s3:smbd: remove deprecated handling of "profile acls = yes"
[Samba.git] / ctdb / server / ctdb_lock.c
bloba559e2444710c2319c3ccb72f1a0cbe660d948d0
1 /*
2 ctdb lock handling
3 provide API to do non-blocking locks for single or all databases
5 Copyright (C) Amitay Isaacs 2012
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, see <http://www.gnu.org/licenses/>.
20 #include "replace.h"
21 #include "system/filesys.h"
22 #include "system/network.h"
24 #include <talloc.h>
25 #include <tevent.h>
27 #include "lib/tdb_wrap/tdb_wrap.h"
28 #include "lib/util/dlinklist.h"
29 #include "lib/util/debug.h"
30 #include "lib/util/samba_util.h"
31 #include "lib/util/sys_rw.h"
33 #include "ctdb_private.h"
35 #include "common/common.h"
36 #include "common/logging.h"
39 * Non-blocking Locking API
41 * 1. Create a child process to do blocking locks.
42 * 2. Once the locks are obtained, signal parent process via fd.
43 * 3. Invoke registered callback routine with locking status.
44 * 4. If the child process cannot get locks within certain time,
45 * execute an external script to debug.
47 * ctdb_lock_record() - get a lock on a record
48 * ctdb_lock_db() - get a lock on a DB
50 * auto_mark - whether to mark/unmark DBs in before/after callback
51 * = false is used for freezing databases for
52 * recovery since the recovery cannot start till
53 * databases are locked on all the nodes.
54 * = true is used for record locks.
57 enum lock_type {
58 LOCK_RECORD,
59 LOCK_DB,
62 static const char * const lock_type_str[] = {
63 "lock_record",
64 "lock_db",
67 struct lock_request;
69 /* lock_context is the common part for a lock request */
70 struct lock_context {
71 struct lock_context *next, *prev;
72 enum lock_type type;
73 struct ctdb_context *ctdb;
74 struct ctdb_db_context *ctdb_db;
75 TDB_DATA key;
76 uint32_t priority;
77 bool auto_mark;
78 struct lock_request *request;
79 pid_t child;
80 int fd[2];
81 struct tevent_fd *tfd;
82 struct tevent_timer *ttimer;
83 struct timeval start_time;
84 uint32_t key_hash;
85 bool can_schedule;
88 /* lock_request is the client specific part for a lock request */
89 struct lock_request {
90 struct lock_context *lctx;
91 void (*callback)(void *, bool);
92 void *private_data;
96 int ctdb_db_iterator(struct ctdb_context *ctdb, ctdb_db_handler_t handler,
97 void *private_data)
99 struct ctdb_db_context *ctdb_db;
100 int ret;
102 for (ctdb_db = ctdb->db_list; ctdb_db; ctdb_db = ctdb_db->next) {
103 ret = handler(ctdb_db, private_data);
104 if (ret != 0) {
105 return -1;
109 return 0;
113 * lock all databases - mark only
115 static int db_lock_mark_handler(struct ctdb_db_context *ctdb_db,
116 void *private_data)
118 int tdb_transaction_write_lock_mark(struct tdb_context *);
120 DEBUG(DEBUG_INFO, ("marking locked database %s\n", ctdb_db->db_name));
122 if (tdb_transaction_write_lock_mark(ctdb_db->ltdb->tdb) != 0) {
123 DEBUG(DEBUG_ERR, ("Failed to mark (transaction lock) database %s\n",
124 ctdb_db->db_name));
125 return -1;
128 if (tdb_lockall_mark(ctdb_db->ltdb->tdb) != 0) {
129 DEBUG(DEBUG_ERR, ("Failed to mark (all lock) database %s\n",
130 ctdb_db->db_name));
131 return -1;
134 return 0;
137 int ctdb_lockdb_mark(struct ctdb_db_context *ctdb_db)
