?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ocfs2: fix locking for res->tracking and dlm->tracking_list
[linux/fpc-iii.git] / fs / ocfs2 / dlm / dlmdomain.c
blobc55a9c47ac17455c680f81fe8765ee15c590fe78
1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * dlmdomain.c
5 *
6 * defines domain join / leave apis
7 *
8 * Copyright (C) 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 *
25 */
26
27 #include <linux/module.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/init.h>
32 #include <linux/spinlock.h>
33 #include <linux/delay.h>
34 #include <linux/err.h>
35 #include <linux/debugfs.h>
36
37 #include "cluster/heartbeat.h"
38 #include "cluster/nodemanager.h"
39 #include "cluster/tcp.h"
40
41 #include "dlmapi.h"
42 #include "dlmcommon.h"
43 #include "dlmdomain.h"
44 #include "dlmdebug.h"
45
46 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_DOMAIN)
47 #include "cluster/masklog.h"
48
49 /*
50 * ocfs2 node maps are array of long int, which limits to send them freely
51 * across the wire due to endianness issues. To workaround this, we convert
52 * long ints to byte arrays. Following 3 routines are helper functions to
53 * set/test/copy bits within those array of bytes
54 */
55 static inline void byte_set_bit(u8 nr, u8 map[])
56 {
57 map[nr >> 3] |= (1UL << (nr & 7));
58 }
59
60 static inline int byte_test_bit(u8 nr, u8 map[])
61 {
62 return ((1UL << (nr & 7)) & (map[nr >> 3])) != 0;
63 }
64
65 static inline void byte_copymap(u8 dmap[], unsigned long smap[],
66 unsigned int sz)
67 {
68 unsigned int nn;
69
70 if (!sz)
71 return;
72
73 memset(dmap, 0, ((sz + 7) >> 3));
74 for (nn = 0 ; nn < sz; nn++)
75 if (test_bit(nn, smap))
76 byte_set_bit(nn, dmap);
77 }
78
79 static void dlm_free_pagevec(void **vec, int pages)
80 {
81 while (pages--)
82 free_page((unsigned long)vec[pages]);
83 kfree(vec);
84 }
85
86 static void **dlm_alloc_pagevec(int pages)
87 {
88 void **vec = kmalloc(pages * sizeof(void *), GFP_KERNEL);
89 int i;
90
91 if (!vec)
92 return NULL;
93
94 for (i = 0; i < pages; i++)
95 if (!(vec[i] = (void *)__get_free_page(GFP_KERNEL)))
96 goto out_free;
97
98 mlog(0, "Allocated DLM hash pagevec; %d pages (%lu expected), %lu buckets per page\n",
99 pages, (unsigned long)DLM_HASH_PAGES,
100 (unsigned long)DLM_BUCKETS_PER_PAGE);
101 return vec;
102 out_free:
103 dlm_free_pagevec(vec, i);
104 return NULL;
105 }
106
107 /*
108 *
109 * spinlock lock ordering: if multiple locks are needed, obey this ordering:
110 * dlm_domain_lock
111 * struct dlm_ctxt->spinlock
112 * struct dlm_lock_resource->spinlock
113 * struct dlm_ctxt->master_lock
114 * struct dlm_ctxt->ast_lock
115 * dlm_master_list_entry->spinlock
116 * dlm_lock->spinlock
117 *
118 */
119
120 DEFINE_SPINLOCK(dlm_domain_lock);
121 LIST_HEAD(dlm_domains);
122 static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events);
123
124 /*
125 * The supported protocol version for DLM communication. Running domains
126 * will have a negotiated version with the same major number and a minor
127 * number equal or smaller. The dlm_ctxt->dlm_locking_proto field should
128 * be used to determine what a running domain is actually using.
129 *
130 * New in version 1.1:
131 * - Message DLM_QUERY_REGION added to support global heartbeat
132 * - Message DLM_QUERY_NODEINFO added to allow online node removes
133 * New in version 1.2:
134 * - Message DLM_BEGIN_EXIT_DOMAIN_MSG added to mark start of exit domain
135 */
136 static const struct dlm_protocol_version dlm_protocol = {
137 .pv_major = 1,
138 .pv_minor = 2,
139 };
140
141 #define DLM_DOMAIN_BACKOFF_MS 200
142
143 static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
144 void **ret_data);
145 static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
146 void **ret_data);
147 static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
148 void **ret_data);
149 static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
150 void *data, void **ret_data);
151 static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
152 void **ret_data);
153 static int dlm_protocol_compare(struct dlm_protocol_version *existing,
154 struct dlm_protocol_version *request);
155
156 static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm);
157
158 void __dlm_unhash_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
159 {
160 if (hlist_unhashed(&res->hash_node))
161 return;
162
163 mlog(0, "%s: Unhash res %.*s\n", dlm->name, res->lockname.len,
164 res->lockname.name);
165 hlist_del_init(&res->hash_node);
166 dlm_lockres_put(res);
167 }
168
169 void __dlm_insert_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
170 {
171 struct hlist_head *bucket;
172 struct qstr *q;
173
174 assert_spin_locked(&dlm->spinlock);
175
176 q = &res->lockname;
177 bucket = dlm_lockres_hash(dlm, q->hash);
178
179 /* get a reference for our hashtable */
180 dlm_lockres_get(res);
181
182 hlist_add_head(&res->hash_node, bucket);
183
184 mlog(0, "%s: Hash res %.*s\n", dlm->name, res->lockname.len,
185 res->lockname.name);
186 }
187
188 struct dlm_lock_resource * __dlm_lookup_lockres_full(struct dlm_ctxt *dlm,
189 const char *name,
190 unsigned int len,
191 unsigned int hash)
192 {
193 struct hlist_head *bucket;
194 struct dlm_lock_resource *res;
195
196 mlog(0, "%.*s\n", len, name);
197
198 assert_spin_locked(&dlm->spinlock);
199
200 bucket = dlm_lockres_hash(dlm, hash);
201
202 hlist_for_each_entry(res, bucket, hash_node) {
203 if (res->lockname.name[0] != name[0])
204 continue;
205 if (unlikely(res->lockname.len != len))
206 continue;
207 if (memcmp(res->lockname.name + 1, name + 1, len - 1))
208 continue;
209 dlm_lockres_get(res);
210 return res;
211 }
212 return NULL;
213 }
214
215 /* intended to be called by functions which do not care about lock
216 * resources which are being purged (most net _handler functions).
217 * this will return NULL for any lock resource which is found but
218 * currently in the process of dropping its mastery reference.
219 * use __dlm_lookup_lockres_full when you need the lock resource
220 * regardless (e.g. dlm_get_lock_resource) */
221 struct dlm_lock_resource * __dlm_lookup_lockres(struct dlm_ctxt *dlm,
222 const char *name,
223 unsigned int len,
224 unsigned int hash)
225 {
226 struct dlm_lock_resource *res = NULL;
227
228 mlog(0, "%.*s\n", len, name);
229
230 assert_spin_locked(&dlm->spinlock);
231
232 res = __dlm_lookup_lockres_full(dlm, name, len, hash);
233 if (res) {
234 spin_lock(&res->spinlock);
235 if (res->state & DLM_LOCK_RES_DROPPING_REF) {
236 spin_unlock(&res->spinlock);
237 dlm_lockres_put(res);
238 return NULL;
239 }
240 spin_unlock(&res->spinlock);
241 }
242
243 return res;
244 }
245
246 struct dlm_lock_resource * dlm_lookup_lockres(struct dlm_ctxt *dlm,
247 const char *name,
248 unsigned int len)
249 {
250 struct dlm_lock_resource *res;
251 unsigned int hash = dlm_lockid_hash(name, len);
252
253 spin_lock(&dlm->spinlock);
254 res = __dlm_lookup_lockres(dlm, name, len, hash);
255 spin_unlock(&dlm->spinlock);
256 return res;
257 }
258
259 static struct dlm_ctxt * __dlm_lookup_domain_full(const char *domain, int len)
260 {
261 struct dlm_ctxt *tmp;
262
263 assert_spin_locked(&dlm_domain_lock);
264
265 /* tmp->name here is always NULL terminated,
266 * but domain may not be! */
267 list_for_each_entry(tmp, &dlm_domains, list) {
268 if (strlen(tmp->name) == len &&
269 memcmp(tmp->name, domain, len)==0)
270 return tmp;
271 }
272
273 return NULL;
274 }
275
276 /* For null terminated domain strings ONLY */
277 static struct dlm_ctxt * __dlm_lookup_domain(const char *domain)
278 {
279 assert_spin_locked(&dlm_domain_lock);
280
281 return __dlm_lookup_domain_full(domain, strlen(domain));
282 }
283
284
285 /* returns true on one of two conditions:
286 * 1) the domain does not exist
287 * 2) the domain exists and it's state is "joined" */
288 static int dlm_wait_on_domain_helper(const char *domain)
289 {
290 int ret = 0;
291 struct dlm_ctxt *tmp = NULL;
292
293 spin_lock(&dlm_domain_lock);
294
295 tmp = __dlm_lookup_domain(domain);
296 if (!tmp)
297 ret = 1;
298 else if (tmp->dlm_state == DLM_CTXT_JOINED)
299 ret = 1;
300
301 spin_unlock(&dlm_domain_lock);
302 return ret;
303 }
304
305 static void dlm_free_ctxt_mem(struct dlm_ctxt *dlm)
306 {
307 dlm_destroy_debugfs_subroot(dlm);
308
309 if (dlm->lockres_hash)
310 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES);
311
312 if (dlm->master_hash)
313 dlm_free_pagevec((void **)dlm->master_hash, DLM_HASH_PAGES);
314
315 kfree(dlm->name);
316 kfree(dlm);
317 }
318
319 /* A little strange - this function will be called while holding
320 * dlm_domain_lock and is expected to be holding it on the way out. We
