projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
fs/ocfs2/cluster/tcp.c: fix possible null pointer dereferences
[deliverable/linux.git] / fs / ocfs2 / dlm / dlmmaster.c
CommitLineData
6714d8e8
KH		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * dlmmod.c
5 *
6 * standalone DLM module
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
28#include <linux/module.h>
29#include <linux/fs.h>
30#include <linux/types.h>
31#include <linux/slab.h>
32#include <linux/highmem.h>
6714d8e8
KH		 33#include <linux/init.h>
34#include <linux/sysctl.h>
35#include <linux/random.h>
36#include <linux/blkdev.h>
37#include <linux/socket.h>
38#include <linux/inet.h>
39#include <linux/spinlock.h>
40#include <linux/delay.h>
41
42
43#include "cluster/heartbeat.h"
44#include "cluster/nodemanager.h"
45#include "cluster/tcp.h"
46
47#include "dlmapi.h"
48#include "dlmcommon.h"
82353b59 49#include "dlmdomain.h"
e5a0334c 50#include "dlmdebug.h"
6714d8e8
KH		 51
52#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_MASTER)
53#include "cluster/masklog.h"
54
6714d8e8
KH		 55static void dlm_mle_node_down(struct dlm_ctxt *dlm,
56 struct dlm_master_list_entry *mle,
57 struct o2nm_node *node,
58 int idx);
59static void dlm_mle_node_up(struct dlm_ctxt *dlm,
60 struct dlm_master_list_entry *mle,
61 struct o2nm_node *node,
62 int idx);
63
64static void dlm_assert_master_worker(struct dlm_work_item *item, void *data);
ba2bf218
KH		 65static int dlm_do_assert_master(struct dlm_ctxt *dlm,
66 struct dlm_lock_resource *res,
67 void *nodemap, u32 flags);
f3f85464 68static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data);
6714d8e8
KH		 69
70static inline int dlm_mle_equal(struct dlm_ctxt *dlm,
71 struct dlm_master_list_entry *mle,
72 const char *name,
73 unsigned int namelen)
74{
6714d8e8
KH		 75 if (dlm != mle->dlm)
76 return 0;
77
7141514b
SM		 78 if (namelen != mle->mnamelen ||
79 memcmp(name, mle->mname, namelen) != 0)
f77a9a78
SM		 80 return 0;
81
6714d8e8
KH		 82 return 1;
83}
84
724bdca9
SM		 85static struct kmem_cache *dlm_lockres_cache = NULL;
86static struct kmem_cache *dlm_lockname_cache = NULL;
e18b890b 87static struct kmem_cache *dlm_mle_cache = NULL;
6714d8e8 88
6714d8e8
KH		 89static void dlm_mle_release(struct kref *kref);
90static void dlm_init_mle(struct dlm_master_list_entry *mle,
91 enum dlm_mle_type type,
92 struct dlm_ctxt *dlm,
93 struct dlm_lock_resource *res,
94 const char *name,
95 unsigned int namelen);
96static void dlm_put_mle(struct dlm_master_list_entry *mle);
97static void __dlm_put_mle(struct dlm_master_list_entry *mle);
98static int dlm_find_mle(struct dlm_ctxt *dlm,
99 struct dlm_master_list_entry **mle,
100 char *name, unsigned int namelen);
101
ba2bf218
KH		 102static int dlm_do_master_request(struct dlm_lock_resource *res,
103 struct dlm_master_list_entry *mle, int to);
6714d8e8
KH		 104
105
106static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
107 struct dlm_lock_resource *res,
108 struct dlm_master_list_entry *mle,
109 int *blocked);
110static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
111 struct dlm_lock_resource *res,
112 struct dlm_master_list_entry *mle,
113 int blocked);
114static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
115 struct dlm_lock_resource *res,
116 struct dlm_master_list_entry *mle,
117 struct dlm_master_list_entry **oldmle,
118 const char *name, unsigned int namelen,
119 u8 new_master, u8 master);
120
121static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
122 struct dlm_lock_resource *res);
123static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
124 struct dlm_lock_resource *res);
125static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
126 struct dlm_lock_resource *res,
127 u8 target);
c03872f5
KH		 128static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
129 struct dlm_lock_resource *res);
6714d8e8
KH		 130
131
132int dlm_is_host_down(int errno)
133{
134 switch (errno) {
135 case -EBADF:
136 case -ECONNREFUSED:
137 case -ENOTCONN:
138 case -ECONNRESET:
139 case -EPIPE:
140 case -EHOSTDOWN:
141 case -EHOSTUNREACH:
142 case -ETIMEDOUT:
143 case -ECONNABORTED:
144 case -ENETDOWN:
145 case -ENETUNREACH:
146 case -ENETRESET:
147 case -ESHUTDOWN:
148 case -ENOPROTOOPT:
149 case -EINVAL: /* if returned from our tcp code,
150 this means there is no socket */
151 return 1;
152 }
153 return 0;
154}
155
156
157/*
158 * MASTER LIST FUNCTIONS
159 */
160
161
162/*
163 * regarding master list entries and heartbeat callbacks:
164 *
165 * in order to avoid sleeping and allocation that occurs in
166 * heartbeat, master list entries are simply attached to the
167 * dlm's established heartbeat callbacks. the mle is attached
168 * when it is created, and since the dlm->spinlock is held at
169 * that time, any heartbeat event will be properly discovered
170 * by the mle. the mle needs to be detached from the
171 * dlm->mle_hb_events list as soon as heartbeat events are no
172 * longer useful to the mle, and before the mle is freed.
173 *
174 * as a general rule, heartbeat events are no longer needed by
175 * the mle once an "answer" regarding the lock master has been
176 * received.
177 */
178static inline void __dlm_mle_attach_hb_events(struct dlm_ctxt *dlm,
179 struct dlm_master_list_entry *mle)
180{
181 assert_spin_locked(&dlm->spinlock);
182
183 list_add_tail(&mle->hb_events, &dlm->mle_hb_events);
184}
185
186
187static inline void __dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
188 struct dlm_master_list_entry *mle)
189{
190 if (!list_empty(&mle->hb_events))
191 list_del_init(&mle->hb_events);
192}
193
194
195static inline void dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
196 struct dlm_master_list_entry *mle)
197{
198 spin_lock(&dlm->spinlock);
199 __dlm_mle_detach_hb_events(dlm, mle);
200 spin_unlock(&dlm->spinlock);
201}
202
a2bf0477
KH		 203static void dlm_get_mle_inuse(struct dlm_master_list_entry *mle)
204{
205 struct dlm_ctxt *dlm;
206 dlm = mle->dlm;
207
208 assert_spin_locked(&dlm->spinlock);
209 assert_spin_locked(&dlm->master_lock);
210 mle->inuse++;
211 kref_get(&mle->mle_refs);
212}
213
214static void dlm_put_mle_inuse(struct dlm_master_list_entry *mle)
215{
216 struct dlm_ctxt *dlm;
217 dlm = mle->dlm;
218
219 spin_lock(&dlm->spinlock);
220 spin_lock(&dlm->master_lock);
221 mle->inuse--;
222 __dlm_put_mle(mle);
223 spin_unlock(&dlm->master_lock);
224 spin_unlock(&dlm->spinlock);
225
226}
227
6714d8e8
KH		 228/* remove from list and free */
229static void __dlm_put_mle(struct dlm_master_list_entry *mle)
230{
231 struct dlm_ctxt *dlm;
232 dlm = mle->dlm;
233
234 assert_spin_locked(&dlm->spinlock);
235 assert_spin_locked(&dlm->master_lock);
aa852354
KH		 236 if (!atomic_read(&mle->mle_refs.refcount)) {
237 /* this may or may not crash, but who cares.
238 * it's a BUG. */
239 mlog(ML_ERROR, "bad mle: %p\n", mle);
240 dlm_print_one_mle(mle);
241 BUG();
242 } else
243 kref_put(&mle->mle_refs, dlm_mle_release);
6714d8e8
KH		 244}
245
246
247/* must not have any spinlocks coming in */
248static void dlm_put_mle(struct dlm_master_list_entry *mle)
249{
250 struct dlm_ctxt *dlm;
251 dlm = mle->dlm;
252
253 spin_lock(&dlm->spinlock);
254 spin_lock(&dlm->master_lock);
255 __dlm_put_mle(mle);
256 spin_unlock(&dlm->master_lock);
257 spin_unlock(&dlm->spinlock);
258}
259
260static inline void dlm_get_mle(struct dlm_master_list_entry *mle)
261{
262 kref_get(&mle->mle_refs);
263}
264
265static void dlm_init_mle(struct dlm_master_list_entry *mle,
266 enum dlm_mle_type type,
267 struct dlm_ctxt *dlm,
268 struct dlm_lock_resource *res,
269 const char *name,
270 unsigned int namelen)
271{
272 assert_spin_locked(&dlm->spinlock);
273
274 mle->dlm = dlm;
275 mle->type = type;
2ed6c750 276 INIT_HLIST_NODE(&mle->master_hash_node);
6714d8e8
KH		 277 INIT_LIST_HEAD(&mle->hb_events);
278 memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
279 spin_lock_init(&mle->spinlock);
280 init_waitqueue_head(&mle->wq);
281 atomic_set(&mle->woken, 0);
282 kref_init(&mle->mle_refs);
283 memset(mle->response_map, 0, sizeof(mle->response_map));
284 mle->master = O2NM_MAX_NODES;
285 mle->new_master = O2NM_MAX_NODES;
a2bf0477 286 mle->inuse = 0;
6714d8e8 287
f77a9a78
SM		 288 BUG_ON(mle->type != DLM_MLE_BLOCK &&
289 mle->type != DLM_MLE_MASTER &&
290 mle->type != DLM_MLE_MIGRATION);
291
6714d8e8
KH		 292 if (mle->type == DLM_MLE_MASTER) {
293 BUG_ON(!res);
7141514b
SM		 294 mle->mleres = res;
295 memcpy(mle->mname, res->lockname.name, res->lockname.len);
296 mle->mnamelen = res->lockname.len;
297 mle->mnamehash = res->lockname.hash;
f77a9a78 298 } else {
6714d8e8 299 BUG_ON(!name);
7141514b
SM		 300 mle->mleres = NULL;
301 memcpy(mle->mname, name, namelen);
302 mle->mnamelen = namelen;
303 mle->mnamehash = dlm_lockid_hash(name, namelen);
6714d8e8
KH		 304 }
305
2041d8fd
SM		 306 atomic_inc(&dlm->mle_tot_count[mle->type]);
307 atomic_inc(&dlm->mle_cur_count[mle->type]);
308
6714d8e8
KH		 309 /* copy off the node_map and register hb callbacks on our copy */
310 memcpy(mle->node_map, dlm->domain_map, sizeof(mle->node_map));
311 memcpy(mle->vote_map, dlm->domain_map, sizeof(mle->vote_map));
312 clear_bit(dlm->node_num, mle->vote_map);
313 clear_bit(dlm->node_num, mle->node_map);
314
315 /* attach the mle to the domain node up/down events */
316 __dlm_mle_attach_hb_events(dlm, mle);
317}
318
1c084577
SM		 319void __dlm_unlink_mle(struct dlm_ctxt *dlm, struct dlm_master_list_entry *mle)
320{
321 assert_spin_locked(&dlm->spinlock);
322 assert_spin_locked(&dlm->master_lock);
323
2ed6c750
SM		 324 if (!hlist_unhashed(&mle->master_hash_node))
325 hlist_del_init(&mle->master_hash_node);
1c084577
SM		 326}
327
328void __dlm_insert_mle(struct dlm_ctxt *dlm, struct dlm_master_list_entry *mle)
329{
2ed6c750 330 struct hlist_head *bucket;
2ed6c750 331
1c084577
SM		 332 assert_spin_locked(&dlm->master_lock);
333
7141514b 334 bucket = dlm_master_hash(dlm, mle->mnamehash);
2ed6c750 335 hlist_add_head(&mle->master_hash_node, bucket);
1c084577 336}
6714d8e8
KH		 337
338/* returns 1 if found, 0 if not */
339static int dlm_find_mle(struct dlm_ctxt *dlm,
340 struct dlm_master_list_entry **mle,
341 char *name, unsigned int namelen)
342{
343 struct dlm_master_list_entry *tmpmle;
2ed6c750
SM		 344 struct hlist_head *bucket;
345 struct hlist_node *list;
346 unsigned int hash;
6714d8e8
KH		 347
348 assert_spin_locked(&dlm->master_lock);
349
2ed6c750
SM		 350 hash = dlm_lockid_hash(name, namelen);
351 bucket = dlm_master_hash(dlm, hash);
352 hlist_for_each(list, bucket) {
353 tmpmle = hlist_entry(list, struct dlm_master_list_entry,
354 master_hash_node);
6714d8e8
KH		 355 if (!dlm_mle_equal(dlm, tmpmle, name, namelen))
356 continue;
357 dlm_get_mle(tmpmle);
358 *mle = tmpmle;
359 return 1;
360 }
361 return 0;
362}
363
364void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up)
365{
366 struct dlm_master_list_entry *mle;
6714d8e8
KH		 367
368 assert_spin_locked(&dlm->spinlock);
2bd63216 369
800deef3 370 list_for_each_entry(mle, &dlm->mle_hb_events, hb_events) {
6714d8e8
KH		 371 if (node_up)
372 dlm_mle_node_up(dlm, mle, NULL, idx);
373 else
374 dlm_mle_node_down(dlm, mle, NULL, idx);
375 }
376}
377
378static void dlm_mle_node_down(struct dlm_ctxt *dlm,
379 struct dlm_master_list_entry *mle,
380 struct o2nm_node *node, int idx)
381{
382 spin_lock(&mle->spinlock);
383
384 if (!test_bit(idx, mle->node_map))
385 mlog(0, "node %u already removed from nodemap!\n", idx);
386 else
387 clear_bit(idx, mle->node_map);
388
389 spin_unlock(&mle->spinlock);
390}
391
392static void dlm_mle_node_up(struct dlm_ctxt *dlm,
393 struct dlm_master_list_entry *mle,
394 struct o2nm_node *node, int idx)
395{
396 spin_lock(&mle->spinlock);
397
398 if (test_bit(idx, mle->node_map))
399 mlog(0, "node %u already in node map!\n", idx);
400 else
401 set_bit(idx, mle->node_map);
402
403 spin_unlock(&mle->spinlock);
404}
405
406
407int dlm_init_mle_cache(void)
408{
12eb0035 409 dlm_mle_cache = kmem_cache_create("o2dlm_mle",
6714d8e8
KH		 410 sizeof(struct dlm_master_list_entry),
411 0, SLAB_HWCACHE_ALIGN,
20c2df83 412 NULL);
6714d8e8
KH		 413 if (dlm_mle_cache == NULL)
414 return -ENOMEM;
415 return 0;
416}
417
418void dlm_destroy_mle_cache(void)
419{
420 if (dlm_mle_cache)
421 kmem_cache_destroy(dlm_mle_cache);
422}
423
424static void dlm_mle_release(struct kref *kref)
425{
426 struct dlm_master_list_entry *mle;
427 struct dlm_ctxt *dlm;
428
6714d8e8
KH		 429 mle = container_of(kref, struct dlm_master_list_entry, mle_refs);
430 dlm = mle->dlm;
431
6714d8e8
KH		 432 assert_spin_locked(&dlm->spinlock);
433 assert_spin_locked(&dlm->master_lock);
434
7141514b
SM		 435 mlog(0, "Releasing mle for %.*s, type %d\n", mle->mnamelen, mle->mname,
436 mle->type);
2ed6c750 437
6714d8e8 438 /* remove from list if not already */
1c084577 439 __dlm_unlink_mle(dlm, mle);
6714d8e8
KH		 440
441 /* detach the mle from the domain node up/down events */
442 __dlm_mle_detach_hb_events(dlm, mle);
443
2041d8fd
SM		 444 atomic_dec(&dlm->mle_cur_count[mle->type]);
445
6714d8e8
KH		 446 /* NOTE: kfree under spinlock here.