139 if (!ctdb_db_frozen(ctdb_db)) {
140 DEBUG(DEBUG_ERR,
141 ("Attempt to mark database locked when not frozen\n"));
142 return -1;
145 return db_lock_mark_handler(ctdb_db, NULL);
149 * lock all databases - unmark only
151 static int db_lock_unmark_handler(struct ctdb_db_context *ctdb_db,
152 void *private_data)
154 int tdb_transaction_write_lock_unmark(struct tdb_context *);
156 DEBUG(DEBUG_INFO, ("unmarking locked database %s\n", ctdb_db->db_name));
158 if (tdb_transaction_write_lock_unmark(ctdb_db->ltdb->tdb) != 0) {
159 DEBUG(DEBUG_ERR, ("Failed to unmark (transaction lock) database %s\n",
160 ctdb_db->db_name));
161 return -1;
164 if (tdb_lockall_unmark(ctdb_db->ltdb->tdb) != 0) {
165 DEBUG(DEBUG_ERR, ("Failed to unmark (all lock) database %s\n",
166 ctdb_db->db_name));
167 return -1;
170 return 0;
173 int ctdb_lockdb_unmark(struct ctdb_db_context *ctdb_db)
175 if (!ctdb_db_frozen(ctdb_db)) {
176 DEBUG(DEBUG_ERR,
177 ("Attempt to unmark database locked when not frozen\n"));
178 return -1;
181 return db_lock_unmark_handler(ctdb_db, NULL);
184 static void ctdb_lock_schedule(struct ctdb_context *ctdb);
187 * Destructor to kill the child locking process
189 static int ctdb_lock_context_destructor(struct lock_context *lock_ctx)
191 if (lock_ctx->request) {
192 lock_ctx->request->lctx = NULL;
194 if (lock_ctx->child > 0) {
195 ctdb_kill(lock_ctx->ctdb, lock_ctx->child, SIGTERM);
196 if (lock_ctx->type == LOCK_RECORD) {
197 DLIST_REMOVE(lock_ctx->ctdb_db->lock_current, lock_ctx);
198 } else {
199 DLIST_REMOVE(lock_ctx->ctdb->lock_current, lock_ctx);
201 if (lock_ctx->ctdb_db) {
202 lock_ctx->ctdb_db->lock_num_current--;
204 CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_current);
205 if (lock_ctx->ctdb_db) {
206 CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_current);
208 } else {
209 if (lock_ctx->type == LOCK_RECORD) {
210 DLIST_REMOVE(lock_ctx->ctdb_db->lock_pending, lock_ctx);
211 } else {
212 DLIST_REMOVE(lock_ctx->ctdb->lock_pending, lock_ctx);
214 CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_pending);
215 if (lock_ctx->ctdb_db) {
216 CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending);
220 ctdb_lock_schedule(lock_ctx->ctdb);
222 return 0;
227 * Destructor to remove lock request
229 static int ctdb_lock_request_destructor(struct lock_request *lock_request)
231 if (lock_request->lctx == NULL) {
232 return 0;
235 lock_request->lctx->request = NULL;
236 TALLOC_FREE(lock_request->lctx);
238 return 0;
242 * Process all the callbacks waiting for lock
244 * If lock has failed, callback is executed with locked=false
246 static void process_callbacks(struct lock_context *lock_ctx, bool locked)
248 struct lock_request *request;
249 bool auto_mark = lock_ctx->auto_mark;
251 if (auto_mark && locked) {
252 switch (lock_ctx->type) {
253 case LOCK_RECORD:
254 tdb_chainlock_mark(lock_ctx->ctdb_db->ltdb->tdb, lock_ctx->key);
255 break;
257 case LOCK_DB:
258 ctdb_lockdb_mark(lock_ctx->ctdb_db);
259 break;
263 request = lock_ctx->request;
264 if (auto_mark) {
265 /* Since request may be freed in the callback, unset the lock
266 * context, so request destructor will not free lock context.
