/* ctdb lock handling provide API to do non-blocking locks for single or all databases Copyright (C) Amitay Isaacs 2012 This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, see . */ #include "replace.h" #include "system/filesys.h" #include "system/network.h" #include #include #include "lib/tdb_wrap/tdb_wrap.h" #include "lib/util/dlinklist.h" #include "lib/util/debug.h" #include "lib/util/samba_util.h" #include "ctdb_private.h" #include "common/system.h" #include "common/common.h" #include "common/logging.h" /* * Non-blocking Locking API * * 1. Create a child process to do blocking locks. * 2. Once the locks are obtained, signal parent process via fd. * 3. Invoke registered callback routine with locking status. * 4. If the child process cannot get locks within certain time, * execute an external script to debug. * * ctdb_lock_record() - get a lock on a record * ctdb_lock_db() - get a lock on a DB * * auto_mark - whether to mark/unmark DBs in before/after callback * = false is used for freezing databases for * recovery since the recovery cannot start till * databases are locked on all the nodes. * = true is used for record locks. */ enum lock_type { LOCK_RECORD, LOCK_DB, }; static const char * const lock_type_str[] = { "lock_record", "lock_db", }; struct lock_request; /* lock_context is the common part for a lock request */ struct lock_context { struct lock_context *next, *prev; enum lock_type type; struct ctdb_context *ctdb; struct ctdb_db_context *ctdb_db; TDB_DATA key; uint32_t priority; bool auto_mark; struct lock_request *request; pid_t child; int fd[2]; struct tevent_fd *tfd; struct tevent_timer *ttimer; struct timeval start_time; uint32_t key_hash; bool can_schedule; }; /* lock_request is the client specific part for a lock request */ struct lock_request { struct lock_context *lctx; void (*callback)(void *, bool); void *private_data; }; int ctdb_db_iterator(struct ctdb_context *ctdb, ctdb_db_handler_t handler, void *private_data) { struct ctdb_db_context *ctdb_db; int ret; for (ctdb_db = ctdb->db_list; ctdb_db; ctdb_db = ctdb_db->next) { ret = handler(ctdb_db, private_data); if (ret != 0) { return -1; } } return 0; } /* * lock all databases - mark only */ static int db_lock_mark_handler(struct ctdb_db_context *ctdb_db, void *private_data) { int tdb_transaction_write_lock_mark(struct tdb_context *); DEBUG(DEBUG_INFO, ("marking locked database %s\n", ctdb_db->db_name)); if (tdb_transaction_write_lock_mark(ctdb_db->ltdb->tdb) != 0) { DEBUG(DEBUG_ERR, ("Failed to mark (transaction lock) database %s\n", ctdb_db->db_name)); return -1; } if (tdb_lockall_mark(ctdb_db->ltdb->tdb) != 0) { DEBUG(DEBUG_ERR, ("Failed to mark (all lock) database %s\n", ctdb_db->db_name)); return -1; } return 0; } int ctdb_lockdb_mark(struct ctdb_db_context *ctdb_db) { if (!ctdb_db_frozen(ctdb_db)) { DEBUG(DEBUG_ERR, ("Attempt to mark database locked when not frozen\n")); return -1; } return db_lock_mark_handler(ctdb_db, NULL); } /* * lock all databases - unmark only */ static int db_lock_unmark_handler(struct ctdb_db_context *ctdb_db, void *private_data) { int tdb_transaction_write_lock_unmark(struct tdb_context *); DEBUG(DEBUG_INFO, ("unmarking locked database %s\n", ctdb_db->db_name)); if (tdb_transaction_write_lock_unmark(ctdb_db->ltdb->tdb) != 0) { DEBUG(DEBUG_ERR, ("Failed to unmark (transaction lock) database %s\n", ctdb_db->db_name)); return -1; } if (tdb_lockall_unmark(ctdb_db->ltdb->tdb) != 0) { DEBUG(DEBUG_ERR, ("Failed to unmark (all lock) database %s\n", ctdb_db->db_name)); return -1; } return 0; } int