447 * if this is bad, we can move this to a freelist. */
448 kmem_cache_free(dlm_mle_cache, mle);
449}
450
451
452/*
453 * LOCK RESOURCE FUNCTIONS
454 */
455
724bdca9
SM		 456int dlm_init_master_caches(void)
457{
458 dlm_lockres_cache = kmem_cache_create("o2dlm_lockres",
459 sizeof(struct dlm_lock_resource),
460 0, SLAB_HWCACHE_ALIGN, NULL);
461 if (!dlm_lockres_cache)
462 goto bail;
463
464 dlm_lockname_cache = kmem_cache_create("o2dlm_lockname",
465 DLM_LOCKID_NAME_MAX, 0,
466 SLAB_HWCACHE_ALIGN, NULL);
467 if (!dlm_lockname_cache)
468 goto bail;
469
470 return 0;
471bail:
472 dlm_destroy_master_caches();
473 return -ENOMEM;
474}
475
476void dlm_destroy_master_caches(void)
477{
478 if (dlm_lockname_cache)
479 kmem_cache_destroy(dlm_lockname_cache);
480
481 if (dlm_lockres_cache)
482 kmem_cache_destroy(dlm_lockres_cache);
483}
484
6714d8e8
KH		 485static void dlm_lockres_release(struct kref *kref)
486{
487 struct dlm_lock_resource *res;
b0d4f817 488 struct dlm_ctxt *dlm;
6714d8e8
KH		 489
490 res = container_of(kref, struct dlm_lock_resource, refs);
b0d4f817 491 dlm = res->dlm;
6714d8e8
KH		 492
493 /* This should not happen -- all lockres' have a name
494 * associated with them at init time. */
495 BUG_ON(!res->lockname.name);
496
497 mlog(0, "destroying lockres %.*s\n", res->lockname.len,
498 res->lockname.name);
499
b0d4f817 500 spin_lock(&dlm->track_lock);
29576f8b
SM		 501 if (!list_empty(&res->tracking))
502 list_del_init(&res->tracking);
503 else {
504 mlog(ML_ERROR, "Resource %.*s not on the Tracking list\n",
505 res->lockname.len, res->lockname.name);
506 dlm_print_one_lock_resource(res);
507 }
b0d4f817
SM		 508 spin_unlock(&dlm->track_lock);
509
6800791a
SM		 510 atomic_dec(&dlm->res_cur_count);
511
a7f90d83
KH		 512 if (!hlist_unhashed(&res->hash_node) ||
513 !list_empty(&res->granted) ||
514 !list_empty(&res->converting) ||
515 !list_empty(&res->blocked) ||
516 !list_empty(&res->dirty) ||
517 !list_empty(&res->recovering) ||
518 !list_empty(&res->purge)) {
519 mlog(ML_ERROR,
520 "Going to BUG for resource %.*s."
521 " We're on a list! [%c%c%c%c%c%c%c]\n",
522 res->lockname.len, res->lockname.name,
523 !hlist_unhashed(&res->hash_node) ? 'H' : ' ',
524 !list_empty(&res->granted) ? 'G' : ' ',
525 !list_empty(&res->converting) ? 'C' : ' ',
526 !list_empty(&res->blocked) ? 'B' : ' ',
527 !list_empty(&res->dirty) ? 'D' : ' ',
528 !list_empty(&res->recovering) ? 'R' : ' ',
529 !list_empty(&res->purge) ? 'P' : ' ');
530
531 dlm_print_one_lock_resource(res);
532 }
533
6714d8e8
KH		 534 /* By the time we're ready to blow this guy away, we shouldn't
535 * be on any lists. */
81f2094a 536 BUG_ON(!hlist_unhashed(&res->hash_node));
6714d8e8
KH		 537 BUG_ON(!list_empty(&res->granted));
538 BUG_ON(!list_empty(&res->converting));
539 BUG_ON(!list_empty(&res->blocked));
540 BUG_ON(!list_empty(&res->dirty));
541 BUG_ON(!list_empty(&res->recovering));
542 BUG_ON(!list_empty(&res->purge));
543
724bdca9 544 kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
6714d8e8 545
724bdca9 546 kmem_cache_free(dlm_lockres_cache, res);
6714d8e8
KH		 547}
548
6714d8e8
KH		 549void dlm_lockres_put(struct dlm_lock_resource *res)
550{
551 kref_put(&res->refs, dlm_lockres_release);
552}
553
554static void dlm_init_lockres(struct dlm_ctxt *dlm,
555 struct dlm_lock_resource *res,
556 const char *name, unsigned int namelen)
557{
558 char *qname;
559
560 /* If we memset here, we lose our reference to the kmalloc'd
561 * res->lockname.name, so be sure to init every field
562 * correctly! */
563
564 qname = (char *) res->lockname.name;
565 memcpy(qname, name, namelen);
566
567 res->lockname.len = namelen;
a3d33291 568 res->lockname.hash = dlm_lockid_hash(name, namelen);
6714d8e8
KH		 569
570 init_waitqueue_head(&res->wq);
571 spin_lock_init(&res->spinlock);
81f2094a 572 INIT_HLIST_NODE(&res->hash_node);
6714d8e8
KH		 573 INIT_LIST_HEAD(&res->granted);
574 INIT_LIST_HEAD(&res->converting);
575 INIT_LIST_HEAD(&res->blocked);
576 INIT_LIST_HEAD(&res->dirty);
577 INIT_LIST_HEAD(&res->recovering);
578 INIT_LIST_HEAD(&res->purge);
29576f8b 579 INIT_LIST_HEAD(&res->tracking);
6714d8e8
KH		 580 atomic_set(&res->asts_reserved, 0);
581 res->migration_pending = 0;
ba2bf218 582 res->inflight_locks = 0;
6714d8e8 583
b0d4f817
SM		 584 res->dlm = dlm;
585
6714d8e8
KH		 586 kref_init(&res->refs);
587
6800791a
SM		 588 atomic_inc(&dlm->res_tot_count);
589 atomic_inc(&dlm->res_cur_count);
590
6714d8e8
KH		 591 /* just for consistency */
592 spin_lock(&res->spinlock);
593 dlm_set_lockres_owner(dlm, res, DLM_LOCK_RES_OWNER_UNKNOWN);
594 spin_unlock(&res->spinlock);
595
596 res->state = DLM_LOCK_RES_IN_PROGRESS;
597
598 res->last_used = 0;
599
18c6ac38 600 spin_lock(&dlm->spinlock);
29576f8b 601 list_add_tail(&res->tracking, &dlm->tracking_list);
18c6ac38 602 spin_unlock(&dlm->spinlock);
29576f8b 603
6714d8e8 604 memset(res->lvb, 0, DLM_LVB_LEN);
ba2bf218 605 memset(res->refmap, 0, sizeof(res->refmap));
6714d8e8
KH		 606}
607
608struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
609 const char *name,
610 unsigned int namelen)
611{
724bdca9 612 struct dlm_lock_resource *res = NULL;
6714d8e8 613
3914ed0c 614 res = kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
6714d8e8 615 if (!res)
724bdca9 616 goto error;
6714d8e8 617
3914ed0c 618 res->lockname.name = kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
724bdca9
SM		 619 if (!res->lockname.name)
620 goto error;
6714d8e8
KH		 621
622 dlm_init_lockres(dlm, res, name, namelen);
623 return res;
724bdca9
SM		 624
625error:
626 if (res && res->lockname.name)
627 kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
628
629 if (res)
630 kmem_cache_free(dlm_lockres_cache, res);
631 return NULL;
6714d8e8
KH		 632}
633
8d400b81
SM		 634void dlm_lockres_set_refmap_bit(struct dlm_ctxt *dlm,
635 struct dlm_lock_resource *res, int bit)
ba2bf218 636{
8d400b81
SM		 637 assert_spin_locked(&res->spinlock);
638
639 mlog(0, "res %.*s, set node %u, %ps()\n", res->lockname.len,
640 res->lockname.name, bit, __builtin_return_address(0));
641
642 set_bit(bit, res->refmap);
643}
644
645void dlm_lockres_clear_refmap_bit(struct dlm_ctxt *dlm,
646 struct dlm_lock_resource *res, int bit)
647{
648 assert_spin_locked(&res->spinlock);
649
650 mlog(0, "res %.*s, clr node %u, %ps()\n", res->lockname.len,
651 res->lockname.name, bit, __builtin_return_address(0));
652
653 clear_bit(bit, res->refmap);
654}
655
656
657void dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm,
658 struct dlm_lock_resource *res)
659{
660 assert_spin_locked(&res->spinlock);
ba2bf218 661
ba2bf218 662 res->inflight_locks++;
ff0a522e 663
8d400b81
SM		 664 mlog(0, "%s: res %.*s, inflight++: now %u, %ps()\n", dlm->name,
665 res->lockname.len, res->lockname.name, res->inflight_locks,
666 __builtin_return_address(0));
ba2bf218
KH		 667}
668
8d400b81
SM		 669void dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm,
670 struct dlm_lock_resource *res)
ba2bf218
KH		 671{
672 assert_spin_locked(&res->spinlock);
673
674 BUG_ON(res->inflight_locks == 0);
8d400b81 675
ba2bf218 676 res->inflight_locks--;
ff0a522e 677
8d400b81
SM		 678 mlog(0, "%s: res %.*s, inflight--: now %u, %ps()\n", dlm->name,
679 res->lockname.len, res->lockname.name, res->inflight_locks,
680 __builtin_return_address(0));
681
ba2bf218
KH		 682 wake_up(&res->wq);
683}
684
6714d8e8
KH		 685/*
686 * lookup a lock resource by name.
687 * may already exist in the hashtable.
688 * lockid is null terminated
689 *
690 * if not, allocate enough for the lockres and for
691 * the temporary structure used in doing the mastering.
692 *
693 * also, do a lookup in the dlm->master_list to see
694 * if another node has begun mastering the same lock.
695 * if so, there should be a block entry in there
696 * for this name, and we should *not* attempt to master
697 * the lock here. need to wait around for that node
698 * to assert_master (or die).
699 *
700 */
701struct dlm_lock_resource * dlm_get_lock_resource(struct dlm_ctxt *dlm,
702 const char *lockid,
3384f3df 703 int namelen,
6714d8e8
KH		 704 int flags)
705{
706 struct dlm_lock_resource *tmpres=NULL, *res=NULL;
707 struct dlm_master_list_entry *mle = NULL;
708 struct dlm_master_list_entry *alloc_mle = NULL;
709 int blocked = 0;
710 int ret, nodenum;
711 struct dlm_node_iter iter;
3384f3df 712 unsigned int hash;
6714d8e8 713 int tries = 0;
c03872f5 714 int bit, wait_on_recovery = 0;
6714d8e8
KH		 715
716 BUG_ON(!lockid);
717
a3d33291 718 hash = dlm_lockid_hash(lockid, namelen);
6714d8e8
KH		 719
720 mlog(0, "get lockres %s (len %d)\n", lockid, namelen);
721
722lookup:
723 spin_lock(&dlm->spinlock);
ba2bf218 724 tmpres = __dlm_lookup_lockres_full(dlm, lockid, namelen, hash);
6714d8e8 725 if (tmpres) {
7b791d68 726 spin_unlock(&dlm->spinlock);
ba2bf218 727 spin_lock(&tmpres->spinlock);
ff0a522e 728 /* Wait on the thread that is mastering the resource */
7b791d68
SM		 729 if (tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
730 __dlm_wait_on_lockres(tmpres);
731 BUG_ON(tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN);
ff0a522e
SM		 732 spin_unlock(&tmpres->spinlock);
733 dlm_lockres_put(tmpres);
734 tmpres = NULL;
735 goto lookup;
7b791d68
SM		 736 }
737
ff0a522e
SM		 738 /* Wait on the resource purge to complete before continuing */
739 if (tmpres->state & DLM_LOCK_RES_DROPPING_REF) {
740 BUG_ON(tmpres->owner == dlm->node_num);
741 __dlm_wait_on_lockres_flags(tmpres,
742 DLM_LOCK_RES_DROPPING_REF);
ba2bf218
KH		 743 spin_unlock(&tmpres->spinlock);
744 dlm_lockres_put(tmpres);
745 tmpres = NULL;
746 goto lookup;
747 }
748
ff0a522e
SM		 749 /* Grab inflight ref to pin the resource */
750 dlm_lockres_grab_inflight_ref(dlm, tmpres);
751
752 spin_unlock(&tmpres->spinlock);
6714d8e8
KH		 753 if (res)
754 dlm_lockres_put(res);
755 res = tmpres;
756 goto leave;
757 }
758
759 if (!res) {
760 spin_unlock(&dlm->spinlock);
761 mlog(0, "allocating a new resource\n");
762 /* nothing found and we need to allocate one. */
3914ed0c 763 alloc_mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
6714d8e8
KH		 764 if (!alloc_mle)
765 goto leave;
766 res = dlm_new_lockres(dlm, lockid, namelen);
767 if (!res)
768 goto leave;
769 goto lookup;
770 }
771
772 mlog(0, "no lockres found, allocated our own: %p\n", res);
773
774 if (flags & LKM_LOCAL) {
775 /* caller knows it's safe to assume it's not mastered elsewhere
776 * DONE! return right away */
777 spin_lock(&res->spinlock);
778 dlm_change_lockres_owner(dlm, res, dlm->node_num);
779 __dlm_insert_lockres(dlm, res);
ba2bf218 780 dlm_lockres_grab_inflight_ref(dlm, res);
6714d8e8
KH		 781 spin_unlock(&res->spinlock);
782 spin_unlock(&dlm->spinlock);
783 /* lockres still marked IN_PROGRESS */
784 goto wake_waiters;
785 }
786
787 /* check master list to see if another node has started mastering it */
788 spin_lock(&dlm->master_lock);
789
790 /* if we found a block, wait for lock to be mastered by another node */
791 blocked = dlm_find_mle(dlm, &mle, (char *)lockid, namelen);
792 if (blocked) {
ba2bf218 793 int mig;
6714d8e8
KH		 794 if (mle->type == DLM_MLE_MASTER) {
795 mlog(ML_ERROR, "master entry for nonexistent lock!\n");
796 BUG();
ba2bf218
KH		 797 }
798 mig = (mle->type == DLM_MLE_MIGRATION);
799 /* if there is a migration in progress, let the migration
800 * finish before continuing. we can wait for the absence
801 * of the MIGRATION mle: either the migrate finished or
802 * one of the nodes died and the mle was cleaned up.
803 * if there is a BLOCK here, but it already has a master
804 * set, we are too late. the master does not have a ref
805 * for us in the refmap. detach the mle and drop it.