321 * will however drop and reacquire it multiple times */
322 static void dlm_ctxt_release(struct kref *kref)
323 {
324 struct dlm_ctxt *dlm;
325
326 dlm = container_of(kref, struct dlm_ctxt, dlm_refs);
327
328 BUG_ON(dlm->num_joins);
329 BUG_ON(dlm->dlm_state == DLM_CTXT_JOINED);
330
331 /* we may still be in the list if we hit an error during join. */
332 list_del_init(&dlm->list);
333
334 spin_unlock(&dlm_domain_lock);
335
336 mlog(0, "freeing memory from domain %s\n", dlm->name);
337
338 wake_up(&dlm_domain_events);
339
340 dlm_free_ctxt_mem(dlm);
341
342 spin_lock(&dlm_domain_lock);
343 }
344
345 void dlm_put(struct dlm_ctxt *dlm)
346 {
347 spin_lock(&dlm_domain_lock);
348 kref_put(&dlm->dlm_refs, dlm_ctxt_release);
349 spin_unlock(&dlm_domain_lock);
350 }
351
352 static void __dlm_get(struct dlm_ctxt *dlm)
353 {
354 kref_get(&dlm->dlm_refs);
355 }
356
357 /* given a questionable reference to a dlm object, gets a reference if
358 * it can find it in the list, otherwise returns NULL in which case
359 * you shouldn't trust your pointer. */
360 struct dlm_ctxt *dlm_grab(struct dlm_ctxt *dlm)
361 {
362 struct dlm_ctxt *target;
363 struct dlm_ctxt *ret = NULL;
364
365 spin_lock(&dlm_domain_lock);
366
367 list_for_each_entry(target, &dlm_domains, list) {
368 if (target == dlm) {
369 __dlm_get(target);
370 ret = target;
371 break;
372 }
373 }
374
375 spin_unlock(&dlm_domain_lock);
376
377 return ret;
378 }
379
380 int dlm_domain_fully_joined(struct dlm_ctxt *dlm)
381 {
382 int ret;
383
384 spin_lock(&dlm_domain_lock);
385 ret = (dlm->dlm_state == DLM_CTXT_JOINED) ||
386 (dlm->dlm_state == DLM_CTXT_IN_SHUTDOWN);
387 spin_unlock(&dlm_domain_lock);
388
389 return ret;
390 }
391
392 static void dlm_destroy_dlm_worker(struct dlm_ctxt *dlm)
393 {
394 if (dlm->dlm_worker) {
395 flush_workqueue(dlm->dlm_worker);
396 destroy_workqueue(dlm->dlm_worker);
397 dlm->dlm_worker = NULL;
398 }
399 }
400
401 static void dlm_complete_dlm_shutdown(struct dlm_ctxt *dlm)
402 {
403 dlm_unregister_domain_handlers(dlm);
404 dlm_debug_shutdown(dlm);
405 dlm_complete_thread(dlm);
406 dlm_complete_recovery_thread(dlm);
407 dlm_destroy_dlm_worker(dlm);
408
409 /* We've left the domain. Now we can take ourselves out of the
410 * list and allow the kref stuff to help us free the
411 * memory. */
412 spin_lock(&dlm_domain_lock);
413 list_del_init(&dlm->list);
414 spin_unlock(&dlm_domain_lock);
415
416 /* Wake up anyone waiting for us to remove this domain */
417 wake_up(&dlm_domain_events);
418 }
419
420 static int dlm_migrate_all_locks(struct dlm_ctxt *dlm)
421 {
422 int i, num, n, ret = 0;
423 struct dlm_lock_resource *res;
424 struct hlist_node *iter;
425 struct hlist_head *bucket;
426 int dropped;
427
428 mlog(0, "Migrating locks from domain %s\n", dlm->name);
429
430 num = 0;
431 spin_lock(&dlm->spinlock);
432 for (i = 0; i < DLM_HASH_BUCKETS; i++) {
433 redo_bucket:
434 n = 0;
435 bucket = dlm_lockres_hash(dlm, i);
436 iter = bucket->first;
437 while (iter) {
438 n++;
439 res = hlist_entry(iter, struct dlm_lock_resource,
440 hash_node);
441 dlm_lockres_get(res);
442 /* migrate, if necessary. this will drop the dlm
443 * spinlock and retake it if it does migration. */
444 dropped = dlm_empty_lockres(dlm, res);
445
446 spin_lock(&res->spinlock);
447 if (dropped)
448 __dlm_lockres_calc_usage(dlm, res);
449 else
450 iter = res->hash_node.next;
451 spin_unlock(&res->spinlock);
452
453 dlm_lockres_put(res);
454
455 if (dropped) {
456 cond_resched_lock(&dlm->spinlock);
457 goto redo_bucket;
458 }
459 }
460 cond_resched_lock(&dlm->spinlock);
461 num += n;
462 }
463 spin_unlock(&dlm->spinlock);
464 wake_up(&dlm->dlm_thread_wq);
465
466 /* let the dlm thread take care of purging, keep scanning until
467 * nothing remains in the hash */
468 if (num) {
469 mlog(0, "%s: %d lock resources in hash last pass\n",
470 dlm->name, num);
471 ret = -EAGAIN;
472 }
473 mlog(0, "DONE Migrating locks from domain %s\n", dlm->name);
474 return ret;
475 }
476
477 static int dlm_no_joining_node(struct dlm_ctxt *dlm)
478 {
479 int ret;
480
481 spin_lock(&dlm->spinlock);
482 ret = dlm->joining_node == DLM_LOCK_RES_OWNER_UNKNOWN;
483 spin_unlock(&dlm->spinlock);
484
485 return ret;
486 }
487
488 static int dlm_begin_exit_domain_handler(struct o2net_msg *msg, u32 len,
489 void *data, void **ret_data)
490 {
491 struct dlm_ctxt *dlm = data;
492 unsigned int node;
493 struct dlm_exit_domain *exit_msg = (struct dlm_exit_domain *) msg->buf;
494
495 if (!dlm_grab(dlm))
496 return 0;
497
498 node = exit_msg->node_idx;
499 mlog(0, "%s: Node %u sent a begin exit domain message\n", dlm->name, node);
500
501 spin_lock(&dlm->spinlock);
502 set_bit(node, dlm->exit_domain_map);
503 spin_unlock(&dlm->spinlock);
504
505 dlm_put(dlm);
506
507 return 0;
508 }
509
510 static void dlm_mark_domain_leaving(struct dlm_ctxt *dlm)
511 {
512 /* Yikes, a double spinlock! I need domain_lock for the dlm
513 * state and the dlm spinlock for join state... Sorry! */
514 again:
515 spin_lock(&dlm_domain_lock);
516 spin_lock(&dlm->spinlock);
517
518 if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) {
519 mlog(0, "Node %d is joining, we wait on it.\n",
520 dlm->joining_node);
521 spin_unlock(&dlm->spinlock);
522 spin_unlock(&dlm_domain_lock);
523
524 wait_event(dlm->dlm_join_events, dlm_no_joining_node(dlm));
525 goto again;
526 }
527
528 dlm->dlm_state = DLM_CTXT_LEAVING;
529 spin_unlock(&dlm->spinlock);
530 spin_unlock(&dlm_domain_lock);
531 }
532
533 static void __dlm_print_nodes(struct dlm_ctxt *dlm)
534 {
535 int node = -1, num = 0;
536
537 assert_spin_locked(&dlm->spinlock);
538
539 printk("( ");
540 while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES,
541 node + 1)) < O2NM_MAX_NODES) {
542 printk("%d ", node);
543 ++num;
544 }
545 printk(") %u nodes\n", num);
546 }
547
548 static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
549 void **ret_data)
550 {
551 struct dlm_ctxt *dlm = data;
552 unsigned int node;
553 struct dlm_exit_domain *exit_msg = (struct dlm_exit_domain *) msg->buf;
554
555 mlog(0, "%p %u %p", msg, len, data);
556
557 if (!dlm_grab(dlm))
558 return 0;
559
560 node = exit_msg->node_idx;
561
562 spin_lock(&dlm->spinlock);
563 clear_bit(node, dlm->domain_map);
564 clear_bit(node, dlm->exit_domain_map);
565 printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s ", node, dlm->name);
566 __dlm_print_nodes(dlm);
567
568 /* notify anything attached to the heartbeat events */
569 dlm_hb_event_notify_attached(dlm, node, 0);
570
571 spin_unlock(&dlm->spinlock);
572
573 dlm_put(dlm);
574
575 return 0;
576 }
577
578 static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm, u32 msg_type,
579 unsigned int node)
580 {
581 int status;
582 struct dlm_exit_domain leave_msg;
583
584 mlog(0, "%s: Sending domain exit message %u to node %u\n", dlm->name,
585 msg_type, node);
586
587 memset(&leave_msg, 0, sizeof(leave_msg));
588 leave_msg.node_idx = dlm->node_num;
589
590 status = o2net_send_message(msg_type, dlm->key, &leave_msg,
591 sizeof(leave_msg), node, NULL);
592 if (status < 0)
593 mlog(ML_ERROR, "Error %d sending domain exit message %u "
594 "to node %u on domain %s\n", status, msg_type, node,
595 dlm->name);
596
597 return status;
598 }
599
600 static void dlm_begin_exit_domain(struct dlm_ctxt *dlm)
601 {
602 int node = -1;
603
604 /* Support for begin exit domain was added in 1.2 */
605 if (dlm->dlm_locking_proto.pv_major == 1 &&
606 dlm->dlm_locking_proto.pv_minor < 2)
607 return;
608
609 /*
610 * Unlike DLM_EXIT_DOMAIN_MSG, DLM_BEGIN_EXIT_DOMAIN_MSG is purely
611 * informational. Meaning if a node does not receive the message,
612 * so be it.
613 */
614 spin_lock(&dlm->spinlock);
615 while (1) {
616 node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES, node + 1);
617 if (node >= O2NM_MAX_NODES)
618 break;
619 if (node == dlm->node_num)
620 continue;
621
622 spin_unlock(&dlm->spinlock);
623 dlm_send_one_domain_exit(dlm, DLM_BEGIN_EXIT_DOMAIN_MSG, node);
624 spin_lock(&dlm->spinlock);
625 }
626 spin_unlock(&dlm->spinlock);
627 }
628
629 static void dlm_leave_domain(struct dlm_ctxt *dlm)
630 {
631 int node, clear_node, status;
632
633 /* At this point we've migrated away all our locks and won't
634 * accept mastership of new ones. The dlm is responsible for
635 * almost nothing now. We make sure not to confuse any joining
636 * nodes and then commence shutdown procedure. */
637
638 spin_lock(&dlm->spinlock);
639 /* Clear ourselves from the domain map */
640 clear_bit(dlm->node_num, dlm->domain_map);
641 while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES,
642 0)) < O2NM_MAX_NODES) {
643 /* Drop the dlm spinlock. This is safe wrt the domain_map.