ctdb_lockdb_unmark(struct ctdb_db_context *ctdb_db) { if (!ctdb_db_frozen(ctdb_db)) { DEBUG(DEBUG_ERR, ("Attempt to unmark database locked when not frozen\n")); return -1; } return db_lock_unmark_handler(ctdb_db, NULL); } static void ctdb_lock_schedule(struct ctdb_context *ctdb); /* * Destructor to kill the child locking process */ static int ctdb_lock_context_destructor(struct lock_context *lock_ctx) { if (lock_ctx->request) { lock_ctx->request->lctx = NULL; } if (lock_ctx->child > 0) { ctdb_kill(lock_ctx->ctdb, lock_ctx->child, SIGKILL); if (lock_ctx->type == LOCK_RECORD) { DLIST_REMOVE(lock_ctx->ctdb_db->lock_current, lock_ctx); } else { DLIST_REMOVE(lock_ctx->ctdb->lock_current, lock_ctx); } if (lock_ctx->ctdb_db) { lock_ctx->ctdb_db->lock_num_current--; } CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_current); if (lock_ctx->ctdb_db) { CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_current); } } else { if (lock_ctx->type == LOCK_RECORD) { DLIST_REMOVE(lock_ctx->ctdb_db->lock_pending, lock_ctx); } else { DLIST_REMOVE(lock_ctx->ctdb->lock_pending, lock_ctx); } CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_pending); if (lock_ctx->ctdb_db) { CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending); } } ctdb_lock_schedule(lock_ctx->ctdb); return 0; } /* * Destructor to remove lock request */ static int ctdb_lock_request_destructor(struct lock_request *lock_request) { if (lock_request->lctx == NULL) { return 0; } lock_request->lctx->request = NULL; TALLOC_FREE(lock_request->lctx); return 0; } /* * Process all the callbacks waiting for lock * * If lock has failed, callback is executed with locked=false */ static void process_callbacks(struct lock_context *lock_ctx, bool locked) { struct lock_request *request; bool auto_mark = lock_ctx->auto_mark; if (auto_mark && locked) { switch (lock_ctx->type) { case LOCK_RECORD: tdb_chainlock_mark(lock_ctx->ctdb_db->ltdb->tdb, lock_ctx->key); break; case LOCK_DB: ctdb_lockdb_mark(lock_ctx->ctdb_db); break; } } request = lock_ctx->request; if (auto_mark) { /* Since request may be freed in the callback, unset the lock * context, so request destructor will not free lock context. */ request->lctx = NULL; } /* Since request may be freed in the callback, unset the request */ lock_ctx->request = NULL; request->callback(request->private_data, locked); if (!auto_mark) { return; } if (locked) { switch (lock_ctx->type) { case LOCK_RECORD: tdb_chainlock_unmark(lock_ctx->ctdb_db->ltdb->tdb, lock_ctx->key); break; case LOCK_DB: ctdb_lockdb_unmark(lock_ctx->ctdb_db); break; } } talloc_free(lock_ctx); } static int lock_bucket_id(double t) { double ms = 1.e-3, s = 1; int id; if (t < 1*ms) { id = 0; } else if (t < 10*ms) { id = 1; } else if (t < 100*ms) { id = 2; } else if (t < 1*s) { id = 3; } else if (t < 2*s) { id = 4; } else if (t < 4*s) { id = 5; } else if (t < 8*s) { id = 6; } else if (t < 16*s) { id = 7; } else if (t < 32*s) { id = 8; } else if (t < 64*s) { id = 9; } else { id = 10; } return id; } /* * Callback routine when the required locks are obtained. * Called from parent context */ static void ctdb_lock_handler(struct tevent_context *ev, struct tevent_fd *tfd, uint16_t flags, void *private_data) { struct lock_context *lock_ctx; char c; bool locked; double t; int id; lock_ctx = talloc_get_type_abort(private_data, struct lock_context); /* cancel the timeout event */ TALLOC_FREE(lock_ctx->ttimer); t = timeval_elapsed(&lock_ctx->start_time); id = lock_bucket_id(t); /* Read the status from the child process */ if (sys_read(lock_ctx->fd[0], &c, 