806 * either way, go back to the top and start over. */
807 if (mig || mle->master != O2NM_MAX_NODES) {
808 BUG_ON(mig && mle->master == dlm->node_num);
809 /* we arrived too late. the master does not
810 * have a ref for us. retry. */
811 mlog(0, "%s:%.*s: late on %s\n",
812 dlm->name, namelen, lockid,
813 mig ? "MIGRATION" : "BLOCK");
6714d8e8 814 spin_unlock(&dlm->master_lock);
6714d8e8
KH		 815 spin_unlock(&dlm->spinlock);
816
817 /* master is known, detach */
ba2bf218
KH		 818 if (!mig)
819 dlm_mle_detach_hb_events(dlm, mle);
6714d8e8
KH		 820 dlm_put_mle(mle);
821 mle = NULL;
25985edc 822 /* this is lame, but we can't wait on either
ba2bf218
KH		 823 * the mle or lockres waitqueue here */
824 if (mig)
825 msleep(100);
826 goto lookup;
6714d8e8
KH		 827 }
828 } else {
829 /* go ahead and try to master lock on this node */
830 mle = alloc_mle;
831 /* make sure this does not get freed below */
832 alloc_mle = NULL;
833 dlm_init_mle(mle, DLM_MLE_MASTER, dlm, res, NULL, 0);
834 set_bit(dlm->node_num, mle->maybe_map);
1c084577 835 __dlm_insert_mle(dlm, mle);
c03872f5
KH		 836
837 /* still holding the dlm spinlock, check the recovery map
2bd63216 838 * to see if there are any nodes that still need to be
c03872f5
KH		 839 * considered. these will not appear in the mle nodemap
840 * but they might own this lockres. wait on them. */
841 bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
842 if (bit < O2NM_MAX_NODES) {
8decab3c
SM		 843 mlog(0, "%s: res %.*s, At least one node (%d) "
844 "to recover before lock mastery can begin\n",
c03872f5
KH		 845 dlm->name, namelen, (char *)lockid, bit);
846 wait_on_recovery = 1;
847 }
6714d8e8
KH		 848 }
849
850 /* at this point there is either a DLM_MLE_BLOCK or a
851 * DLM_MLE_MASTER on the master list, so it's safe to add the
852 * lockres to the hashtable. anyone who finds the lock will
853 * still have to wait on the IN_PROGRESS. */
854
855 /* finally add the lockres to its hash bucket */
856 __dlm_insert_lockres(dlm, res);
8d400b81 857
ff0a522e 858 /* Grab inflight ref to pin the resource */
8d400b81
SM		 859 spin_lock(&res->spinlock);
860 dlm_lockres_grab_inflight_ref(dlm, res);
861 spin_unlock(&res->spinlock);
ba2bf218 862
6714d8e8
KH		 863 /* get an extra ref on the mle in case this is a BLOCK
864 * if so, the creator of the BLOCK may try to put the last
865 * ref at this time in the assert master handler, so we
866 * need an extra one to keep from a bad ptr deref. */
a2bf0477 867 dlm_get_mle_inuse(mle);
6714d8e8
KH		 868 spin_unlock(&dlm->master_lock);
869 spin_unlock(&dlm->spinlock);
870
e7e69eb3 871redo_request:
c03872f5
KH		 872 while (wait_on_recovery) {
873 /* any cluster changes that occurred after dropping the
874 * dlm spinlock would be detectable be a change on the mle,
875 * so we only need to clear out the recovery map once. */
876 if (dlm_is_recovery_lock(lockid, namelen)) {
8decab3c
SM		 877 mlog(0, "%s: Recovery map is not empty, but must "
878 "master $RECOVERY lock now\n", dlm->name);
c03872f5
KH		 879 if (!dlm_pre_master_reco_lockres(dlm, res))
880 wait_on_recovery = 0;
881 else {
882 mlog(0, "%s: waiting 500ms for heartbeat state "
883 "change\n", dlm->name);
884 msleep(500);
885 }
886 continue;
2bd63216 887 }
c03872f5
KH		 888
889 dlm_kick_recovery_thread(dlm);
aa087b84 890 msleep(1000);
c03872f5
KH		 891 dlm_wait_for_recovery(dlm);
892
893 spin_lock(&dlm->spinlock);
894 bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
895 if (bit < O2NM_MAX_NODES) {
8decab3c
SM		 896 mlog(0, "%s: res %.*s, At least one node (%d) "
897 "to recover before lock mastery can begin\n",
c03872f5
KH		 898 dlm->name, namelen, (char *)lockid, bit);
899 wait_on_recovery = 1;
900 } else
901 wait_on_recovery = 0;
902 spin_unlock(&dlm->spinlock);
b7084ab5
KH		 903
904 if (wait_on_recovery)
905 dlm_wait_for_node_recovery(dlm, bit, 10000);
c03872f5
KH		 906 }
907
6714d8e8
KH		 908 /* must wait for lock to be mastered elsewhere */
909 if (blocked)
910 goto wait;
911
6714d8e8
KH		 912 ret = -EINVAL;
913 dlm_node_iter_init(mle->vote_map, &iter);
914 while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
ba2bf218 915 ret = dlm_do_master_request(res, mle, nodenum);
6714d8e8
KH		 916 if (ret < 0)
917 mlog_errno(ret);
918 if (mle->master != O2NM_MAX_NODES) {
919 /* found a master ! */
9c6510a5
KH		 920 if (mle->master <= nodenum)
921 break;
922 /* if our master request has not reached the master
923 * yet, keep going until it does. this is how the
924 * master will know that asserts are needed back to
925 * the lower nodes. */
8decab3c
SM		 926 mlog(0, "%s: res %.*s, Requests only up to %u but "
927 "master is %u, keep going\n", dlm->name, namelen,
9c6510a5 928 lockid, nodenum, mle->master);
6714d8e8
KH		 929 }
930 }
931
932wait:
933 /* keep going until the response map includes all nodes */
934 ret = dlm_wait_for_lock_mastery(dlm, res, mle, &blocked);
935 if (ret < 0) {
e7e69eb3 936 wait_on_recovery = 1;
8decab3c
SM		 937 mlog(0, "%s: res %.*s, Node map changed, redo the master "
938 "request now, blocked=%d\n", dlm->name, res->lockname.len,
6714d8e8
KH		 939 res->lockname.name, blocked);
940 if (++tries > 20) {
8decab3c
SM		 941 mlog(ML_ERROR, "%s: res %.*s, Spinning on "
942 "dlm_wait_for_lock_mastery, blocked = %d\n",
2bd63216 943 dlm->name, res->lockname.len,
6714d8e8
KH		 944 res->lockname.name, blocked);
945 dlm_print_one_lock_resource(res);
8a9343fa 946 dlm_print_one_mle(mle);
6714d8e8
KH		 947 tries = 0;
948 }
949 goto redo_request;
950 }
951
8decab3c
SM		 952 mlog(0, "%s: res %.*s, Mastered by %u\n", dlm->name, res->lockname.len,
953 res->lockname.name, res->owner);
6714d8e8
KH		 954 /* make sure we never continue without this */
955 BUG_ON(res->owner == O2NM_MAX_NODES);
956
957 /* master is known, detach if not already detached */
958 dlm_mle_detach_hb_events(dlm, mle);
959 dlm_put_mle(mle);
960 /* put the extra ref */
a2bf0477 961 dlm_put_mle_inuse(mle);
6714d8e8
KH		 962
963wake_waiters:
964 spin_lock(&res->spinlock);
965 res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
966 spin_unlock(&res->spinlock);
967 wake_up(&res->wq);
968
969leave:
970 /* need to free the unused mle */
971 if (alloc_mle)
972 kmem_cache_free(dlm_mle_cache, alloc_mle);
973
974 return res;
975}
976
977
978#define DLM_MASTERY_TIMEOUT_MS 5000
979
980static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
981 struct dlm_lock_resource *res,
982 struct dlm_master_list_entry *mle,
983 int *blocked)
984{
985 u8 m;
986 int ret, bit;
987 int map_changed, voting_done;
988 int assert, sleep;
989
990recheck:
991 ret = 0;
992 assert = 0;
993
994 /* check if another node has already become the owner */
995 spin_lock(&res->spinlock);
996 if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
9c6510a5
KH		 997 mlog(0, "%s:%.*s: owner is suddenly %u\n", dlm->name,
998 res->lockname.len, res->lockname.name, res->owner);
6714d8e8 999 spin_unlock(&res->spinlock);
9c6510a5
KH		 1000 /* this will cause the master to re-assert across
1001 * the whole cluster, freeing up mles */
588e0090 1002 if (res->owner != dlm->node_num) {
ba2bf218 1003 ret = dlm_do_master_request(res, mle, res->owner);
588e0090
KH		 1004 if (ret < 0) {
1005 /* give recovery a chance to run */
1006 mlog(ML_ERROR, "link to %u went down?: %d\n", res->owner, ret);
1007 msleep(500);
1008 goto recheck;
1009 }
9c6510a5
KH		 1010 }
1011 ret = 0;
6714d8e8
KH		 1012 goto leave;
1013 }
1014 spin_unlock(&res->spinlock);
1015
1016 spin_lock(&mle->spinlock);
1017 m = mle->master;
1018 map_changed = (memcmp(mle->vote_map, mle->node_map,
1019 sizeof(mle->vote_map)) != 0);
1020 voting_done = (memcmp(mle->vote_map, mle->response_map,
1021 sizeof(mle->vote_map)) == 0);
1022
1023 /* restart if we hit any errors */
1024 if (map_changed) {
1025 int b;
1026 mlog(0, "%s: %.*s: node map changed, restarting\n",
1027 dlm->name, res->lockname.len, res->lockname.name);
1028 ret = dlm_restart_lock_mastery(dlm, res, mle, *blocked);
1029 b = (mle->type == DLM_MLE_BLOCK);
1030 if ((*blocked && !b) || (!*blocked && b)) {
2bd63216 1031 mlog(0, "%s:%.*s: status change: old=%d new=%d\n",
6714d8e8
KH		 1032 dlm->name, res->lockname.len, res->lockname.name,
1033 *blocked, b);
1034 *blocked = b;
1035 }
1036 spin_unlock(&mle->spinlock);
1037 if (ret < 0) {
1038 mlog_errno(ret);
1039 goto leave;
1040 }
1041 mlog(0, "%s:%.*s: restart lock mastery succeeded, "
1042 "rechecking now\n", dlm->name, res->lockname.len,
1043 res->lockname.name);
1044 goto recheck;
aa852354
KH		 1045 } else {
1046 if (!voting_done) {
1047 mlog(0, "map not changed and voting not done "
1048 "for %s:%.*s\n", dlm->name, res->lockname.len,
1049 res->lockname.name);
1050 }
6714d8e8
KH		 1051 }
1052
1053 if (m != O2NM_MAX_NODES) {
1054 /* another node has done an assert!
1055 * all done! */
1056 sleep = 0;
1057 } else {
1058 sleep = 1;
1059 /* have all nodes responded? */
1060 if (voting_done && !*blocked) {
1061 bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
1062 if (dlm->node_num <= bit) {
1063 /* my node number is lowest.
1064 * now tell other nodes that I am
1065 * mastering this. */
1066 mle->master = dlm->node_num;
ba2bf218
KH		 1067 /* ref was grabbed in get_lock_resource
1068 * will be dropped in dlmlock_master */
6714d8e8
KH		 1069 assert = 1;
1070 sleep = 0;
1071 }
1072 /* if voting is done, but we have not received
1073 * an assert master yet, we must sleep */
1074 }
1075 }
1076
1077 spin_unlock(&mle->spinlock);
1078
1079 /* sleep if we haven't finished voting yet */
1080 if (sleep) {
1081 unsigned long timeo = msecs_to_jiffies(DLM_MASTERY_TIMEOUT_MS);
1082
1083 /*
1084 if (atomic_read(&mle->mle_refs.refcount) < 2)
1085 mlog(ML_ERROR, "mle (%p) refs=%d, name=%.*s\n", mle,
1086 atomic_read(&mle->mle_refs.refcount),
1087 res->lockname.len, res->lockname.name);
1088 */
1089 atomic_set(&mle->woken, 0);
1090 (void)wait_event_timeout(mle->wq,
1091 (atomic_read(&mle->woken) == 1),
1092 timeo);
1093 if (res->owner == O2NM_MAX_NODES) {
ba2bf218
KH		 1094 mlog(0, "%s:%.*s: waiting again\n", dlm->name,
1095 res->lockname.len, res->lockname.name);
6714d8e8
KH		 1096 goto recheck;
1097 }
1098 mlog(0, "done waiting, master is %u\n", res->owner);
1099 ret = 0;
1100 goto leave;
1101 }
1102
1103 ret = 0; /* done */
1104 if (assert) {
1105 m = dlm->node_num;
1106 mlog(0, "about to master %.*s here, this=%u\n",
1107 res->lockname.len, res->lockname.name, m);
ba2bf218 1108 ret = dlm_do_assert_master(dlm, res, mle->vote_map, 0);
6714d8e8
KH		 1109 if (ret) {
1110 /* This is a failure in the network path,
1111 * not in the response to the assert_master
1112 * (any nonzero response is a BUG on this node).
1113 * Most likely a socket just got disconnected
1114 * due to node death. */
1115 mlog_errno(ret);
1116 }
1117 /* no longer need to restart lock mastery.
1118 * all living nodes have been contacted. */
1119 ret = 0;
1120 }
1121
1122 /* set the lockres owner */
1123 spin_lock(&res->spinlock);
ba2bf218
KH		 1124 /* mastery reference obtained either during
1125 * assert_master_handler or in get_lock_resource */
6714d8e8
KH		 1126 dlm_change_lockres_owner(dlm, res, m);
1127 spin_unlock(&res->spinlock);
1128
1129leave:
1130 return ret;
1131}
1132
1133struct dlm_bitmap_diff_iter
1134{
1135 int curnode;
1136 unsigned long *orig_bm;
1137 unsigned long *cur_bm;
1138 unsigned long diff_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
1139};
1140
1141enum dlm_node_state_change
1142{
1143 NODE_DOWN = -1,
1144 NODE_NO_CHANGE = 0,
1145 NODE_UP
1146};
1147
1148static void dlm_bitmap_diff_iter_init(struct dlm_bitmap_diff_iter *iter,
1149 unsigned long *orig_bm,
1150 unsigned long *cur_bm)
1151{
1152 unsigned long p1, p2;
1153 int i;
1154
1155 iter->curnode = -1;
1156 iter->orig_bm = orig_bm;
1157 iter->cur_bm = cur_bm;
1158
1159 for (i = 0; i < BITS_TO_LONGS(O2NM_MAX_NODES); i++) {
1160 p1 = *(iter->orig_bm + i);
1161 p2 = *(iter->cur_bm + i);
1162 iter->diff_bm[i] = (p1 & ~p2) | (p2 & ~p1);
1163 }
1164}
1165
1166static int dlm_bitmap_diff_iter_next(struct dlm_bitmap_diff_iter *iter,
1167 enum dlm_node_state_change *state)
1168{
1169 int bit;
1170
1171 if (iter->curnode >= O2NM_MAX_NODES)
1172 return -ENOENT;
1173
1174 bit = find_next_bit(iter->diff_bm, O2NM_MAX_NODES,
1175 iter->curnode+1);
1176 if (bit >= O2NM_MAX_NODES) {
1177 iter->curnode = O2NM_MAX_NODES;
1178 return -ENOENT;
1179 }
1180
1181 /* if it was there in the original then this node died */
1182 if (test_bit(bit, iter->orig_bm))
1183 *state = NODE_DOWN;
1184 else
1185 *state = NODE_UP;
1186
1187 iter->curnode = bit;
1188 return bit;
1189}
1190
1191
1192static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
1193 struct dlm_lock_resource *res,
1194 struct dlm_master_list_entry *mle,
1195 int blocked)
1196{
1197 struct dlm_bitmap_diff_iter bdi;
1198 enum dlm_node_state_change sc;
1199 int node;
1200 int ret = 0;
1201
1202 mlog(0, "something happened such that the "
1203 "master process may need to be restarted!\n");
1204
1205 assert_spin_locked(&mle->spinlock);
1206
1207 dlm_bitmap_diff_iter_init(&bdi, mle->vote_map, mle->node_map);
1208 node = dlm_bitmap_diff_iter_next(&bdi, &sc);
1209 while (node >= 0) {
1210 if (sc == NODE_UP) {
e2faea4c
KH		 1211 /* a node came up. clear any old vote from
1212 * the response map and set it in the vote map
1213 * then restart the mastery. */
1214 mlog(ML_NOTICE, "node %d up while restarting\n", node);
6714d8e8
KH		 1215
1216 /* redo the master request, but only for the new node */
1217 mlog(0, "sending request to new node\n");
1218 clear_bit(node, mle->response_map);
1219 set_bit(node, mle->vote_map);
1220 } else {
1221 mlog(ML_ERROR, "node down! %d\n", node);
6714d8e8
KH		 1222 if (blocked) {
1223 int lowest = find_next_bit(mle->maybe_map,
1224 O2NM_MAX_NODES, 0);
1225
1226 /* act like it was never there */
1227 clear_bit(node, mle->maybe_map);
1228
e7e69eb3
KH		 1229 if (node == lowest) {
1230 mlog(0, "expected master %u died"
1231 " while this node was blocked "
1232 "waiting on it!\n", node);
1233 lowest = find_next_bit(mle->maybe_map,
1234 O2NM_MAX_NODES,
1235 lowest+1);
1236 if (lowest < O2NM_MAX_NODES) {
1237 mlog(0, "%s:%.*s:still "
1238 "blocked. waiting on %u "
1239 "now\n", dlm->name,
1240 res->lockname.len,
1241 res->lockname.name,
1242 lowest);
1243 } else {
1244 /* mle is an MLE_BLOCK, but
1245 * there is now nothing left to
1246 * block on. we need to return
1247 * all the way back out and try
1248 * again with an MLE_MASTER.