644 * -nodes cannot be added now as the
645 * query_join_handlers knows to respond with OK_NO_MAP
646 * -we catch the right network errors if a node is
647 * removed from the map while we're sending him the
648 * exit message. */
649 spin_unlock(&dlm->spinlock);
650
651 clear_node = 1;
652
653 status = dlm_send_one_domain_exit(dlm, DLM_EXIT_DOMAIN_MSG,
654 node);
655 if (status < 0 &&
656 status != -ENOPROTOOPT &&
657 status != -ENOTCONN) {
658 mlog(ML_NOTICE, "Error %d sending domain exit message "
659 "to node %d\n", status, node);
660
661 /* Not sure what to do here but lets sleep for
662 * a bit in case this was a transient
663 * error... */
664 msleep(DLM_DOMAIN_BACKOFF_MS);
665 clear_node = 0;
666 }
667
668 spin_lock(&dlm->spinlock);
669 /* If we're not clearing the node bit then we intend
670 * to loop back around to try again. */
671 if (clear_node)
672 clear_bit(node, dlm->domain_map);
673 }
674 spin_unlock(&dlm->spinlock);
675 }
676
677 void dlm_unregister_domain(struct dlm_ctxt *dlm)
678 {
679 int leave = 0;
680 struct dlm_lock_resource *res;
681
682 spin_lock(&dlm_domain_lock);
683 BUG_ON(dlm->dlm_state != DLM_CTXT_JOINED);
684 BUG_ON(!dlm->num_joins);
685
686 dlm->num_joins--;
687 if (!dlm->num_joins) {
688 /* We mark it "in shutdown" now so new register
689 * requests wait until we've completely left the
690 * domain. Don't use DLM_CTXT_LEAVING yet as we still
691 * want new domain joins to communicate with us at
692 * least until we've completed migration of our
693 * resources. */
694 dlm->dlm_state = DLM_CTXT_IN_SHUTDOWN;
695 leave = 1;
696 }
697 spin_unlock(&dlm_domain_lock);
698
699 if (leave) {
700 mlog(0, "shutting down domain %s\n", dlm->name);
701 dlm_begin_exit_domain(dlm);
702
703 /* We changed dlm state, notify the thread */
704 dlm_kick_thread(dlm, NULL);
705
706 while (dlm_migrate_all_locks(dlm)) {
707 /* Give dlm_thread time to purge the lockres' */
708 msleep(500);
709 mlog(0, "%s: more migration to do\n", dlm->name);
710 }
711
712 /* This list should be empty. If not, print remaining lockres */
713 if (!list_empty(&dlm->tracking_list)) {
714 mlog(ML_ERROR, "Following lockres' are still on the "
715 "tracking list:\n");
716 list_for_each_entry(res, &dlm->tracking_list, tracking)
717 dlm_print_one_lock_resource(res);
718 }
719
720 dlm_mark_domain_leaving(dlm);
721 dlm_leave_domain(dlm);
722 printk(KERN_NOTICE "o2dlm: Leaving domain %s\n", dlm->name);
723 dlm_force_free_mles(dlm);
724 dlm_complete_dlm_shutdown(dlm);
725 }
726 dlm_put(dlm);
727 }
728 EXPORT_SYMBOL_GPL(dlm_unregister_domain);
729
730 static int dlm_query_join_proto_check(char *proto_type, int node,
731 struct dlm_protocol_version *ours,
732 struct dlm_protocol_version *request)
733 {
734 int rc;
735 struct dlm_protocol_version proto = *request;
736
737 if (!dlm_protocol_compare(ours, &proto)) {
738 mlog(0,
739 "node %u wanted to join with %s locking protocol "
740 "%u.%u, we respond with %u.%u\n",
741 node, proto_type,
742 request->pv_major,
743 request->pv_minor,
744 proto.pv_major, proto.pv_minor);
745 request->pv_minor = proto.pv_minor;
746 rc = 0;
747 } else {
748 mlog(ML_NOTICE,
749 "Node %u wanted to join with %s locking "
750 "protocol %u.%u, but we have %u.%u, disallowing\n",
751 node, proto_type,
752 request->pv_major,
753 request->pv_minor,
754 ours->pv_major,
755 ours->pv_minor);
756 rc = 1;
757 }
758
759 return rc;
760 }
761
762 /*
763 * struct dlm_query_join_packet is made up of four one-byte fields. They
764 * are effectively in big-endian order already. However, little-endian
765 * machines swap them before putting the packet on the wire (because
766 * query_join's response is a status, and that status is treated as a u32
767 * on the wire). Thus, a big-endian and little-endian machines will treat
768 * this structure differently.
769 *
770 * The solution is to have little-endian machines swap the structure when
771 * converting from the structure to the u32 representation. This will
772 * result in the structure having the correct format on the wire no matter
773 * the host endian format.
774 */
775 static void dlm_query_join_packet_to_wire(struct dlm_query_join_packet *packet,
776 u32 *wire)
777 {
778 union dlm_query_join_response response;
779
780 response.packet = *packet;
781 *wire = be32_to_cpu(response.intval);
782 }
783
784 static void dlm_query_join_wire_to_packet(u32 wire,
785 struct dlm_query_join_packet *packet)
786 {
787 union dlm_query_join_response response;
788
789 response.intval = cpu_to_be32(wire);
790 *packet = response.packet;
791 }
792
793 static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
794 void **ret_data)
795 {
796 struct dlm_query_join_request *query;
797 struct dlm_query_join_packet packet = {
798 .code = JOIN_DISALLOW,
799 };
800 struct dlm_ctxt *dlm = NULL;
801 u32 response;
802 u8 nodenum;
803
804 query = (struct dlm_query_join_request *) msg->buf;
805
806 mlog(0, "node %u wants to join domain %s\n", query->node_idx,
807 query->domain);
808
809 /*
810 * If heartbeat doesn't consider the node live, tell it
811 * to back off and try again. This gives heartbeat a chance
812 * to catch up.
813 */
814 if (!o2hb_check_node_heartbeating_no_sem(query->node_idx)) {
815 mlog(0, "node %u is not in our live map yet\n",
816 query->node_idx);
817
818 packet.code = JOIN_DISALLOW;
819 goto respond;
820 }
821
822 packet.code = JOIN_OK_NO_MAP;
823
824 spin_lock(&dlm_domain_lock);
825 dlm = __dlm_lookup_domain_full(query->domain, query->name_len);
826 if (!dlm)
827 goto unlock_respond;
828
829 /*
830 * There is a small window where the joining node may not see the
831 * node(s) that just left but still part of the cluster. DISALLOW
832 * join request if joining node has different node map.
833 */
834 nodenum=0;
835 while (nodenum < O2NM_MAX_NODES) {
836 if (test_bit(nodenum, dlm->domain_map)) {
837 if (!byte_test_bit(nodenum, query->node_map)) {
838 mlog(0, "disallow join as node %u does not "
839 "have node %u in its nodemap\n",
840 query->node_idx, nodenum);
841 packet.code = JOIN_DISALLOW;
842 goto unlock_respond;
843 }
844 }
845 nodenum++;
846 }
847
848 /* Once the dlm ctxt is marked as leaving then we don't want
849 * to be put in someone's domain map.
850 * Also, explicitly disallow joining at certain troublesome
851 * times (ie. during recovery). */
852 if (dlm->dlm_state != DLM_CTXT_LEAVING) {
853 int bit = query->node_idx;
854 spin_lock(&dlm->spinlock);
855
856 if (dlm->dlm_state == DLM_CTXT_NEW &&
857 dlm->joining_node == DLM_LOCK_RES_OWNER_UNKNOWN) {
858 /*If this is a brand new context and we
859 * haven't started our join process yet, then
860 * the other node won the race. */
861 packet.code = JOIN_OK_NO_MAP;
862 } else if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) {
863 /* Disallow parallel joins. */
864 packet.code = JOIN_DISALLOW;
865 } else if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) {
866 mlog(0, "node %u trying to join, but recovery "
867 "is ongoing.\n", bit);
868 packet.code = JOIN_DISALLOW;
869 } else if (test_bit(bit, dlm->recovery_map)) {
870 mlog(0, "node %u trying to join, but it "
871 "still needs recovery.\n", bit);
872 packet.code = JOIN_DISALLOW;
873 } else if (test_bit(bit, dlm->domain_map)) {
874 mlog(0, "node %u trying to join, but it "
875 "is still in the domain! needs recovery?\n",
876 bit);
877 packet.code = JOIN_DISALLOW;
878 } else {
879 /* Alright we're fully a part of this domain
880 * so we keep some state as to who's joining
881 * and indicate to him that needs to be fixed
882 * up. */
883
884 /* Make sure we speak compatible locking protocols. */
885 if (dlm_query_join_proto_check("DLM", bit,
886 &dlm->dlm_locking_proto,
887 &query->dlm_proto)) {
888 packet.code = JOIN_PROTOCOL_MISMATCH;
889 } else if (dlm_query_join_proto_check("fs", bit,
890 &dlm->fs_locking_proto,
891 &query->fs_proto)) {
892 packet.code = JOIN_PROTOCOL_MISMATCH;
893 } else {
894 packet.dlm_minor = query->dlm_proto.pv_minor;
895 packet.fs_minor = query->fs_proto.pv_minor;
896 packet.code = JOIN_OK;
897 __dlm_set_joining_node(dlm, query->node_idx);
898 }
899 }
900
901 spin_unlock(&dlm->spinlock);
902 }
903 unlock_respond:
904 spin_unlock(&dlm_domain_lock);
905
906 respond:
907 mlog(0, "We respond with %u\n", packet.code);
908
909 dlm_query_join_packet_to_wire(&packet, &response);
910 return response;
911 }
912
913 static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