268 request->lctx = NULL;
271 /* Since request may be freed in the callback, unset the request */
272 lock_ctx->request = NULL;
274 request->callback(request->private_data, locked);
276 if (!auto_mark) {
277 return;
280 if (locked) {
281 switch (lock_ctx->type) {
282 case LOCK_RECORD:
283 tdb_chainlock_unmark(lock_ctx->ctdb_db->ltdb->tdb, lock_ctx->key);
284 break;
286 case LOCK_DB:
287 ctdb_lockdb_unmark(lock_ctx->ctdb_db);
288 break;
292 talloc_free(lock_ctx);
296 static int lock_bucket_id(double t)
298 double ms = 1.e-3, s = 1;
299 int id;
301 if (t < 1*ms) {
302 id = 0;
303 } else if (t < 10*ms) {
304 id = 1;
305 } else if (t < 100*ms) {
306 id = 2;
307 } else if (t < 1*s) {
308 id = 3;
309 } else if (t < 2*s) {
310 id = 4;
311 } else if (t < 4*s) {
312 id = 5;
313 } else if (t < 8*s) {
314 id = 6;
315 } else if (t < 16*s) {
316 id = 7;
317 } else if (t < 32*s) {
318 id = 8;
319 } else if (t < 64*s) {
320 id = 9;
321 } else {
322 id = 10;
325 return id;
329 * Callback routine when the required locks are obtained.
330 * Called from parent context
332 static void ctdb_lock_handler(struct tevent_context *ev,
333 struct tevent_fd *tfd,
334 uint16_t flags,
335 void *private_data)
337 struct lock_context *lock_ctx;
338 char c;
339 bool locked;
340 double t;
341 int id;
343 lock_ctx = talloc_get_type_abort(private_data, struct lock_context);
345 /* cancel the timeout event */
346 TALLOC_FREE(lock_ctx->ttimer);
348 t = timeval_elapsed(&lock_ctx->start_time);
349 id = lock_bucket_id(t);
351 /* Read the status from the child process */
352 if (sys_read(lock_ctx->fd[0], &c, 1) != 1) {
353 locked = false;
354 } else {
355 locked = (c == 0 ? true : false);
358 /* Update statistics */
359 CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_calls);
360 if (lock_ctx->ctdb_db) {
361 CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_calls);
364 if (locked) {
365 if (lock_ctx->ctdb_db) {
366 CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.buckets[id]);
367 CTDB_UPDATE_LATENCY(lock_ctx->ctdb, lock_ctx->ctdb_db,
368 lock_type_str[lock_ctx->type], locks.latency,
369 lock_ctx->start_time);
371 CTDB_UPDATE_DB_LATENCY(lock_ctx->ctdb_db, lock_type_str[lock_ctx->type], locks.latency, t);
372 CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.buckets[id]);
374 } else {
375 CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_failed);
376 if (lock_ctx->ctdb_db) {
377 CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_failed);
381 process_callbacks(lock_ctx, locked);
384 struct lock_log_entry {
385 struct db_hash_context *lock_log;
386 TDB_DATA key;
387 unsigned long log_sec;
388 struct tevent_timer *timer;
391 static int lock_log_fetch_parser(uint8_t *keybuf, size_t keylen,
392 uint8_t *databuf, size_t datalen,
393 void *private_data)
395 struct lock_log_entry **entry =
396 (struct lock_log_entry **)private_data;
398 if (datalen != sizeof(struct lock_log_entry *)) {
399 return EINVAL;
402 *entry = talloc_get_type_abort(*(void **)databuf,
403 struct lock_log_entry);
404 return 0;
407 static void lock_log_cleanup(struct tevent_context *ev,
408 struct tevent_timer *ttimer,
409 struct timeval current_time,
410 void *private_data)
412 struct lock_log_entry *entry = talloc_get_type_abort(
413 private_data, struct lock_log_entry);