1249 * dlm_do_local_recovery_cleanup
1250 * has already run, so the mle
1251 * refcount is ok */
1252 mlog(0, "%s:%.*s: no "
1253 "longer blocking. try to "
1254 "master this here\n",
1255 dlm->name,
1256 res->lockname.len,
1257 res->lockname.name);
1258 mle->type = DLM_MLE_MASTER;
7141514b 1259 mle->mleres = res;
e7e69eb3 1260 }
6714d8e8 1261 }
6714d8e8
KH		 1262 }
1263
e7e69eb3
KH		 1264 /* now blank out everything, as if we had never
1265 * contacted anyone */
1266 memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
1267 memset(mle->response_map, 0, sizeof(mle->response_map));
1268 /* reset the vote_map to the current node_map */
1269 memcpy(mle->vote_map, mle->node_map,
1270 sizeof(mle->node_map));
1271 /* put myself into the maybe map */
1272 if (mle->type != DLM_MLE_BLOCK)
1273 set_bit(dlm->node_num, mle->maybe_map);
6714d8e8
KH		 1274 }
1275 ret = -EAGAIN;
6714d8e8
KH		 1276 node = dlm_bitmap_diff_iter_next(&bdi, &sc);
1277 }
1278 return ret;
1279}
1280
1281
1282/*
1283 * DLM_MASTER_REQUEST_MSG
1284 *
1285 * returns: 0 on success,
1286 * -errno on a network error
1287 *
1288 * on error, the caller should assume the target node is "dead"
1289 *
1290 */
1291
ba2bf218
KH		 1292static int dlm_do_master_request(struct dlm_lock_resource *res,
1293 struct dlm_master_list_entry *mle, int to)
6714d8e8
KH		 1294{
1295 struct dlm_ctxt *dlm = mle->dlm;
1296 struct dlm_master_request request;
1297 int ret, response=0, resend;
1298
1299 memset(&request, 0, sizeof(request));
1300 request.node_idx = dlm->node_num;
1301
1302 BUG_ON(mle->type == DLM_MLE_MIGRATION);
1303
7141514b
SM		 1304 request.namelen = (u8)mle->mnamelen;
1305 memcpy(request.name, mle->mname, request.namelen);
6714d8e8
KH		 1306
1307again:
1308 ret = o2net_send_message(DLM_MASTER_REQUEST_MSG, dlm->key, &request,
1309 sizeof(request), to, &response);
1310 if (ret < 0) {
1311 if (ret == -ESRCH) {
1312 /* should never happen */
1313 mlog(ML_ERROR, "TCP stack not ready!\n");
1314 BUG();
1315 } else if (ret == -EINVAL) {
1316 mlog(ML_ERROR, "bad args passed to o2net!\n");
1317 BUG();
1318 } else if (ret == -ENOMEM) {
1319 mlog(ML_ERROR, "out of memory while trying to send "
1320 "network message! retrying\n");
1321 /* this is totally crude */
1322 msleep(50);
1323 goto again;
1324 } else if (!dlm_is_host_down(ret)) {
1325 /* not a network error. bad. */
1326 mlog_errno(ret);
1327 mlog(ML_ERROR, "unhandled error!");
1328 BUG();
1329 }
1330 /* all other errors should be network errors,
1331 * and likely indicate node death */
1332 mlog(ML_ERROR, "link to %d went down!\n", to);
1333 goto out;
1334 }
1335
1336 ret = 0;
1337 resend = 0;
1338 spin_lock(&mle->spinlock);
1339 switch (response) {
1340 case DLM_MASTER_RESP_YES:
1341 set_bit(to, mle->response_map);
1342 mlog(0, "node %u is the master, response=YES\n", to);
ba2bf218
KH		 1343 mlog(0, "%s:%.*s: master node %u now knows I have a "
1344 "reference\n", dlm->name, res->lockname.len,
1345 res->lockname.name, to);
6714d8e8
KH		 1346 mle->master = to;
1347 break;
1348 case DLM_MASTER_RESP_NO:
1349 mlog(0, "node %u not master, response=NO\n", to);
1350 set_bit(to, mle->response_map);
1351 break;
1352 case DLM_MASTER_RESP_MAYBE:
1353 mlog(0, "node %u not master, response=MAYBE\n", to);
1354 set_bit(to, mle->response_map);
1355 set_bit(to, mle->maybe_map);
1356 break;
1357 case DLM_MASTER_RESP_ERROR:
1358 mlog(0, "node %u hit an error, resending\n", to);
1359 resend = 1;
1360 response = 0;
1361 break;
1362 default:
1363 mlog(ML_ERROR, "bad response! %u\n", response);
1364 BUG();
1365 }
1366 spin_unlock(&mle->spinlock);
1367 if (resend) {
1368 /* this is also totally crude */
1369 msleep(50);
1370 goto again;
1371 }
1372
1373out:
1374 return ret;
1375}
1376
1377/*
1378 * locks that can be taken here:
1379 * dlm->spinlock
1380 * res->spinlock
1381 * mle->spinlock
1382 * dlm->master_list
1383 *
1384 * if possible, TRIM THIS DOWN!!!
1385 */
d74c9803
KH		 1386int dlm_master_request_handler(struct o2net_msg *msg, u32 len, void *data,
1387 void **ret_data)
6714d8e8
KH		 1388{
1389 u8 response = DLM_MASTER_RESP_MAYBE;
1390 struct dlm_ctxt *dlm = data;
9c6510a5 1391 struct dlm_lock_resource *res = NULL;
6714d8e8
KH		 1392 struct dlm_master_request *request = (struct dlm_master_request *) msg->buf;
1393 struct dlm_master_list_entry *mle = NULL, *tmpmle = NULL;
1394 char *name;
a3d33291 1395 unsigned int namelen, hash;
6714d8e8
KH		 1396 int found, ret;
1397 int set_maybe;
9c6510a5 1398 int dispatch_assert = 0;
6714d8e8
KH		 1399
1400 if (!dlm_grab(dlm))
1401 return DLM_MASTER_RESP_NO;
1402
1403 if (!dlm_domain_fully_joined(dlm)) {
1404 response = DLM_MASTER_RESP_NO;
1405 goto send_response;
1406 }
1407
1408 name = request->name;
1409 namelen = request->namelen;
a3d33291 1410 hash = dlm_lockid_hash(name, namelen);
6714d8e8
KH		 1411
1412 if (namelen > DLM_LOCKID_NAME_MAX) {
1413 response = DLM_IVBUFLEN;
1414 goto send_response;
1415 }
1416
1417way_up_top:
1418 spin_lock(&dlm->spinlock);
a3d33291 1419 res = __dlm_lookup_lockres(dlm, name, namelen, hash);
6714d8e8
KH		 1420 if (res) {
1421 spin_unlock(&dlm->spinlock);
1422
1423 /* take care of the easy cases up front */
1424 spin_lock(&res->spinlock);
1cd04dbe
KH		 1425 if (res->state & (DLM_LOCK_RES_RECOVERING|
1426 DLM_LOCK_RES_MIGRATING)) {
6714d8e8
KH		 1427 spin_unlock(&res->spinlock);
1428 mlog(0, "returning DLM_MASTER_RESP_ERROR since res is "
1cd04dbe 1429 "being recovered/migrated\n");
6714d8e8
KH		 1430 response = DLM_MASTER_RESP_ERROR;
1431 if (mle)
1432 kmem_cache_free(dlm_mle_cache, mle);
1433 goto send_response;
1434 }
1435
1436 if (res->owner == dlm->node_num) {
8d400b81 1437 dlm_lockres_set_refmap_bit(dlm, res, request->node_idx);
6714d8e8 1438 spin_unlock(&res->spinlock);
6714d8e8
KH		 1439 response = DLM_MASTER_RESP_YES;
1440 if (mle)
1441 kmem_cache_free(dlm_mle_cache, mle);
1442
1443 /* this node is the owner.
1444 * there is some extra work that needs to
1445 * happen now. the requesting node has
1446 * caused all nodes up to this one to
1447 * create mles. this node now needs to
1448 * go back and clean those up. */
9c6510a5 1449 dispatch_assert = 1;
6714d8e8
KH		 1450 goto send_response;
1451 } else if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
1452 spin_unlock(&res->spinlock);
1453 // mlog(0, "node %u is the master\n", res->owner);
1454 response = DLM_MASTER_RESP_NO;
1455 if (mle)
1456 kmem_cache_free(dlm_mle_cache, mle);
1457 goto send_response;
1458 }
1459
1460 /* ok, there is no owner. either this node is
1461 * being blocked, or it is actively trying to
1462 * master this lock. */
1463 if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
1464 mlog(ML_ERROR, "lock with no owner should be "
1465 "in-progress!\n");
1466 BUG();
1467 }
1468
1469 // mlog(0, "lockres is in progress...\n");
1470 spin_lock(&dlm->master_lock);
1471 found = dlm_find_mle(dlm, &tmpmle, name, namelen);
1472 if (!found) {
1473 mlog(ML_ERROR, "no mle found for this lock!\n");
1474 BUG();
1475 }
1476 set_maybe = 1;
1477 spin_lock(&tmpmle->spinlock);
1478 if (tmpmle->type == DLM_MLE_BLOCK) {
1479 // mlog(0, "this node is waiting for "
1480 // "lockres to be mastered\n");
1481 response = DLM_MASTER_RESP_NO;
1482 } else if (tmpmle->type == DLM_MLE_MIGRATION) {
1483 mlog(0, "node %u is master, but trying to migrate to "
1484 "node %u.\n", tmpmle->master, tmpmle->new_master);
1485 if (tmpmle->master == dlm->node_num) {
6714d8e8
KH		 1486 mlog(ML_ERROR, "no owner on lockres, but this "
1487 "node is trying to migrate it to %u?!\n",
1488 tmpmle->new_master);
1489 BUG();
1490 } else {
1491 /* the real master can respond on its own */
1492 response = DLM_MASTER_RESP_NO;
1493 }
1494 } else if (tmpmle->master != DLM_LOCK_RES_OWNER_UNKNOWN) {
1495 set_maybe = 0;
9c6510a5 1496 if (tmpmle->master == dlm->node_num) {
6714d8e8 1497 response = DLM_MASTER_RESP_YES;
9c6510a5
KH		 1498 /* this node will be the owner.
1499 * go back and clean the mles on any
1500 * other nodes */
1501 dispatch_assert = 1;
8d400b81
SM		 1502 dlm_lockres_set_refmap_bit(dlm, res,
1503 request->node_idx);
9c6510a5 1504 } else
6714d8e8
KH		 1505 response = DLM_MASTER_RESP_NO;
1506 } else {
1507 // mlog(0, "this node is attempting to "
1508 // "master lockres\n");
1509 response = DLM_MASTER_RESP_MAYBE;
1510 }
1511 if (set_maybe)
1512 set_bit(request->node_idx, tmpmle->maybe_map);
1513 spin_unlock(&tmpmle->spinlock);
1514
1515 spin_unlock(&dlm->master_lock);
1516 spin_unlock(&res->spinlock);
1517
1518 /* keep the mle attached to heartbeat events */
1519 dlm_put_mle(tmpmle);
1520 if (mle)
1521 kmem_cache_free(dlm_mle_cache, mle);
1522 goto send_response;
1523 }
1524
1525 /*
1526 * lockres doesn't exist on this node
1527 * if there is an MLE_BLOCK, return NO
1528 * if there is an MLE_MASTER, return MAYBE
1529 * otherwise, add an MLE_BLOCK, return NO
1530 */
1531 spin_lock(&dlm->master_lock);
1532 found = dlm_find_mle(dlm, &tmpmle, name, namelen);
1533 if (!found) {
1534 /* this lockid has never been seen on this node yet */
1535 // mlog(0, "no mle found\n");
1536 if (!mle) {
1537 spin_unlock(&dlm->master_lock);
1538 spin_unlock(&dlm->spinlock);
1539
3914ed0c 1540 mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
6714d8e8 1541 if (!mle) {
6714d8e8 1542 response = DLM_MASTER_RESP_ERROR;
9c6510a5 1543 mlog_errno(-ENOMEM);
6714d8e8
KH		 1544 goto send_response;
1545 }
6714d8e8
KH		 1546 goto way_up_top;
1547 }
1548
1549 // mlog(0, "this is second time thru, already allocated, "
1550 // "add the block.\n");
41b8c8a1 1551 dlm_init_mle(mle, DLM_MLE_BLOCK, dlm, NULL, name, namelen);
6714d8e8 1552 set_bit(request->node_idx, mle->maybe_map);
1c084577 1553 __dlm_insert_mle(dlm, mle);
6714d8e8
KH		 1554 response = DLM_MASTER_RESP_NO;
1555 } else {
1556 // mlog(0, "mle was found\n");
1557 set_maybe = 1;
1558 spin_lock(&tmpmle->spinlock);
9c6510a5
KH		 1559 if (tmpmle->master == dlm->node_num) {
1560 mlog(ML_ERROR, "no lockres, but an mle with this node as master!\n");
1561 BUG();
1562 }
6714d8e8
KH		 1563 if (tmpmle->type == DLM_MLE_BLOCK)
1564 response = DLM_MASTER_RESP_NO;
1565 else if (tmpmle->type == DLM_MLE_MIGRATION) {
1566 mlog(0, "migration mle was found (%u->%u)\n",
1567 tmpmle->master, tmpmle->new_master);
6714d8e8
KH		 1568 /* real master can respond on its own */
1569 response = DLM_MASTER_RESP_NO;
9c6510a5
KH		 1570 } else
1571 response = DLM_MASTER_RESP_MAYBE;
6714d8e8
KH		 1572 if (set_maybe)
1573 set_bit(request->node_idx, tmpmle->maybe_map);
1574 spin_unlock(&tmpmle->spinlock);
1575 }
1576 spin_unlock(&dlm->master_lock);
1577 spin_unlock(&dlm->spinlock);
1578
1579 if (found) {
1580 /* keep the mle attached to heartbeat events */
1581 dlm_put_mle(tmpmle);
1582 }
1583send_response:
b31cfc02
SM		 1584 /*
1585 * __dlm_lookup_lockres() grabbed a reference to this lockres.
1586 * The reference is released by dlm_assert_master_worker() under
1587 * the call to dlm_dispatch_assert_master(). If
1588 * dlm_assert_master_worker() isn't called, we drop it here.
1589 */
9c6510a5
KH		 1590 if (dispatch_assert) {
1591 if (response != DLM_MASTER_RESP_YES)
1592 mlog(ML_ERROR, "invalid response %d\n", response);
1593 if (!res) {
1594 mlog(ML_ERROR, "bad lockres while trying to assert!\n");
1595 BUG();
1596 }
1597 mlog(0, "%u is the owner of %.*s, cleaning everyone else\n",
1598 dlm->node_num, res->lockname.len, res->lockname.name);
2bd63216 1599 ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx,
9c6510a5
KH		 1600 DLM_ASSERT_MASTER_MLE_CLEANUP);
1601 if (ret < 0) {
1602 mlog(ML_ERROR, "failed to dispatch assert master work\n");
1603 response = DLM_MASTER_RESP_ERROR;
b31cfc02 1604 dlm_lockres_put(res);
9c6510a5 1605 }
b31cfc02
SM		 1606 } else {
1607 if (res)
1608 dlm_lockres_put(res);
9c6510a5
KH		 1609 }
1610
6714d8e8
KH		 1611 dlm_put(dlm);
1612 return response;
1613}
1614
1615/*
1616 * DLM_ASSERT_MASTER_MSG
1617 */
1618
1619
1620/*
1621 * NOTE: this can be used for debugging
1622 * can periodically run all locks owned by this node
1623 * and re-assert across the cluster...
1624 */
05488bbe
AB		 1625static int dlm_do_assert_master(struct dlm_ctxt *dlm,
1626 struct dlm_lock_resource *res,
1627 void *nodemap, u32 flags)
6714d8e8
KH		 1628{
1629 struct dlm_assert_master assert;
1630 int to, tmpret;
1631 struct dlm_node_iter iter;
1632 int ret = 0;
9c6510a5 1633 int reassert;
ba2bf218
KH		 1634 const char *lockname = res->lockname.name;
1635 unsigned int namelen = res->lockname.len;
6714d8e8
KH		 1636
1637 BUG_ON(namelen > O2NM_MAX_NAME_LEN);
f3f85464
SM		 1638
1639 spin_lock(&res->spinlock);
1640 res->state |= DLM_LOCK_RES_SETREF_INPROG;
1641 spin_unlock(&res->spinlock);
1642
9c6510a5
KH		 1643again:
1644 reassert = 0;
6714d8e8
KH		 1645
1646 /* note that if this nodemap is empty, it returns 0 */
1647 dlm_node_iter_init(nodemap, &iter);
1648 while ((to = dlm_node_iter_next(&iter)) >= 0) {
1649 int r = 0;
a9ee4c8a
KH		 1650 struct dlm_master_list_entry *mle = NULL;
1651
6714d8e8
KH		 1652 mlog(0, "sending assert master to %d (%.*s)\n", to,
1653 namelen, lockname);
1654 memset(&assert, 0, sizeof(assert));
1655 assert.node_idx = dlm->node_num;
1656 assert.namelen = namelen;
1657 memcpy(assert.name, lockname, namelen);
1658 assert.flags = cpu_to_be32(flags);
1659
1660 tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key,
1661 &assert, sizeof(assert), to, &r);
1662 if (tmpret < 0) {
a5196ec5
WW		 1663 mlog(ML_ERROR, "Error %d when sending message %u (key "
1664 "0x%x) to node %u\n", tmpret,
1665 DLM_ASSERT_MASTER_MSG, dlm->key, to);
6714d8e8 1666 if (!dlm_is_host_down(tmpret)) {
3b3b84a8 1667 mlog(ML_ERROR, "unhandled error=%d!\n", tmpret);
6714d8e8
KH		 1668 BUG();
1669 }
1670 /* a node died. finish out the rest of the nodes. */
3b3b84a8 1671 mlog(0, "link to %d went down!\n", to);
6714d8e8
KH		 1672 /* any nonzero status return will do */
1673 ret = tmpret;
ba2bf218 1674 r = 0;
6714d8e8
KH		 1675 } else if (r < 0) {
1676 /* ok, something horribly messed. kill thyself. */
1677 mlog(ML_ERROR,"during assert master of %.*s to %u, "
1678 "got %d.\n", namelen, lockname, to, r);
a9ee4c8a
KH		 1679 spin_lock(&dlm->spinlock);
1680 spin_lock(&dlm->master_lock);
1681 if (dlm_find_mle(dlm, &mle, (char *)lockname,
1682 namelen)) {
1683 dlm_print_one_mle(mle);
1684 __dlm_put_mle(mle);
1685 }
1686 spin_unlock(&dlm->master_lock);
1687 spin_unlock(&dlm->spinlock);
6714d8e8 1688 BUG();
ba2bf218
KH		 1689 }
1690
1691 if (r & DLM_ASSERT_RESPONSE_REASSERT &&
1692 !(r & DLM_ASSERT_RESPONSE_MASTERY_REF)) {
1693 mlog(ML_ERROR, "%.*s: very strange, "
1694 "master MLE but no lockres on %u\n",
1695 namelen, lockname, to);
1696 }
1697
1698 if (r & DLM_ASSERT_RESPONSE_REASSERT) {
9c6510a5 1699 mlog(0, "%.*s: node %u create mles on other "
2bd63216 1700 "nodes and requests a re-assert\n",
9c6510a5
KH		 1701 namelen, lockname, to);
1702 reassert = 1;
6714d8e8 1703 }
ba2bf218
KH		 1704 if (r & DLM_ASSERT_RESPONSE_MASTERY_REF) {
1705 mlog(0, "%.*s: node %u has a reference to this "
1706 "lockres, set the bit in the refmap\n",
1707 namelen, lockname, to);
1708 spin_lock(&res->spinlock);
8d400b81 1709 dlm_lockres_set_refmap_bit(dlm, res, to);
ba2bf218
KH		 1710 spin_unlock(&res->spinlock);
1711 }
6714d8e8
KH		 1712 }
1713
9c6510a5
KH		 1714 if (reassert)
1715 goto again;
1716
f3f85464
SM		 1717 spin_lock(&res->spinlock);
1718 res->state &= ~DLM_LOCK_RES_SETREF_INPROG;
1719 spin_unlock(&res->spinlock);
1720 wake_up(&res->wq);
1721
6714d8e8
KH		 1722 return ret;
1723}
1724
1725/*
1726 * locks that can be taken here:
1727 * dlm->spinlock
1728 * res->spinlock
1729 * mle->spinlock
1730 * dlm->master_list
1731 *
1732 * if possible, TRIM THIS DOWN!!!