914 void **ret_data)
915 {
916 struct dlm_assert_joined *assert;
917 struct dlm_ctxt *dlm = NULL;
918
919 assert = (struct dlm_assert_joined *) msg->buf;
920
921 mlog(0, "node %u asserts join on domain %s\n", assert->node_idx,
922 assert->domain);
923
924 spin_lock(&dlm_domain_lock);
925 dlm = __dlm_lookup_domain_full(assert->domain, assert->name_len);
926 /* XXX should we consider no dlm ctxt an error? */
927 if (dlm) {
928 spin_lock(&dlm->spinlock);
929
930 /* Alright, this node has officially joined our
931 * domain. Set him in the map and clean up our
932 * leftover join state. */
933 BUG_ON(dlm->joining_node != assert->node_idx);
934
935 if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) {
936 mlog(0, "dlm recovery is ongoing, disallow join\n");
937 spin_unlock(&dlm->spinlock);
938 spin_unlock(&dlm_domain_lock);
939 return -EAGAIN;
940 }
941
942 set_bit(assert->node_idx, dlm->domain_map);
943 clear_bit(assert->node_idx, dlm->exit_domain_map);
944 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
945
946 printk(KERN_NOTICE "o2dlm: Node %u joins domain %s ",
947 assert->node_idx, dlm->name);
948 __dlm_print_nodes(dlm);
949
950 /* notify anything attached to the heartbeat events */
951 dlm_hb_event_notify_attached(dlm, assert->node_idx, 1);
952
953 spin_unlock(&dlm->spinlock);
954 }
955 spin_unlock(&dlm_domain_lock);
956
957 return 0;
958 }
959
960 static int dlm_match_regions(struct dlm_ctxt *dlm,
961 struct dlm_query_region *qr,
962 char *local, int locallen)
963 {
964 char *remote = qr->qr_regions;
965 char *l, *r;
966 int localnr, i, j, foundit;
967 int status = 0;
968
969 if (!o2hb_global_heartbeat_active()) {
970 if (qr->qr_numregions) {
971 mlog(ML_ERROR, "Domain %s: Joining node %d has global "
972 "heartbeat enabled but local node %d does not\n",
973 qr->qr_domain, qr->qr_node, dlm->node_num);
974 status = -EINVAL;
975 }
976 goto bail;
977 }
978
979 if (o2hb_global_heartbeat_active() && !qr->qr_numregions) {
980 mlog(ML_ERROR, "Domain %s: Local node %d has global "
981 "heartbeat enabled but joining node %d does not\n",
982 qr->qr_domain, dlm->node_num, qr->qr_node);
983 status = -EINVAL;
984 goto bail;
985 }
986
987 r = remote;
988 for (i = 0; i < qr->qr_numregions; ++i) {
989 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, r);
990 r += O2HB_MAX_REGION_NAME_LEN;
991 }
992
993 localnr = min(O2NM_MAX_REGIONS, locallen/O2HB_MAX_REGION_NAME_LEN);
994 localnr = o2hb_get_all_regions(local, (u8)localnr);
995
996 /* compare local regions with remote */
997 l = local;
998 for (i = 0; i < localnr; ++i) {
999 foundit = 0;
1000 r = remote;
1001 for (j = 0; j <= qr->qr_numregions; ++j) {
1002 if (!memcmp(l, r, O2HB_MAX_REGION_NAME_LEN)) {
1003 foundit = 1;
1004 break;
1005 }
1006 r += O2HB_MAX_REGION_NAME_LEN;
1007 }
1008 if (!foundit) {
1009 status = -EINVAL;
1010 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
1011 "in local node %d but not in joining node %d\n",
1012 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, l,
1013 dlm->node_num, qr->qr_node);
1014 goto bail;
1015 }
1016 l += O2HB_MAX_REGION_NAME_LEN;
1017 }
1018
1019 /* compare remote with local regions */
1020 r = remote;
1021 for (i = 0; i < qr->qr_numregions; ++i) {
1022 foundit = 0;
1023 l = local;
1024 for (j = 0; j < localnr; ++j) {
1025 if (!memcmp(r, l, O2HB_MAX_REGION_NAME_LEN)) {
1026 foundit = 1;
1027 break;
1028 }
1029 l += O2HB_MAX_REGION_NAME_LEN;
1030 }
1031 if (!foundit) {
1032 status = -EINVAL;
1033 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered "
1034 "in joining node %d but not in local node %d\n",
1035 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, r,
1036 qr->qr_node, dlm->node_num);
1037 goto bail;
1038 }
1039 r += O2HB_MAX_REGION_NAME_LEN;
1040 }
1041
1042 bail:
1043 return status;
1044 }
1045
1046 static int dlm_send_regions(struct dlm_ctxt *dlm, unsigned long *node_map)
1047 {
1048 struct dlm_query_region *qr = NULL;
1049 int status, ret = 0, i;
1050 char *p;
1051
1052 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES)
1053 goto bail;
1054
1055 qr = kzalloc(sizeof(struct dlm_query_region), GFP_KERNEL);
1056 if (!qr) {
1057 ret = -ENOMEM;
1058 mlog_errno(ret);
1059 goto bail;
1060 }
1061
1062 qr->qr_node = dlm->node_num;
1063 qr->qr_namelen = strlen(dlm->name);
1064 memcpy(qr->qr_domain, dlm->name, qr->qr_namelen);
1065 /* if local hb, the numregions will be zero */
1066 if (o2hb_global_heartbeat_active())
1067 qr->qr_numregions = o2hb_get_all_regions(qr->qr_regions,
1068 O2NM_MAX_REGIONS);
1069
1070 p = qr->qr_regions;
1071 for (i = 0; i < qr->qr_numregions; ++i, p += O2HB_MAX_REGION_NAME_LEN)
1072 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, p);
1073
1074 i = -1;
1075 while ((i = find_next_bit(node_map, O2NM_MAX_NODES,
1076 i + 1)) < O2NM_MAX_NODES) {
1077 if (i == dlm->node_num)
1078 continue;
1079
1080 mlog(0, "Sending regions to node %d\n", i);
1081
1082 ret = o2net_send_message(DLM_QUERY_REGION, DLM_MOD_KEY, qr,
1083 sizeof(struct dlm_query_region),
1084 i, &status);
1085 if (ret >= 0)
1086 ret = status;
1087 if (ret) {
1088 mlog(ML_ERROR, "Region mismatch %d, node %d\n",
1089 ret, i);
1090 break;
1091 }
1092 }
1093
1094 bail:
1095 kfree(qr);
1096 return ret;
1097 }
1098
1099 static int dlm_query_region_handler(struct o2net_msg *msg, u32 len,
1100 void *data, void **ret_data)
1101 {
1102 struct dlm_query_region *qr;
1103 struct dlm_ctxt *dlm = NULL;
1104 char *local = NULL;
1105 int status = 0;
1106
1107 qr = (struct dlm_query_region *) msg->buf;
1108
1109 mlog(0, "Node %u queries hb regions on domain %s\n", qr->qr_node,
1110 qr->qr_domain);
1111
1112 /* buffer used in dlm_mast_regions() */
1113 local = kmalloc(sizeof(qr->qr_regions), GFP_KERNEL);
1114 if (!local)
1115 return -ENOMEM;
1116
1117 status = -EINVAL;
1118
1119 spin_lock(&dlm_domain_lock);
1120 dlm = __dlm_lookup_domain_full(qr->qr_domain, qr->qr_namelen);
1121 if (!dlm) {
1122 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1123 "before join domain\n", qr->qr_node, qr->qr_domain);
1124 goto out_domain_lock;
1125 }
1126
1127 spin_lock(&dlm->spinlock);
1128 if (dlm->joining_node != qr->qr_node) {
1129 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1130 "but joining node is %d\n", qr->qr_node, qr->qr_domain,
1131 dlm->joining_node);
1132 goto out_dlm_lock;
1133 }
1134
1135 /* Support for global heartbeat was added in 1.1 */
1136 if (dlm->dlm_locking_proto.pv_major == 1 &&
1137 dlm->dlm_locking_proto.pv_minor == 0) {
1138 mlog(ML_ERROR, "Node %d queried hb regions on domain %s "
1139 "but active dlm protocol is %d.%d\n", qr->qr_node,
1140 qr->qr_domain, dlm->dlm_locking_proto.pv_major,
1141 dlm->dlm_locking_proto.pv_minor);
1142 goto out_dlm_lock;
1143 }
1144
1145 status = dlm_match_regions(dlm, qr, local, sizeof(qr->qr_regions));
1146
1147 out_dlm_lock:
1148 spin_unlock(&dlm->spinlock);
1149
1150 out_domain_lock:
1151 spin_unlock(&dlm_domain_lock);
1152
1153 kfree(local);
1154
1155 return status;
1156 }
1157
1158 static int dlm_match_nodes(struct dlm_ctxt *dlm, struct dlm_query_nodeinfo *qn)
1159 {
1160 struct o2nm_node *local;
1161 struct dlm_node_info *remote;
1162 int i, j;
1163 int status = 0;
1164
1165 for (j = 0; j < qn->qn_numnodes; ++j)
1166 mlog(0, "Node %3d, %pI4:%u\n", qn->qn_nodes[j].ni_nodenum,
1167 &(qn->qn_nodes[j].ni_ipv4_address),
1168 ntohs(qn->qn_nodes[j].ni_ipv4_port));
1169
1170 for (i = 0; i < O2NM_MAX_NODES && !status; ++i) {
1171 local = o2nm_get_node_by_num(i);
1172 remote = NULL;
1173 for (j = 0; j < qn->qn_numnodes; ++j) {
1174 if (qn->qn_nodes[j].ni_nodenum == i) {
1175 remote = &(qn->qn_nodes[j]);
1176 break;
1177 }
1178 }
1179
1180 if (!local && !remote)
1181 continue;
1182
1183 if ((local && !remote) || (!local && remote))
1184 status = -EINVAL;
1185
1186 if (!status &&
1187 ((remote->ni_nodenum != local->nd_num) ||
1188 (remote->ni_ipv4_port != local->nd_ipv4_port) ||
1189 (remote->ni_ipv4_address != local->nd_ipv4_address)))
1190 status = -EINVAL;
1191
1192 if (status) {
1193 if (remote && !local)
1194 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) "
1195 "registered in joining node %d but not in "
1196 "local node %d\n", qn->qn_domain,
1197 remote->ni_nodenum,
1198 &(remote->ni_ipv4_address),
1199 ntohs(remote->ni_ipv4_port),
1200 qn->qn_nodenum, dlm->node_num);
1201 if (local && !remote)
1202 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) "