414 int ret;
416 entry->timer = NULL;
418 ret = db_hash_delete(entry->lock_log, entry->key.dptr,
419 entry->key.dsize);
420 if (ret != 0) {
421 return;
423 talloc_free(entry);
426 static bool lock_log_skip(struct tevent_context *ev,
427 struct db_hash_context *lock_log,
428 TDB_DATA key, unsigned long elapsed_sec)
430 struct lock_log_entry *entry = NULL;
431 int ret;
433 ret = db_hash_fetch(lock_log, key.dptr, key.dsize,
434 lock_log_fetch_parser, &entry);
435 if (ret == ENOENT) {
437 entry = talloc_zero(lock_log, struct lock_log_entry);
438 if (entry == NULL) {
439 goto fail;
442 entry->lock_log = lock_log;
444 entry->key.dptr = talloc_memdup(entry, key.dptr, key.dsize);
445 if (entry->key.dptr == NULL) {
446 talloc_free(entry);
447 goto fail;
449 entry->key.dsize = key.dsize;
451 entry->log_sec = elapsed_sec;
452 entry->timer = tevent_add_timer(ev, entry,
453 timeval_current_ofs(30, 0),
454 lock_log_cleanup, entry);
455 if (entry->timer == NULL) {
456 talloc_free(entry);
457 goto fail;
460 ret = db_hash_add(lock_log, key.dptr, key.dsize,
461 (uint8_t *)&entry,
462 sizeof(struct lock_log_entry *));
463 if (ret != 0) {
464 talloc_free(entry);
465 goto fail;
468 return false;
470 } else if (ret == EINVAL) {
472 ret = db_hash_delete(lock_log, key.dptr, key.dsize);
473 if (ret != 0) {
474 goto fail;
477 return false;
479 } else if (ret == 0) {
481 if (elapsed_sec <= entry->log_sec) {
482 return true;
485 entry->log_sec = elapsed_sec;
487 TALLOC_FREE(entry->timer);
488 entry->timer = tevent_add_timer(ev, entry,
489 timeval_current_ofs(30, 0),
490 lock_log_cleanup, entry);
491 if (entry->timer == NULL) {
492 ret = db_hash_delete(lock_log, key.dptr, key.dsize);
493 if (ret != 0) {
494 goto fail;
496 talloc_free(entry);
499 return false;
503 fail:
504 return false;
509 * Callback routine when required locks are not obtained within timeout
510 * Called from parent context
512 static void ctdb_lock_timeout_handler(struct tevent_context *ev,
513 struct tevent_timer *ttimer,
514 struct timeval current_time,
515 void *private_data)
517 static char debug_locks[PATH_MAX+1] = "";
518 struct lock_context *lock_ctx;
519 struct ctdb_context *ctdb;
520 pid_t pid;
521 double elapsed_time;
522 bool skip;
523 char *keystr;
525 lock_ctx = talloc_get_type_abort(private_data, struct lock_context);
526 ctdb = lock_ctx->ctdb;
528 elapsed_time = timeval_elapsed(&lock_ctx->start_time);
530 /* For database locks, always log */
531 if (lock_ctx->type == LOCK_DB) {
532 DEBUG(DEBUG_WARNING,
533 ("Unable to get DB lock on database %s for "
534 "%.0lf seconds\n",
535 lock_ctx->ctdb_db->db_name, elapsed_time));
536 goto lock_debug;
539 /* For record locks, check if we have already logged */
540 skip = lock_log_skip(ev, lock_ctx->ctdb_db->lock_log,
541 lock_ctx->key, (unsigned long)elapsed_time);
542 if (skip) {
543 goto skip_lock_debug;
546 keystr = hex_encode_talloc(lock_ctx, lock_ctx->key.dptr,
547 lock_ctx->key.dsize);
548 DEBUG(DEBUG_WARNING,
549 ("Unable to get RECORD lock on database %s for %.0lf seconds"
550 " (key %s)\n",
551 lock_ctx->ctdb_db->db_name, elapsed_time,
552 keystr ? keystr : ""));
553 TALLOC_FREE(keystr);
555 /* If a node stopped/banned, don't spam the logs */
556 if (ctdb->nodes[ctdb->pnn]->flags & NODE_FLAGS_INACTIVE) {
