1 /******************************************************************************
2 *******************************************************************************
4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
5 ** Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
7 ** This copyrighted material is made available to anyone wishing to use,
8 ** modify, copy, or redistribute it subject to the terms and conditions
9 ** of the GNU General Public License v.2.
11 *******************************************************************************
12 ******************************************************************************/
14 #include "dlm_internal.h"
15 #include "lockspace.h"
28 * Recovery waiting routines: these functions wait for a particular reply from
29 * a remote node, or for the remote node to report a certain status. They need
30 * to abort if the lockspace is stopped indicating a node has failed (perhaps
31 * the one being waited for).
35 * Wait until given function returns non-zero or lockspace is stopped
36 * (LS_RECOVERY_STOP set due to failure of a node in ls_nodes). When another
37 * function thinks it could have completed the waited-on task, they should wake
38 * up ls_wait_general to get an immediate response rather than waiting for the
39 * timer to detect the result. A timer wakes us up periodically while waiting
40 * to see if we should abort due to a node failure. This should only be called
41 * by the dlm_recoverd thread.
44 static void dlm_wait_timer_fn(unsigned long data)
46 struct dlm_ls *ls = (struct dlm_ls *) data;
47 mod_timer(&ls->ls_timer, jiffies + (dlm_config.ci_recover_timer * HZ));
48 wake_up(&ls->ls_wait_general);
51 int dlm_wait_function(struct dlm_ls *ls, int (*testfn) (struct dlm_ls *ls))
55 init_timer(&ls->ls_timer);
56 ls->ls_timer.function = dlm_wait_timer_fn;
57 ls->ls_timer.data = (long) ls;
58 ls->ls_timer.expires = jiffies + (dlm_config.ci_recover_timer * HZ);
59 add_timer(&ls->ls_timer);
61 wait_event(ls->ls_wait_general, testfn(ls) || dlm_recovery_stopped(ls));
62 del_timer_sync(&ls->ls_timer);
64 if (dlm_recovery_stopped(ls)) {
65 log_debug(ls, "dlm_wait_function aborted");
72 * An efficient way for all nodes to wait for all others to have a certain
73 * status. The node with the lowest nodeid polls all the others for their
74 * status (wait_status_all) and all the others poll the node with the low id
75 * for its accumulated result (wait_status_low). When all nodes have set
76 * status flag X, then status flag X_ALL will be set on the low nodeid.
79 uint32_t dlm_recover_status(struct dlm_ls *ls)
82 spin_lock(&ls->ls_recover_lock);
83 status = ls->ls_recover_status;
84 spin_unlock(&ls->ls_recover_lock);
88 void dlm_set_recover_status(struct dlm_ls *ls, uint32_t status)
90 spin_lock(&ls->ls_recover_lock);
91 ls->ls_recover_status |= status;
92 spin_unlock(&ls->ls_recover_lock);
95 static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status)
97 struct dlm_rcom *rc = ls->ls_recover_buf;
98 struct dlm_member *memb;
101 list_for_each_entry(memb, &ls->ls_nodes, list) {
104 if (dlm_recovery_stopped(ls)) {
109 error = dlm_rcom_status(ls, memb->nodeid);
113 if (rc->rc_result & wait_status)
124 static int wait_status_low(struct dlm_ls *ls, uint32_t wait_status)
126 struct dlm_rcom *rc = ls->ls_recover_buf;
127 int error = 0, delay = 0, nodeid = ls->ls_low_nodeid;
130 if (dlm_recovery_stopped(ls)) {
135 error = dlm_rcom_status(ls, nodeid);
139 if (rc->rc_result & wait_status)
149 static int wait_status(struct dlm_ls *ls, uint32_t status)
151 uint32_t status_all = status << 1;
154 if (ls->ls_low_nodeid == dlm_our_nodeid()) {
155 error = wait_status_all(ls, status);
157 dlm_set_recover_status(ls, status_all);
159 error = wait_status_low(ls, status_all);
164 int dlm_recover_members_wait(struct dlm_ls *ls)
166 return wait_status(ls, DLM_RS_NODES);
169 int dlm_recover_directory_wait(struct dlm_ls *ls)
171 return wait_status(ls, DLM_RS_DIR);
174 int dlm_recover_locks_wait(struct dlm_ls *ls)
176 return wait_status(ls, DLM_RS_LOCKS);
179 int dlm_recover_done_wait(struct dlm_ls *ls)
181 return wait_status(ls, DLM_RS_DONE);
185 * The recover_list contains all the rsb's for which we've requested the new
186 * master nodeid. As replies are returned from the resource directories the
187 * rsb's are removed from the list. When the list is empty we're done.