1203 "registered in local node %d but not in "
1204 "joining node %d\n", qn->qn_domain,
1205 local->nd_num, &(local->nd_ipv4_address),
1206 ntohs(local->nd_ipv4_port),
1207 dlm->node_num, qn->qn_nodenum);
1208 BUG_ON((!local && !remote));
1209 }
1210
1211 if (local)
1212 o2nm_node_put(local);
1213 }
1214
1215 return status;
1216 }
1217
1218 static int dlm_send_nodeinfo(struct dlm_ctxt *dlm, unsigned long *node_map)
1219 {
1220 struct dlm_query_nodeinfo *qn = NULL;
1221 struct o2nm_node *node;
1222 int ret = 0, status, count, i;
1223
1224 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES)
1225 goto bail;
1226
1227 qn = kzalloc(sizeof(struct dlm_query_nodeinfo), GFP_KERNEL);
1228 if (!qn) {
1229 ret = -ENOMEM;
1230 mlog_errno(ret);
1231 goto bail;
1232 }
1233
1234 for (i = 0, count = 0; i < O2NM_MAX_NODES; ++i) {
1235 node = o2nm_get_node_by_num(i);
1236 if (!node)
1237 continue;
1238 qn->qn_nodes[count].ni_nodenum = node->nd_num;
1239 qn->qn_nodes[count].ni_ipv4_port = node->nd_ipv4_port;
1240 qn->qn_nodes[count].ni_ipv4_address = node->nd_ipv4_address;
1241 mlog(0, "Node %3d, %pI4:%u\n", node->nd_num,
1242 &(node->nd_ipv4_address), ntohs(node->nd_ipv4_port));
1243 ++count;
1244 o2nm_node_put(node);
1245 }
1246
1247 qn->qn_nodenum = dlm->node_num;
1248 qn->qn_numnodes = count;
1249 qn->qn_namelen = strlen(dlm->name);
1250 memcpy(qn->qn_domain, dlm->name, qn->qn_namelen);
1251
1252 i = -1;
1253 while ((i = find_next_bit(node_map, O2NM_MAX_NODES,
1254 i + 1)) < O2NM_MAX_NODES) {
1255 if (i == dlm->node_num)
1256 continue;
1257
1258 mlog(0, "Sending nodeinfo to node %d\n", i);
1259
1260 ret = o2net_send_message(DLM_QUERY_NODEINFO, DLM_MOD_KEY,
1261 qn, sizeof(struct dlm_query_nodeinfo),
1262 i, &status);
1263 if (ret >= 0)
1264 ret = status;
1265 if (ret) {
1266 mlog(ML_ERROR, "node mismatch %d, node %d\n", ret, i);
1267 break;
1268 }
1269 }
1270
1271 bail:
1272 kfree(qn);
1273 return ret;
1274 }
1275
1276 static int dlm_query_nodeinfo_handler(struct o2net_msg *msg, u32 len,
1277 void *data, void **ret_data)
1278 {
1279 struct dlm_query_nodeinfo *qn;
1280 struct dlm_ctxt *dlm = NULL;
1281 int locked = 0, status = -EINVAL;
1282
1283 qn = (struct dlm_query_nodeinfo *) msg->buf;
1284
1285 mlog(0, "Node %u queries nodes on domain %s\n", qn->qn_nodenum,
1286 qn->qn_domain);
1287
1288 spin_lock(&dlm_domain_lock);
1289 dlm = __dlm_lookup_domain_full(qn->qn_domain, qn->qn_namelen);
1290 if (!dlm) {
1291 mlog(ML_ERROR, "Node %d queried nodes on domain %s before "
1292 "join domain\n", qn->qn_nodenum, qn->qn_domain);
1293 goto bail;
1294 }
1295
1296 spin_lock(&dlm->spinlock);
1297 locked = 1;
1298 if (dlm->joining_node != qn->qn_nodenum) {
1299 mlog(ML_ERROR, "Node %d queried nodes on domain %s but "
1300 "joining node is %d\n", qn->qn_nodenum, qn->qn_domain,
1301 dlm->joining_node);
1302 goto bail;
1303 }
1304
1305 /* Support for node query was added in 1.1 */
1306 if (dlm->dlm_locking_proto.pv_major == 1 &&
1307 dlm->dlm_locking_proto.pv_minor == 0) {
1308 mlog(ML_ERROR, "Node %d queried nodes on domain %s "
1309 "but active dlm protocol is %d.%d\n", qn->qn_nodenum,
1310 qn->qn_domain, dlm->dlm_locking_proto.pv_major,
1311 dlm->dlm_locking_proto.pv_minor);
1312 goto bail;
1313 }
1314
1315 status = dlm_match_nodes(dlm, qn);
1316
1317 bail:
1318 if (locked)
1319 spin_unlock(&dlm->spinlock);
1320 spin_unlock(&dlm_domain_lock);
1321
1322 return status;
1323 }
1324
1325 static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data,
1326 void **ret_data)
1327 {
1328 struct dlm_cancel_join *cancel;
1329 struct dlm_ctxt *dlm = NULL;
1330
1331 cancel = (struct dlm_cancel_join *) msg->buf;
1332
1333 mlog(0, "node %u cancels join on domain %s\n", cancel->node_idx,
1334 cancel->domain);
1335
1336 spin_lock(&dlm_domain_lock);
1337 dlm = __dlm_lookup_domain_full(cancel->domain, cancel->name_len);
1338
1339 if (dlm) {
1340 spin_lock(&dlm->spinlock);
1341
1342 /* Yikes, this guy wants to cancel his join. No
1343 * problem, we simply cleanup our join state. */
1344 BUG_ON(dlm->joining_node != cancel->node_idx);
1345 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
1346
1347 spin_unlock(&dlm->spinlock);
1348 }
1349 spin_unlock(&dlm_domain_lock);
1350
1351 return 0;
1352 }
1353
1354 static int dlm_send_one_join_cancel(struct dlm_ctxt *dlm,
1355 unsigned int node)
1356 {
1357 int status;
1358 struct dlm_cancel_join cancel_msg;
1359
1360 memset(&cancel_msg, 0, sizeof(cancel_msg));
1361 cancel_msg.node_idx = dlm->node_num;
1362 cancel_msg.name_len = strlen(dlm->name);
1363 memcpy(cancel_msg.domain, dlm->name, cancel_msg.name_len);
1364
1365 status = o2net_send_message(DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY,
1366 &cancel_msg, sizeof(cancel_msg), node,
1367 NULL);
1368 if (status < 0) {
1369 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
1370 "node %u\n", status, DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY,
1371 node);
1372 goto bail;
1373 }
1374
1375 bail:
1376 return status;
1377 }
1378
1379 /* map_size should be in bytes. */
1380 static int dlm_send_join_cancels(struct dlm_ctxt *dlm,
1381 unsigned long *node_map,
1382 unsigned int map_size)
1383 {
1384 int status, tmpstat;
1385 unsigned int node;
1386
1387 if (map_size != (BITS_TO_LONGS(O2NM_MAX_NODES) *
1388 sizeof(unsigned long))) {
1389 mlog(ML_ERROR,
1390 "map_size %u != BITS_TO_LONGS(O2NM_MAX_NODES) %u\n",
1391 map_size, (unsigned)BITS_TO_LONGS(O2NM_MAX_NODES));
1392 return -EINVAL;
1393 }
1394
1395 status = 0;
1396 node = -1;
1397 while ((node = find_next_bit(node_map, O2NM_MAX_NODES,
1398 node + 1)) < O2NM_MAX_NODES) {
1399 if (node == dlm->node_num)
1400 continue;
1401
1402 tmpstat = dlm_send_one_join_cancel(dlm, node);
1403 if (tmpstat) {
1404 mlog(ML_ERROR, "Error return %d cancelling join on "
1405 "node %d\n", tmpstat, node);
1406 if (!status)
1407 status = tmpstat;
1408 }
1409 }
1410
1411 if (status)
1412 mlog_errno(status);
1413 return status;
1414 }
1415
1416 static int dlm_request_join(struct dlm_ctxt *dlm,
1417 int node,
1418 enum dlm_query_join_response_code *response)
1419 {
1420 int status;
1421 struct dlm_query_join_request join_msg;
1422 struct dlm_query_join_packet packet;
1423 u32 join_resp;
1424
1425 mlog(0, "querying node %d\n", node);
1426
1427 memset(&join_msg, 0, sizeof(join_msg));
1428 join_msg.node_idx = dlm->node_num;
1429 join_msg.name_len = strlen(dlm->name);
1430 memcpy(join_msg.domain, dlm->name, join_msg.name_len);
1431 join_msg.dlm_proto = dlm->dlm_locking_proto;
1432 join_msg.fs_proto = dlm->fs_locking_proto;
1433
1434 /* copy live node map to join message */
1435 byte_copymap(join_msg.node_map, dlm->live_nodes_map, O2NM_MAX_NODES);
1436
1437 status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg,
1438 sizeof(join_msg), node, &join_resp);
1439 if (status < 0 && status != -ENOPROTOOPT) {
1440 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
1441 "node %u\n", status, DLM_QUERY_JOIN_MSG, DLM_MOD_KEY,
1442 node);
1443 goto bail;
1444 }
1445 dlm_query_join_wire_to_packet(join_resp, &packet);
1446
1447 /* -ENOPROTOOPT from the net code means the other side isn't
1448 listening for our message type -- that's fine, it means
1449 his dlm isn't up, so we can consider him a 'yes' but not
1450 joined into the domain. */
1451 if (status == -ENOPROTOOPT) {
1452 status = 0;
1453 *response = JOIN_OK_NO_MAP;
1454 } else {
1455 *response = packet.code;
1456 switch (packet.code) {
1457 case JOIN_DISALLOW:
1458 case JOIN_OK_NO_MAP:
1459 break;
1460 case JOIN_PROTOCOL_MISMATCH:
1461 mlog(ML_NOTICE,
1462 "This node requested DLM locking protocol %u.%u and "
1463 "filesystem locking protocol %u.%u. At least one of "
1464 "the protocol versions on node %d is not compatible, "
1465 "disconnecting\n",
1466 dlm->dlm_locking_proto.pv_major,
1467 dlm->dlm_locking_proto.pv_minor,
1468 dlm->fs_locking_proto.pv_major,
1469 dlm->fs_locking_proto.pv_minor,
1470 node);
1471 status = -EPROTO;
1472 break;
1473 case JOIN_OK:
1474 /* Use the same locking protocol as the remote node */
1475 dlm->dlm_locking_proto.pv_minor = packet.dlm_minor;
1476 dlm->fs_locking_proto.pv_minor = packet.fs_minor;
1477 mlog(0,
1478 "Node %d responds JOIN_OK with DLM locking protocol "
1479 "%u.%u and fs locking protocol %u.%u\n",
1480 node,
1481 dlm->dlm_locking_proto.pv_major,
1482 dlm->dlm_locking_proto.pv_minor,
1483 dlm->fs_locking_proto.pv_major,