557 goto skip_lock_debug;
560 lock_debug:
562 if (ctdb_set_helper("lock debugging helper",
563 debug_locks, sizeof(debug_locks),
564 "CTDB_DEBUG_LOCKS",
565 getenv("CTDB_BASE"), "debug_locks.sh")) {
566 pid = vfork();
567 if (pid == 0) {
568 execl(debug_locks, debug_locks, NULL);
569 _exit(0);
571 ctdb_track_child(ctdb, pid);
572 } else {
573 DEBUG(DEBUG_WARNING,
574 (__location__
575 " Unable to setup lock debugging\n"));
578 skip_lock_debug:
580 /* reset the timeout timer */
581 // talloc_free(lock_ctx->ttimer);
582 lock_ctx->ttimer = tevent_add_timer(ctdb->ev,
583 lock_ctx,
584 timeval_current_ofs(10, 0),
585 ctdb_lock_timeout_handler,
586 (void *)lock_ctx);
589 static bool lock_helper_args(TALLOC_CTX *mem_ctx,
590 struct lock_context *lock_ctx, int fd,
591 int *argc, const char ***argv)
593 const char **args = NULL;
594 int nargs = 0, i;
596 switch (lock_ctx->type) {
597 case LOCK_RECORD:
598 nargs = 6;
599 break;
601 case LOCK_DB:
602 nargs = 5;
603 break;
606 /* Add extra argument for null termination */
607 nargs++;
609 args = talloc_array(mem_ctx, const char *, nargs);
610 if (args == NULL) {
611 return false;
614 args[0] = talloc_asprintf(args, "%d", getpid());
615 args[1] = talloc_asprintf(args, "%d", fd);
617 switch (lock_ctx->type) {
618 case LOCK_RECORD:
619 args[2] = talloc_strdup(args, "RECORD");
620 args[3] = talloc_strdup(args, lock_ctx->ctdb_db->db_path);
621 args[4] = talloc_asprintf(args, "0x%x",
622 tdb_get_flags(lock_ctx->ctdb_db->ltdb->tdb));
623 if (lock_ctx->key.dsize == 0) {
624 args[5] = talloc_strdup(args, "NULL");
625 } else {
626 args[5] = hex_encode_talloc(args, lock_ctx->key.dptr, lock_ctx->key.dsize);
628 break;
630 case LOCK_DB:
631 args[2] = talloc_strdup(args, "DB");
632 args[3] = talloc_strdup(args, lock_ctx->ctdb_db->db_path);
633 args[4] = talloc_asprintf(args, "0x%x",
634 tdb_get_flags(lock_ctx->ctdb_db->ltdb->tdb));
635 break;
638 /* Make sure last argument is NULL */
639 args[nargs-1] = NULL;
641 for (i=0; i<nargs-1; i++) {
642 if (args[i] == NULL) {
643 talloc_free(args);
644 return false;
648 *argc = nargs;
649 *argv = args;
650 return true;
654 * Find a lock request that can be scheduled
656 static struct lock_context *ctdb_find_lock_context(struct ctdb_context *ctdb)
658 struct lock_context *lock_ctx, *next_ctx;
659 struct ctdb_db_context *ctdb_db;
661 /* First check if there are database lock requests */
663 for (lock_ctx = ctdb->lock_pending; lock_ctx != NULL;
664 lock_ctx = next_ctx) {
666 if (lock_ctx->request != NULL) {
667 /* Found a lock context with a request */
668 return lock_ctx;
671 next_ctx = lock_ctx->next;
673 DEBUG(DEBUG_INFO, ("Removing lock context without lock "
674 "request\n"));
675 DLIST_REMOVE(ctdb->lock_pending, lock_ctx);
676 CTDB_DECREMENT_STAT(ctdb, locks.num_pending);
677 if (lock_ctx->ctdb_db) {
678 CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db,
679 locks.num_pending);
681 talloc_free(lock_ctx);
684 /* Next check database queues */
685 for (ctdb_db = ctdb->db_list; ctdb_db; ctdb_db = ctdb_db->next) {
686 if (ctdb_db->lock_num_current ==
687 ctdb->tunable.lock_processes_per_db) {
688 continue;
691 for (lock_ctx = ctdb_db->lock_pending; lock_ctx != NULL;
692 lock_ctx = next_ctx) {
694 next_ctx = lock_ctx->next;
696 if (lock_ctx->request != NULL) {