189 * The recover_list is later similarly used for all rsb's for which we've sent
190 * new lkb's and need to receive new corresponding lkid's.
192 * We use the address of the rsb struct as a simple local identifier for the
193 * rsb so we can match an rcom reply with the rsb it was sent for.
196 static int recover_list_empty(struct dlm_ls *ls)
200 spin_lock(&ls->ls_recover_list_lock);
201 empty = list_empty(&ls->ls_recover_list);
202 spin_unlock(&ls->ls_recover_list_lock);
207 static void recover_list_add(struct dlm_rsb *r)
209 struct dlm_ls *ls = r->res_ls;
211 spin_lock(&ls->ls_recover_list_lock);
212 if (list_empty(&r->res_recover_list)) {
213 list_add_tail(&r->res_recover_list, &ls->ls_recover_list);
214 ls->ls_recover_list_count++;
217 spin_unlock(&ls->ls_recover_list_lock);
220 static void recover_list_del(struct dlm_rsb *r)
222 struct dlm_ls *ls = r->res_ls;
224 spin_lock(&ls->ls_recover_list_lock);
225 list_del_init(&r->res_recover_list);
226 ls->ls_recover_list_count--;
227 spin_unlock(&ls->ls_recover_list_lock);
232 static struct dlm_rsb *recover_list_find(struct dlm_ls *ls, uint64_t id)
234 struct dlm_rsb *r = NULL;
236 spin_lock(&ls->ls_recover_list_lock);
238 list_for_each_entry(r, &ls->ls_recover_list, res_recover_list) {
239 if (id == (unsigned long) r)
244 spin_unlock(&ls->ls_recover_list_lock);
248 static void recover_list_clear(struct dlm_ls *ls)
250 struct dlm_rsb *r, *s;
252 spin_lock(&ls->ls_recover_list_lock);
253 list_for_each_entry_safe(r, s, &ls->ls_recover_list, res_recover_list) {
254 list_del_init(&r->res_recover_list);
255 r->res_recover_locks_count = 0;
257 ls->ls_recover_list_count--;
260 if (ls->ls_recover_list_count != 0) {
261 log_error(ls, "warning: recover_list_count %d",
262 ls->ls_recover_list_count);
263 ls->ls_recover_list_count = 0;
265 spin_unlock(&ls->ls_recover_list_lock);
269 /* Master recovery: find new master node for rsb's that were
270 mastered on nodes that have been removed.
274 dlm_send_rcom_lookup -> receive_rcom_lookup
276 receive_rcom_lookup_reply <-
277 dlm_recover_master_reply
284 * Set the lock master for all LKBs in a lock queue
285 * If we are the new master of the rsb, we may have received new
286 * MSTCPY locks from other nodes already which we need to ignore
287 * when setting the new nodeid.
290 static void set_lock_master(struct list_head *queue, int nodeid)
294 list_for_each_entry(lkb, queue, lkb_statequeue)
295 if (!(lkb->lkb_flags & DLM_IFL_MSTCPY))
296 lkb->lkb_nodeid = nodeid;
299 static void set_master_lkbs(struct dlm_rsb *r)
301 set_lock_master(&r->res_grantqueue, r->res_nodeid);
302 set_lock_master(&r->res_convertqueue, r->res_nodeid);
303 set_lock_master(&r->res_waitqueue, r->res_nodeid);
307 * Propogate the new master nodeid to locks
308 * The NEW_MASTER flag tells dlm_recover_locks() which rsb's to consider.