1484 dlm->fs_locking_proto.pv_minor);
1485 break;
1486 default:
1487 status = -EINVAL;
1488 mlog(ML_ERROR, "invalid response %d from node %u\n",
1489 packet.code, node);
1490 /* Reset response to JOIN_DISALLOW */
1491 *response = JOIN_DISALLOW;
1492 break;
1493 }
1494 }
1495
1496 mlog(0, "status %d, node %d response is %d\n", status, node,
1497 *response);
1498
1499 bail:
1500 return status;
1501 }
1502
1503 static int dlm_send_one_join_assert(struct dlm_ctxt *dlm,
1504 unsigned int node)
1505 {
1506 int status;
1507 int ret;
1508 struct dlm_assert_joined assert_msg;
1509
1510 mlog(0, "Sending join assert to node %u\n", node);
1511
1512 memset(&assert_msg, 0, sizeof(assert_msg));
1513 assert_msg.node_idx = dlm->node_num;
1514 assert_msg.name_len = strlen(dlm->name);
1515 memcpy(assert_msg.domain, dlm->name, assert_msg.name_len);
1516
1517 status = o2net_send_message(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
1518 &assert_msg, sizeof(assert_msg), node,
1519 &ret);
1520 if (status < 0)
1521 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
1522 "node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
1523 node);
1524 else
1525 status = ret;
1526
1527 return status;
1528 }
1529
1530 static void dlm_send_join_asserts(struct dlm_ctxt *dlm,
1531 unsigned long *node_map)
1532 {
1533 int status, node, live;
1534
1535 status = 0;
1536 node = -1;
1537 while ((node = find_next_bit(node_map, O2NM_MAX_NODES,
1538 node + 1)) < O2NM_MAX_NODES) {
1539 if (node == dlm->node_num)
1540 continue;
1541
1542 do {
1543 /* It is very important that this message be
1544 * received so we spin until either the node
1545 * has died or it gets the message. */
1546 status = dlm_send_one_join_assert(dlm, node);
1547
1548 spin_lock(&dlm->spinlock);
1549 live = test_bit(node, dlm->live_nodes_map);
1550 spin_unlock(&dlm->spinlock);
1551
1552 if (status) {
1553 mlog(ML_ERROR, "Error return %d asserting "
1554 "join on node %d\n", status, node);
1555
1556 /* give us some time between errors... */
1557 if (live)
1558 msleep(DLM_DOMAIN_BACKOFF_MS);
1559 }
1560 } while (status && live);
1561 }
1562 }
1563
1564 struct domain_join_ctxt {
1565 unsigned long live_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1566 unsigned long yes_resp_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1567 };
1568
1569 static int dlm_should_restart_join(struct dlm_ctxt *dlm,
1570 struct domain_join_ctxt *ctxt,
1571 enum dlm_query_join_response_code response)
1572 {
1573 int ret;
1574
1575 if (response == JOIN_DISALLOW) {
1576 mlog(0, "Latest response of disallow -- should restart\n");
1577 return 1;
1578 }
1579
1580 spin_lock(&dlm->spinlock);
1581 /* For now, we restart the process if the node maps have
1582 * changed at all */
1583 ret = memcmp(ctxt->live_map, dlm->live_nodes_map,
1584 sizeof(dlm->live_nodes_map));
1585 spin_unlock(&dlm->spinlock);
1586
1587 if (ret)
1588 mlog(0, "Node maps changed -- should restart\n");
1589
1590 return ret;
1591 }
1592
1593 static int dlm_try_to_join_domain(struct dlm_ctxt *dlm)
1594 {
1595 int status = 0, tmpstat, node;
1596 struct domain_join_ctxt *ctxt;
1597 enum dlm_query_join_response_code response = JOIN_DISALLOW;
1598
1599 mlog(0, "%p", dlm);
1600
1601 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
1602 if (!ctxt) {
1603 status = -ENOMEM;
1604 mlog_errno(status);
1605 goto bail;
1606 }
1607
1608 /* group sem locking should work for us here -- we're already
1609 * registered for heartbeat events so filling this should be
1610 * atomic wrt getting those handlers called. */
1611 o2hb_fill_node_map(dlm->live_nodes_map, sizeof(dlm->live_nodes_map));
1612
1613 spin_lock(&dlm->spinlock);
1614 memcpy(ctxt->live_map, dlm->live_nodes_map, sizeof(ctxt->live_map));
1615
1616 __dlm_set_joining_node(dlm, dlm->node_num);
1617
1618 spin_unlock(&dlm->spinlock);
1619
1620 node = -1;
1621 while ((node = find_next_bit(ctxt->live_map, O2NM_MAX_NODES,
1622 node + 1)) < O2NM_MAX_NODES) {
1623 if (node == dlm->node_num)
1624 continue;
1625
1626 status = dlm_request_join(dlm, node, &response);
1627 if (status < 0) {
1628 mlog_errno(status);
1629 goto bail;
1630 }
1631
1632 /* Ok, either we got a response or the node doesn't have a
1633 * dlm up. */
1634 if (response == JOIN_OK)
1635 set_bit(node, ctxt->yes_resp_map);
1636
1637 if (dlm_should_restart_join(dlm, ctxt, response)) {
1638 status = -EAGAIN;
1639 goto bail;
1640 }
1641 }
1642
1643 mlog(0, "Yay, done querying nodes!\n");
1644
1645 /* Yay, everyone agree's we can join the domain. My domain is
1646 * comprised of all nodes who were put in the
1647 * yes_resp_map. Copy that into our domain map and send a join
1648 * assert message to clean up everyone elses state. */
1649 spin_lock(&dlm->spinlock);
1650 memcpy(dlm->domain_map, ctxt->yes_resp_map,
1651 sizeof(ctxt->yes_resp_map));
1652 set_bit(dlm->node_num, dlm->domain_map);
1653 spin_unlock(&dlm->spinlock);
1654
1655 /* Support for global heartbeat and node info was added in 1.1 */
1656 if (dlm->dlm_locking_proto.pv_major > 1 ||
1657 dlm->dlm_locking_proto.pv_minor > 0) {
1658 status = dlm_send_nodeinfo(dlm, ctxt->yes_resp_map);
1659 if (status) {
1660 mlog_errno(status);
1661 goto bail;
1662 }
1663 status = dlm_send_regions(dlm, ctxt->yes_resp_map);
1664 if (status) {
1665 mlog_errno(status);
1666 goto bail;
1667 }
1668 }
1669
1670 dlm_send_join_asserts(dlm, ctxt->yes_resp_map);
1671
1672 /* Joined state *must* be set before the joining node
1673 * information, otherwise the query_join handler may read no
1674 * current joiner but a state of NEW and tell joining nodes
1675 * we're not in the domain. */
1676 spin_lock(&dlm_domain_lock);
1677 dlm->dlm_state = DLM_CTXT_JOINED;
1678 dlm->num_joins++;
1679 spin_unlock(&dlm_domain_lock);
1680
1681 bail:
1682 spin_lock(&dlm->spinlock);
1683 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
1684 if (!status) {
1685 printk(KERN_NOTICE "o2dlm: Joining domain %s ", dlm->name);
1686 __dlm_print_nodes(dlm);
1687 }
1688 spin_unlock(&dlm->spinlock);
1689
1690 if (ctxt) {
1691 /* Do we need to send a cancel message to any nodes? */
1692 if (status < 0) {
1693 tmpstat = dlm_send_join_cancels(dlm,
1694 ctxt->yes_resp_map,
1695 sizeof(ctxt->yes_resp_map));
1696 if (tmpstat < 0)
1697 mlog_errno(tmpstat);
1698 }
1699 kfree(ctxt);
1700 }
1701
1702 mlog(0, "returning %d\n", status);
1703 return status;
1704 }
1705
1706 static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm)
1707 {
1708 o2hb_unregister_callback(dlm->name, &dlm->dlm_hb_up);
1709 o2hb_unregister_callback(dlm->name, &dlm->dlm_hb_down);
1710 o2net_unregister_handler_list(&dlm->dlm_domain_handlers);
1711 }
1712
1713 static int dlm_register_domain_handlers(struct dlm_ctxt *dlm)
1714 {
1715 int status;
1716
1717 mlog(0, "registering handlers.\n");
1718
1719 o2hb_setup_callback(&dlm->dlm_hb_down, O2HB_NODE_DOWN_CB,
1720 dlm_hb_node_down_cb, dlm, DLM_HB_NODE_DOWN_PRI);
1721 o2hb_setup_callback(&dlm->dlm_hb_up, O2HB_NODE_UP_CB,
1722 dlm_hb_node_up_cb, dlm, DLM_HB_NODE_UP_PRI);
1723
1724 status = o2hb_register_callback(dlm->name, &dlm->dlm_hb_down);
1725 if (status)
1726 goto bail;
1727
1728 status = o2hb_register_callback(dlm->name, &dlm->dlm_hb_up);
1729 if (status)
1730 goto bail;
1731
1732 status = o2net_register_handler(DLM_MASTER_REQUEST_MSG, dlm->key,
1733 sizeof(struct dlm_master_request),
1734 dlm_master_request_handler,
1735 dlm, NULL, &dlm->dlm_domain_handlers);
1736 if (status)
1737 goto bail;
1738
1739 status = o2net_register_handler(DLM_ASSERT_MASTER_MSG, dlm->key,
1740 sizeof(struct dlm_assert_master),
1741 dlm_assert_master_handler,
1742 dlm, dlm_assert_master_post_handler,
1743 &dlm->dlm_domain_handlers);
1744 if (status)
1745 goto bail;
1746
1747 status = o2net_register_handler(DLM_CREATE_LOCK_MSG, dlm->key,
1748 sizeof(struct dlm_create_lock),
1749 dlm_create_lock_handler,
1750 dlm, NULL, &dlm->dlm_domain_handlers);
1751 if (status)
1752 goto bail;
1753
1754 status = o2net_register_handler(DLM_CONVERT_LOCK_MSG, dlm->key,
1755 DLM_CONVERT_LOCK_MAX_LEN,
1756 dlm_convert_lock_handler,
1757 dlm, NULL, &dlm->dlm_domain_handlers);
1758 if (status)
1759 goto bail;
1760
1761 status = o2net_register_handler(DLM_UNLOCK_LOCK_MSG, dlm->key,
1762 DLM_UNLOCK_LOCK_MAX_LEN,
1763 dlm_unlock_lock_handler,
1764 dlm, NULL, &dlm->dlm_domain_handlers);
1765 if (status)
1766 goto bail;
1767
