk__reset_request(&request); } free_xml(msg); return 0; } struct qb_ipcs_service_handlers pacemakerd_ipc_callbacks = { .connection_accept = pcmk_ipc_accept, .connection_created = NULL, .msg_process = pcmk_ipc_dispatch, .connection_closed = pcmk_ipc_closed, .connection_destroyed = pcmk_ipc_destroy }; diff --git a/include/crm/msg_xml.h b/include/crm/msg_xml.h index abeb000c5f..df3f9f2d4c 100644 --- a/include/crm/msg_xml.h +++ b/include/crm/msg_xml.h @@ -1,416 +1,417 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__CRM_MSG_XML__H # define PCMK__CRM_MSG_XML__H # include #if !defined(PCMK_ALLOW_DEPRECATED) || (PCMK_ALLOW_DEPRECATED == 1) #include #endif #ifdef __cplusplus extern "C" { #endif /* This file defines constants for various XML syntax (mainly element and * attribute names). * * For consistency, new constants should start with "PCMK_", followed by "XE" * for XML element names, "XA" for XML attribute names, and "META" for meta * attribute names. Old names that don't follow this policy should eventually be * deprecated and replaced with names that do. * * Symbols should be public if the user may specify them somewhere (especially * the CIB). They should be internal if they're used only internally to * Pacemaker (such as daemon IPC/CPG message XML). * * For meta-attributes that can be specified as either XML attributes or nvpair * names, use "META" unless using both "XA" and "META" constants adds clarity. */ /* * XML elements */ #define PCMK_XE_DATE_EXPRESSION "date_expression" #define PCMK_XE_OP "op" #define PCMK_XE_OP_EXPRESSION "op_expression" #define PCMK_XE_RSC_EXPRESSION "rsc_expression" /* This has been deprecated as a CIB element (an alias for with * PCMK_META_PROMOTABLE set to "true") since 2.0.0. */ #define PCMK_XE_PROMOTABLE_LEGACY "master" /* * XML attributes */ #define PCMK_XA_ADMIN_EPOCH "admin_epoch" #define PCMK_XA_CIB_LAST_WRITTEN "cib-last-written" #define PCMK_XA_CLASS "class" #define PCMK_XA_CRM_DEBUG_ORIGIN "crm-debug-origin" #define PCMK_XA_CRM_FEATURE_SET "crm_feature_set" #define PCMK_XA_CRM_TIMESTAMP "crm-timestamp" #define PCMK_XA_DC_UUID "dc-uuid" #define PCMK_XA_DESCRIPTION "description" #define PCMK_XA_EPOCH "epoch" #define PCMK_XA_FORMAT "format" #define PCMK_XA_HAVE_QUORUM "have-quorum" #define PCMK_XA_ID "id" #define PCMK_XA_ID_REF "id-ref" #define PCMK_XA_NAME "name" #define PCMK_XA_NO_QUORUM_PANIC "no-quorum-panic" #define PCMK_XA_NUM_UPDATES "num_updates" #define PCMK_XA_OP "op" #define PCMK_XA_PROVIDER "provider" +#define PCMK_XA_REFERENCE "reference" #define PCMK_XA_REQUEST "request" #define PCMK_XA_TYPE "type" #define PCMK_XA_UNAME "uname" #define PCMK_XA_UPDATE_CLIENT "update-client" #define PCMK_XA_UPDATE_ORIGIN "update-origin" #define PCMK_XA_UPDATE_USER "update-user" #define PCMK_XA_VALIDATE_WITH "validate-with" #define PCMK_XA_VALUE "value" #define PCMK_XA_VERSION "version" /* * Older constants that don't follow current naming */ # ifndef F_ORIG # define F_ORIG "src" # endif # ifndef F_SEQ # define F_SEQ "seq" # endif # ifndef F_SUBTYPE # define F_SUBTYPE "subt" # endif # ifndef F_TYPE # define F_TYPE "t" # endif # ifndef F_CLIENTNAME # define F_CLIENTNAME "cn" # endif # ifndef F_XML_TAGNAME # define F_XML_TAGNAME "__name__" # endif # ifndef T_CRM # define T_CRM "crmd" # endif # ifndef T_ATTRD # define T_ATTRD "attrd" # endif # define CIB_OPTIONS_FIRST "cib-bootstrap-options" # define F_CRM_DATA "crm_xml" # define F_CRM_TASK "crm_task" # define F_CRM_HOST_TO "crm_host_to" # define F_CRM_MSG_TYPE F_SUBTYPE # define F_CRM_SYS_TO "crm_sys_to" # define F_CRM_SYS_FROM "crm_sys_from" # define F_CRM_HOST_FROM F_ORIG -# define F_CRM_REFERENCE XML_ATTR_REFERENCE +# define F_CRM_REFERENCE PCMK_XA_REFERENCE # define F_CRM_VERSION PCMK_XA_VERSION # define F_CRM_ORIGIN "origin" # define F_CRM_USER "crm_user" # define F_CRM_JOIN_ID "join_id" # define F_CRM_DC_LEAVING "dc-leaving" # define F_CRM_ELECTION_ID "election-id" # define F_CRM_ELECTION_AGE_S "election-age-sec" # define F_CRM_ELECTION_AGE_US "election-age-nano-sec" # define F_CRM_ELECTION_OWNER "election-owner" # define F_CRM_TGRAPH "crm-tgraph-file" # define F_CRM_TGRAPH_INPUT "crm-tgraph-in" # define F_CRM_THROTTLE_MODE "crm-limit-mode" # define F_CRM_THROTTLE_MAX "crm-limit-max" /*---- Common tags/attrs */ # define XML_DIFF_MARKER "__crm_diff_marker__" # define XML_TAG_CIB "cib" # define XML_TAG_FAILED "failed" # define XML_ATTR_TIMEOUT "timeout" # define XML_BOOLEAN_TRUE "true" # define XML_BOOLEAN_FALSE "false" # define XML_TAG_OPTIONS "options" /*---- top level tags/attrs */ # define XML_ATTR_REFERENCE "reference" # define XML_CRM_TAG_PING "ping_response" # define XML_PING_ATTR_STATUS "result" # define XML_PING_ATTR_SYSFROM "crm_subsystem" # define XML_PING_ATTR_CRMDSTATE "crmd_state" # define XML_PING_ATTR_PACEMAKERDSTATE "pacemakerd_state" # define XML_PING_ATTR_PACEMAKERDSTATE_INIT "init" # define XML_PING_ATTR_PACEMAKERDSTATE_STARTINGDAEMONS "starting_daemons" # define XML_PING_ATTR_PACEMAKERDSTATE_WAITPING "wait_for_ping" # define XML_PING_ATTR_PACEMAKERDSTATE_RUNNING "running" # define XML_PING_ATTR_PACEMAKERDSTATE_SHUTTINGDOWN "shutting_down" # define XML_PING_ATTR_PACEMAKERDSTATE_SHUTDOWNCOMPLETE "shutdown_complete" # define XML_PING_ATTR_PACEMAKERDSTATE_REMOTE "remote" # define XML_FAIL_TAG_CIB "failed_update" # define XML_FAILCIB_ATTR_OBJTYPE "object_type" # define XML_FAILCIB_ATTR_OP "operation" # define XML_FAILCIB_ATTR_REASON "reason" /*---- CIB specific tags/attrs */ # define XML_CIB_TAG_SECTION_ALL "all" # define XML_CIB_TAG_CONFIGURATION "configuration" # define XML_CIB_TAG_STATUS "status" # define XML_CIB_TAG_RESOURCES "resources" # define XML_CIB_TAG_NODES "nodes" # define XML_CIB_TAG_CONSTRAINTS "constraints" # define XML_CIB_TAG_CRMCONFIG "crm_config" # define XML_CIB_TAG_OPCONFIG "op_defaults" # define XML_CIB_TAG_RSCCONFIG "rsc_defaults" # define XML_CIB_TAG_ACLS "acls" # define XML_CIB_TAG_ALERTS "alerts" # define XML_CIB_TAG_ALERT "alert" # define XML_CIB_TAG_ALERT_RECIPIENT "recipient" # define XML_CIB_TAG_ALERT_SELECT "select" # define XML_CIB_TAG_ALERT_ATTRIBUTES "select_attributes" # define XML_CIB_TAG_ALERT_FENCING "select_fencing" # define XML_CIB_TAG_ALERT_NODES "select_nodes" # define XML_CIB_TAG_ALERT_RESOURCES "select_resources" # define XML_CIB_TAG_ALERT_ATTR "attribute" # define XML_CIB_TAG_STATE "node_state" # define XML_CIB_TAG_NODE "node" # define XML_CIB_TAG_NVPAIR "nvpair" # define XML_CIB_TAG_PROPSET "cluster_property_set" # define XML_TAG_ATTR_SETS "instance_attributes" # define XML_TAG_META_SETS "meta_attributes" # define XML_TAG_ATTRS "attributes" # define XML_TAG_PARAMS "parameters" # define XML_TAG_PARAM "param" # define XML_TAG_UTILIZATION "utilization" # define XML_TAG_RESOURCE_REF "resource_ref" # define XML_CIB_TAG_RESOURCE "primitive" # define XML_CIB_TAG_GROUP "group" # define XML_CIB_TAG_INCARNATION "clone" # define XML_CIB_TAG_CONTAINER "bundle" # define XML_CIB_TAG_RSC_TEMPLATE "template" # define XML_OP_ATTR_ON_FAIL "on-fail" # define XML_OP_ATTR_START_DELAY "start-delay" # define XML_OP_ATTR_ORIGIN "interval-origin" # define XML_OP_ATTR_PENDING "record-pending" # define XML_OP_ATTR_DIGESTS_ALL "digests-all" # define XML_OP_ATTR_DIGESTS_SECURE "digests-secure" # define XML_CIB_TAG_LRM "lrm" # define XML_LRM_TAG_RESOURCES "lrm_resources" # define XML_LRM_TAG_RESOURCE "lrm_resource" # define XML_LRM_TAG_RSC_OP "lrm_rsc_op" //! \deprecated Do not use (will be removed in a future release) # define XML_CIB_ATTR_REPLACE "replace" # define XML_CIB_ATTR_PRIORITY "priority" # define XML_NODE_IS_REMOTE "remote_node" # define XML_NODE_IS_FENCED "node_fenced" # define XML_NODE_IS_MAINTENANCE "node_in_maintenance" # define XML_CIB_ATTR_SHUTDOWN "shutdown" /* Aside from being an old name for the executor, LRM is a misnomer here because * the controller and scheduler use these to track actions, which are not always * executor operations. */ // XML attribute that takes interval specification (user-facing configuration) # define XML_LRM_ATTR_INTERVAL "interval" // XML attribute that takes interval in milliseconds (daemon APIs) // (identical value as above, but different constant allows clearer code intent) # define XML_LRM_ATTR_INTERVAL_MS XML_LRM_ATTR_INTERVAL # define XML_LRM_ATTR_TASK "operation" # define XML_LRM_ATTR_TASK_KEY "operation_key" # define XML_LRM_ATTR_TARGET "on_node" # define XML_LRM_ATTR_TARGET_UUID "on_node_uuid" /*! Actions to be executed on Pacemaker Remote nodes are routed through the * controller on the cluster node hosting the remote connection. That cluster * node is considered the router node for the action. */ # define XML_LRM_ATTR_ROUTER_NODE "router_node" # define XML_LRM_ATTR_RSCID "rsc-id" # define