1) != 1) { locked = false; } else { locked = (c == 0 ? true : false); } /* Update statistics */ CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_calls); if (lock_ctx->ctdb_db) { CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_calls); } if (locked) { if (lock_ctx->ctdb_db) { CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.buckets[id]); CTDB_UPDATE_LATENCY(lock_ctx->ctdb, lock_ctx->ctdb_db, lock_type_str[lock_ctx->type], locks.latency, lock_ctx->start_time); CTDB_UPDATE_DB_LATENCY(lock_ctx->ctdb_db, lock_type_str[lock_ctx->type], locks.latency, t); CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.buckets[id]); } } else { CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_failed); if (lock_ctx->ctdb_db) { CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_failed); } } process_callbacks(lock_ctx, locked); } /* * Callback routine when required locks are not obtained within timeout * Called from parent context */ static void ctdb_lock_timeout_handler(struct tevent_context *ev, struct tevent_timer *ttimer, struct timeval current_time, void *private_data) { static char debug_locks[PATH_MAX+1] = ""; struct lock_context *lock_ctx; struct ctdb_context *ctdb; pid_t pid; double elapsed_time; int new_timer; lock_ctx = talloc_get_type_abort(private_data, struct lock_context); ctdb = lock_ctx->ctdb; /* If a node stopped/banned, don't spam the logs */ if (ctdb->nodes[ctdb->pnn]->flags & NODE_FLAGS_INACTIVE) { lock_ctx->ttimer = NULL; return; } elapsed_time = timeval_elapsed(&lock_ctx->start_time); if (lock_ctx->ctdb_db) { DEBUG(DEBUG_WARNING, ("Unable to get %s lock on database %s for %.0lf seconds\n", (lock_ctx->type == LOCK_RECORD ? "RECORD" : "DB"), lock_ctx->ctdb_db->db_name, elapsed_time)); } else { DEBUG(DEBUG_WARNING, ("Unable to get ALLDB locks for %.0lf seconds\n", elapsed_time)); } if (ctdb_set_helper("lock debugging helper", debug_locks, sizeof(debug_locks), "CTDB_DEBUG_LOCKS", getenv("CTDB_BASE"), "debug_locks.sh")) { pid = vfork(); if (pid == 0) { execl(debug_locks, debug_locks, NULL); _exit(0); } ctdb_track_child(ctdb, pid); } else { DEBUG(DEBUG_WARNING, (__location__ " Unable to setup lock debugging\n")); } /* Back-off logging if lock is not obtained for a long time */ if (elapsed_time < 100.0) { new_timer = 10; } else if (elapsed_time < 1000.0) { new_timer = 100; } else { new_timer = 1000; } /* reset the timeout timer */ // talloc_free(lock_ctx->ttimer); lock_ctx->ttimer = tevent_add_timer(ctdb->ev, lock_ctx, timeval_current_ofs(new_timer, 0), ctdb_lock_timeout_handler, (void *)lock_ctx); } static int db_flags(struct ctdb_db_context *ctdb_db) { int tdb_flags = TDB_DEFAULT; #ifdef TDB_MUTEX_LOCKING if (!ctdb_db->persistent && ctdb_db->ctdb->tunable.mutex_enabled) { tdb_flags = (TDB_MUTEX_LOCKING | TDB_CLEAR_IF_FIRST); } #endif return tdb_flags; } static bool lock_helper_args(TALLOC_CTX *mem_ctx, struct lock_context *lock_ctx, int fd, int *argc, const char ***argv) { const char **args = NULL; int nargs = 0, i; switch (lock_ctx->type) { case LOCK_RECORD: nargs = 6; break; case LOCK_DB: nargs = 5; break; } /* Add extra argument for null termination */ nargs++; args = talloc_array(mem_ctx, const char *, nargs); if (args == NULL) { return false; } args[0] = talloc_asprintf(args, "%d", getpid()); args[1] = talloc_asprintf(args, "%d", fd); switch (lock_ctx->type) { case LOCK_RECORD: args[2] = talloc_strdup(args, "RECORD"); args[3] = talloc_strdup(args, lock_ctx->ctdb_db->db_path); args[4] = talloc_asprintf(args, "0x%x", db_flags(lock_ctx->ctdb_db)); if (lock_ctx->key.dsize == 0) { args[5] = talloc_strdup(args, "NULL"); } else { args[5] = hex_encode_talloc(args, lock_ctx->key.dptr, lock_ctx->key.dsize); } break; case LOCK_DB: args[2] = talloc_strdup(args, "DB"); args[3] = talloc_strdup(args, lock_ctx->ctdb_db->db_path); args[4] = talloc_asprintf(args, "0x%x", db_flags(lock_ctx->ctdb_db)); break; } /* Make sure last argument is NULL */ args[nargs-1] = NULL; for (i=0; ilock_pending; lock_ctx != NULL; lock_ctx = next_ctx) { if (lock_ctx->request != NULL) { /* Found a lock context with a request */ return lock_ctx; } next_ctx = lock_ctx->next; DEBUG(DEBUG_INFO, ("Removing lock context without lock " "request\n")); DLIST_REMOVE(ctdb->lock_pending, lock_ctx); CTDB_DECREMENT_STAT(ctdb, locks.num_pending); if (lock_ctx->ctdb_db) { CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending); } talloc_free(lock_ctx); } /* Next check database queues */ for (ctdb_db = ctdb->db_list; ctdb_db; ctdb_db = ctdb_db->next) { if (ctdb_db->lock_num_current == ctdb->tunable.lock_processes_per_db) { continue; } for (lock_ctx = ctdb_db->lock_pending; lock_ctx != NULL; lock_ctx = next_ctx) { next_ctx = lock_ctx->next; if (lock_ctx->request != NULL) { return lock_ctx; } DEBUG(DEBUG_INFO, ("Removing lock context without " "lock request\n")); DLIST_REMOVE(ctdb_db->lock_pending, lock_ctx); CTDB_DECREMENT_STAT(ctdb, locks.num_pending); CTDB_DECREMENT_DB_STAT(ctdb_db, locks.num_pending); talloc_free(lock_ctx); } } return NULL; } /* * Schedule a new lock child process * Set up callback handler and timeout handler */ static void ctdb_lock_schedule(struct ctdb_context *ctdb) { struct lock_context *lock_ctx; int ret, argc; TALLOC_CTX *tmp_ctx; static char prog[PATH_MAX+1] = ""; const char **args; if (!ctdb_set_helper("lock helper", prog, sizeof(prog), "CTDB_LOCK_HELPER", CTDB_HELPER_BINDIR, "ctdb_lock_helper")) { ctdb_die(ctdb, __location__ " Unable to set lock helper\n"); } /* Find a lock context with requests */ lock_ctx = ctdb_find_lock_context(ctdb); if (lock_ctx == NULL) { return; } lock_ctx->child = -1; ret = pipe(lock_ctx->fd); if (ret != 0) { DEBUG(DEBUG_ERR, ("Failed to create pipe in ctdb_lock_schedule\n")); return; } set_close_on_exec(lock_ctx->fd[0]); /* Create data for child process */ tmp_ctx = talloc_new(lock_ctx); if (tmp_ctx == NULL) { DEBUG(DEBUG_ERR, ("Failed to allocate memory for helper args\n")); close(lock_ctx->fd[0]); close(lock_ctx->fd[1]); return; } if (! ctdb->do_setsched) { ret = setenv("CTDB_NOSETSCHED", "1", 1); if (ret != 0) { DEBUG(DEBUG_WARNING, ("Failed to set CTDB_NOSETSCHED variable\n")); } } /* Create arguments for lock helper */ if (!lock_helper_args(tmp_ctx, lock_ctx, lock_ctx->fd[1], &argc, &args)) { DEBUG(DEBUG_ERR, ("Failed to create lock helper args\n")); close(lock_ctx->fd[0]); close(lock_ctx->fd[1]); talloc_free(tmp_ctx); return; } if (!ctdb_vfork_with_logging(lock_ctx, ctdb, "lock_helper", prog, argc, (const char **)args, NULL, NULL, &lock_ctx->child)) { DEBUG(DEBUG_ERR, ("Failed to create a child in ctdb_lock_schedule\n")); close(lock_ctx->fd[0]); close(lock_ctx->fd[1]); talloc_free(tmp_ctx); return; } /* Parent process */ close(lock_ctx->fd[1]); talloc_free(tmp_ctx); /* Set up timeout handler */ lock_ctx->ttimer = tevent_add_timer(ctdb->ev, lock_ctx, timeval_current_ofs(10, 0), ctdb_lock_timeout_handler, (void *)lock_ctx); if (lock_ctx->ttimer == NULL) { ctdb_kill(ctdb, lock_ctx->child, SIGKILL); lock_ctx->child = -1; close(lock_ctx->fd[0]); return; } /* Set up callback */ lock_ctx->tfd = tevent_add_fd(ctdb->ev, lock_ctx, lock_ctx->fd[0], TEVENT_FD_READ, ctdb_lock_handler, (void *)lock_ctx); if (lock_ctx->tfd == NULL) { TALLOC_FREE(lock_ctx->ttimer); ctdb_kill(ctdb, lock_ctx->child, SIGKILL); lock_ctx->child = -1; close(lock_ctx->fd[0]); return; } tevent_fd_set_auto_close(lock_ctx->tfd); /* Move the context from pending to current */ if (lock_ctx->type == LOCK_RECORD) { DLIST_REMOVE(lock_ctx->ctdb_db->lock_pending, lock_ctx); DLIST_ADD_END(lock_ctx->ctdb_db->lock_current, lock_ctx); } else { DLIST_REMOVE(ctdb->lock_pending, lock_ctx); DLIST_ADD_END(ctdb->lock_current, lock_ctx); } CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_pending); CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_current); if (lock_ctx->ctdb_db) { lock_ctx->ctdb_db->lock_num_current++; CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending); CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_current); } } /* * Lock record / db depending on type */ static struct lock_request *ctdb_lock_internal(TALLOC_CTX *mem_ctx, struct ctdb_context *ctdb, struct ctdb_db_context *ctdb_db, TDB_DATA key, uint32_t priority, void (*callback)(void *, bool), void *private_data, enum lock_type type, bool auto_mark) { struct lock_context *lock_ctx = NULL; struct lock_request *request; if (callback == NULL) { DEBUG(DEBUG_WARNING, ("No callback function specified, not locking\n")); return NULL; } lock_ctx = talloc_zero(ctdb, struct lock_context); if (lock_ctx == NULL) { DEBUG(DEBUG_ERR, ("Failed to create a new lock context\n")); return NULL; } if ((request = talloc_zero(mem_ctx, struct lock_request)) == NULL) { talloc_free(lock_ctx); return NULL; } lock_ctx->type = type; lock_ctx->ctdb = ctdb; lock_ctx->ctdb_db = ctdb_db; lock_ctx->key.dsize = key.dsize; if (key.dsize > 0) { lock_ctx->key.dptr = talloc_memdup(lock_ctx, key.dptr, key.dsize); if (lock_ctx->key.dptr == NULL) { DEBUG(DEBUG_ERR, (__location__ "Memory allocation error\n")); talloc_free(lock_ctx); talloc_free(request); return NULL; } lock_ctx->key_hash = ctdb_hash(&key); } else { lock_ctx->key.dptr = NULL; } lock_ctx->priority = priority; lock_ctx->auto_mark = auto_mark; lock_ctx->request = request; lock_ctx->child = -1; /* Non-record locks are required by recovery and should be scheduled * immediately, so keep them at the head of the pending queue. */ if (lock_ctx->type == LOCK_RECORD) { DLIST_ADD_END(ctdb_db->lock_pending, lock_ctx); } else { DLIST_ADD_END(ctdb->lock_pending, lock_ctx); } CTDB_INCREMENT_STAT(ctdb, locks.num_pending); if (ctdb_db) { CTDB_INCREMENT_DB_STAT(ctdb_db, locks.num_pending); } /* Start the timer when we activate the context */ lock_ctx->start_time = timeval_current(); request->lctx = lock_ctx; request->callback = callback; request->private_data = private_data; talloc_set_destructor(request, ctdb_lock_request_destructor); talloc_set_destructor(lock_ctx, ctdb_lock_context_destructor); ctdb_lock_schedule(ctdb); return request; } /* * obtain a lock on a record in a database */ struct lock_request *ctdb_lock_record(TALLOC_CTX *mem_ctx, struct ctdb_db_context *ctdb_db, TDB_DATA key, bool auto_mark, void (*callback)(void *, bool), void *private_data) { return ctdb_lock_internal(mem_ctx, ctdb_db->ctdb, ctdb_db, key, 0, callback, private_data, LOCK_RECORD, auto_mark); } /* * obtain a lock on a database */ struct lock_request *ctdb_lock_db(TALLOC_CTX *mem_ctx, struct ctdb_db_context *ctdb_db, bool auto_mark, void (*callback)(void *, bool), void *private_data) { return ctdb_lock_internal(mem_ctx, ctdb_db->ctdb, ctdb_db, tdb_null, 0, callback, private_data, LOCK_DB, auto_mark); }