1768 status = o2net_register_handler(DLM_PROXY_AST_MSG, dlm->key,
1769 DLM_PROXY_AST_MAX_LEN,
1770 dlm_proxy_ast_handler,
1771 dlm, NULL, &dlm->dlm_domain_handlers);
1772 if (status)
1773 goto bail;
1774
1775 status = o2net_register_handler(DLM_EXIT_DOMAIN_MSG, dlm->key,
1776 sizeof(struct dlm_exit_domain),
1777 dlm_exit_domain_handler,
1778 dlm, NULL, &dlm->dlm_domain_handlers);
1779 if (status)
1780 goto bail;
1781
1782 status = o2net_register_handler(DLM_DEREF_LOCKRES_MSG, dlm->key,
1783 sizeof(struct dlm_deref_lockres),
1784 dlm_deref_lockres_handler,
1785 dlm, NULL, &dlm->dlm_domain_handlers);
1786 if (status)
1787 goto bail;
1788
1789 status = o2net_register_handler(DLM_MIGRATE_REQUEST_MSG, dlm->key,
1790 sizeof(struct dlm_migrate_request),
1791 dlm_migrate_request_handler,
1792 dlm, NULL, &dlm->dlm_domain_handlers);
1793 if (status)
1794 goto bail;
1795
1796 status = o2net_register_handler(DLM_MIG_LOCKRES_MSG, dlm->key,
1797 DLM_MIG_LOCKRES_MAX_LEN,
1798 dlm_mig_lockres_handler,
1799 dlm, NULL, &dlm->dlm_domain_handlers);
1800 if (status)
1801 goto bail;
1802
1803 status = o2net_register_handler(DLM_MASTER_REQUERY_MSG, dlm->key,
1804 sizeof(struct dlm_master_requery),
1805 dlm_master_requery_handler,
1806 dlm, NULL, &dlm->dlm_domain_handlers);
1807 if (status)
1808 goto bail;
1809
1810 status = o2net_register_handler(DLM_LOCK_REQUEST_MSG, dlm->key,
1811 sizeof(struct dlm_lock_request),
1812 dlm_request_all_locks_handler,
1813 dlm, NULL, &dlm->dlm_domain_handlers);
1814 if (status)
1815 goto bail;
1816
1817 status = o2net_register_handler(DLM_RECO_DATA_DONE_MSG, dlm->key,
1818 sizeof(struct dlm_reco_data_done),
1819 dlm_reco_data_done_handler,
1820 dlm, NULL, &dlm->dlm_domain_handlers);
1821 if (status)
1822 goto bail;
1823
1824 status = o2net_register_handler(DLM_BEGIN_RECO_MSG, dlm->key,
1825 sizeof(struct dlm_begin_reco),
1826 dlm_begin_reco_handler,
1827 dlm, NULL, &dlm->dlm_domain_handlers);
1828 if (status)
1829 goto bail;
1830
1831 status = o2net_register_handler(DLM_FINALIZE_RECO_MSG, dlm->key,
1832 sizeof(struct dlm_finalize_reco),
1833 dlm_finalize_reco_handler,
1834 dlm, NULL, &dlm->dlm_domain_handlers);
1835 if (status)
1836 goto bail;
1837
1838 status = o2net_register_handler(DLM_BEGIN_EXIT_DOMAIN_MSG, dlm->key,
1839 sizeof(struct dlm_exit_domain),
1840 dlm_begin_exit_domain_handler,
1841 dlm, NULL, &dlm->dlm_domain_handlers);
1842
1843 bail:
1844 if (status)
1845 dlm_unregister_domain_handlers(dlm);
1846
1847 return status;
1848 }
1849
1850 static int dlm_join_domain(struct dlm_ctxt *dlm)
1851 {
1852 int status;
1853 unsigned int backoff;
1854 unsigned int total_backoff = 0;
1855 char wq_name[O2NM_MAX_NAME_LEN];
1856
1857 BUG_ON(!dlm);
1858
1859 mlog(0, "Join domain %s\n", dlm->name);
1860
1861 status = dlm_register_domain_handlers(dlm);
1862 if (status) {
1863 mlog_errno(status);
1864 goto bail;
1865 }
1866
1867 status = dlm_launch_thread(dlm);
1868 if (status < 0) {
1869 mlog_errno(status);
1870 goto bail;
1871 }
1872
1873 status = dlm_launch_recovery_thread(dlm);
1874 if (status < 0) {
1875 mlog_errno(status);
1876 goto bail;
1877 }
1878
1879 status = dlm_debug_init(dlm);
1880 if (status < 0) {
1881 mlog_errno(status);
1882 goto bail;
1883 }
1884
1885 snprintf(wq_name, O2NM_MAX_NAME_LEN, "dlm_wq-%s", dlm->name);
1886 dlm->dlm_worker = create_singlethread_workqueue(wq_name);
1887 if (!dlm->dlm_worker) {
1888 status = -ENOMEM;
1889 mlog_errno(status);
1890 goto bail;
1891 }
1892
1893 do {
1894 status = dlm_try_to_join_domain(dlm);
1895
1896 /* If we're racing another node to the join, then we
1897 * need to back off temporarily and let them
1898 * complete. */
1899 #define DLM_JOIN_TIMEOUT_MSECS 90000
1900 if (status == -EAGAIN) {
1901 if (signal_pending(current)) {
1902 status = -ERESTARTSYS;
1903 goto bail;
1904 }
1905
1906 if (total_backoff > DLM_JOIN_TIMEOUT_MSECS) {
1907 status = -ERESTARTSYS;
1908 mlog(ML_NOTICE, "Timed out joining dlm domain "
1909 "%s after %u msecs\n", dlm->name,
1910 total_backoff);
1911 goto bail;
1912 }
1913
1914 /*
1915 * <chip> After you!
1916 * <dale> No, after you!
1917 * <chip> I insist!
1918 * <dale> But you first!
1919 * ...
1920 */
1921 backoff = (unsigned int)(jiffies & 0x3);
1922 backoff *= DLM_DOMAIN_BACKOFF_MS;
1923 total_backoff += backoff;
1924 mlog(0, "backoff %d\n", backoff);
1925 msleep(backoff);
1926 }
1927 } while (status == -EAGAIN);
1928
1929 if (status < 0) {
1930 mlog_errno(status);
1931 goto bail;
1932 }
1933
1934 status = 0;
1935 bail:
1936 wake_up(&dlm_domain_events);
1937
1938 if (status) {
1939 dlm_unregister_domain_handlers(dlm);
1940 dlm_debug_shutdown(dlm);
1941 dlm_complete_thread(dlm);
1942 dlm_complete_recovery_thread(dlm);
1943 dlm_destroy_dlm_worker(dlm);
1944 }
1945
1946 return status;
1947 }
1948
1949 static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
1950 u32 key)
1951 {
1952 int i;
1953 int ret;
1954 struct dlm_ctxt *dlm = NULL;
1955
1956 dlm = kzalloc(sizeof(*dlm), GFP_KERNEL);
1957 if (!dlm) {
1958 ret = -ENOMEM;
1959 mlog_errno(ret);
1960 goto leave;
1961 }
1962
1963 dlm->name = kstrdup(domain, GFP_KERNEL);
1964 if (dlm->name == NULL) {
1965 ret = -ENOMEM;
1966 mlog_errno(ret);
1967 goto leave;
1968 }
1969
1970 dlm->lockres_hash = (struct hlist_head **)dlm_alloc_pagevec(DLM_HASH_PAGES);
1971 if (!dlm->lockres_hash) {
1972 ret = -ENOMEM;
1973 mlog_errno(ret);
1974 goto leave;
1975 }
1976
1977 for (i = 0; i < DLM_HASH_BUCKETS; i++)
1978 INIT_HLIST_HEAD(dlm_lockres_hash(dlm, i));
1979
1980 dlm->master_hash = (struct hlist_head **)
1981 dlm_alloc_pagevec(DLM_HASH_PAGES);
1982 if (!dlm->master_hash) {
1983 ret = -ENOMEM;
1984 mlog_errno(ret);
1985 goto leave;
1986 }
1987
1988 for (i = 0; i < DLM_HASH_BUCKETS; i++)
1989 INIT_HLIST_HEAD(dlm_master_hash(dlm, i));
1990
1991 dlm->key = key;
1992 dlm->node_num = o2nm_this_node();
1993
1994 ret = dlm_create_debugfs_subroot(dlm);
1995 if (ret < 0)
1996 goto leave;
1997
1998 spin_lock_init(&dlm->spinlock);
1999 spin_lock_init(&dlm->master_lock);
2000 spin_lock_init(&dlm->ast_lock);
2001 spin_lock_init(&dlm->track_lock);
2002 INIT_LIST_HEAD(&dlm->list);
2003 INIT_LIST_HEAD(&dlm->dirty_list);
2004 INIT_LIST_HEAD(&dlm->reco.resources);
2005 INIT_LIST_HEAD(&dlm->reco.node_data);
2006 INIT_LIST_HEAD(&dlm->purge_list);
2007 INIT_LIST_HEAD(&dlm->dlm_domain_handlers);
2008 INIT_LIST_HEAD(&dlm->tracking_list);
2009 dlm->reco.state = 0;
2010
2011 INIT_LIST_HEAD(&dlm->pending_asts);
2012 INIT_LIST_HEAD(&dlm->pending_basts);
2013
2014 mlog(0, "dlm->recovery_map=%p, &(dlm->recovery_map[0])=%p\n",
2015 dlm->recovery_map, &(dlm->recovery_map[0]));
2016
2017 memset(dlm->recovery_map, 0, sizeof(dlm->recovery_map));
2018 memset(dlm->live_nodes_map, 0, sizeof(dlm->live_nodes_map));
2019 memset(dlm->domain_map, 0, sizeof(dlm->domain_map));
2020
2021 dlm->dlm_thread_task = NULL;
2022 dlm->dlm_reco_thread_task = NULL;
2023 dlm->dlm_worker = NULL;
2024 init_waitqueue_head(&dlm->dlm_thread_wq);
2025 init_waitqueue_head(&dlm->dlm_reco_thread_wq);
2026 init_waitqueue_head(&dlm->reco.event);
2027 init_waitqueue_head(&dlm->ast_wq);
2028 init_waitqueue_head(&dlm->migration_wq);
2029 INIT_LIST_HEAD(&dlm->mle_hb_events);
2030
2031 dlm->joining_node = DLM_LOCK_RES_OWNER_UNKNOWN;
2032 init_waitqueue_head(&dlm->dlm_join_events);
2033
2034 dlm->reco.new_master = O2NM_INVALID_NODE_NUM;
2035 dlm->reco.dead_node = O2NM_INVALID_NODE_NUM;
2036
2037 atomic_set(&dlm->res_tot_count, 0);
2038 atomic_set(&dlm->res_cur_count, 0);
2039 for (i = 0; i < DLM_MLE_NUM_TYPES; ++i) {
2040 atomic_set(&dlm->mle_tot_count[i], 0);
2041 atomic_set(&dlm->mle_cur_count[i], 0);
2042 }
2043
2044 spin_lock_init(&dlm->work_lock);
2045 INIT_LIST_HEAD(&dlm->work_list);
2046 INIT_WORK(&dlm->dispatched_work, dlm_dispatch_work);
2047
2048 kref_init(&dlm->dlm_refs);
2049 dlm->dlm_state = DLM_CTXT_NEW;
2050
2051 INIT_LIST_HEAD(&dlm->dlm_eviction_callbacks);
2052
2053 mlog(0, "context init: refcount %u\n",
2054 atomic_read(&dlm->dlm_refs.refcount));
2055
2056 leave:
2057 if (ret < 0 && dlm) {
2058 if (dlm->master_hash)
2059 dlm_free_pagevec((void **)dlm->master_hash,
2060 DLM_HASH_PAGES);
2061
2062 if (dlm->lockres_hash)
2063 dlm_free_pagevec((void **)dlm->lockres_hash,
2064 DLM_HASH_PAGES);
2065
2066 kfree(dlm->name);
2067 kfree(dlm);
2068 dlm = NULL;
2069 }
2070 return dlm;
2071 }
2072
2073 /*
2074 * Compare a requested locking protocol version against the current one.