XML_LRM_ATTR_OPSTATUS "op-status" # define XML_LRM_ATTR_RC "rc-code" # define XML_LRM_ATTR_CALLID "call-id" # define XML_LRM_ATTR_OP_DIGEST "op-digest" # define XML_LRM_ATTR_OP_RESTART "op-force-restart" # define XML_LRM_ATTR_OP_SECURE "op-secure-params" # define XML_LRM_ATTR_RESTART_DIGEST "op-restart-digest" # define XML_LRM_ATTR_SECURE_DIGEST "op-secure-digest" # define XML_LRM_ATTR_EXIT_REASON "exit-reason" # define XML_RSC_OP_LAST_CHANGE "last-rc-change" # define XML_RSC_OP_T_EXEC "exec-time" # define XML_RSC_OP_T_QUEUE "queue-time" # define XML_LRM_ATTR_MIGRATE_SOURCE "migrate_source" # define XML_LRM_ATTR_MIGRATE_TARGET "migrate_target" # define XML_TAG_GRAPH "transition_graph" # define XML_GRAPH_TAG_RSC_OP "rsc_op" # define XML_GRAPH_TAG_PSEUDO_EVENT "pseudo_event" # define XML_GRAPH_TAG_CRM_EVENT "crm_event" # define XML_GRAPH_TAG_DOWNED "downed" # define XML_GRAPH_TAG_MAINTENANCE "maintenance" # define XML_TAG_RULE "rule" # define XML_RULE_ATTR_SCORE "score" # define XML_RULE_ATTR_SCORE_ATTRIBUTE "score-attribute" # define XML_RULE_ATTR_ROLE "role" # define XML_RULE_ATTR_BOOLEAN_OP "boolean-op" # define XML_TAG_EXPRESSION "expression" # define XML_EXPR_ATTR_ATTRIBUTE "attribute" # define XML_EXPR_ATTR_OPERATION "operation" # define XML_EXPR_ATTR_VALUE_SOURCE "value-source" # define XML_CONS_TAG_RSC_DEPEND "rsc_colocation" # define XML_CONS_TAG_RSC_ORDER "rsc_order" # define XML_CONS_TAG_RSC_LOCATION "rsc_location" # define XML_CONS_TAG_RSC_TICKET "rsc_ticket" # define XML_CONS_TAG_RSC_SET "resource_set" # define XML_CONS_ATTR_SYMMETRICAL "symmetrical" # define XML_LOCATION_ATTR_DISCOVERY "resource-discovery" # define XML_COLOC_ATTR_SOURCE "rsc" # define XML_COLOC_ATTR_SOURCE_ROLE "rsc-role" # define XML_COLOC_ATTR_TARGET "with-rsc" # define XML_COLOC_ATTR_TARGET_ROLE "with-rsc-role" # define XML_COLOC_ATTR_NODE_ATTR "node-attribute" # define XML_COLOC_ATTR_INFLUENCE "influence" //! \deprecated Deprecated since 2.1.5 # define XML_COLOC_ATTR_SOURCE_INSTANCE "rsc-instance" //! \deprecated Deprecated since 2.1.5 # define XML_COLOC_ATTR_TARGET_INSTANCE "with-rsc-instance" # define XML_LOC_ATTR_SOURCE "rsc" # define XML_LOC_ATTR_SOURCE_PATTERN "rsc-pattern" # define XML_ORDER_ATTR_FIRST "first" # define XML_ORDER_ATTR_THEN "then" # define XML_ORDER_ATTR_FIRST_ACTION "first-action" # define XML_ORDER_ATTR_THEN_ACTION "then-action" # define XML_ORDER_ATTR_KIND "kind" //! \deprecated Deprecated since 2.1.5 # define XML_ORDER_ATTR_FIRST_INSTANCE "first-instance" //! \deprecated Deprecated since 2.1.5 # define XML_ORDER_ATTR_THEN_INSTANCE "then-instance" # define XML_TICKET_ATTR_TICKET "ticket" # define XML_TICKET_ATTR_LOSS_POLICY "loss-policy" # define XML_NODE_ATTR_RSC_DISCOVERY "resource-discovery-enabled" # define XML_ALERT_ATTR_PATH "path" # define XML_ALERT_ATTR_TIMEOUT "timeout" # define XML_ALERT_ATTR_TSTAMP_FORMAT "timestamp-format" # define XML_CIB_TAG_GENERATION_TUPPLE "generation_tuple" # define XML_ATTR_TRANSITION_MAGIC "transition-magic" # define XML_ATTR_TRANSITION_KEY "transition-key" # define XML_ATTR_TE_NOWAIT "op_no_wait" # define XML_ATTR_TE_TARGET_RC "op_target_rc" # define XML_TAG_TRANSIENT_NODEATTRS "transient_attributes" //! \deprecated Do not use (will be removed in a future release) # define XML_TAG_DIFF_ADDED "diff-added" //! \deprecated Do not use (will be removed in a future release) # define XML_TAG_DIFF_REMOVED "diff-removed" # define XML_ACL_TAG_USER "acl_target" # define XML_ACL_TAG_USERv1 "acl_user" # define XML_ACL_TAG_GROUP "acl_group" # define XML_ACL_TAG_ROLE "acl_role" # define XML_ACL_TAG_PERMISSION "acl_permission" # define XML_ACL_TAG_ROLE_REF "role" # define XML_ACL_TAG_ROLE_REFv1 "role_ref" # define XML_ACL_ATTR_KIND "kind" # define XML_ACL_TAG_READ "read" # define XML_ACL_TAG_WRITE "write" # define XML_ACL_TAG_DENY "deny" # define XML_ACL_ATTR_REF "reference" # define XML_ACL_ATTR_REFv1 "ref" # define XML_ACL_ATTR_TAG "object-type" # define XML_ACL_ATTR_TAGv1 "tag" # define XML_ACL_ATTR_XPATH "xpath" # define XML_ACL_ATTR_ATTRIBUTE "attribute" # define XML_CIB_TAG_TICKETS "tickets" # define XML_CIB_TAG_TICKET_STATE "ticket_state" # define XML_CIB_TAG_TAGS "tags" # define XML_CIB_TAG_TAG "tag" # define XML_CIB_TAG_OBJ_REF "obj_ref" # define XML_TAG_FENCING_TOPOLOGY "fencing-topology" # define XML_TAG_FENCING_LEVEL "fencing-level" # define XML_ATTR_STONITH_INDEX "index" # define XML_ATTR_STONITH_TARGET "target" # define XML_ATTR_STONITH_TARGET_VALUE "target-value" # define XML_ATTR_STONITH_TARGET_PATTERN "target-pattern" # define XML_ATTR_STONITH_TARGET_ATTRIBUTE "target-attribute" # define XML_ATTR_STONITH_DEVICES "devices" # define XML_TAG_DIFF "diff" # define XML_DIFF_VERSION "version" # define XML_DIFF_VSOURCE "source" # define XML_DIFF_VTARGET "target" # define XML_DIFF_CHANGE "change" # define XML_DIFF_LIST "change-list" # define XML_DIFF_ATTR "change-attr" # define XML_DIFF_RESULT "change-result" # define XML_DIFF_OP "operation" # define XML_DIFF_PATH "path" # define XML_DIFF_POSITION "position" # define ID(x) crm_element_value(x, PCMK_XA_ID) #ifdef __cplusplus } #endif #endif diff --git a/lib/common/acl.c b/lib/common/acl.c index 2091decccd..0c8b2024e2 100644 --- a/lib/common/acl.c +++ b/lib/common/acl.c @@ -1,857 +1,857 @@ /* * Copyright 2004-2024 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 "crmcommon_private.h" typedef struct xml_acl_s { enum xml_private_flags mode; gchar *xpath; } xml_acl_t; static void free_acl(void *data) { if (data) { xml_acl_t *acl = data; g_free(acl->xpath); free(acl); } } void pcmk__free_acls(GList *acls) { g_list_free_full(acls, free_acl); } static GList * create_acl(const xmlNode *xml, GList *acls, enum xml_private_flags mode) { xml_acl_t *acl = NULL; const char *tag = crm_element_value(xml, XML_ACL_ATTR_TAG); - const char *ref = crm_element_value(xml, XML_ACL_ATTR_REF); + const char *ref = crm_element_value(xml, PCMK_XA_REFERENCE); const char *xpath = crm_element_value(xml, XML_ACL_ATTR_XPATH); const char *attr = crm_element_value(xml, XML_ACL_ATTR_ATTRIBUTE); if (tag == NULL) { // @COMPAT rolling upgrades <=1.1.11 tag = crm_element_value(xml, XML_ACL_ATTR_TAGv1); } if (ref == NULL) { // @COMPAT rolling upgrades <=1.1.11 ref = crm_element_value(xml, XML_ACL_ATTR_REFv1); } if ((tag == NULL) && (ref == NULL) && (xpath == NULL)) { // Schema should prevent this, but to be safe ... crm_trace("Ignoring ACL <%s> element without selection criteria", xml->name); return NULL; } acl = calloc(1, sizeof (xml_acl_t)); CRM_ASSERT(acl != NULL); acl->mode = mode; if (xpath) { acl->xpath = g_strdup(xpath); crm_trace("Unpacked ACL <%s> element using xpath: %s", xml->name, acl->xpath); } else { GString *buf = g_string_sized_new(128); if ((ref != NULL) && (attr != NULL)) { // NOTE: schema currently does not allow this pcmk__g_strcat(buf, "//", pcmk__s(tag, "*"), "[@" PCMK_XA_ID "='", ref, "' and @", attr, "]", NULL); } else if (ref != NULL) { pcmk__g_strcat(buf, "//", pcmk__s(tag, "*"), "[@" PCMK_XA_ID "='", ref, "']", NULL); } else if (attr != NULL) { pcmk__g_strcat(buf, "//", pcmk__s(tag, "*"), "[@", attr, "]", NULL); } else { pcmk__g_strcat(buf, "//", pcmk__s(tag, "*"), NULL); } acl->xpath = buf->str; g_string_free(buf, FALSE); crm_trace("Unpacked ACL <%s> element as xpath: %s", xml->name, acl->xpath); } return g_list_append(acls, acl); } /*! * \internal * \brief Unpack a user, group, or role subtree of the ACLs section * * \param[in] acl_top XML of entire ACLs section * \param[in] acl_entry XML of ACL element being unpacked * \param[in,out] acls List of ACLs unpacked so far * * \return New head of (possibly modified) acls * * \note This function is recursive */ static GList * parse_acl_entry(const xmlNode *acl_top, const xmlNode *acl_entry, GList *acls) { xmlNode *child = NULL; for (child = pcmk__xe_first_child(acl_entry); child; child = pcmk__xe_next(child)) { const char *tag = (const char *) child->name; const char *kind = crm_element_value(child, XML_ACL_ATTR_KIND); if (pcmk__xe_is(child, XML_ACL_TAG_PERMISSION)) { CRM_ASSERT(kind != NULL); crm_trace("Unpacking ACL <%s> element of kind '%s'", tag, kind); tag = kind; } else { crm_trace("Unpacking ACL <%s> element", tag); } if (strcmp(XML_ACL_TAG_ROLE_REF, tag) == 0 || strcmp(XML_ACL_TAG_ROLE_REFv1, tag) == 0) { const char *ref_role = crm_element_value(child, PCMK_XA_ID); if (ref_role) { xmlNode *role = NULL; for (role = pcmk__xe_first_child(acl_top); role; role = pcmk__xe_next(role)) { if (!strcmp(XML_ACL_TAG_ROLE, (const char *) role->name)) { const char *role_id = crm_element_value(role, PCMK_XA_ID); if (role_id && strcmp(ref_role, role_id) == 0) { crm_trace("Unpacking referenced role '%s' in ACL <%s> element", role_id, acl_entry->name); acls = parse_acl_entry(acl_top, role, acls); break; } } } } } else if (strcmp(XML_ACL_TAG_READ, tag) == 0) { acls = create_acl(child, acls, pcmk__xf_acl_read); } else if (strcmp(XML_ACL_TAG_WRITE, tag) == 0) { acls = create_acl(child, acls, pcmk__xf_acl_write); } else if (strcmp(XML_ACL_TAG_DENY, tag) == 0) { acls = create_acl(child, acls, pcmk__xf_acl_deny); } else { crm_warn("Ignoring unknown ACL %s '%s'", (kind? "kind" : "element"), tag); } } return acls; } /* */ static const char * acl_to_text(enum xml_private_flags flags) { if (pcmk_is_set(flags, pcmk__xf_acl_deny)) { return "deny"; } else if (pcmk_any_flags_set(flags, pcmk__xf_acl_write|pcmk__xf_acl_create)) { return "read/write"; } else if (pcmk_is_set(flags, pcmk__xf_acl_read)) { return "read"; } return "none"; } void pcmk__apply_acl(xmlNode *xml) { GList *aIter = NULL; xml_doc_private_t *docpriv = xml->doc->_private; xml_node_private_t *nodepriv; xmlXPathObjectPtr xpathObj = NULL; if (!xml_acl_enabled(xml)) { crm_trace("Skipping ACLs for user '%s' because not enabled for this XML", docpriv->user); return; } for (aIter = docpriv->acls; aIter != NULL; aIter = aIter->next) { int max = 0, lpc = 0; xml_acl_t *acl = aIter->data; xpathObj = xpath_search(xml, acl->xpath); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); nodepriv = match->_private; pcmk__set_xml_flags(nodepriv, acl->mode); // Build a GString only if tracing is enabled pcmk__if_tracing( { GString *path = pcmk__element_xpath(match); crm_trace("Applying %s ACL to %s matched by %s", acl_to_text(acl->mode), path->str, acl->xpath); g_string_free(path, TRUE); }, {} ); } crm_trace("Applied %s ACL %s (%d match%s)", acl_to_text(acl->mode), acl->xpath, max, ((max == 1)? "" : "es")); freeXpathObject(xpathObj); } } /*! * \internal * \brief Unpack ACLs for a given user into the * metadata of the target XML tree * * Taking the description of ACLs from the source XML tree and * marking up the target XML tree with access information for the * given user by tacking it onto the relevant nodes * * \param[in] source XML with ACL definitions * \param[in,out] target XML that ACLs will be applied to * \param[in] user Username whose ACLs need to be unpacked */ void pcmk__unpack_acl(xmlNode *source, xmlNode *target, const char *user) { xml_doc_private_t *docpriv = NULL; if ((target == NULL) || (target->doc == NULL) || (target->doc->_private == NULL)) { return; } docpriv = target->doc->_private; if (!pcmk_acl_required(user)) { crm_trace("Not unpacking ACLs because not required for user '%s'", user); } else if (docpriv->acls == NULL) { xmlNode *acls = get_xpath_object("//" XML_CIB_TAG_ACLS, source, LOG_NEVER); pcmk__str_update(&docpriv->user, user); if (acls) { xmlNode *child = NULL; for (child = pcmk__xe_first_child(acls); child; child = pcmk__xe_next(child)) { if (pcmk__xe_is(child, XML_ACL_TAG_USER) || pcmk__xe_is(child, XML_ACL_TAG_USERv1)) { const char *id = crm_element_value(child, PCMK_XA_NAME); if (id == NULL) { id = crm_element_value(child, PCMK_XA_ID); } if (id && strcmp(id, user) == 0) { crm_debug("Unpacking ACLs for user '%s'", id); docpriv->acls = parse_acl_entry(acls, child, docpriv->acls); } } else if (pcmk__xe_is(child, XML_ACL_TAG_GROUP)) { const char *id = crm_element_value(child, PCMK_XA_NAME); if (id == NULL) { id = crm_element_value(child, PCMK_XA_ID); } if (id && pcmk__is_user_in_group(user,id)) { crm_debug("Unpacking ACLs for group '%s'", id); docpriv->acls = parse_acl_entry(acls, child, docpriv->acls); } } } } } } /*! * \internal * \brief Copy source to target and set xf_acl_enabled flag in target * * \param[in] acl_source XML with ACL definitions * \param[in,out] target XML that ACLs will be applied to * \param[in] user Username whose ACLs need to be set */ void pcmk__enable_acl(xmlNode *acl_source, xmlNode *target, const char *user) { pcmk__unpack_acl(acl_source, target, user); pcmk__set_xml_doc_flag(target, pcmk__xf_acl_enabled); pcmk__apply_acl(target); } static inline bool test_acl_mode(enum xml_private_flags allowed, enum xml_private_flags requested) { if (pcmk_is_set(allowed, pcmk__xf_acl_deny)) { return false; } else if (pcmk_all_flags_set(allowed, requested)) { return true; } else if (pcmk_is_set(requested, pcmk__xf_acl_read) && pcmk_is_set(allowed, pcmk__xf_acl_write)) { return true; } else if (pcmk_is_set(requested, pcmk__xf_acl_create) && pcmk_any_flags_set(allowed, pcmk__xf_acl_write|pcmk__xf_created)) { return true; } return false; } /*! * \internal * \brief Rid XML tree of all unreadable nodes and node properties * * \param[in,out] xml Root XML node to be purged of attributes * * \return true if this node or any of its children are readable * if false is returned, xml will be freed * * \note This function is recursive */ static bool purge_xml_attributes(xmlNode *xml) { xmlNode *child = NULL; xmlAttr *xIter = NULL; bool readable_children = false; xml_node_private_t *nodepriv = xml->_private; if (test_acl_mode(nodepriv->flags, pcmk__xf_acl_read)) { crm_trace("%s[@" PCMK_XA_ID "=%s] is readable", xml->name, ID(xml)); return true; } xIter = xml->properties; while (xIter != NULL) { xmlAttr *tmp = xIter; const char *prop_name = (const char *)xIter->name; xIter = xIter->next; if (strcmp(prop_name, PCMK_XA_ID) == 0) { continue; } xmlUnsetProp(xml, tmp->name); } child = pcmk__xml_first_child(xml); while ( child != NULL ) { xmlNode *tmp = child; child = pcmk__xml_next(child); readable_children |= purge_xml_attributes(tmp); } if (!readable_children) { free_xml(xml); /* Nothing readable under here, purge completely */ } return readable_children; } /*! * \brief Copy ACL-allowed portions of specified XML * * \param[in] user Username whose ACLs should be used * \param[in] acl_source XML containing ACLs * \param[in] xml XML to be copied * \param[out] result Copy of XML portions readable via ACLs * * \return true if xml exists and ACLs are required for user, false otherwise * \note If this returns true, caller should use \p result rather than \p xml */ bool xml_acl_filtered_copy(const char *user, xmlNode *acl_source, xmlNode *xml, xmlNode **result) { GList *aIter = NULL; xmlNode *target = NULL; xml_doc_private_t *docpriv = NULL; *result = NULL; if ((xml == NULL) || !pcmk_acl_required(user)) { crm_trace("Not filtering XML because ACLs not required for user '%s'", user); return false; } crm_trace("Filtering XML copy using user '%s' ACLs", user); target = copy_xml(xml); if (target == NULL) { return true; } pcmk__enable_acl(acl_source, target, user); docpriv = target->doc->_private; for(aIter = docpriv->acls; aIter != NULL && target; aIter = aIter->next) { int max = 0; xml_acl_t *acl = aIter->data; if (acl->mode != pcmk__xf_acl_deny) { /* Nothing to do */ } else if (acl->xpath) { int lpc = 0; xmlXPathObjectPtr xpathObj = xpath_search(target, acl->xpath); max = numXpathResults(xpathObj); for(lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); if (!purge_xml_attributes(match) && (match == target)) { crm_trace("ACLs deny user '%s' access to entire XML document", user); freeXpathObject(xpathObj); return true; } } crm_trace("ACLs deny user '%s' access to %s (%d %s)", user, acl->xpath, max, pcmk__plural_alt(max, "match", "matches")); freeXpathObject(xpathObj); } } if (!purge_xml_attributes(target)) { crm_trace("ACLs deny user '%s' access to entire XML document", user); return true; } if (docpriv->acls) { g_list_free_full(docpriv->acls, free_acl); docpriv->acls = NULL; } else { crm_trace("User '%s' without ACLs denied access to entire XML document", user); free_xml(target); target = NULL; } if (target) { *result = target; } return true; } /*! * \internal * \brief Check whether creation of an XML element is implicitly allowed * * Check whether XML is a "scaffolding" element whose creation is implicitly * allowed regardless of ACLs (that is, it is not in the ACL section and has * no attributes other than \c PCMK_XA_ID). * * \param[in] xml XML element to check * * \return true if XML element is implicitly allowed, false otherwise */ static bool implicitly_allowed(const xmlNode *xml) { GString *path = NULL; for (xmlAttr *prop = xml->properties; prop != NULL; prop = prop->next) { if (strcmp((const char *) prop->name, PCMK_XA_ID) != 0) { return false; } } path = pcmk__element_xpath(xml); CRM_ASSERT(path != NULL); if (strstr((const char *) path->str, "/" XML_CIB_TAG_ACLS "/") != NULL) { g_string_free(path, TRUE); return false; } g_string_free(path, TRUE); return true; } #define display_id(xml) (ID(xml)? ID(xml) : "") /*! * \internal * \brief Drop XML nodes created in violation of ACLs * * Given an XML element, free all of its descendant nodes created in violation * of ACLs, with the exception of allowing "scaffolding" elements (i.e. those * that aren't in the ACL section and don't