309 * The NEW_MASTER2 flag tells recover_lvb() and set_locks_purged() which
313 static void set_new_master(struct dlm_rsb *r, int nodeid)
316 r->res_nodeid = nodeid;
318 rsb_set_flag(r, RSB_NEW_MASTER);
319 rsb_set_flag(r, RSB_NEW_MASTER2);
324 * We do async lookups on rsb's that need new masters. The rsb's
325 * waiting for a lookup reply are kept on the recover_list.
328 static int recover_master(struct dlm_rsb *r)
330 struct dlm_ls *ls = r->res_ls;
331 int error, dir_nodeid, ret_nodeid, our_nodeid = dlm_our_nodeid();
333 dir_nodeid = dlm_dir_nodeid(r);
335 if (dir_nodeid == our_nodeid) {
336 error = dlm_dir_lookup(ls, our_nodeid, r->res_name,
337 r->res_length, &ret_nodeid);
339 log_error(ls, "recover dir lookup error %d", error);
341 if (ret_nodeid == our_nodeid)
343 set_new_master(r, ret_nodeid);
346 error = dlm_send_rcom_lookup(r, dir_nodeid);
353 * When not using a directory, most resource names will hash to a new static
354 * master nodeid and the resource will need to be remastered.
357 static int recover_master_static(struct dlm_rsb *r)
359 int master = dlm_dir_nodeid(r);
361 if (master == dlm_our_nodeid())
364 if (r->res_nodeid != master) {
366 dlm_purge_mstcpy_locks(r);
367 set_new_master(r, master);
374 * Go through local root resources and for each rsb which has a master which
375 * has departed, get the new master nodeid from the directory. The dir will
376 * assign mastery to the first node to look up the new master. That means
377 * we'll discover in this lookup if we're the new master of any rsb's.
379 * We fire off all the dir lookup requests individually and asynchronously to
380 * the correct dir node.
383 int dlm_recover_masters(struct dlm_ls *ls)
386 int error = 0, count = 0;
388 log_debug(ls, "dlm_recover_masters");
390 down_read(&ls->ls_root_sem);
391 list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
392 if (dlm_recovery_stopped(ls)) {
393 up_read(&ls->ls_root_sem);
398 if (dlm_no_directory(ls))
399 count += recover_master_static(r);
400 else if (!is_master(r) &&
401 (dlm_is_removed(ls, r->res_nodeid) ||
402 rsb_flag(r, RSB_NEW_MASTER))) {
409 up_read(&ls->ls_root_sem);
411 log_debug(ls, "dlm_recover_masters %d resources", count);
413 error = dlm_wait_function(ls, &recover_list_empty);
416 recover_list_clear(ls);
420 int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
425 r = recover_list_find(ls, rc->rc_id);
427 log_error(ls, "dlm_recover_master_reply no id %llx",
428 (unsigned long long)rc->rc_id);
432 nodeid = rc->rc_result;
433 if (nodeid == dlm_our_nodeid())
436 set_new_master(r, nodeid);
439 if (recover_list_empty(ls))
440 wake_up(&ls->ls_wait_general);
446 /* Lock recovery: rebuild the process-copy locks we hold on a
447 remastered rsb on the new rsb master.