1733 */
d74c9803
KH		 1734int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data,
1735 void **ret_data)
6714d8e8
KH		 1736{
1737 struct dlm_ctxt *dlm = data;
1738 struct dlm_master_list_entry *mle = NULL;
1739 struct dlm_assert_master *assert = (struct dlm_assert_master *)msg->buf;
1740 struct dlm_lock_resource *res = NULL;
1741 char *name;
a3d33291 1742 unsigned int namelen, hash;
6714d8e8 1743 u32 flags;
ba2bf218 1744 int master_request = 0, have_lockres_ref = 0;
9c6510a5 1745 int ret = 0;
6714d8e8
KH		 1746
1747 if (!dlm_grab(dlm))
1748 return 0;
1749
1750 name = assert->name;
1751 namelen = assert->namelen;
a3d33291 1752 hash = dlm_lockid_hash(name, namelen);
6714d8e8
KH		 1753 flags = be32_to_cpu(assert->flags);
1754
1755 if (namelen > DLM_LOCKID_NAME_MAX) {
1756 mlog(ML_ERROR, "Invalid name length!");
1757 goto done;
1758 }
1759
1760 spin_lock(&dlm->spinlock);
1761
1762 if (flags)
1763 mlog(0, "assert_master with flags: %u\n", flags);
1764
1765 /* find the MLE */
1766 spin_lock(&dlm->master_lock);
1767 if (!dlm_find_mle(dlm, &mle, name, namelen)) {
1768 /* not an error, could be master just re-asserting */
1769 mlog(0, "just got an assert_master from %u, but no "
1770 "MLE for it! (%.*s)\n", assert->node_idx,
1771 namelen, name);
1772 } else {
1773 int bit = find_next_bit (mle->maybe_map, O2NM_MAX_NODES, 0);
1774 if (bit >= O2NM_MAX_NODES) {
1775 /* not necessarily an error, though less likely.
1776 * could be master just re-asserting. */
aa852354 1777 mlog(0, "no bits set in the maybe_map, but %u "
6714d8e8
KH		 1778 "is asserting! (%.*s)\n", assert->node_idx,
1779 namelen, name);
1780 } else if (bit != assert->node_idx) {
1781 if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
1782 mlog(0, "master %u was found, %u should "
1783 "back off\n", assert->node_idx, bit);
1784 } else {
1785 /* with the fix for bug 569, a higher node
1786 * number winning the mastery will respond
1787 * YES to mastery requests, but this node
1788 * had no way of knowing. let it pass. */
aa852354 1789 mlog(0, "%u is the lowest node, "
6714d8e8
KH		 1790 "%u is asserting. (%.*s) %u must "
1791 "have begun after %u won.\n", bit,
1792 assert->node_idx, namelen, name, bit,
1793 assert->node_idx);
1794 }
1795 }
2d1a868c
KH		 1796 if (mle->type == DLM_MLE_MIGRATION) {
1797 if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
1798 mlog(0, "%s:%.*s: got cleanup assert"
1799 " from %u for migration\n",
1800 dlm->name, namelen, name,
1801 assert->node_idx);
1802 } else if (!(flags & DLM_ASSERT_MASTER_FINISH_MIGRATION)) {
1803 mlog(0, "%s:%.*s: got unrelated assert"
1804 " from %u for migration, ignoring\n",
1805 dlm->name, namelen, name,
1806 assert->node_idx);
1807 __dlm_put_mle(mle);
1808 spin_unlock(&dlm->master_lock);
1809 spin_unlock(&dlm->spinlock);
1810 goto done;
2bd63216 1811 }
2d1a868c 1812 }
6714d8e8
KH		 1813 }
1814 spin_unlock(&dlm->master_lock);
1815
1816 /* ok everything checks out with the MLE
1817 * now check to see if there is a lockres */
a3d33291 1818 res = __dlm_lookup_lockres(dlm, name, namelen, hash);
6714d8e8
KH		 1819 if (res) {
1820 spin_lock(&res->spinlock);
1821 if (res->state & DLM_LOCK_RES_RECOVERING) {
1822 mlog(ML_ERROR, "%u asserting but %.*s is "
1823 "RECOVERING!\n", assert->node_idx, namelen, name);
1824 goto kill;
1825 }
1826 if (!mle) {
dc2ed195
KH		 1827 if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN &&
1828 res->owner != assert->node_idx) {
53ecd25e
SM		 1829 mlog(ML_ERROR, "DIE! Mastery assert from %u, "
1830 "but current owner is %u! (%.*s)\n",
1831 assert->node_idx, res->owner, namelen,
1832 name);
1833 __dlm_print_one_lock_resource(res);
1834 BUG();
6714d8e8
KH		 1835 }
1836 } else if (mle->type != DLM_MLE_MIGRATION) {
1837 if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
1838 /* owner is just re-asserting */
1839 if (res->owner == assert->node_idx) {
1840 mlog(0, "owner %u re-asserting on "
1841 "lock %.*s\n", assert->node_idx,
1842 namelen, name);
1843 goto ok;
1844 }
1845 mlog(ML_ERROR, "got assert_master from "
1846 "node %u, but %u is the owner! "
1847 "(%.*s)\n", assert->node_idx,
1848 res->owner, namelen, name);
1849 goto kill;
1850 }
1851 if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
1852 mlog(ML_ERROR, "got assert from %u, but lock "
1853 "with no owner should be "
1854 "in-progress! (%.*s)\n",
1855 assert->node_idx,
1856 namelen, name);
1857 goto kill;
1858 }
1859 } else /* mle->type == DLM_MLE_MIGRATION */ {
1860 /* should only be getting an assert from new master */
1861 if (assert->node_idx != mle->new_master) {
1862 mlog(ML_ERROR, "got assert from %u, but "
1863 "new master is %u, and old master "
1864 "was %u (%.*s)\n",
1865 assert->node_idx, mle->new_master,
1866 mle->master, namelen, name);
1867 goto kill;
1868 }
1869
1870 }
1871ok:
1872 spin_unlock(&res->spinlock);
1873 }
6714d8e8
KH		 1874
1875 // mlog(0, "woo! got an assert_master from node %u!\n",
1876 // assert->node_idx);
1877 if (mle) {
9c6510a5
KH		 1878 int extra_ref = 0;
1879 int nn = -1;
a2bf0477 1880 int rr, err = 0;
2bd63216 1881
6714d8e8 1882 spin_lock(&mle->spinlock);
9c6510a5
KH		 1883 if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION)
1884 extra_ref = 1;
1885 else {
1886 /* MASTER mle: if any bits set in the response map
1887 * then the calling node needs to re-assert to clear
1888 * up nodes that this node contacted */
2bd63216 1889 while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES,
9c6510a5
KH		 1890 nn+1)) < O2NM_MAX_NODES) {
1891 if (nn != dlm->node_num && nn != assert->node_idx)
1892 master_request = 1;
1893 }
1894 }
6714d8e8
KH		 1895 mle->master = assert->node_idx;
1896 atomic_set(&mle->woken, 1);
1897 wake_up(&mle->wq);
1898 spin_unlock(&mle->spinlock);
1899
a2bf0477 1900 if (res) {
a6fa3640 1901 int wake = 0;
6714d8e8 1902 spin_lock(&res->spinlock);
a2bf0477
KH		 1903 if (mle->type == DLM_MLE_MIGRATION) {
1904 mlog(0, "finishing off migration of lockres %.*s, "
1905 "from %u to %u\n",
1906 res->lockname.len, res->lockname.name,
1907 dlm->node_num, mle->new_master);
1908 res->state &= ~DLM_LOCK_RES_MIGRATING;
a6fa3640 1909 wake = 1;
a2bf0477
KH		 1910 dlm_change_lockres_owner(dlm, res, mle->new_master);
1911 BUG_ON(res->state & DLM_LOCK_RES_DIRTY);
1912 } else {
1913 dlm_change_lockres_owner(dlm, res, mle->master);
1914 }
6714d8e8 1915 spin_unlock(&res->spinlock);
ba2bf218 1916 have_lockres_ref = 1;
a6fa3640
KH		 1917 if (wake)
1918 wake_up(&res->wq);
6714d8e8 1919 }
a2bf0477
KH		 1920
1921 /* master is known, detach if not already detached.
1922 * ensures that only one assert_master call will happen
1923 * on this mle. */
a2bf0477
KH		 1924 spin_lock(&dlm->master_lock);
1925
1926 rr = atomic_read(&mle->mle_refs.refcount);
1927 if (mle->inuse > 0) {
1928 if (extra_ref && rr < 3)
1929 err = 1;
1930 else if (!extra_ref && rr < 2)
1931 err = 1;
1932 } else {
1933 if (extra_ref && rr < 2)
1934 err = 1;
1935 else if (!extra_ref && rr < 1)
1936 err = 1;
1937 }
1938 if (err) {
1939 mlog(ML_ERROR, "%s:%.*s: got assert master from %u "
1940 "that will mess up this node, refs=%d, extra=%d, "
1941 "inuse=%d\n", dlm->name, namelen, name,
1942 assert->node_idx, rr, extra_ref, mle->inuse);
1943 dlm_print_one_mle(mle);
1944 }
1c084577 1945 __dlm_unlink_mle(dlm, mle);
a2bf0477
KH		 1946 __dlm_mle_detach_hb_events(dlm, mle);
1947 __dlm_put_mle(mle);
6714d8e8
KH		 1948 if (extra_ref) {
1949 /* the assert master message now balances the extra
1950 * ref given by the master / migration request message.
1951 * if this is the last put, it will be removed
1952 * from the list. */
a2bf0477
KH		 1953 __dlm_put_mle(mle);
1954 }
1955 spin_unlock(&dlm->master_lock);
a2bf0477
KH		 1956 } else if (res) {
1957 if (res->owner != assert->node_idx) {
1958 mlog(0, "assert_master from %u, but current "
1959 "owner is %u (%.*s), no mle\n", assert->node_idx,
1960 res->owner, namelen, name);
6714d8e8
KH		 1961 }
1962 }
14741472 1963 spin_unlock(&dlm->spinlock);
6714d8e8
KH		 1964
1965done:
9c6510a5 1966 ret = 0;
3b8118cf
KH		 1967 if (res) {
1968 spin_lock(&res->spinlock);
1969 res->state |= DLM_LOCK_RES_SETREF_INPROG;
1970 spin_unlock(&res->spinlock);
1971 *ret_data = (void *)res;
1972 }
6714d8e8 1973 dlm_put(dlm);
9c6510a5
KH		 1974 if (master_request) {
1975 mlog(0, "need to tell master to reassert\n");
ba2bf218
KH		 1976 /* positive. negative would shoot down the node. */
1977 ret |= DLM_ASSERT_RESPONSE_REASSERT;
1978 if (!have_lockres_ref) {
1979 mlog(ML_ERROR, "strange, got assert from %u, MASTER "
1980 "mle present here for %s:%.*s, but no lockres!\n",
1981 assert->node_idx, dlm->name, namelen, name);
1982 }
1983 }
1984 if (have_lockres_ref) {
1985 /* let the master know we have a reference to the lockres */
1986 ret |= DLM_ASSERT_RESPONSE_MASTERY_REF;
1987 mlog(0, "%s:%.*s: got assert from %u, need a ref\n",
1988 dlm->name, namelen, name, assert->node_idx);
9c6510a5
KH		 1989 }
1990 return ret;
6714d8e8
KH		 1991
1992kill:
1993 /* kill the caller! */
a9ee4c8a
KH		 1994 mlog(ML_ERROR, "Bad message received from another node. Dumping state "
1995 "and killing the other node now! This node is OK and can continue.\n");
1996 __dlm_print_one_lock_resource(res);
6714d8e8
KH		 1997 spin_unlock(&res->spinlock);
1998 spin_unlock(&dlm->spinlock);
2bd63216 1999 *ret_data = (void *)res;
6714d8e8
KH		 2000 dlm_put(dlm);
2001 return -EINVAL;
2002}
2003
3b8118cf
KH		 2004void dlm_assert_master_post_handler(int status, void *data, void *ret_data)
2005{
2006 struct dlm_lock_resource *res = (struct dlm_lock_resource *)ret_data;
2007
2008 if (ret_data) {
2009 spin_lock(&res->spinlock);
2010 res->state &= ~DLM_LOCK_RES_SETREF_INPROG;
2011 spin_unlock(&res->spinlock);
2012 wake_up(&res->wq);
2013 dlm_lockres_put(res);
2014 }
2015 return;
2016}
2017
6714d8e8
KH		 2018int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
2019 struct dlm_lock_resource *res,
2020 int ignore_higher, u8 request_from, u32 flags)
2021{
2022 struct dlm_work_item *item;
b24ae0b5 2023 item = kzalloc(sizeof(*item), GFP_ATOMIC);
6714d8e8
KH		 2024 if (!item)
2025 return -ENOMEM;
2026
2027
2028 /* queue up work for dlm_assert_master_worker */
2029 dlm_grab(dlm); /* get an extra ref for the work item */
2030 dlm_init_work_item(dlm, item, dlm_assert_master_worker, NULL);
2031 item->u.am.lockres = res; /* already have a ref */
2032 /* can optionally ignore node numbers higher than this node */
2033 item->u.am.ignore_higher = ignore_higher;
2034 item->u.am.request_from = request_from;
2035 item->u.am.flags = flags;
2036
2bd63216
SM		 2037 if (ignore_higher)
2038 mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len,
9c6510a5 2039 res->lockname.name);
2bd63216 2040
6714d8e8
KH		 2041 spin_lock(&dlm->work_lock);
2042 list_add_tail(&item->list, &dlm->work_list);
2043 spin_unlock(&dlm->work_lock);
2044
3156d267 2045 queue_work(dlm->dlm_worker, &dlm->dispatched_work);
6714d8e8
KH		 2046 return 0;
2047}
2048
2049static void dlm_assert_master_worker(struct dlm_work_item *item, void *data)
2050{
2051 struct dlm_ctxt *dlm = data;
2052 int ret = 0;
2053 struct dlm_lock_resource *res;
2054 unsigned long nodemap[BITS_TO_LONGS(O2NM_MAX_NODES)];
2055 int ignore_higher;
2056 int bit;
2057 u8 request_from;
2058 u32 flags;
2059
2060 dlm = item->dlm;
2061 res = item->u.am.lockres;
2062 ignore_higher = item->u.am.ignore_higher;
2063 request_from = item->u.am.request_from;
2064 flags = item->u.am.flags;
2065
2066 spin_lock(&dlm->spinlock);
2067 memcpy(nodemap, dlm->domain_map, sizeof(nodemap));
2068 spin_unlock(&dlm->spinlock);
2069
2070 clear_bit(dlm->node_num, nodemap);
2071 if (ignore_higher) {
2072 /* if is this just to clear up mles for nodes below
2073 * this node, do not send the message to the original
2074 * caller or any node number higher than this */
2075 clear_bit(request_from, nodemap);
2076 bit = dlm->node_num;
2077 while (1) {
2078 bit = find_next_bit(nodemap, O2NM_MAX_NODES,
2079 bit+1);
2080 if (bit >= O2NM_MAX_NODES)
2081 break;
2082 clear_bit(bit, nodemap);
2083 }
2084 }
2085
36407488
KH		 2086 /*
2087 * If we're migrating this lock to someone else, we are no
2088 * longer allowed to assert out own mastery. OTOH, we need to
2089 * prevent migration from starting while we're still asserting
2090 * our dominance. The reserved ast delays migration.