have any attributes other than * \c PCMK_XA_ID). * * \param[in,out] xml XML to check * \param[in] check_top Whether to apply checks to argument itself * (if true, xml might get freed) * * \note This function is recursive */ void pcmk__apply_creation_acl(xmlNode *xml, bool check_top) { xml_node_private_t *nodepriv = xml->_private; if (pcmk_is_set(nodepriv->flags, pcmk__xf_created)) { if (implicitly_allowed(xml)) { crm_trace("Creation of <%s> scaffolding with id=\"%s\"" " is implicitly allowed", xml->name, display_id(xml)); } else if (pcmk__check_acl(xml, NULL, pcmk__xf_acl_write)) { crm_trace("ACLs allow creation of <%s> with id=\"%s\"", xml->name, display_id(xml)); } else if (check_top) { crm_trace("ACLs disallow creation of <%s> with id=\"%s\"", xml->name, display_id(xml)); pcmk_free_xml_subtree(xml); return; } else { crm_notice("ACLs would disallow creation of %s<%s> with id=\"%s\"", ((xml == xmlDocGetRootElement(xml->doc))? "root element " : ""), xml->name, display_id(xml)); } } for (xmlNode *cIter = pcmk__xml_first_child(xml); cIter != NULL; ) { xmlNode *child = cIter; cIter = pcmk__xml_next(cIter); /* In case it is free'd */ pcmk__apply_creation_acl(child, true); } } /*! * \brief Check whether or not an XML node is ACL-denied * * \param[in] xml node to check * * \return true if XML node exists and is ACL-denied, false otherwise */ bool xml_acl_denied(const xmlNode *xml) { if (xml && xml->doc && xml->doc->_private){ xml_doc_private_t *docpriv = xml->doc->_private; return pcmk_is_set(docpriv->flags, pcmk__xf_acl_denied); } return false; } void xml_acl_disable(xmlNode *xml) { if (xml_acl_enabled(xml)) { xml_doc_private_t *docpriv = xml->doc->_private; /* Catch anything that was created but shouldn't have been */ pcmk__apply_acl(xml); pcmk__apply_creation_acl(xml, false); pcmk__clear_xml_flags(docpriv, pcmk__xf_acl_enabled); } } /*! * \brief Check whether or not an XML node is ACL-enabled * * \param[in] xml node to check * * \return true if XML node exists and is ACL-enabled, false otherwise */ bool xml_acl_enabled(const xmlNode *xml) { if (xml && xml->doc && xml->doc->_private){ xml_doc_private_t *docpriv = xml->doc->_private; return pcmk_is_set(docpriv->flags, pcmk__xf_acl_enabled); } return false; } bool pcmk__check_acl(xmlNode *xml, const char *name, enum xml_private_flags mode) { CRM_ASSERT(xml); CRM_ASSERT(xml->doc); CRM_ASSERT(xml->doc->_private); if (pcmk__tracking_xml_changes(xml, false) && xml_acl_enabled(xml)) { xmlNode *parent = xml; xml_doc_private_t *docpriv = xml->doc->_private; GString *xpath = NULL; if (docpriv->acls == NULL) { pcmk__set_xml_doc_flag(xml, pcmk__xf_acl_denied); pcmk__if_tracing({}, return false); xpath = pcmk__element_xpath(xml); if (name != NULL) { pcmk__g_strcat(xpath, "[@", name, "]", NULL); } qb_log_from_external_source(__func__, __FILE__, "User '%s' without ACLs denied %s " "access to %s", LOG_TRACE, __LINE__, 0, docpriv->user, acl_to_text(mode), (const char *) xpath->str); g_string_free(xpath, TRUE); return false; } /* Walk the tree upwards looking for xml_acl_* flags * - Creating an attribute requires write permissions for the node * - Creating a child requires write permissions for the parent */ if (name) { xmlAttr *attr = xmlHasProp(xml, (pcmkXmlStr) name); if (attr && mode == pcmk__xf_acl_create) { mode = pcmk__xf_acl_write; } } while (parent && parent->_private) { xml_node_private_t *nodepriv = parent->_private; if (test_acl_mode(nodepriv->flags, mode)) { return true; } else if (pcmk_is_set(nodepriv->flags, pcmk__xf_acl_deny)) { pcmk__set_xml_doc_flag(xml, pcmk__xf_acl_denied); pcmk__if_tracing({}, return false); xpath = pcmk__element_xpath(xml); if (name != NULL) { pcmk__g_strcat(xpath, "[@", name, "]", NULL); } qb_log_from_external_source(__func__, __FILE__, "%sACL denies user '%s' %s access " "to %s", LOG_TRACE, __LINE__, 0, (parent != xml)? "Parent ": "", docpriv->user, acl_to_text(mode), (const char *) xpath->str); g_string_free(xpath, TRUE); return false; } parent = parent->parent; } pcmk__set_xml_doc_flag(xml, pcmk__xf_acl_denied); pcmk__if_tracing({}, return false); xpath = pcmk__element_xpath(xml); if (name != NULL) { pcmk__g_strcat(xpath, "[@", name, "]", NULL); } qb_log_from_external_source(__func__, __FILE__, "Default ACL denies user '%s' %s access to " "%s", LOG_TRACE, __LINE__, 0, docpriv->user, acl_to_text(mode), (const char *) xpath->str); g_string_free(xpath, TRUE); return false; } return true; } /*! * \brief Check whether ACLs are required for a given user * * \param[in] User name to check * * \return true if the user requires ACLs, false otherwise */ bool pcmk_acl_required(const char *user) { if (pcmk__str_empty(user)) { crm_trace("ACLs not required because no user set"); return false; } else if (!strcmp(user, CRM_DAEMON_USER) || !strcmp(user, "root")) { crm_trace("ACLs not required for privileged user %s", user); return false; } crm_trace("ACLs required for %s", user); return true; } char * pcmk__uid2username(uid_t uid) { char *result = NULL; struct passwd *pwent = getpwuid(uid); if (pwent == NULL) { crm_perror(LOG_INFO, "Cannot get user details for user ID %d", uid); return NULL; } pcmk__str_update(&result, pwent->pw_name); return result; } /*! * \internal * \brief Set the ACL user field properly on an XML request * * Multiple user names are potentially involved in an XML request: the effective * user of the current process; the user name known from an IPC client * connection; and the user name obtained from the request itself, whether by * the current standard XML attribute name or an older legacy attribute name. * This function chooses the appropriate one that should be used for ACLs, sets * it in the request (using the standard attribute name, and the legacy name if * given), and returns it. * * \param[in,out] request XML request to update * \param[in] field Alternate name for ACL user name XML attribute * \param[in] peer_user User name as known from IPC connection * * \return ACL user name actually used */ const char * pcmk__update_acl_user(xmlNode *request, const char *field, const char *peer_user) { static const char *effective_user = NULL; const char *requested_user = NULL; const char *user = NULL; if (effective_user == NULL) { effective_user = pcmk__uid2username(geteuid()); if (effective_user == NULL) { effective_user = strdup("#unprivileged"); CRM_CHECK(effective_user != NULL, return NULL); crm_err("Unable to determine effective user, assuming unprivileged for ACLs"); } } requested_user = crm_element_value(request, XML_ACL_TAG_USER); if (requested_user == NULL) { /* @COMPAT rolling upgrades <=1.1.11 * * field is checked for backward compatibility with older versions that * did not use XML_ACL_TAG_USER. */ requested_user = crm_element_value(request, field); } if (!pcmk__is_privileged(effective_user)) { /* We're not running as a privileged user, set or overwrite any existing * value for $XML_ACL_TAG_USER */ user = effective_user; } else if (peer_user == NULL && requested_user == NULL) { /* No user known or requested, use 'effective_user' and make sure one is * set for the request */ user = effective_user; } else if (peer_user == NULL) { /* No user known, trusting 'requested_user' */ user = requested_user; } else if (!pcmk__is_privileged(peer_user)) { /* The peer is not a privileged user, set or overwrite any existing * value for $XML_ACL_TAG_USER */ user = peer_user; } else if (requested_user == NULL) { /* Even if we're privileged, make sure there is always a value set */ user = peer_user; } else { /* Legal delegation to 'requested_user' */ user = requested_user; } // This requires pointer comparison, not string comparison if (user != crm_element_value(request, XML_ACL_TAG_USER)) { crm_xml_add(request, XML_ACL_TAG_USER, user); } if (field != NULL && user != crm_element_value(request, field)) { crm_xml_add(request, field, user); } return requested_user; } diff --git a/lib/common/ipc_controld.c b/lib/common/ipc_controld.c index 524bb443e8..765b6f4507 100644 --- a/lib/common/ipc_controld.c +++ b/lib/common/ipc_controld.c @@ -1,657 +1,657 @@ /* * Copyright 2020-2024 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 "crmcommon_private.h" struct controld_api_private_s { char *client_uuid; unsigned int replies_expected; }; /*! * \internal * \brief Get a string representation of a controller API reply type * * \param[in] reply Controller API reply type * * \return String representation of a controller API reply type */ const char * pcmk__controld_api_reply2str(enum pcmk_controld_api_reply reply) { switch (reply) { case pcmk_controld_reply_reprobe: return "reprobe"; case pcmk_controld_reply_info: return "info"; case pcmk_controld_reply_resource: return "resource"; case pcmk_controld_reply_ping: return "ping"; case pcmk_controld_reply_nodes: return "nodes"; default: return "unknown"; } } // \return Standard Pacemaker return code static int new_data(pcmk_ipc_api_t *api) { struct controld_api_private_s *private = NULL; api->api_data = calloc(1, sizeof(struct controld_api_private_s)); if (api->api_data == NULL) { return errno; } private = api->api_data; /* This is set to the PID because that's how it was always done, but PIDs * are not unique because clients can be remote. The value appears to be * unused other than as part of F_CRM_SYS_FROM in IPC requests, which is * only compared against the internal system names (CRM_SYSTEM_TENGINE, * etc.), so it shouldn't be a problem. */ private->client_uuid = pcmk__getpid_s(); /* @TODO Implement a call ID model similar to the CIB, executor, and fencer * IPC APIs, so that requests and replies can be matched, and * duplicate replies can be discarded. */ return pcmk_rc_ok; } static void free_data(void *data) { free(((struct controld_api_private_s *) data)->client_uuid); free(data); } // \return Standard Pacemaker return code static int post_connect(pcmk_ipc_api_t *api) { /* The controller currently requires clients to register via a hello * request, but does not reply back. */ struct controld_api_private_s *private = api->api_data; const char *client_name = crm_system_name? crm_system_name : "client"; xmlNode *hello; int rc; hello = create_hello_message(private->client_uuid, client_name, PCMK__CONTROLD_API_MAJOR, PCMK__CONTROLD_API_MINOR); rc = pcmk__send_ipc_request(api, hello); free_xml(hello); if (rc != pcmk_rc_ok) { crm_info("Could not send IPC hello to %s: %s " CRM_XS " rc=%s", pcmk_ipc_name(api, true), pcmk_rc_str(rc), rc); } else { crm_debug("Sent IPC hello to %s", pcmk_ipc_name(api, true)); } return rc; } static void set_node_info_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_info; if (msg_data == NULL) { return; } data->data.node_info.have_quorum = pcmk__xe_attr_is_true(msg_data, PCMK_XA_HAVE_QUORUM); data->data.node_info.is_remote = pcmk__xe_attr_is_true(msg_data, XML_NODE_IS_REMOTE); /* Integer node_info.id is currently valid only for Corosync nodes. * * @TODO: Improve handling after crm_node_t is refactored to handle layer- * specific data better. */ crm_element_value_int(msg_data, PCMK_XA_ID, &(data->data.node_info.id)); data->data.node_info.uuid = crm_element_value(msg_data, PCMK_XA_ID); data->data.node_info.uname = crm_element_value(msg_data, PCMK_XA_UNAME); data->data.node_info.state = crm_element_value(msg_data, PCMK__XA_CRMD); } static void set_ping_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_ping; if (msg_data == NULL) { return; } data->data.ping.sys_from = crm_element_value(msg_data, XML_PING_ATTR_SYSFROM); data->data.ping.fsa_state = crm_element_value(msg_data, XML_PING_ATTR_CRMDSTATE); data->data.ping.result = crm_element_value(msg_data, XML_PING_ATTR_STATUS); } static void set_nodes_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { pcmk_controld_api_node_t *node_info; data->reply_type = pcmk_controld_reply_nodes; for (xmlNode *node = first_named_child(msg_data, XML_CIB_TAG_NODE); node != NULL; node = crm_next_same_xml(node)) { long long id_ll = 0; node_info = calloc(1, sizeof(pcmk_controld_api_node_t)); crm_element_value_ll(node, PCMK_XA_ID, &id_ll); if (id_ll > 0) { node_info->id = id_ll; } node_info->uname = crm_element_value(node, PCMK_XA_UNAME); node_info->state = crm_element_value(node, PCMK__XA_IN_CCM); data->data.nodes = g_list_prepend(data->data.nodes, node_info); } } static bool reply_expected(pcmk_ipc_api_t *api, const xmlNode *request) { // We only need to handle commands that API functions can send return pcmk__str_any_of(crm_element_value(request, F_CRM_TASK), PCMK__CONTROLD_CMD_NODES, CRM_OP_LRM_DELETE, CRM_OP_LRM_FAIL, CRM_OP_NODE_INFO, CRM_OP_PING, CRM_OP_REPROBE, CRM_OP_RM_NODE_CACHE, NULL); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { struct controld_api_private_s *private = api->api_data; crm_exit_t status = CRM_EX_OK; xmlNode *msg_data = NULL; const char *value = NULL; pcmk_controld_api_reply_t reply_data = { pcmk_controld_reply_unknown, NULL, NULL, }; if (pcmk__xe_is(reply, "ack")) { /* ACKs are trivial responses that do not count toward expected replies, * and do not have all the fields that validation requires, so skip that * processing. */ return private->replies_expected > 0; } if (private->replies_expected > 0) { private->replies_expected--; } // Do some basic validation of the reply /* @TODO We should be able to verify that value is always a response, but * currently the controller doesn't always properly set the type. Even * if we fix the controller, we'll still need to handle replies from * old versions (feature set could be used to differentiate). */ value = crm_element_value(reply, F_CRM_MSG_TYPE); if (!pcmk__str_any_of(value, PCMK__VALUE_REQUEST, PCMK__VALUE_RESPONSE, NULL)) { crm_info("Unrecognizable message from controller: " "invalid message type '%s'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } - if (pcmk__str_empty(crm_element_value(reply, XML_ATTR_REFERENCE))) { + if (pcmk__str_empty(crm_element_value(reply, PCMK_XA_REFERENCE))) { crm_info("Unrecognizable message from controller: no reference"); status = CRM_EX_PROTOCOL; goto done; } value = crm_element_value(reply, F_CRM_TASK); if (pcmk__str_empty(value)) { crm_info("Unrecognizable message from controller: no command name"); status = CRM_EX_PROTOCOL; goto done; } // Parse useful info from reply reply_data.feature_set = crm_element_value(reply, PCMK_XA_VERSION); reply_data.host_from = crm_element_value(reply, F_CRM_HOST_FROM); msg_data = get_message_xml(reply, F_CRM_DATA); if (!strcmp(value, CRM_OP_REPROBE)) { reply_data.reply_type = pcmk_controld_reply_reprobe; } else if (!strcmp(value, CRM_OP_NODE_INFO)) { set_node_info_data(&reply_data, msg_data); } else if (!strcmp(value, CRM_OP_INVOKE_LRM)) { reply_data.reply_type = pcmk_controld_reply_resource; reply_data.data.resource.node_state = msg_data; } else if (!strcmp(value, CRM_OP_PING)) { set_ping_data(&reply_data, msg_data); } else if (!strcmp(value, PCMK__CONTROLD_CMD_NODES)) { set_nodes_data(&reply_data, msg_data); } else { crm_info("Unrecognizable message from controller: unknown command '%s'", value); status = CRM_EX_PROTOCOL; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); // Free any reply data that was allocated if (pcmk__str_eq(value, PCMK__CONTROLD_CMD_NODES, pcmk__str_casei)) { g_list_free_full(reply_data.data.nodes, free); } return false; // No further replies needed } pcmk__ipc_methods_t * pcmk__controld_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = new_data; cmds->free_data = free_data; cmds->post_connect = post_connect; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; } return cmds; } /*! * \internal * \brief Create XML for a controller IPC request * * \param[in] api Controller connection * \param[in] op Controller IPC command name * \param[in] node Node name to set as destination host * \param[in] msg_data XML to attach to request as message data * * \return Newly allocated XML for request */ static xmlNode * create_controller_request(const pcmk_ipc_api_t *api, const char *op, const char *node, xmlNode *msg_data) { struct controld_api_private_s *private = NULL; const char *sys_to = NULL; if (api == NULL) { return NULL; } private = api->api_data; if ((node == NULL) && !strcmp(op, CRM_OP_PING)) { sys_to = CRM_SYSTEM_DC; } else { sys_to = CRM_SYSTEM_CRMD; } return create_request(op, msg_data, node, sys_to, (crm_system_name? crm_system_name : "client"), private->client_uuid); } // \return Standard Pacemaker return code static int send_controller_request(pcmk_ipc_api_t *api, const xmlNode *request, bool reply_is_expected) { - if (crm_element_value(request, XML_ATTR_REFERENCE) == NULL) { + if (crm_element_value(request, PCMK_XA_REFERENCE) == NULL) { return EINVAL; } if (reply_is_expected) { struct controld_api_private_s *private = api->api_data; private->replies_expected++; } return pcmk__send_ipc_request(api, request); } static xmlNode * create_reprobe_message_data(const char *target_node, const char *router_node) { xmlNode *msg_data; msg_data = create_xml_node(NULL, "data_for_" CRM_OP_REPROBE); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if ((router_node != NULL) && !pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } return msg_data; } /*! * \brief Send a reprobe controller operation * * \param[in,out] api Controller connection * \param[in] target_node Name of node to reprobe * \param[in] router_node Router node for host * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_reprobe. */ int pcmk_controld_api_reprobe(pcmk_ipc_api_t *api, const char *target_node, const char *router_node) { xmlNode *request; xmlNode *msg_data; int rc = pcmk_rc_ok; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } crm_debug("Sending %s IPC request to reprobe %s via %s", pcmk_ipc_name(api, true), pcmk__s(target_node, "local node"), pcmk__s(router_node, "local node")); msg_data = create_reprobe_message_data(target_node, router_node); request = create_controller_request(api, CRM_OP_REPROBE, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Send a "node info" controller operation * * \param[in,out] api Controller connection * \param[in] nodeid ID of node to get info for (or 0 for local node) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_info. */ int pcmk_controld_api_node_info(pcmk_ipc_api_t *api, uint32_t nodeid) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_NODE_INFO, NULL, NULL); if (request == NULL) { return EINVAL; } if (nodeid > 0) { crm_xml_set_id(request, "%lu", (unsigned long) nodeid); } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for status * * \param[in,out] api Controller connection * \param[in] node_name Name of node whose status is desired (NULL for DC) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_ping. */ int pcmk_controld_api_ping(pcmk_ipc_api_t *api, const char *node_name) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_PING, node_name, NULL); if (request == NULL) { return EINVAL; } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for cluster information * * \param[in,out] api Controller connection * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_nodes. */ int pcmk_controld_api_list_nodes(pcmk_ipc_api_t *api) { xmlNode *request; int rc = EINVAL; request = create_controller_request(api, PCMK__CONTROLD_CMD_NODES, NULL, NULL); if (request != NULL) { rc = send_controller_request(api, request, true); free_xml(request); } return rc; } // \return Standard Pacemaker return code static int controller_resource_op(pcmk_ipc_api_t *api, const char *op, const char *target_node, const char *router_node, bool cib_only, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { int rc = pcmk_rc_ok; char *key; xmlNode *request, *msg_data, *xml_rsc, *params; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } msg_data = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); /* The controller logs the transition key from resource op requests, so we * need to have *something* for it. * @TODO don't use "crm-resource" */ key = pcmk__transition_key(0, getpid(), 0, "xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx"); crm_xml_add(msg_data, XML_ATTR_TRANSITION_KEY, key); free(key); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if (!pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } if (cib_only) { // Indicate that only the CIB needs to be cleaned crm_xml_add(msg_data, PCMK__XA_MODE, XML_TAG_CIB); } xml_rsc = create_xml_node(msg_data, XML_CIB_TAG_RESOURCE); crm_xml_add(xml_rsc, PCMK_XA_ID, rsc_id); crm_xml_add(xml_rsc, PCMK__XA_LONG_ID, rsc_long_id); crm_xml_add(xml_rsc, PCMK_XA_CLASS, standard); crm_xml_add(xml_rsc, PCMK_XA_PROVIDER, provider); crm_xml_add(xml_rsc, PCMK_XA_TYPE, type); params = create_xml_node(msg_data, XML_TAG_ATTRS); crm_xml_add(params, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET); // The controller parses the timeout from the request key = crm_meta_name(XML_ATTR_TIMEOUT); crm_xml_add(params, key, "60000"); /* 1 minute */ //@TODO pass as arg free(key); request = create_controller_request(api, op, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Ask the controller to fail a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to fail * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource to fail * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_fail(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { crm_debug("Sending %s IPC request to fail %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_FAIL, target_node, router_node, false, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Ask the controller to refresh a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to refresh * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource * \param[in] cib_only If true, clean resource from CIB only * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_refresh(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type, bool cib_only) { crm_debug("Sending %s IPC request to refresh %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_DELETE, target_node, router_node, cib_only, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Get the number of IPC replies currently expected from the controller * * \param[in] api Controller IPC API connection * * \return Number of replies expected */ unsigned int pcmk_controld_api_replies_expected(const pcmk_ipc_api_t *api) { struct controld_api_private_s *private = api->api_data; return private->replies_expected; } /*! * \brief Create XML for a controller IPC "hello" message * * \deprecated This function is deprecated as part of the public C API. */ // \todo make this static to this file when breaking API backward compatibility xmlNode * create_hello_message(const char *uuid, const char *client_name, const char *major_version, const char *minor_version) { xmlNode *hello_node = NULL; xmlNode *hello = NULL; if (pcmk__str_empty(uuid) || pcmk__str_empty(client_name) || pcmk__str_empty(major_version) || pcmk__str_empty(minor_version)) { crm_err("Could not create IPC hello message from %s (UUID %s): " "missing information", client_name? client_name : "unknown client", uuid? uuid : "unknown"); return NULL; } hello_node = create_xml_node(NULL, XML_TAG_OPTIONS); if (hello_node == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Message data creation failed", client_name, uuid); return NULL; } crm_xml_add(hello_node, "major_version", major_version); crm_xml_add(hello_node, "minor_version", minor_version); crm_xml_add(hello_node, "client_name", client_name); crm_xml_add(hello_node, "client_uuid", uuid); hello = create_request(CRM_OP_HELLO, hello_node, NULL, NULL, client_name, uuid); if (hello == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Request creation failed", client_name, uuid); return NULL; } free_xml(hello_node); crm_trace("Created hello message from %s (UUID %s)", client_name, uuid); return hello; } diff --git a/lib/common/ipc_pacemakerd.c b/lib/common/ipc_pacemakerd.c index 8994a6a152..9c5fd779d1 100644 --- a/lib/common/ipc_pacemakerd.c +++ b/lib/common/ipc_pacemakerd.c @@ -1,316 +1,316 @@ /* * Copyright 2020-2024 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 "crmcommon_private.h" typedef struct pacemakerd_api_private_s { enum pcmk_pacemakerd_state state; char *client_uuid; } pacemakerd_api_private_t; static const char *pacemakerd_state_str[] = { XML_PING_ATTR_PACEMAKERDSTATE_INIT, XML_PING_ATTR_PACEMAKERDSTATE_STARTINGDAEMONS, XML_PING_ATTR_PACEMAKERDSTATE_WAITPING, XML_PING_ATTR_PACEMAKERDSTATE_RUNNING, XML_PING_ATTR_PACEMAKERDSTATE_SHUTTINGDOWN, XML_PING_ATTR_PACEMAKERDSTATE_SHUTDOWNCOMPLETE, XML_PING_ATTR_PACEMAKERDSTATE_REMOTE, }; enum pcmk_pacemakerd_state pcmk_pacemakerd_api_daemon_state_text2enum(const char *state) { int i; if (state == NULL) { return pcmk_pacemakerd_state_invalid; } for (i=pcmk_pacemakerd_state_init; i <= pcmk_pacemakerd_state_max; i++) { if (pcmk__str_eq(state, pacemakerd_state_str[i], pcmk__str_none)) { return i; } } return pcmk_pacemakerd_state_invalid; } const char * pcmk_pacemakerd_api_daemon_state_enum2text( enum pcmk_pacemakerd_state state) { if ((state >= pcmk_pacemakerd_state_init) && (state <= pcmk_pacemakerd_state_max)) { return pacemakerd_state_str[state]; } return "invalid"; } /*! * \internal * \brief Return a friendly string representation of a \p pacemakerd state * * \param[in] state \p pacemakerd state * * \return A user-friendly string representation of \p state, or * "Invalid pacemakerd state" */ const char * pcmk__pcmkd_state_enum2friendly(enum pcmk_pacemakerd_state state) { switch (state) { case pcmk_pacemakerd_state_init: return "Initializing pacemaker"; case pcmk_pacemakerd_state_starting_daemons: return "Pacemaker daemons are starting"; case pcmk_pacemakerd_state_wait_for_ping: return "Waiting for startup trigger from SBD"; case pcmk_pacemakerd_state_running: return "Pacemaker is running"; case pcmk_pacemakerd_state_shutting_down: return "Pacemaker daemons are shutting down"; case pcmk_pacemakerd_state_shutdown_complete: /* Assuming pacemakerd won't process messages while in * shutdown_complete state unless reporting to SBD */ return "Pacemaker daemons are shut down (reporting to SBD)"; case pcmk_pacemakerd_state_remote: return "pacemaker-remoted is running (on a Pacemaker Remote node)"; default: return "Invalid pacemakerd state"; } } /*! * \internal * \brief Get a string representation