452 dlm_send_rcom_lock -> receive_rcom_lock
453 dlm_recover_master_copy
454 receive_rcom_lock_reply <-
455 dlm_recover_process_copy
460 * keep a count of the number of lkb's we send to the new master; when we get
461 * an equal number of replies then recovery for the rsb is done
464 static int recover_locks_queue(struct dlm_rsb *r, struct list_head *head)
469 list_for_each_entry(lkb, head, lkb_statequeue) {
470 error = dlm_send_rcom_lock(r, lkb);
473 r->res_recover_locks_count++;
479 static int recover_locks(struct dlm_rsb *r)
485 DLM_ASSERT(!r->res_recover_locks_count, dlm_dump_rsb(r););
487 error = recover_locks_queue(r, &r->res_grantqueue);
490 error = recover_locks_queue(r, &r->res_convertqueue);
493 error = recover_locks_queue(r, &r->res_waitqueue);
497 if (r->res_recover_locks_count)
500 rsb_clear_flag(r, RSB_NEW_MASTER);
506 int dlm_recover_locks(struct dlm_ls *ls)
509 int error, count = 0;
511 log_debug(ls, "dlm_recover_locks");
513 down_read(&ls->ls_root_sem);
514 list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
516 rsb_clear_flag(r, RSB_NEW_MASTER);
520 if (!rsb_flag(r, RSB_NEW_MASTER))
523 if (dlm_recovery_stopped(ls)) {
525 up_read(&ls->ls_root_sem);
529 error = recover_locks(r);
531 up_read(&ls->ls_root_sem);
535 count += r->res_recover_locks_count;
537 up_read(&ls->ls_root_sem);
539 log_debug(ls, "dlm_recover_locks %d locks", count);
541 error = dlm_wait_function(ls, &recover_list_empty);
544 recover_list_clear(ls);
546 dlm_set_recover_status(ls, DLM_RS_LOCKS);
550 void dlm_recovered_lock(struct dlm_rsb *r)
552 DLM_ASSERT(rsb_flag(r, RSB_NEW_MASTER), dlm_dump_rsb(r););
554 r->res_recover_locks_count--;
555 if (!r->res_recover_locks_count) {
556 rsb_clear_flag(r, RSB_NEW_MASTER);
560 if (recover_list_empty(r->res_ls))
561 wake_up(&r->res_ls->ls_wait_general);
565 * The lvb needs to be recovered on all master rsb's. This includes setting
566 * the VALNOTVALID flag if necessary, and determining the correct lvb contents
567 * based on the lvb's of the locks held on the rsb.
569 * RSB_VALNOTVALID is set if there are only NL/CR locks on the rsb. If it
570 * was already set prior to recovery, it's not cleared, regardless of locks.
572 * The LVB contents are only considered for changing when this is a new master
573 * of the rsb (NEW_MASTER2). Then, the rsb's lvb is taken from any lkb with
574 * mode > CR. If no lkb's exist with mode above CR, the lvb contents are taken
575 * from the lkb with the largest lvb sequence number.
578 static void recover_lvb(struct dlm_rsb *r)
580 struct dlm_lkb *lkb, *high_lkb = NULL;
581 uint32_t high_seq = 0;
582 int lock_lvb_exists = 0;
583 int big_lock_exists = 0;
584 int lvblen = r->res_ls->ls_lvblen;
586 list_for_each_entry(lkb, &r->res_grantqueue, lkb_statequeue) {
587 if (!(lkb->lkb_exflags & DLM_LKF_VALBLK))
592 if (lkb->lkb_grmode > DLM_LOCK_CR) {
597 if (((int)lkb->lkb_lvbseq - (int)high_seq) >= 0) {
599 high_seq = lkb->lkb_lvbseq;
603 list_for_each_entry(lkb, &r->res_convertqueue, lkb_statequeue) {
604 if (!(lkb->lkb_exflags & DLM_LKF_VALBLK))
609 if (lkb->lkb_grmode > DLM_LOCK_CR) {
614 if (((int)lkb->lkb_lvbseq - (int)high_seq) >= 0) {
616 high_seq = lkb->lkb_lvbseq;
621 if (!lock_lvb_exists)
624 if (!big_lock_exists)
625 rsb_set_flag(r, RSB_VALNOTVALID);
627 /* don't mess with the lvb unless we're the new master */
628 if (!rsb_flag(r, RSB_NEW_MASTER2))
631 if (!r->res_lvbptr) {
632 r->res_lvbptr = dlm_allocate_lvb(r->res_ls);