2091 */
2092 spin_lock(&res->spinlock);
2093 if (res->state & DLM_LOCK_RES_MIGRATING) {
2094 mlog(0, "Someone asked us to assert mastery, but we're "
2095 "in the middle of migration. Skipping assert, "
2096 "the new master will handle that.\n");
2097 spin_unlock(&res->spinlock);
2098 goto put;
2099 } else
2100 __dlm_lockres_reserve_ast(res);
2101 spin_unlock(&res->spinlock);
2102
6714d8e8
KH		 2103 /* this call now finishes out the nodemap
2104 * even if one or more nodes die */
2105 mlog(0, "worker about to master %.*s here, this=%u\n",
2106 res->lockname.len, res->lockname.name, dlm->node_num);
ba2bf218 2107 ret = dlm_do_assert_master(dlm, res, nodemap, flags);
6714d8e8
KH		 2108 if (ret < 0) {
2109 /* no need to restart, we are done */
3b3b84a8
KH		 2110 if (!dlm_is_host_down(ret))
2111 mlog_errno(ret);
6714d8e8
KH		 2112 }
2113
36407488
KH		 2114 /* Ok, we've asserted ourselves. Let's let migration start. */
2115 dlm_lockres_release_ast(dlm, res);
2116
2117put:
6714d8e8
KH		 2118 dlm_lockres_put(res);
2119
2120 mlog(0, "finished with dlm_assert_master_worker\n");
2121}
2122
c03872f5
KH		 2123/* SPECIAL CASE for the $RECOVERY lock used by the recovery thread.
2124 * We cannot wait for node recovery to complete to begin mastering this
2125 * lockres because this lockres is used to kick off recovery! ;-)
2126 * So, do a pre-check on all living nodes to see if any of those nodes
2127 * think that $RECOVERY is currently mastered by a dead node. If so,
2128 * we wait a short time to allow that node to get notified by its own
2129 * heartbeat stack, then check again. All $RECOVERY lock resources
2bd63216 2130 * mastered by dead nodes are purged when the hearbeat callback is
c03872f5
KH		 2131 * fired, so we can know for sure that it is safe to continue once
2132 * the node returns a live node or no node. */
2133static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
2134 struct dlm_lock_resource *res)
2135{
2136 struct dlm_node_iter iter;
2137 int nodenum;
2138 int ret = 0;
2139 u8 master = DLM_LOCK_RES_OWNER_UNKNOWN;
2140
2141 spin_lock(&dlm->spinlock);
2142 dlm_node_iter_init(dlm->domain_map, &iter);
2143 spin_unlock(&dlm->spinlock);
2144
2145 while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
2146 /* do not send to self */
2147 if (nodenum == dlm->node_num)
2148 continue;
2149 ret = dlm_do_master_requery(dlm, res, nodenum, &master);
2150 if (ret < 0) {
2151 mlog_errno(ret);
2152 if (!dlm_is_host_down(ret))
2153 BUG();
2154 /* host is down, so answer for that node would be
2155 * DLM_LOCK_RES_OWNER_UNKNOWN. continue. */
f42a100b 2156 ret = 0;
c03872f5
KH		 2157 }
2158
2159 if (master != DLM_LOCK_RES_OWNER_UNKNOWN) {
2160 /* check to see if this master is in the recovery map */
2161 spin_lock(&dlm->spinlock);
2162 if (test_bit(master, dlm->recovery_map)) {
2163 mlog(ML_NOTICE, "%s: node %u has not seen "
2164 "node %u go down yet, and thinks the "
2165 "dead node is mastering the recovery "
2166 "lock. must wait.\n", dlm->name,
2167 nodenum, master);
2168 ret = -EAGAIN;
2169 }
2170 spin_unlock(&dlm->spinlock);
2bd63216 2171 mlog(0, "%s: reco lock master is %u\n", dlm->name,
c03872f5
KH		 2172 master);
2173 break;
2174 }
2175 }
2176 return ret;
2177}
2178
ba2bf218
KH		 2179/*
2180 * DLM_DEREF_LOCKRES_MSG
2181 */
2182
2183int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
2184{
2185 struct dlm_deref_lockres deref;
2186 int ret = 0, r;
2187 const char *lockname;
2188 unsigned int namelen;
2189
2190 lockname = res->lockname.name;
2191 namelen = res->lockname.len;
2192 BUG_ON(namelen > O2NM_MAX_NAME_LEN);
2193
ba2bf218
KH		 2194 memset(&deref, 0, sizeof(deref));
2195 deref.node_idx = dlm->node_num;
2196 deref.namelen = namelen;
2197 memcpy(deref.name, lockname, namelen);
2198
2199 ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key,
2200 &deref, sizeof(deref), res->owner, &r);
2201 if (ret < 0)
8decab3c
SM		 2202 mlog(ML_ERROR, "%s: res %.*s, error %d send DEREF to node %u\n",
2203 dlm->name, namelen, lockname, ret, res->owner);
ba2bf218
KH		 2204 else if (r < 0) {
2205 /* BAD. other node says I did not have a ref. */
8decab3c
SM		 2206 mlog(ML_ERROR, "%s: res %.*s, DEREF to node %u got %d\n",
2207 dlm->name, namelen, lockname, res->owner, r);
ba2bf218
KH		 2208 dlm_print_one_lock_resource(res);
2209 BUG();
2210 }
2211 return ret;
2212}
2213
d74c9803
KH		 2214int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
2215 void **ret_data)
ba2bf218
KH		 2216{
2217 struct dlm_ctxt *dlm = data;
2218 struct dlm_deref_lockres *deref = (struct dlm_deref_lockres *)msg->buf;
2219 struct dlm_lock_resource *res = NULL;
2220 char *name;
2221 unsigned int namelen;
2222 int ret = -EINVAL;
2223 u8 node;
2224 unsigned int hash;
f3f85464
SM		 2225 struct dlm_work_item *item;
2226 int cleared = 0;
2227 int dispatch = 0;
ba2bf218
KH		 2228
2229 if (!dlm_grab(dlm))
2230 return 0;
2231
2232 name = deref->name;
2233 namelen = deref->namelen;
2234 node = deref->node_idx;
2235
2236 if (namelen > DLM_LOCKID_NAME_MAX) {
2237 mlog(ML_ERROR, "Invalid name length!");
2238 goto done;
2239 }
2240 if (deref->node_idx >= O2NM_MAX_NODES) {
2241 mlog(ML_ERROR, "Invalid node number: %u\n", node);
2242 goto done;
2243 }
2244
2245 hash = dlm_lockid_hash(name, namelen);
2246
2247 spin_lock(&dlm->spinlock);
2248 res = __dlm_lookup_lockres_full(dlm, name, namelen, hash);
2249 if (!res) {
2250 spin_unlock(&dlm->spinlock);
2251 mlog(ML_ERROR, "%s:%.*s: bad lockres name\n",
2252 dlm->name, namelen, name);
2253 goto done;
2254 }
2255 spin_unlock(&dlm->spinlock);
2256
2257 spin_lock(&res->spinlock);
f3f85464
SM		 2258 if (res->state & DLM_LOCK_RES_SETREF_INPROG)
2259 dispatch = 1;
2260 else {
2261 BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF);
2262 if (test_bit(node, res->refmap)) {
8d400b81 2263 dlm_lockres_clear_refmap_bit(dlm, res, node);
f3f85464
SM		 2264 cleared = 1;
2265 }
ba2bf218
KH		 2266 }
2267 spin_unlock(&res->spinlock);
2268
f3f85464
SM		 2269 if (!dispatch) {
2270 if (cleared)
2271 dlm_lockres_calc_usage(dlm, res);
2272 else {
2273 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2274 "but it is already dropped!\n", dlm->name,
2275 res->lockname.len, res->lockname.name, node);
2af37ce8 2276 dlm_print_one_lock_resource(res);
f3f85464
SM		 2277 }
2278 ret = 0;
2279 goto done;
2280 }
2281
2282 item = kzalloc(sizeof(*item), GFP_NOFS);
2283 if (!item) {
2284 ret = -ENOMEM;
2285 mlog_errno(ret);
2286 goto done;
2287 }
2288
2289 dlm_init_work_item(dlm, item, dlm_deref_lockres_worker, NULL);
2290 item->u.dl.deref_res = res;
2291 item->u.dl.deref_node = node;
2292
2293 spin_lock(&dlm->work_lock);
2294 list_add_tail(&item->list, &dlm->work_list);
2295 spin_unlock(&dlm->work_lock);
2296
2297 queue_work(dlm->dlm_worker, &dlm->dispatched_work);
2298 return 0;
2299
ba2bf218
KH		 2300done:
2301 if (res)
2302 dlm_lockres_put(res);
2303 dlm_put(dlm);
f3f85464 2304
ba2bf218
KH		 2305 return ret;
2306}
2307
f3f85464
SM		 2308static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
2309{
2310 struct dlm_ctxt *dlm;
2311 struct dlm_lock_resource *res;
2312 u8 node;
2313 u8 cleared = 0;
2314
2315 dlm = item->dlm;
2316 res = item->u.dl.deref_res;
2317 node = item->u.dl.deref_node;
2318
2319 spin_lock(&res->spinlock);
2320 BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF);
2321 if (test_bit(node, res->refmap)) {
2322 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
8d400b81 2323 dlm_lockres_clear_refmap_bit(dlm, res, node);
f3f85464
SM		 2324 cleared = 1;
2325 }
2326 spin_unlock(&res->spinlock);
2327
2328 if (cleared) {
2329 mlog(0, "%s:%.*s node %u ref dropped in dispatch\n",
2330 dlm->name, res->lockname.len, res->lockname.name, node);
2331 dlm_lockres_calc_usage(dlm, res);
2332 } else {
2333 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2334 "but it is already dropped!\n", dlm->name,
2335 res->lockname.len, res->lockname.name, node);
2af37ce8 2336 dlm_print_one_lock_resource(res);
f3f85464
SM		 2337 }
2338
2339 dlm_lockres_put(res);
2340}
2341
9f62e960
SM		 2342/*
2343 * A migrateable resource is one that is :
2344 * 1. locally mastered, and,
2345 * 2. zero local locks, and,
2346 * 3. one or more non-local locks, or, one or more references
2347 * Returns 1 if yes, 0 if not.
2f5bf1f2
SM		 2348 */
2349static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
9f62e960 2350 struct dlm_lock_resource *res)
2f5bf1f2 2351{
9f62e960
SM		 2352 enum dlm_lockres_list idx;
2353 int nonlocal = 0, node_ref;
800deef3 2354 struct list_head *queue;
2f5bf1f2 2355 struct dlm_lock *lock;
9f62e960 2356 u64 cookie;
2f5bf1f2
SM		 2357
2358 assert_spin_locked(&res->spinlock);
2359
9f62e960
SM		 2360 if (res->owner != dlm->node_num)
2361 return 0;
2f5bf1f2 2362
9f62e960
SM		 2363 for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) {
2364 queue = dlm_list_idx_to_ptr(res, idx);
800deef3 2365 list_for_each_entry(lock, queue, list) {
9f62e960
SM		 2366 if (lock->ml.node != dlm->node_num) {
2367 nonlocal++;
2368 continue;
2f5bf1f2 2369 }
9f62e960
SM		 2370 cookie = be64_to_cpu(lock->ml.cookie);
2371 mlog(0, "%s: Not migrateable res %.*s, lock %u:%llu on "
2372 "%s list\n", dlm->name, res->lockname.len,
2373 res->lockname.name,
2374 dlm_get_lock_cookie_node(cookie),
2375 dlm_get_lock_cookie_seq(cookie),
2376 dlm_list_in_text(idx));
2377 return 0;
2f5bf1f2 2378 }
2f5bf1f2
SM		 2379 }
2380
9f62e960
SM		 2381 if (!nonlocal) {
2382 node_ref = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
2383 if (node_ref >= O2NM_MAX_NODES)
2384 return 0;
2385 }
388c4bcb 2386
9f62e960
SM		 2387 mlog(0, "%s: res %.*s, Migrateable\n", dlm->name, res->lockname.len,
2388 res->lockname.name);
2f5bf1f2 2389
9f62e960 2390 return 1;
2f5bf1f2 2391}
6714d8e8
KH		 2392
2393/*
2394 * DLM_MIGRATE_LOCKRES
2395 */
2396
2397
faf0ec9f 2398static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
66effd3c 2399 struct dlm_lock_resource *res, u8 target)
6714d8e8
KH		 2400{
2401 struct dlm_master_list_entry *mle = NULL;
2402 struct dlm_master_list_entry *oldmle = NULL;
2403 struct dlm_migratable_lockres *mres = NULL;
2f5bf1f2 2404 int ret = 0;
6714d8e8
KH		 2405 const char *name;
2406 unsigned int namelen;
2407 int mle_added = 0;
2f5bf1f2 2408 int wake = 0;
6714d8e8
KH		 2409
2410 if (!dlm_grab(dlm))
2411 return -EINVAL;
2412
66effd3c
SM		 2413 BUG_ON(target == O2NM_MAX_NODES);
2414
6714d8e8
KH		 2415 name = res->lockname.name;
2416 namelen = res->lockname.len;
2417
66effd3c
SM		 2418 mlog(0, "%s: Migrating %.*s to node %u\n", dlm->name, namelen, name,
2419 target);
6714d8e8 2420
66effd3c 2421 /* preallocate up front. if this fails, abort */
6714d8e8 2422 ret = -ENOMEM;
ad8100e0 2423 mres = (struct dlm_migratable_lockres *) __get_free_page(GFP_NOFS);
6714d8e8
KH		 2424 if (!mres) {
2425 mlog_errno(ret);
2426 goto leave;
2427 }
2428
3914ed0c 2429 mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
6714d8e8
KH		 2430 if (!mle) {
2431 mlog_errno(ret);
2432 goto leave;
2433 }
2434 ret = 0;
2435
6714d8e8
KH		 2436 /*
2437 * clear any existing master requests and
2438 * add the migration mle to the list
2439 */
66effd3c 2440 spin_lock(&dlm->spinlock);
6714d8e8
KH		 2441 spin_lock(&dlm->master_lock);
2442 ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, name,
2443 namelen, target, dlm->node_num);
2444 spin_unlock(&dlm->master_lock);
2445 spin_unlock(&dlm->spinlock);
2446
2447 if (ret == -EEXIST) {
2448 mlog(0, "another process is already migrating it\n");
2449 goto fail;
2450 }
2451 mle_added = 1;
2452
2453 /*
2454 * set the MIGRATING flag and flush asts
2455 * if we fail after this we need to re-dirty the lockres
2456 */
2457 if (dlm_mark_lockres_migrating(dlm, res, target) < 0) {
2458 mlog(ML_ERROR, "tried to migrate %.*s to %u, but "
2459 "the target went down.\n", res->lockname.len,
2460 res->lockname.name, target);
2461 spin_lock(&res->spinlock);
2462 res->state &= ~DLM_LOCK_RES_MIGRATING;
a6fa3640 2463 wake = 1;
6714d8e8
KH		 2464 spin_unlock(&res->spinlock);
2465 ret = -EINVAL;
2466 }
2467
2468fail:
2469 if (oldmle) {
2470 /* master is known, detach if not already detached */
2471 dlm_mle_detach_hb_events(dlm, oldmle);
2472 dlm_put_mle(oldmle);
2473 }
2474
2475 if (ret < 0) {
2476 if (mle_added) {
2477 dlm_mle_detach_hb_events(dlm, mle);
2478 dlm_put_mle(mle);
2479 } else if (mle) {
2480 kmem_cache_free(dlm_mle_cache, mle);
66effd3c 2481 mle = NULL;
6714d8e8
KH		 2482 }
2483 goto leave;
2484 }
2485
2486 /*
2487 * at this point, we have a migration target, an mle
2488 * in the master list, and the MIGRATING flag set on
2489 * the lockres
2490 */
2491
1cd04dbe
KH		 2492 /* now that remote nodes are spinning on the MIGRATING flag,
2493 * ensure that all assert_master work is flushed. */
2494 flush_workqueue(dlm->dlm_worker);
6714d8e8
KH		 2495
2496 /* get an extra reference on the mle.
2497 * otherwise the assert_master from the new
2498 * master will destroy this.
2499 * also, make sure that all callers of dlm_get_mle
2500 * take both dlm->spinlock and dlm->master_lock */
2501 spin_lock(&dlm->spinlock);
2502 spin_lock(&dlm->master_lock);
a2bf0477 2503 dlm_get_mle_inuse(mle);
6714d8e8
KH		 2504 spin_unlock(&dlm->master_lock);
2505 spin_unlock(&dlm->spinlock);
2506
2507 /* notify new node and send all lock state */
2508 /* call send_one_lockres with migration flag.