of a \p pacemakerd API reply type * * \param[in] reply \p pacemakerd API reply type * * \return String representation of a \p pacemakerd API reply type */ const char * pcmk__pcmkd_api_reply2str(enum pcmk_pacemakerd_api_reply reply) { switch (reply) { case pcmk_pacemakerd_reply_ping: return "ping"; case pcmk_pacemakerd_reply_shutdown: return "shutdown"; default: return "unknown"; } } // \return Standard Pacemaker return code static int new_data(pcmk_ipc_api_t *api) { struct pacemakerd_api_private_s *private = NULL; api->api_data = calloc(1, sizeof(struct pacemakerd_api_private_s)); if (api->api_data == NULL) { return errno; } private = api->api_data; private->state = pcmk_pacemakerd_state_invalid; /* other as with cib, controld, ... we are addressing pacemakerd just from the local node -> pid is unique and thus sufficient as an ID */ private->client_uuid = pcmk__getpid_s(); return pcmk_rc_ok; } static void free_data(void *data) { free(((struct pacemakerd_api_private_s *) data)->client_uuid); free(data); } // \return Standard Pacemaker return code static int post_connect(pcmk_ipc_api_t *api) { struct pacemakerd_api_private_s *private = NULL; if (api->api_data == NULL) { return EINVAL; } private = api->api_data; private->state = pcmk_pacemakerd_state_invalid; return pcmk_rc_ok; } static void post_disconnect(pcmk_ipc_api_t *api) { struct pacemakerd_api_private_s *private = NULL; if (api->api_data == NULL) { return; } private = api->api_data; private->state = pcmk_pacemakerd_state_invalid; return; } static bool reply_expected(pcmk_ipc_api_t *api, const xmlNode *request) { const char *command = crm_element_value(request, F_CRM_TASK); if (command == NULL) { return false; } // We only need to handle commands that functions in this file can send return pcmk__str_any_of(command, CRM_OP_PING, CRM_OP_QUIT, NULL); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { crm_exit_t status = CRM_EX_OK; xmlNode *msg_data = NULL; pcmk_pacemakerd_api_reply_t reply_data = { pcmk_pacemakerd_reply_unknown }; const char *value = NULL; long long value_ll = 0; if (pcmk__str_eq((const char *) reply->name, "ack", pcmk__str_none)) { long long int ack_status = 0; pcmk__scan_ll(crm_element_value(reply, "status"), &ack_status, CRM_EX_OK); return ack_status == CRM_EX_INDETERMINATE; } value = crm_element_value(reply, F_CRM_MSG_TYPE); if (!pcmk__str_eq(value, PCMK__VALUE_RESPONSE, pcmk__str_none)) { crm_info("Unrecognizable message from %s: " "message type '%s' not '" PCMK__VALUE_RESPONSE "'", pcmk_ipc_name(api, true), pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } - if (pcmk__str_empty(crm_element_value(reply, XML_ATTR_REFERENCE))) { + if (pcmk__str_empty(crm_element_value(reply, PCMK_XA_REFERENCE))) { crm_info("Unrecognizable message from %s: no reference", pcmk_ipc_name(api, true)); status = CRM_EX_PROTOCOL; goto done; } value = crm_element_value(reply, F_CRM_TASK); // Parse useful info from reply msg_data = get_message_xml(reply, F_CRM_DATA); crm_element_value_ll(msg_data, PCMK_XA_CRM_TIMESTAMP, &value_ll); if (pcmk__str_eq(value, CRM_OP_PING, pcmk__str_none)) { reply_data.reply_type = pcmk_pacemakerd_reply_ping; reply_data.data.ping.state = pcmk_pacemakerd_api_daemon_state_text2enum( crm_element_value(msg_data, XML_PING_ATTR_PACEMAKERDSTATE)); reply_data.data.ping.status = pcmk__str_eq(crm_element_value(msg_data, XML_PING_ATTR_STATUS), "ok", pcmk__str_casei)?pcmk_rc_ok:pcmk_rc_error; reply_data.data.ping.last_good = (value_ll < 0)? 0 : (time_t) value_ll; reply_data.data.ping.sys_from = crm_element_value(msg_data, XML_PING_ATTR_SYSFROM); } else if (pcmk__str_eq(value, CRM_OP_QUIT, pcmk__str_none)) { reply_data.reply_type = pcmk_pacemakerd_reply_shutdown; reply_data.data.shutdown.status = atoi(crm_element_value(msg_data, XML_LRM_ATTR_OPSTATUS)); } else { crm_info("Unrecognizable message from %s: unknown command '%s'", pcmk_ipc_name(api, true), pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); return false; } pcmk__ipc_methods_t * pcmk__pacemakerd_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = new_data; cmds->free_data = free_data; cmds->post_connect = post_connect; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; cmds->post_disconnect = post_disconnect; } return cmds; } static int do_pacemakerd_api_call(pcmk_ipc_api_t *api, const char *ipc_name, const char *task) { pacemakerd_api_private_t *private; xmlNode *cmd; int rc; if (api == NULL) { return EINVAL; } private = api->api_data; CRM_ASSERT(private != NULL); cmd = create_request(task, NULL, NULL, CRM_SYSTEM_MCP, pcmk__ipc_sys_name(ipc_name, "client"), private->client_uuid); if (cmd) { rc = pcmk__send_ipc_request(api, cmd); if (rc != pcmk_rc_ok) { crm_debug("Couldn't send request to %s: %s rc=%d", pcmk_ipc_name(api, true), pcmk_rc_str(rc), rc); } free_xml(cmd); } else { rc = ENOMSG; } return rc; } int pcmk_pacemakerd_api_ping(pcmk_ipc_api_t *api, const char *ipc_name) { return do_pacemakerd_api_call(api, ipc_name, CRM_OP_PING); } int pcmk_pacemakerd_api_shutdown(pcmk_ipc_api_t *api, const char *ipc_name) { return do_pacemakerd_api_call(api, ipc_name, CRM_OP_QUIT); } diff --git a/lib/common/ipc_schedulerd.c b/lib/common/ipc_schedulerd.c index 65033991c5..179d94cbef 100644 --- a/lib/common/ipc_schedulerd.c +++ b/lib/common/ipc_schedulerd.c @@ -1,180 +1,181 @@ /* - * Copyright 2021-2023 the Pacemaker project contributors + * Copyright 2021-2024 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 "crmcommon_private.h" typedef struct schedulerd_api_private_s { char *client_uuid; } schedulerd_api_private_t; // \return Standard Pacemaker return code static int new_data(pcmk_ipc_api_t *api) { struct schedulerd_api_private_s *private = NULL; api->api_data = calloc(1, sizeof(struct schedulerd_api_private_s)); if (api->api_data == NULL) { return errno; } private = api->api_data; /* See comments in ipc_pacemakerd.c. */ private->client_uuid = pcmk__getpid_s(); return pcmk_rc_ok; } static void free_data(void *data) { free(((struct schedulerd_api_private_s *) data)->client_uuid); free(data); } // \return Standard Pacemaker return code static int post_connect(pcmk_ipc_api_t *api) { if (api->api_data == NULL) { return EINVAL; } return pcmk_rc_ok; } static bool reply_expected(pcmk_ipc_api_t *api, const xmlNode *request) { const char *command = crm_element_value(request, F_CRM_TASK); if (command == NULL) { return false; } // We only need to handle commands that functions in this file can send return pcmk__str_any_of(command, CRM_OP_PECALC, NULL); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { crm_exit_t status = CRM_EX_OK; xmlNode *msg_data = NULL; pcmk_schedulerd_api_reply_t reply_data = { pcmk_schedulerd_reply_unknown }; const char *value = NULL; if (pcmk__str_eq((const char *) reply->name, "ack", pcmk__str_casei)) { return false; } value = crm_element_value(reply, F_CRM_MSG_TYPE); if (!pcmk__str_eq(value, PCMK__VALUE_RESPONSE, pcmk__str_none)) { crm_info("Unrecognizable message from schedulerd: " "message type '%s' not '" PCMK__VALUE_RESPONSE "'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } - if (pcmk__str_empty(crm_element_value(reply, XML_ATTR_REFERENCE))) { + if (pcmk__str_empty(crm_element_value(reply, PCMK_XA_REFERENCE))) { crm_info("Unrecognizable message from schedulerd: no reference"); status = CRM_EX_PROTOCOL; goto done; } // Parse useful info from reply msg_data = get_message_xml(reply, F_CRM_DATA); value = crm_element_value(reply, F_CRM_TASK); if (pcmk__str_eq(value, CRM_OP_PECALC, pcmk__str_none)) { reply_data.reply_type = pcmk_schedulerd_reply_graph; - reply_data.data.graph.reference = crm_element_value(reply, XML_ATTR_REFERENCE); + reply_data.data.graph.reference = crm_element_value(reply, + PCMK_XA_REFERENCE); reply_data.data.graph.input = crm_element_value(reply, F_CRM_TGRAPH_INPUT); reply_data.data.graph.tgraph = msg_data; } else { crm_info("Unrecognizable message from schedulerd: " "unknown command '%s'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); return false; } pcmk__ipc_methods_t * pcmk__schedulerd_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = new_data; cmds->free_data = free_data; cmds->post_connect = post_connect; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; } return cmds; } static int do_schedulerd_api_call(pcmk_ipc_api_t *api, const char *task, xmlNode *cib, char **ref) { schedulerd_api_private_t *private; xmlNode *cmd = NULL; int rc; if (!pcmk_ipc_is_connected(api)) { return ENOTCONN; } private = api->api_data; CRM_ASSERT(private != NULL); cmd = create_request(task, cib, NULL, CRM_SYSTEM_PENGINE, crm_system_name? crm_system_name : "client", private->client_uuid); if (cmd) { rc = pcmk__send_ipc_request(api, cmd); if (rc != pcmk_rc_ok) { crm_debug("Couldn't send request to