697 return lock_ctx;
700 DEBUG(DEBUG_INFO, ("Removing lock context without "
701 "lock request\n"));
702 DLIST_REMOVE(ctdb_db->lock_pending, lock_ctx);
703 CTDB_DECREMENT_STAT(ctdb, locks.num_pending);
704 CTDB_DECREMENT_DB_STAT(ctdb_db, locks.num_pending);
705 talloc_free(lock_ctx);
709 return NULL;
713 * Schedule a new lock child process
714 * Set up callback handler and timeout handler
716 static void ctdb_lock_schedule(struct ctdb_context *ctdb)
718 struct lock_context *lock_ctx;
719 int ret, argc;
720 TALLOC_CTX *tmp_ctx;
721 static char prog[PATH_MAX+1] = "";
722 const char **args;
724 if (!ctdb_set_helper("lock helper",
725 prog, sizeof(prog),
726 "CTDB_LOCK_HELPER",
727 CTDB_HELPER_BINDIR, "ctdb_lock_helper")) {
728 ctdb_die(ctdb, __location__
729 " Unable to set lock helper\n");
732 /* Find a lock context with requests */
733 lock_ctx = ctdb_find_lock_context(ctdb);
734 if (lock_ctx == NULL) {
735 return;
738 lock_ctx->child = -1;
739 ret = pipe(lock_ctx->fd);
740 if (ret != 0) {
741 DEBUG(DEBUG_ERR, ("Failed to create pipe in ctdb_lock_schedule\n"));
742 return;
745 set_close_on_exec(lock_ctx->fd[0]);
747 /* Create data for child process */
748 tmp_ctx = talloc_new(lock_ctx);
749 if (tmp_ctx == NULL) {
750 DEBUG(DEBUG_ERR, ("Failed to allocate memory for helper args\n"));
751 close(lock_ctx->fd[0]);
752 close(lock_ctx->fd[1]);
753 return;
756 if (! ctdb->do_setsched) {
757 ret = setenv("CTDB_NOSETSCHED", "1", 1);
758 if (ret != 0) {
759 DEBUG(DEBUG_WARNING,
760 ("Failed to set CTDB_NOSETSCHED variable\n"));
764 /* Create arguments for lock helper */
765 if (!lock_helper_args(tmp_ctx, lock_ctx, lock_ctx->fd[1],
766 &argc, &args)) {
767 DEBUG(DEBUG_ERR, ("Failed to create lock helper args\n"));
768 close(lock_ctx->fd[0]);
769 close(lock_ctx->fd[1]);
770 talloc_free(tmp_ctx);
771 return;
774 lock_ctx->child = ctdb_vfork_exec(lock_ctx, ctdb, prog, argc,
775 (const char **)args);
776 if (lock_ctx->child == -1) {
777 DEBUG(DEBUG_ERR, ("Failed to create a child in ctdb_lock_schedule\n"));
778 close(lock_ctx->fd[0]);
779 close(lock_ctx->fd[1]);
780 talloc_free(tmp_ctx);
781 return;
784 /* Parent process */
785 close(lock_ctx->fd[1]);
787 talloc_free(tmp_ctx);
789 /* Set up timeout handler */
790 lock_ctx->ttimer = tevent_add_timer(ctdb->ev,
791 lock_ctx,
792 timeval_current_ofs(10, 0),
793 ctdb_lock_timeout_handler,
794 (void *)lock_ctx);
795 if (lock_ctx->ttimer == NULL) {
796 ctdb_kill(ctdb, lock_ctx->child, SIGTERM);
797 lock_ctx->child = -1;
798 close(lock_ctx->fd[0]);
799 return;
802 /* Set up callback */
803 lock_ctx->tfd = tevent_add_fd(ctdb->ev,
804 lock_ctx,
805 lock_ctx->fd[0],
806 TEVENT_FD_READ,
807 ctdb_lock_handler,
808 (void *)lock_ctx);
809 if (lock_ctx->tfd == NULL) {
810 TALLOC_FREE(lock_ctx->ttimer);
811 ctdb_kill(ctdb, lock_ctx->child, SIGTERM);
812 lock_ctx->child = -1;
813 close(lock_ctx->fd[0]);
814 return;
816 tevent_fd_set_auto_close(lock_ctx->tfd);
818 /* Move the context from pending to current */
819 if (lock_ctx->type == LOCK_RECORD) {
820 DLIST_REMOVE(lock_ctx->ctdb_db->lock_pending, lock_ctx);
821 DLIST_ADD_END(lock_ctx->ctdb_db->lock_current, lock_ctx);
822 } else {
823 DLIST_REMOVE(ctdb->lock_pending, lock_ctx);
824 DLIST_ADD_END(ctdb->lock_current, lock_ctx);