2509 * this serves as notice to the target node that a
2510 * migration is starting. */
2511 ret = dlm_send_one_lockres(dlm, res, mres, target,
2512 DLM_MRES_MIGRATION);
2513
2514 if (ret < 0) {
2515 mlog(0, "migration to node %u failed with %d\n",
2516 target, ret);
2517 /* migration failed, detach and clean up mle */
2518 dlm_mle_detach_hb_events(dlm, mle);
2519 dlm_put_mle(mle);
a2bf0477
KH		 2520 dlm_put_mle_inuse(mle);
2521 spin_lock(&res->spinlock);
2522 res->state &= ~DLM_LOCK_RES_MIGRATING;
a6fa3640 2523 wake = 1;
a2bf0477 2524 spin_unlock(&res->spinlock);
df016c66
SM		 2525 if (dlm_is_host_down(ret))
2526 dlm_wait_for_node_death(dlm, target,
2527 DLM_NODE_DEATH_WAIT_MAX);
6714d8e8
KH		 2528 goto leave;
2529 }
2530
2531 /* at this point, the target sends a message to all nodes,
2532 * (using dlm_do_migrate_request). this node is skipped since
2533 * we had to put an mle in the list to begin the process. this
2534 * node now waits for target to do an assert master. this node
2535 * will be the last one notified, ensuring that the migration
2536 * is complete everywhere. if the target dies while this is
2537 * going on, some nodes could potentially see the target as the
2538 * master, so it is important that my recovery finds the migration
af901ca1 2539 * mle and sets the master to UNKNOWN. */
6714d8e8
KH		 2540
2541
2542 /* wait for new node to assert master */
2543 while (1) {
2544 ret = wait_event_interruptible_timeout(mle->wq,
2545 (atomic_read(&mle->woken) == 1),
2546 msecs_to_jiffies(5000));
2547
2548 if (ret >= 0) {
2549 if (atomic_read(&mle->woken) == 1 ||
2550 res->owner == target)
2551 break;
2552
1cd04dbe
KH		 2553 mlog(0, "%s:%.*s: timed out during migration\n",
2554 dlm->name, res->lockname.len, res->lockname.name);
2bd63216 2555 /* avoid hang during shutdown when migrating lockres
e2faea4c
KH		 2556 * to a node which also goes down */
2557 if (dlm_is_node_dead(dlm, target)) {
aa852354
KH		 2558 mlog(0, "%s:%.*s: expected migration "
2559 "target %u is no longer up, restarting\n",
e2faea4c
KH		 2560 dlm->name, res->lockname.len,
2561 res->lockname.name, target);
1cd04dbe
KH		 2562 ret = -EINVAL;
2563 /* migration failed, detach and clean up mle */
2564 dlm_mle_detach_hb_events(dlm, mle);
2565 dlm_put_mle(mle);
2566 dlm_put_mle_inuse(mle);
2567 spin_lock(&res->spinlock);
2568 res->state &= ~DLM_LOCK_RES_MIGRATING;
a6fa3640 2569 wake = 1;
1cd04dbe
KH		 2570 spin_unlock(&res->spinlock);
2571 goto leave;
e2faea4c 2572 }
1cd04dbe
KH		 2573 } else
2574 mlog(0, "%s:%.*s: caught signal during migration\n",
2575 dlm->name, res->lockname.len, res->lockname.name);
6714d8e8
KH		 2576 }
2577
2578 /* all done, set the owner, clear the flag */
2579 spin_lock(&res->spinlock);
2580 dlm_set_lockres_owner(dlm, res, target);
2581 res->state &= ~DLM_LOCK_RES_MIGRATING;
2582 dlm_remove_nonlocal_locks(dlm, res);
2583 spin_unlock(&res->spinlock);
2584 wake_up(&res->wq);
2585
2586 /* master is known, detach if not already detached */
2587 dlm_mle_detach_hb_events(dlm, mle);
a2bf0477 2588 dlm_put_mle_inuse(mle);
6714d8e8
KH		 2589 ret = 0;
2590
2591 dlm_lockres_calc_usage(dlm, res);
2592
2593leave:
2594 /* re-dirty the lockres if we failed */
2595 if (ret < 0)
2596 dlm_kick_thread(dlm, res);
2597
a6fa3640
KH		 2598 /* wake up waiters if the MIGRATING flag got set
2599 * but migration failed */
2600 if (wake)
2601 wake_up(&res->wq);
2602
6714d8e8
KH		 2603 if (mres)
2604 free_page((unsigned long)mres);
2605
2606 dlm_put(dlm);
2607
9f62e960
SM		 2608 mlog(0, "%s: Migrating %.*s to %u, returns %d\n", dlm->name, namelen,
2609 name, target, ret);
6714d8e8
KH		 2610 return ret;
2611}
6714d8e8 2612
ba2bf218
KH		 2613#define DLM_MIGRATION_RETRY_MS 100
2614
9f62e960
SM		 2615/*
2616 * Should be called only after beginning the domain leave process.
ba2bf218
KH		 2617 * There should not be any remaining locks on nonlocal lock resources,
2618 * and there should be no local locks left on locally mastered resources.
2619 *
2620 * Called with the dlm spinlock held, may drop it to do migration, but
2621 * will re-acquire before exit.
2622 *
9f62e960
SM		 2623 * Returns: 1 if dlm->spinlock was dropped/retaken, 0 if never dropped
2624 */
ba2bf218
KH		 2625int dlm_empty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
2626{
66effd3c 2627 int ret;
ba2bf218 2628 int lock_dropped = 0;
66effd3c 2629 u8 target = O2NM_MAX_NODES;
ba2bf218 2630
9f62e960 2631 assert_spin_locked(&dlm->spinlock);
2f5bf1f2 2632
9f62e960 2633 spin_lock(&res->spinlock);
66effd3c
SM		 2634 if (dlm_is_lockres_migrateable(dlm, res))
2635 target = dlm_pick_migration_target(dlm, res);
b36c3f84 2636 spin_unlock(&res->spinlock);
66effd3c
SM		 2637
2638 if (target == O2NM_MAX_NODES)
9f62e960 2639 goto leave;
ba2bf218
KH		 2640
2641 /* Wheee! Migrate lockres here! Will sleep so drop spinlock. */
2642 spin_unlock(&dlm->spinlock);
2643 lock_dropped = 1;
66effd3c
SM		 2644 ret = dlm_migrate_lockres(dlm, res, target);
2645 if (ret)
2646 mlog(0, "%s: res %.*s, Migrate to node %u failed with %d\n",
2647 dlm->name, res->lockname.len, res->lockname.name,
2648 target, ret);
ba2bf218
KH		 2649 spin_lock(&dlm->spinlock);
2650leave:
2651 return lock_dropped;
2652}
2653
6714d8e8
KH		 2654int dlm_lock_basts_flushed(struct dlm_ctxt *dlm, struct dlm_lock *lock)
2655{
2656 int ret;
2657 spin_lock(&dlm->ast_lock);
2658 spin_lock(&lock->spinlock);
2659 ret = (list_empty(&lock->bast_list) && !lock->bast_pending);
2660 spin_unlock(&lock->spinlock);
2661 spin_unlock(&dlm->ast_lock);
2662 return ret;
2663}
2664
2665static int dlm_migration_can_proceed(struct dlm_ctxt *dlm,
2666 struct dlm_lock_resource *res,
2667 u8 mig_target)
2668{
2669 int can_proceed;
2670 spin_lock(&res->spinlock);
2671 can_proceed = !!(res->state & DLM_LOCK_RES_MIGRATING);
2672 spin_unlock(&res->spinlock);
2673
2bd63216 2674 /* target has died, so make the caller break out of the
6714d8e8
KH		 2675 * wait_event, but caller must recheck the domain_map */
2676 spin_lock(&dlm->spinlock);
2677 if (!test_bit(mig_target, dlm->domain_map))
2678 can_proceed = 1;
2679 spin_unlock(&dlm->spinlock);
2680 return can_proceed;
2681}
2682
faf0ec9f
AB		 2683static int dlm_lockres_is_dirty(struct dlm_ctxt *dlm,
2684 struct dlm_lock_resource *res)
6714d8e8
KH		 2685{
2686 int ret;
2687 spin_lock(&res->spinlock);
2688 ret = !!(res->state & DLM_LOCK_RES_DIRTY);
2689 spin_unlock(&res->spinlock);
2690 return ret;
2691}
2692
2693
2694static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
2695 struct dlm_lock_resource *res,
2696 u8 target)
2697{
2698 int ret = 0;
2699
2700 mlog(0, "dlm_mark_lockres_migrating: %.*s, from %u to %u\n",
2701 res->lockname.len, res->lockname.name, dlm->node_num,
2702 target);
2703 /* need to set MIGRATING flag on lockres. this is done by
2704 * ensuring that all asts have been flushed for this lockres. */
2705 spin_lock(&res->spinlock);
2706 BUG_ON(res->migration_pending);
2707 res->migration_pending = 1;
2708 /* strategy is to reserve an extra ast then release
2709 * it below, letting the release do all of the work */
2710 __dlm_lockres_reserve_ast(res);
2711 spin_unlock(&res->spinlock);
2712
ddc09c8d 2713 /* now flush all the pending asts */
6714d8e8 2714 dlm_kick_thread(dlm, res);
ddc09c8d
KH		 2715 /* before waiting on DIRTY, block processes which may
2716 * try to dirty the lockres before MIGRATING is set */
2717 spin_lock(&res->spinlock);
2718 BUG_ON(res->state & DLM_LOCK_RES_BLOCK_DIRTY);
2719 res->state |= DLM_LOCK_RES_BLOCK_DIRTY;
2720 spin_unlock(&res->spinlock);
2721 /* now wait on any pending asts and the DIRTY state */
6714d8e8
KH		 2722 wait_event(dlm->ast_wq, !dlm_lockres_is_dirty(dlm, res));
2723 dlm_lockres_release_ast(dlm, res);
2724
2725 mlog(0, "about to wait on migration_wq, dirty=%s\n",
2726 res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
2727 /* if the extra ref we just put was the final one, this
2728 * will pass thru immediately. otherwise, we need to wait
2729 * for the last ast to finish. */
2730again:
2731 ret = wait_event_interruptible_timeout(dlm->migration_wq,
2732 dlm_migration_can_proceed(dlm, res, target),
2733 msecs_to_jiffies(1000));
2734 if (ret < 0) {
2735 mlog(0, "woken again: migrating? %s, dead? %s\n",
2736 res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
2737 test_bit(target, dlm->domain_map) ? "no":"yes");
2738 } else {
2739 mlog(0, "all is well: migrating? %s, dead? %s\n",
2740 res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
2741 test_bit(target, dlm->domain_map) ? "no":"yes");
2742 }
2743 if (!dlm_migration_can_proceed(dlm, res, target)) {
2744 mlog(0, "trying again...\n");
2745 goto again;
2746 }
2747
a39953dd 2748 ret = 0;
6714d8e8
KH		 2749 /* did the target go down or die? */
2750 spin_lock(&dlm->spinlock);
2751 if (!test_bit(target, dlm->domain_map)) {
2752 mlog(ML_ERROR, "aha. migration target %u just went down\n",
2753 target);
2754 ret = -EHOSTDOWN;
2755 }
2756 spin_unlock(&dlm->spinlock);
2757
a39953dd
WW		 2758 /*
2759 * if target is down, we need to clear DLM_LOCK_RES_BLOCK_DIRTY for
2760 * another try; otherwise, we are sure the MIGRATING state is there,
2761 * drop the unneded state which blocked threads trying to DIRTY
2762 */
2763 spin_lock(&res->spinlock);
2764 BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY));
2765 res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY;
2766 if (!ret)
2767 BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING));
2768 spin_unlock(&res->spinlock);
2769
6714d8e8
KH		 2770 /*
2771 * at this point:
2772 *
a39953dd 2773 * o the DLM_LOCK_RES_MIGRATING flag is set if target not down
6714d8e8
KH		 2774 * o there are no pending asts on this lockres
2775 * o all processes trying to reserve an ast on this
2776 * lockres must wait for the MIGRATING flag to clear
2777 */
2778 return ret;
2779}
2780
2781/* last step in the migration process.
2782 * original master calls this to free all of the dlm_lock
2783 * structures that used to be for other nodes. */
2784static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
2785 struct dlm_lock_resource *res)
2786{
6714d8e8 2787 struct list_head *queue = &res->granted;
ba2bf218 2788 int i, bit;
800deef3 2789 struct dlm_lock *lock, *next;
6714d8e8
KH		 2790
2791 assert_spin_locked(&res->spinlock);
2792
2793 BUG_ON(res->owner == dlm->node_num);
2794
2795 for (i=0; i<3; i++) {
800deef3 2796 list_for_each_entry_safe(lock, next, queue, list) {
6714d8e8
KH		 2797 if (lock->ml.node != dlm->node_num) {
2798 mlog(0, "putting lock for node %u\n",
2799 lock->ml.node);
2800 /* be extra careful */
2801 BUG_ON(!list_empty(&lock->ast_list));
2802 BUG_ON(!list_empty(&lock->bast_list));
2803 BUG_ON(lock->ast_pending);
2804 BUG_ON(lock->bast_pending);
8d400b81
SM		 2805 dlm_lockres_clear_refmap_bit(dlm, res,
2806 lock->ml.node);
6714d8e8
KH		 2807 list_del_init(&lock->list);
2808 dlm_lock_put(lock);
2c5c54ac
SM		 2809 /* In a normal unlock, we would have added a
2810 * DLM_UNLOCK_FREE_LOCK action. Force it. */
2811 dlm_lock_put(lock);
6714d8e8
KH		 2812 }
2813 }
2814 queue++;
2815 }
ba2bf218
KH		 2816 bit = 0;
2817 while (1) {
2818 bit = find_next_bit(res->refmap, O2NM_MAX_NODES, bit);
2819 if (bit >= O2NM_MAX_NODES)
2820 break;
2821 /* do not clear the local node reference, if there is a
2822 * process holding this, let it drop the ref itself */
2823 if (bit != dlm->node_num) {
2824 mlog(0, "%s:%.*s: node %u had a ref to this "
2825 "migrating lockres, clearing\n", dlm->name,
2826 res->lockname.len, res->lockname.name, bit);
8d400b81 2827 dlm_lockres_clear_refmap_bit(dlm, res, bit);
ba2bf218
KH		 2828 }
2829 bit++;
2830 }
6714d8e8
KH		 2831}
2832
66effd3c
SM		 2833/*
2834 * Pick a node to migrate the lock resource to. This function selects a
2835 * potential target based first on the locks and then on refmap. It skips
2836 * nodes that are in the process of exiting the domain.