2075 *
2076 * If the major numbers are different, they are incompatible.
2077 * If the current minor is greater than the request, they are incompatible.
2078 * If the current minor is less than or equal to the request, they are
2079 * compatible, and the requester should run at the current minor version.
2080 */
2081 static int dlm_protocol_compare(struct dlm_protocol_version *existing,
2082 struct dlm_protocol_version *request)
2083 {
2084 if (existing->pv_major != request->pv_major)
2085 return 1;
2086
2087 if (existing->pv_minor > request->pv_minor)
2088 return 1;
2089
2090 if (existing->pv_minor < request->pv_minor)
2091 request->pv_minor = existing->pv_minor;
2092
2093 return 0;
2094 }
2095
2096 /*
2097 * dlm_register_domain: one-time setup per "domain".
2098 *
2099 * The filesystem passes in the requested locking version via proto.
2100 * If registration was successful, proto will contain the negotiated
2101 * locking protocol.
2102 */
2103 struct dlm_ctxt * dlm_register_domain(const char *domain,
2104 u32 key,
2105 struct dlm_protocol_version *fs_proto)
2106 {
2107 int ret;
2108 struct dlm_ctxt *dlm = NULL;
2109 struct dlm_ctxt *new_ctxt = NULL;
2110
2111 if (strlen(domain) >= O2NM_MAX_NAME_LEN) {
2112 ret = -ENAMETOOLONG;
2113 mlog(ML_ERROR, "domain name length too long\n");
2114 goto leave;
2115 }
2116
2117 mlog(0, "register called for domain \"%s\"\n", domain);
2118
2119 retry:
2120 dlm = NULL;
2121 if (signal_pending(current)) {
2122 ret = -ERESTARTSYS;
2123 mlog_errno(ret);
2124 goto leave;
2125 }
2126
2127 spin_lock(&dlm_domain_lock);
2128
2129 dlm = __dlm_lookup_domain(domain);
2130 if (dlm) {
2131 if (dlm->dlm_state != DLM_CTXT_JOINED) {
2132 spin_unlock(&dlm_domain_lock);
2133
2134 mlog(0, "This ctxt is not joined yet!\n");
2135 wait_event_interruptible(dlm_domain_events,
2136 dlm_wait_on_domain_helper(
2137 domain));
2138 goto retry;
2139 }
2140
2141 if (dlm_protocol_compare(&dlm->fs_locking_proto, fs_proto)) {
2142 spin_unlock(&dlm_domain_lock);
2143 mlog(ML_ERROR,
2144 "Requested locking protocol version is not "
2145 "compatible with already registered domain "
2146 "\"%s\"\n", domain);
2147 ret = -EPROTO;
2148 goto leave;
2149 }
2150
2151 __dlm_get(dlm);
2152 dlm->num_joins++;
2153
2154 spin_unlock(&dlm_domain_lock);
2155
2156 ret = 0;
2157 goto leave;
2158 }
2159
2160 /* doesn't exist */
2161 if (!new_ctxt) {
2162 spin_unlock(&dlm_domain_lock);
2163
2164 new_ctxt = dlm_alloc_ctxt(domain, key);
2165 if (new_ctxt)
2166 goto retry;
2167
2168 ret = -ENOMEM;
2169 mlog_errno(ret);
2170 goto leave;
2171 }
2172
2173 /* a little variable switch-a-roo here... */
2174 dlm = new_ctxt;
2175 new_ctxt = NULL;
2176
2177 /* add the new domain */
2178 list_add_tail(&dlm->list, &dlm_domains);
2179 spin_unlock(&dlm_domain_lock);
2180
2181 /*
2182 * Pass the locking protocol version into the join. If the join
2183 * succeeds, it will have the negotiated protocol set.
2184 */
2185 dlm->dlm_locking_proto = dlm_protocol;
2186 dlm->fs_locking_proto = *fs_proto;
2187
2188 ret = dlm_join_domain(dlm);
2189 if (ret) {
2190 mlog_errno(ret);
2191 dlm_put(dlm);
2192 goto leave;
2193 }
2194
2195 /* Tell the caller what locking protocol we negotiated */
2196 *fs_proto = dlm->fs_locking_proto;
2197
2198 ret = 0;
2199 leave:
2200 if (new_ctxt)
2201 dlm_free_ctxt_mem(new_ctxt);
2202
2203 if (ret < 0)
2204 dlm = ERR_PTR(ret);
2205
2206 return dlm;
2207 }
2208 EXPORT_SYMBOL_GPL(dlm_register_domain);
2209
2210 static LIST_HEAD(dlm_join_handlers);
2211
2212 static void dlm_unregister_net_handlers(void)
2213 {
2214 o2net_unregister_handler_list(&dlm_join_handlers);
2215 }
2216
2217 static int dlm_register_net_handlers(void)
2218 {
2219 int status = 0;
2220
2221 status = o2net_register_handler(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY,
2222 sizeof(struct dlm_query_join_request),
2223 dlm_query_join_handler,
2224 NULL, NULL, &dlm_join_handlers);
2225 if (status)
2226 goto bail;
2227
2228 status = o2net_register_handler(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
2229 sizeof(struct dlm_assert_joined),
2230 dlm_assert_joined_handler,
2231 NULL, NULL, &dlm_join_handlers);
2232 if (status)
2233 goto bail;
2234
2235 status = o2net_register_handler(DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY,
2236 sizeof(struct dlm_cancel_join),
2237 dlm_cancel_join_handler,
2238 NULL, NULL, &dlm_join_handlers);
2239 if (status)
2240 goto bail;
2241
2242 status = o2net_register_handler(DLM_QUERY_REGION, DLM_MOD_KEY,
2243 sizeof(struct dlm_query_region),
2244 dlm_query_region_handler,
2245 NULL, NULL, &dlm_join_handlers);
2246
2247 if (status)
2248 goto bail;
2249
2250 status = o2net_register_handler(DLM_QUERY_NODEINFO, DLM_MOD_KEY,
2251 sizeof(struct dlm_query_nodeinfo),
2252 dlm_query_nodeinfo_handler,
2253 NULL, NULL, &dlm_join_handlers);
2254 bail:
2255 if (status < 0)
2256 dlm_unregister_net_handlers();
2257
2258 return status;
2259 }
2260
2261 /* Domain eviction callback handling.
2262 *
2263 * The file system requires notification of node death *before* the
2264 * dlm completes it's recovery work, otherwise it may be able to
2265 * acquire locks on resources requiring recovery. Since the dlm can
2266 * evict a node from it's domain *before* heartbeat fires, a similar
2267 * mechanism is required. */
2268
2269 /* Eviction is not expected to happen often, so a per-domain lock is
2270 * not necessary. Eviction callbacks are allowed to sleep for short
2271 * periods of time. */
2272 static DECLARE_RWSEM(dlm_callback_sem);
2273
2274 void dlm_fire_domain_eviction_callbacks(struct dlm_ctxt *dlm,
2275 int node_num)
2276 {
2277 struct dlm_eviction_cb *cb;
2278
2279 down_read(&dlm_callback_sem);
2280 list_for_each_entry(cb, &dlm->dlm_eviction_callbacks, ec_item) {
2281 cb->ec_func(node_num, cb->ec_data);
2282 }
2283 up_read(&dlm_callback_sem);
2284 }
2285
2286 void dlm_setup_eviction_cb(struct dlm_eviction_cb *cb,
2287 dlm_eviction_func *f,
2288 void *data)
2289 {
2290 INIT_LIST_HEAD(&cb->ec_item);
2291 cb->ec_func = f;
2292 cb->ec_data = data;
2293 }
2294 EXPORT_SYMBOL_GPL(dlm_setup_eviction_cb);
2295
2296 void dlm_register_eviction_cb(struct dlm_ctxt *dlm,
2297 struct dlm_eviction_cb *cb)
2298 {
2299 down_write(&dlm_callback_sem);
2300 list_add_tail(&cb->ec_item, &dlm->dlm_eviction_callbacks);
2301 up_write(&dlm_callback_sem);
2302 }
2303 EXPORT_SYMBOL_GPL(dlm_register_eviction_cb);
2304
2305 void dlm_unregister_eviction_cb(struct dlm_eviction_cb *cb)
2306 {
2307 down_write(&dlm_callback_sem);
2308 list_del_init(&cb->ec_item);
2309 up_write(&dlm_callback_sem);
2310 }
2311 EXPORT_SYMBOL_GPL(dlm_unregister_eviction_cb);
2312
2313 static int __init dlm_init(void)
2314 {
2315 int status;
2316
2317 status = dlm_init_mle_cache();
2318 if (status) {
2319 mlog(ML_ERROR, "Could not create o2dlm_mle slabcache\n");
2320 goto error;
2321 }
2322
2323 status = dlm_init_master_caches();
2324 if (status) {
2325 mlog(ML_ERROR, "Could not create o2dlm_lockres and "
2326 "o2dlm_lockname slabcaches\n");
2327 goto error;
2328 }
2329
2330 status = dlm_init_lock_cache();
2331 if (status) {
2332 mlog(ML_ERROR, "Count not create o2dlm_lock slabcache\n");
2333 goto error;
2334 }
2335
2336 status = dlm_register_net_handlers();
2337 if (status) {
2338 mlog(ML_ERROR, "Unable to register network handlers\n");
2339 goto error;
2340 }
2341
2342 status = dlm_create_debugfs_root();
2343 if (status)
2344 goto error;
2345
2346 return 0;
2347 error:
2348 dlm_unregister_net_handlers();
2349 dlm_destroy_lock_cache();
2350 dlm_destroy_master_caches();
2351 dlm_destroy_mle_cache();
2352 return -1;
2353 }
2354
2355 static void __exit dlm_exit (void)
2356 {
2357 dlm_destroy_debugfs_root();
2358 dlm_unregister_net_handlers();
2359 dlm_destroy_lock_cache();
2360 dlm_destroy_master_caches();
2361 dlm_destroy_mle_cache();
2362 }
2363
2364 MODULE_AUTHOR("Oracle");
2365 MODULE_LICENSE("GPL");
2366 MODULE_DESCRIPTION("OCFS2 Distributed Lock Management");
2367
2368 module_init(dlm_init);
2369 module_exit(dlm_exit);
Linux with added FPC-III support