schedulerd: %s rc=%d", pcmk_rc_str(rc), rc); } - *ref = strdup(crm_element_value(cmd, F_CRM_REFERENCE)); + *ref = strdup(crm_element_value(cmd, PCMK_XA_REFERENCE)); free_xml(cmd); } else { rc = ENOMSG; } return rc; } int pcmk_schedulerd_api_graph(pcmk_ipc_api_t *api, xmlNode *cib, char **ref) { return do_schedulerd_api_call(api, CRM_OP_PECALC, cib, ref); } diff --git a/lib/common/messages.c b/lib/common/messages.c index f227753ccf..99c64e0b6e 100644 --- a/lib/common/messages.c +++ b/lib/common/messages.c @@ -1,293 +1,294 @@ /* * Copyright 2004-2024 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 /*! * \brief Create a Pacemaker request (for IPC or cluster layer) * * \param[in] task What to set as the request's task * \param[in] msg_data What to add as the request's data contents * \param[in] host_to What to set as the request's destination host * \param[in] sys_to What to set as the request's destination system * \param[in] sys_from If not NULL, set as request's origin system * \param[in] uuid_from If not NULL, use in request's origin system * \param[in] origin Name of function that called this one * * \return XML of new request * * \note One of sys_from or uuid_from must be non-NULL * \note This function should not be called directly, but via the * create_request() wrapper. * \note The caller is responsible for freeing the result using free_xml(). */ xmlNode * create_request_adv(const char *task, xmlNode *msg_data, const char *host_to, const char *sys_to, const char *sys_from, const char *uuid_from, const char *origin) { static uint ref_counter = 0; char *true_from = NULL; xmlNode *request = NULL; char *reference = crm_strdup_printf("%s-%s-%lld-%u", (task? task : "_empty_"), (sys_from? sys_from : "_empty_"), (long long) time(NULL), ref_counter++); if (uuid_from != NULL) { true_from = crm_strdup_printf("%s_%s", uuid_from, (sys_from? sys_from : "none")); } else if (sys_from != NULL) { true_from = strdup(sys_from); } else { crm_err("Cannot create IPC request: No originating system specified"); } // host_from will get set for us if necessary by the controller when routed request = create_xml_node(NULL, __func__); crm_xml_add(request, F_CRM_ORIGIN, origin); crm_xml_add(request, F_TYPE, T_CRM); crm_xml_add(request, F_CRM_VERSION, CRM_FEATURE_SET); crm_xml_add(request, F_CRM_MSG_TYPE, PCMK__VALUE_REQUEST); - crm_xml_add(request, F_CRM_REFERENCE, reference); + crm_xml_add(request, PCMK_XA_REFERENCE, reference); crm_xml_add(request, F_CRM_TASK, task); crm_xml_add(request, F_CRM_SYS_TO, sys_to); crm_xml_add(request, F_CRM_SYS_FROM, true_from); /* HOSTTO will be ignored if it is to the DC anyway. */ if (host_to != NULL && strlen(host_to) > 0) { crm_xml_add(request, F_CRM_HOST_TO, host_to); } if (msg_data != NULL) { add_message_xml(request, F_CRM_DATA, msg_data); } free(reference); free(true_from); return request; } /*! * \brief Create a Pacemaker reply (for IPC or cluster layer) * * \param[in] original_request XML of request this is a reply to * \param[in] xml_response_data XML to copy as data section of reply * \param[in] origin Name of function that called this one * * \return XML of new reply * * \note This function should not be called directly, but via the * create_reply() wrapper. * \note The caller is responsible for freeing the result using free_xml(). */ xmlNode * create_reply_adv(const xmlNode *original_request, xmlNode *xml_response_data, const char *origin) { xmlNode *reply = NULL; const char *host_from = crm_element_value(original_request, F_CRM_HOST_FROM); const char *sys_from = crm_element_value(original_request, F_CRM_SYS_FROM); const char *sys_to = crm_element_value(original_request, F_CRM_SYS_TO); const char *type = crm_element_value(original_request, F_CRM_MSG_TYPE); const char *operation = crm_element_value(original_request, F_CRM_TASK); - const char *crm_msg_reference = crm_element_value(original_request, F_CRM_REFERENCE); + const char *crm_msg_reference = crm_element_value(original_request, + PCMK_XA_REFERENCE); if (type == NULL) { crm_err("Cannot create new_message, no message type in original message"); CRM_ASSERT(type != NULL); return NULL; } if (strcmp(type, PCMK__VALUE_REQUEST) != 0) { /* Replies should only be generated for request messages, but it's possible * we expect replies to other messages right now so this can't be enforced. */ crm_trace("Creating a reply for a non-request original message"); } reply = create_xml_node(NULL, __func__); if (reply == NULL) { crm_err("Cannot create new_message, malloc failed"); return NULL; } crm_xml_add(reply, F_CRM_ORIGIN, origin); crm_xml_add(reply, F_TYPE, T_CRM); crm_xml_add(reply, F_CRM_VERSION, CRM_FEATURE_SET); crm_xml_add(reply, F_CRM_MSG_TYPE, PCMK__VALUE_RESPONSE); - crm_xml_add(reply, F_CRM_REFERENCE, crm_msg_reference); + crm_xml_add(reply, PCMK_XA_REFERENCE, crm_msg_reference); crm_xml_add(reply, F_CRM_TASK, operation); /* since this is a reply, we reverse the from and to */ crm_xml_add(reply, F_CRM_SYS_TO, sys_from); crm_xml_add(reply, F_CRM_SYS_FROM, sys_to); /* HOSTTO will be ignored if it is to the DC anyway. */ if (host_from != NULL && strlen(host_from) > 0) { crm_xml_add(reply, F_CRM_HOST_TO, host_from); } if (xml_response_data != NULL) { add_message_xml(reply, F_CRM_DATA, xml_response_data); } return reply; } xmlNode * get_message_xml(const xmlNode *msg, const char *field) { return pcmk__xml_first_child(first_named_child(msg, field)); } gboolean add_message_xml(xmlNode *msg, const char *field, xmlNode *xml) { xmlNode *holder = create_xml_node(msg, field); add_node_copy(holder, xml); return TRUE; } /*! * \brief Get name to be used as identifier for cluster messages * * \param[in] name Actual system name to check * * \return Non-NULL cluster message identifier corresponding to name * * \note The Pacemaker daemons were renamed in version 2.0.0, but the old names * must continue to be used as the identifier for cluster messages, so * that mixed-version clusters are possible during a rolling upgrade. */ const char * pcmk__message_name(const char *name) { if (name == NULL) { return "unknown"; } else if (!strcmp(name, "pacemaker-attrd")) { return "attrd"; } else if (!strcmp(name, "pacemaker-based")) { return CRM_SYSTEM_CIB; } else if (!strcmp(name, "pacemaker-controld")) { return CRM_SYSTEM_CRMD; } else if (!strcmp(name, "pacemaker-execd")) { return CRM_SYSTEM_LRMD; } else if (!strcmp(name, "pacemaker-fenced")) { return "stonith-ng"; } else if (!strcmp(name, "pacemaker-schedulerd")) { return CRM_SYSTEM_PENGINE; } else { return name; } } /*! * \internal * \brief Register handlers for server commands * * \param[in] handlers Array of handler functions for supported server commands * (the final entry must have a NULL command name, and if * it has a handler it will be used as the default handler * for unrecognized commands) * * \return Newly created hash table with commands and handlers * \note The caller is responsible for freeing the return value with * g_hash_table_destroy(). */ GHashTable * pcmk__register_handlers(const pcmk__server_command_t handlers[]) { GHashTable *commands = g_hash_table_new(g_str_hash, g_str_equal); if (handlers != NULL) { int i; for (i = 0; handlers[i].command != NULL; ++i) { g_hash_table_insert(commands, (gpointer) handlers[i].command, handlers[i].handler); } if (handlers[i].handler != NULL) { // g_str_hash() can't handle NULL, so use empty string for default g_hash_table_insert(commands, (gpointer) "", handlers[i].handler); } } return commands; } /*! * \internal * \brief Process an incoming request * * \param[in,out] request Request to process * \param[in] handlers Command table created by pcmk__register_handlers() * * \return XML to send as reply (or NULL if no reply is needed) */ xmlNode * pcmk__process_request(pcmk__request_t *request, GHashTable *handlers) { xmlNode *(*handler)(pcmk__request_t *request) = NULL; CRM_CHECK((request != NULL) && (request->op != NULL) && (handlers != NULL), return NULL); if (pcmk_is_set(request->flags, pcmk__request_sync) && (request->ipc_client != NULL)) { CRM_CHECK(request->ipc_client->request_id == request->ipc_id, return NULL); } handler = g_hash_table_lookup(handlers, request->op); if (handler == NULL) { handler = g_hash_table_lookup(handlers, ""); // Default handler if (handler == NULL) { crm_info("Ignoring %s request from %s %s with no handler", request->op, pcmk__request_origin_type(request), pcmk__request_origin(request)); return NULL; } } return (*handler)(request); } /*! * \internal * \brief Free memory used within a request (but not the request itself) * * \param[in,out] request Request to reset */ void pcmk__reset_request(pcmk__request_t *request) { free(request->op); request->op = NULL; pcmk__reset_result(&(request->result)); }