637 if (big_lock_exists) {
638 r->res_lvbseq = lkb->lkb_lvbseq;
639 memcpy(r->res_lvbptr, lkb->lkb_lvbptr, lvblen);
640 } else if (high_lkb) {
641 r->res_lvbseq = high_lkb->lkb_lvbseq;
642 memcpy(r->res_lvbptr, high_lkb->lkb_lvbptr, lvblen);
645 memset(r->res_lvbptr, 0, lvblen);
651 /* All master rsb's flagged RECOVER_CONVERT need to be looked at. The locks
652 converting PR->CW or CW->PR need to have their lkb_grmode set. */
654 static void recover_conversion(struct dlm_rsb *r)
659 list_for_each_entry(lkb, &r->res_grantqueue, lkb_statequeue) {
660 if (lkb->lkb_grmode == DLM_LOCK_PR ||
661 lkb->lkb_grmode == DLM_LOCK_CW) {
662 grmode = lkb->lkb_grmode;
667 list_for_each_entry(lkb, &r->res_convertqueue, lkb_statequeue) {
668 if (lkb->lkb_grmode != DLM_LOCK_IV)
671 lkb->lkb_grmode = lkb->lkb_rqmode;
673 lkb->lkb_grmode = grmode;
677 /* We've become the new master for this rsb and waiting/converting locks may
678 need to be granted in dlm_grant_after_purge() due to locks that may have
679 existed from a removed node. */
681 static void set_locks_purged(struct dlm_rsb *r)
683 if (!list_empty(&r->res_waitqueue) || !list_empty(&r->res_convertqueue))
684 rsb_set_flag(r, RSB_LOCKS_PURGED);
687 void dlm_recover_rsbs(struct dlm_ls *ls)
692 log_debug(ls, "dlm_recover_rsbs");
694 down_read(&ls->ls_root_sem);
695 list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
698 if (rsb_flag(r, RSB_RECOVER_CONVERT))
699 recover_conversion(r);
700 if (rsb_flag(r, RSB_NEW_MASTER2))
705 rsb_clear_flag(r, RSB_RECOVER_CONVERT);
706 rsb_clear_flag(r, RSB_NEW_MASTER2);
709 up_read(&ls->ls_root_sem);
711 log_debug(ls, "dlm_recover_rsbs %d rsbs", count);
714 /* Create a single list of all root rsb's to be used during recovery */
716 int dlm_create_root_list(struct dlm_ls *ls)
721 down_write(&ls->ls_root_sem);
722 if (!list_empty(&ls->ls_root_list)) {
723 log_error(ls, "root list not empty");
728 for (i = 0; i < ls->ls_rsbtbl_size; i++) {
729 spin_lock(&ls->ls_rsbtbl[i].lock);
730 list_for_each_entry(r, &ls->ls_rsbtbl[i].list, res_hashchain) {
731 list_add(&r->res_root_list, &ls->ls_root_list);
735 /* If we're using a directory, add tossed rsbs to the root
736 list; they'll have entries created in the new directory,
737 but no other recovery steps should do anything with them. */
739 if (dlm_no_directory(ls)) {
740 spin_unlock(&ls->ls_rsbtbl[i].lock);
744 list_for_each_entry(r, &ls->ls_rsbtbl[i].toss, res_hashchain) {
745 list_add(&r->res_root_list, &ls->ls_root_list);
748 spin_unlock(&ls->ls_rsbtbl[i].lock);
751 up_write(&ls->ls_root_sem);
755 void dlm_release_root_list(struct dlm_ls *ls)
757 struct dlm_rsb *r, *safe;
759 down_write(&ls->ls_root_sem);
760 list_for_each_entry_safe(r, safe, &ls->ls_root_list, res_root_list) {
761 list_del_init(&r->res_root_list);
764 up_write(&ls->ls_root_sem);
767 /* If not using a directory, clear the entire toss list, there's no benefit to
768 caching the master value since it's fixed. If we are using a dir, keep the
769 rsb's we're the master of. Recovery will add them to the root list and from
770 there they'll be entered in the rebuilt directory. */
772 void dlm_clear_toss_list(struct dlm_ls *ls)
774 struct dlm_rsb *r, *safe;
777 for (i = 0; i < ls->ls_rsbtbl_size; i++) {
778 spin_lock(&ls->ls_rsbtbl[i].lock);
779 list_for_each_entry_safe(r, safe, &ls->ls_rsbtbl[i].toss,
781 if (dlm_no_directory(ls) || !is_master(r)) {
782 list_del(&r->res_hashchain);
786 spin_unlock(&ls->ls_rsbtbl[i].lock);