826 CTDB_DECREMENT_STAT(lock_ctx->ctdb, locks.num_pending);
827 CTDB_INCREMENT_STAT(lock_ctx->ctdb, locks.num_current);
828 if (lock_ctx->ctdb_db) {
829 lock_ctx->ctdb_db->lock_num_current++;
830 CTDB_DECREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_pending);
831 CTDB_INCREMENT_DB_STAT(lock_ctx->ctdb_db, locks.num_current);
837 * Lock record / db depending on type
839 static struct lock_request *ctdb_lock_internal(TALLOC_CTX *mem_ctx,
840 struct ctdb_context *ctdb,
841 struct ctdb_db_context *ctdb_db,
842 TDB_DATA key,
843 uint32_t priority,
844 void (*callback)(void *, bool),
845 void *private_data,
846 enum lock_type type,
847 bool auto_mark)
849 struct lock_context *lock_ctx = NULL;
850 struct lock_request *request;
852 if (callback == NULL) {
853 DEBUG(DEBUG_WARNING, ("No callback function specified, not locking\n"));
854 return NULL;
857 lock_ctx = talloc_zero(ctdb, struct lock_context);
858 if (lock_ctx == NULL) {
859 DEBUG(DEBUG_ERR, ("Failed to create a new lock context\n"));
860 return NULL;
863 if ((request = talloc_zero(mem_ctx, struct lock_request)) == NULL) {
864 talloc_free(lock_ctx);
865 return NULL;
868 lock_ctx->type = type;
869 lock_ctx->ctdb = ctdb;
870 lock_ctx->ctdb_db = ctdb_db;
871 lock_ctx->key.dsize = key.dsize;
872 if (key.dsize > 0) {
873 lock_ctx->key.dptr = talloc_memdup(lock_ctx, key.dptr, key.dsize);
874 if (lock_ctx->key.dptr == NULL) {
875 DEBUG(DEBUG_ERR, (__location__ "Memory allocation error\n"));
876 talloc_free(lock_ctx);
877 talloc_free(request);
878 return NULL;
880 lock_ctx->key_hash = ctdb_hash(&key);
881 } else {
882 lock_ctx->key.dptr = NULL;
884 lock_ctx->priority = priority;
885 lock_ctx->auto_mark = auto_mark;
887 lock_ctx->request = request;
888 lock_ctx->child = -1;
890 /* Non-record locks are required by recovery and should be scheduled
891 * immediately, so keep them at the head of the pending queue.
893 if (lock_ctx->type == LOCK_RECORD) {
894 DLIST_ADD_END(ctdb_db->lock_pending, lock_ctx);
895 } else {
896 DLIST_ADD_END(ctdb->lock_pending, lock_ctx);
898 CTDB_INCREMENT_STAT(ctdb, locks.num_pending);
899 if (ctdb_db) {
900 CTDB_INCREMENT_DB_STAT(ctdb_db, locks.num_pending);
903 /* Start the timer when we activate the context */
904 lock_ctx->start_time = timeval_current();
906 request->lctx = lock_ctx;
907 request->callback = callback;
908 request->private_data = private_data;
910 talloc_set_destructor(request, ctdb_lock_request_destructor);
911 talloc_set_destructor(lock_ctx, ctdb_lock_context_destructor);
913 ctdb_lock_schedule(ctdb);
915 return request;
920 * obtain a lock on a record in a database
922 struct lock_request *ctdb_lock_record(TALLOC_CTX *mem_ctx,
923 struct ctdb_db_context *ctdb_db,
924 TDB_DATA key,
925 bool auto_mark,
926 void (*callback)(void *, bool),
927 void *private_data)
929 return ctdb_lock_internal(mem_ctx,
930 ctdb_db->ctdb,
931 ctdb_db,
932 key,
934 callback,
935 private_data,
936 LOCK_RECORD,
937 auto_mark);
942 * obtain a lock on a database
944 struct lock_request *ctdb_lock_db(TALLOC_CTX *mem_ctx,
945 struct ctdb_db_context *ctdb_db,
946 bool auto_mark,
947 void (*callback)(void *, bool),
948 void *private_data)
950 return ctdb_lock_internal(mem_ctx,
951 ctdb_db->ctdb,
952 ctdb_db,
953 tdb_null,
955 callback,
956 private_data,
957 LOCK_DB,
958 auto_mark);