2837 */
6714d8e8
KH		 2838static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
2839 struct dlm_lock_resource *res)
2840{
66effd3c 2841 enum dlm_lockres_list idx;
6714d8e8 2842 struct list_head *queue = &res->granted;
6714d8e8 2843 struct dlm_lock *lock;
66effd3c
SM		 2844 int noderef;
2845 u8 nodenum = O2NM_MAX_NODES;
6714d8e8
KH		 2846
2847 assert_spin_locked(&dlm->spinlock);
66effd3c 2848 assert_spin_locked(&res->spinlock);
6714d8e8 2849
66effd3c
SM		 2850 /* Go through all the locks */
2851 for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) {
2852 queue = dlm_list_idx_to_ptr(res, idx);
800deef3 2853 list_for_each_entry(lock, queue, list) {
66effd3c
SM		 2854 if (lock->ml.node == dlm->node_num)
2855 continue;
2856 if (test_bit(lock->ml.node, dlm->exit_domain_map))
2857 continue;
2858 nodenum = lock->ml.node;
2859 goto bail;
6714d8e8 2860 }
6714d8e8 2861 }
388c4bcb 2862
66effd3c
SM		 2863 /* Go thru the refmap */
2864 noderef = -1;
6714d8e8 2865 while (1) {
66effd3c
SM		 2866 noderef = find_next_bit(res->refmap, O2NM_MAX_NODES,
2867 noderef + 1);
2868 if (noderef >= O2NM_MAX_NODES)
6714d8e8 2869 break;
66effd3c
SM		 2870 if (noderef == dlm->node_num)
2871 continue;
2872 if (test_bit(noderef, dlm->exit_domain_map))
2873 continue;
2874 nodenum = noderef;
2875 goto bail;
6714d8e8
KH		 2876 }
2877
66effd3c
SM		 2878bail:
2879 return nodenum;
6714d8e8
KH		 2880}
2881
6714d8e8
KH		 2882/* this is called by the new master once all lockres
2883 * data has been received */
2884static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
2885 struct dlm_lock_resource *res,
2886 u8 master, u8 new_master,
2887 struct dlm_node_iter *iter)
2888{
2889 struct dlm_migrate_request migrate;
2b832564 2890 int ret, skip, status = 0;
6714d8e8
KH		 2891 int nodenum;
2892
2893 memset(&migrate, 0, sizeof(migrate));
2894 migrate.namelen = res->lockname.len;
2895 memcpy(migrate.name, res->lockname.name, migrate.namelen);
2896 migrate.new_master = new_master;
2897 migrate.master = master;
2898
2899 ret = 0;
2900
2901 /* send message to all nodes, except the master and myself */
2902 while ((nodenum = dlm_node_iter_next(iter)) >= 0) {
2903 if (nodenum == master ||
2904 nodenum == new_master)
2905 continue;
2906
2b832564
SM		 2907 /* We could race exit domain. If exited, skip. */
2908 spin_lock(&dlm->spinlock);
2909 skip = (!test_bit(nodenum, dlm->domain_map));
2910 spin_unlock(&dlm->spinlock);
2911 if (skip) {
2912 clear_bit(nodenum, iter->node_map);
2913 continue;
2914 }
2915
6714d8e8
KH		 2916 ret = o2net_send_message(DLM_MIGRATE_REQUEST_MSG, dlm->key,
2917 &migrate, sizeof(migrate), nodenum,
2918 &status);
2b832564 2919 if (ret < 0) {
8decab3c
SM		 2920 mlog(ML_ERROR, "%s: res %.*s, Error %d send "
2921 "MIGRATE_REQUEST to node %u\n", dlm->name,
2922 migrate.namelen, migrate.name, ret, nodenum);
2b832564
SM		 2923 if (!dlm_is_host_down(ret)) {
2924 mlog(ML_ERROR, "unhandled error=%d!\n", ret);
2925 BUG();
2926 }
2927 clear_bit(nodenum, iter->node_map);
2928 ret = 0;
2929 } else if (status < 0) {
6714d8e8
KH		 2930 mlog(0, "migrate request (node %u) returned %d!\n",
2931 nodenum, status);
2932 ret = status;
ba2bf218
KH		 2933 } else if (status == DLM_MIGRATE_RESPONSE_MASTERY_REF) {
2934 /* during the migration request we short-circuited
2935 * the mastery of the lockres. make sure we have
2936 * a mastery ref for nodenum */
2937 mlog(0, "%s:%.*s: need ref for node %u\n",
2938 dlm->name, res->lockname.len, res->lockname.name,
2939 nodenum);
2940 spin_lock(&res->spinlock);
8d400b81 2941 dlm_lockres_set_refmap_bit(dlm, res, nodenum);
ba2bf218 2942 spin_unlock(&res->spinlock);
6714d8e8
KH		 2943 }
2944 }
2945
2946 if (ret < 0)
2947 mlog_errno(ret);
2948
2949 mlog(0, "returning ret=%d\n", ret);
2950 return ret;
2951}
2952
2953
2954/* if there is an existing mle for this lockres, we now know who the master is.
2955 * (the one who sent us *this* message) we can clear it up right away.
2956 * since the process that put the mle on the list still has a reference to it,
2957 * we can unhash it now, set the master and wake the process. as a result,
2958 * we will have no mle in the list to start with. now we can add an mle for
2959 * the migration and this should be the only one found for those scanning the
2960 * list. */
d74c9803
KH		 2961int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data,
2962 void **ret_data)
6714d8e8
KH		 2963{
2964 struct dlm_ctxt *dlm = data;
2965 struct dlm_lock_resource *res = NULL;
2966 struct dlm_migrate_request *migrate = (struct dlm_migrate_request *) msg->buf;
2967 struct dlm_master_list_entry *mle = NULL, *oldmle = NULL;
2968 const char *name;
a3d33291 2969 unsigned int namelen, hash;
6714d8e8
KH		 2970 int ret = 0;
2971
2972 if (!dlm_grab(dlm))
2973 return -EINVAL;
2974
2975 name = migrate->name;
2976 namelen = migrate->namelen;
a3d33291 2977 hash = dlm_lockid_hash(name, namelen);
6714d8e8
KH		 2978
2979 /* preallocate.. if this fails, abort */
3914ed0c 2980 mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
6714d8e8
KH		 2981
2982 if (!mle) {
2983 ret = -ENOMEM;
2984 goto leave;
2985 }
2986
2987 /* check for pre-existing lock */
2988 spin_lock(&dlm->spinlock);
a3d33291 2989 res = __dlm_lookup_lockres(dlm, name, namelen, hash);
6714d8e8
KH		 2990 if (res) {
2991 spin_lock(&res->spinlock);
2992 if (res->state & DLM_LOCK_RES_RECOVERING) {
2993 /* if all is working ok, this can only mean that we got
2994 * a migrate request from a node that we now see as
2995 * dead. what can we do here? drop it to the floor? */
2996 spin_unlock(&res->spinlock);
2997 mlog(ML_ERROR, "Got a migrate request, but the "
2998 "lockres is marked as recovering!");
2999 kmem_cache_free(dlm_mle_cache, mle);
3000 ret = -EINVAL; /* need a better solution */
3001 goto unlock;
3002 }
3003 res->state |= DLM_LOCK_RES_MIGRATING;
3004 spin_unlock(&res->spinlock);
3005 }
3006
6d98c3cc 3007 spin_lock(&dlm->master_lock);
6714d8e8
KH		 3008 /* ignore status. only nonzero status would BUG. */
3009 ret = dlm_add_migration_mle(dlm, res, mle, &oldmle,
3010 name, namelen,
3011 migrate->new_master,
3012 migrate->master);
3013
6714d8e8 3014 spin_unlock(&dlm->master_lock);
6d98c3cc 3015unlock:
6714d8e8
KH		 3016 spin_unlock(&dlm->spinlock);
3017
3018 if (oldmle) {
3019 /* master is known, detach if not already detached */
3020 dlm_mle_detach_hb_events(dlm, oldmle);
3021 dlm_put_mle(oldmle);
3022 }
3023
3024 if (res)
3025 dlm_lockres_put(res);
3026leave:
3027 dlm_put(dlm);
3028 return ret;
3029}
3030
3031/* must be holding dlm->spinlock and dlm->master_lock
3032 * when adding a migration mle, we can clear any other mles
3033 * in the master list because we know with certainty that
3034 * the master is "master". so we remove any old mle from
3035 * the list after setting it's master field, and then add
3036 * the new migration mle. this way we can hold with the rule
3037 * of having only one mle for a given lock name at all times. */
3038static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
3039 struct dlm_lock_resource *res,
3040 struct dlm_master_list_entry *mle,
3041 struct dlm_master_list_entry **oldmle,
3042 const char *name, unsigned int namelen,
3043 u8 new_master, u8 master)
3044{
3045 int found;
3046 int ret = 0;
3047
3048 *oldmle = NULL;
3049
6714d8e8
KH		 3050 assert_spin_locked(&dlm->spinlock);
3051 assert_spin_locked(&dlm->master_lock);
3052
3053 /* caller is responsible for any ref taken here on oldmle */
3054 found = dlm_find_mle(dlm, oldmle, (char *)name, namelen);
3055 if (found) {
3056 struct dlm_master_list_entry *tmp = *oldmle;
3057 spin_lock(&tmp->spinlock);
3058 if (tmp->type == DLM_MLE_MIGRATION) {
3059 if (master == dlm->node_num) {
3060 /* ah another process raced me to it */
3061 mlog(0, "tried to migrate %.*s, but some "
3062 "process beat me to it\n",
3063 namelen, name);
3064 ret = -EEXIST;
3065 } else {
3066 /* bad. 2 NODES are trying to migrate! */
3067 mlog(ML_ERROR, "migration error mle: "
3068 "master=%u new_master=%u // request: "
3069 "master=%u new_master=%u // "
3070 "lockres=%.*s\n",
3071 tmp->master, tmp->new_master,
3072 master, new_master,
3073 namelen, name);
3074 BUG();
3075 }
3076 } else {
3077 /* this is essentially what assert_master does */
3078 tmp->master = master;
3079 atomic_set(&tmp->woken, 1);
3080 wake_up(&tmp->wq);
1c084577
SM		 3081 /* remove it so that only one mle will be found */
3082 __dlm_unlink_mle(dlm, tmp);
ba2bf218
KH		 3083 __dlm_mle_detach_hb_events(dlm, tmp);
3084 ret = DLM_MIGRATE_RESPONSE_MASTERY_REF;
3085 mlog(0, "%s:%.*s: master=%u, newmaster=%u, "
3086 "telling master to get ref for cleared out mle "
3087 "during migration\n", dlm->name, namelen, name,
3088 master, new_master);
6714d8e8
KH		 3089 }
3090 spin_unlock(&tmp->spinlock);
3091 }
3092
3093 /* now add a migration mle to the tail of the list */
3094 dlm_init_mle(mle, DLM_MLE_MIGRATION, dlm, res, name, namelen);
3095 mle->new_master = new_master;
ba2bf218
KH		 3096 /* the new master will be sending an assert master for this.
3097 * at that point we will get the refmap reference */
6714d8e8
KH		 3098 mle->master = master;
3099 /* do this for consistency with other mle types */
3100 set_bit(new_master, mle->maybe_map);
1c084577 3101 __dlm_insert_mle(dlm, mle);
6714d8e8
KH		 3102
3103 return ret;
3104}
3105
c2cd4a44
SM		 3106/*
3107 * Sets the owner of the lockres, associated to the mle, to UNKNOWN
3108 */
3109static struct dlm_lock_resource *dlm_reset_mleres_owner(struct dlm_ctxt *dlm,
3110 struct dlm_master_list_entry *mle)
3111{
3112 struct dlm_lock_resource *res;
c2cd4a44
SM		 3113
3114 /* Find the lockres associated to the mle and set its owner to UNK */
7141514b
SM		 3115 res = __dlm_lookup_lockres(dlm, mle->mname, mle->mnamelen,
3116 mle->mnamehash);
c2cd4a44
SM		 3117 if (res) {
3118 spin_unlock(&dlm->master_lock);
3119
3120 /* move lockres onto recovery list */
3121 spin_lock(&res->spinlock);
3122 dlm_set_lockres_owner(dlm, res, DLM_LOCK_RES_OWNER_UNKNOWN);
3123 dlm_move_lockres_to_recovery_list(dlm, res);
3124 spin_unlock(&res->spinlock);
3125 dlm_lockres_put(res);
3126
3127 /* about to get rid of mle, detach from heartbeat */
3128 __dlm_mle_detach_hb_events(dlm, mle);
3129
3130 /* dump the mle */
3131 spin_lock(&dlm->master_lock);
3132 __dlm_put_mle(mle);
3133 spin_unlock(&dlm->master_lock);
3134 }
3135
3136 return res;
3137}
3138
3139static void dlm_clean_migration_mle(struct dlm_ctxt *dlm,
3140 struct dlm_master_list_entry *mle)
3141{
3142 __dlm_mle_detach_hb_events(dlm, mle);
3143
3144 spin_lock(&mle->spinlock);
3145 __dlm_unlink_mle(dlm, mle);
3146 atomic_set(&mle->woken, 1);
3147 spin_unlock(&mle->spinlock);
3148
3149 wake_up(&mle->wq);
3150}
3151
3152static void dlm_clean_block_mle(struct dlm_ctxt *dlm,
3153 struct dlm_master_list_entry *mle, u8 dead_node)
3154{
3155 int bit;
3156
3157 BUG_ON(mle->type != DLM_MLE_BLOCK);
3158
3159 spin_lock(&mle->spinlock);
3160 bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
3161 if (bit != dead_node) {
3162 mlog(0, "mle found, but dead node %u would not have been "
3163 "master\n", dead_node);
3164 spin_unlock(&mle->spinlock);
3165 } else {
3166 /* Must drop the refcount by one since the assert_master will
3167 * never arrive. This may result in the mle being unlinked and
3168 * freed, but there may still be a process waiting in the
3169 * dlmlock path which is fine. */
3170 mlog(0, "node %u was expected master\n", dead_node);
3171 atomic_set(&mle->woken, 1);
3172 spin_unlock(&mle->spinlock);
3173 wake_up(&mle->wq);
3174
3175 /* Do not need events any longer, so detach from heartbeat */
3176 __dlm_mle_detach_hb_events(dlm, mle);
3177 __dlm_put_mle(mle);
3178 }
3179}
6714d8e8
KH		 3180
3181void dlm_clean_master_list(struct dlm_ctxt *dlm, u8 dead_node)
3182{
2ed6c750 3183 struct dlm_master_list_entry *mle;
6714d8e8 3184 struct dlm_lock_resource *res;
2ed6c750
SM		 3185 struct hlist_head *bucket;
3186 struct hlist_node *list;
3187 unsigned int i;
6714d8e8 3188
ef6b689b 3189 mlog(0, "dlm=%s, dead node=%u\n", dlm->name, dead_node);
6714d8e8
KH		 3190top:
3191 assert_spin_locked(&dlm->spinlock);
3192
3193 /* clean the master list */
3194 spin_lock(&dlm->master_lock);
2ed6c750
SM		 3195 for (i = 0; i < DLM_HASH_BUCKETS; i++) {
3196 bucket = dlm_master_hash(dlm, i);
3197 hlist_for_each(list, bucket) {
3198 mle = hlist_entry(list, struct dlm_master_list_entry,
3199 master_hash_node);
3200
67ae1f06
SM		 3201 BUG_ON(mle->type != DLM_MLE_BLOCK &&
3202 mle->type != DLM_MLE_MASTER &&
3203 mle->type != DLM_MLE_MIGRATION);
3204
3205 /* MASTER mles are initiated locally. The waiting
3206 * process will notice the node map change shortly.
3207 * Let that happen as normal. */
3208 if (mle->type == DLM_MLE_MASTER)
3209 continue;
3210
3211 /* BLOCK mles are initiated by other nodes. Need to
3212 * clean up if the dead node would have been the
3213 * master. */
3214 if (mle->type == DLM_MLE_BLOCK) {
3215 dlm_clean_block_mle(dlm, mle, dead_node);
3216 continue;
3217 }
6714d8e8 3218
67ae1f06
SM		 3219 /* Everything else is a MIGRATION mle */
3220
3221 /* The rule for MIGRATION mles is that the master
3222 * becomes UNKNOWN if *either* the original or the new
3223 * master dies. All UNKNOWN lockres' are sent to
3224 * whichever node becomes the recovery master. The new
3225 * master is responsible for determining if there is
3226 * still a master for this lockres, or if he needs to
3227 * take over mastery. Either way, this node should
3228 * expect another message to resolve this. */
3229
3230 if (mle->master != dead_node &&
3231 mle->new_master != dead_node)
3232 continue;
3233
3234 /* If we have reached this point, this mle needs to be
3235 * removed from the list and freed. */
3236 dlm_clean_migration_mle(dlm, mle);
3237
3238 mlog(0, "%s: node %u died during migration from "
3239 "%u to %u!\n", dlm->name, dead_node, mle->master,
3240 mle->new_master);
3241
3242 /* If we find a lockres associated with the mle, we've
3243 * hit this rare case that messes up our lock ordering.
3244 * If so, we need to drop the master lock so that we can
3245 * take the lockres lock, meaning that we will have to
3246 * restart from the head of list. */
3247 res = dlm_reset_mleres_owner(dlm, mle);
3248 if (res)
3249 /* restart */
3250 goto top;
3251
3252 /* This may be the last reference */
3253 __dlm_put_mle(mle);
6714d8e8 3254 }
2ed6c750 3255 }
6714d8e8
KH		 3256 spin_unlock(&dlm->master_lock);
3257}
3258
6714d8e8
KH		 3259int dlm_finish_migration(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
3260 u8 old_master)
3261{
3262 struct dlm_node_iter iter;
3263 int ret = 0;
3264
3265 spin_lock(&dlm->spinlock);
3266 dlm_node_iter_init(dlm->domain_map, &iter);
3267 clear_bit(old_master, iter.node_map);
3268 clear_bit(dlm->node_num, iter.node_map);
3269 spin_unlock(&dlm->spinlock);
3270
ba2bf218
KH		 3271 /* ownership of the lockres is changing. account for the
3272 * mastery reference here since old_master will briefly have
3273 * a reference after the migration completes */
3274 spin_lock(&res->spinlock);
8d400b81 3275 dlm_lockres_set_refmap_bit(dlm, res, old_master);
ba2bf218
KH		 3276 spin_unlock(&res->spinlock);
3277
6714d8e8
KH		 3278 mlog(0, "now time to do a migrate request to other nodes\n");
3279 ret = dlm_do_migrate_request(dlm, res, old_master,
3280 dlm->node_num, &iter);
3281 if (ret < 0) {
3282 mlog_errno(ret);
3283 goto leave;
3284 }
3285
3286 mlog(0, "doing assert master of %.*s to all except the original node\n",
3287 res->lockname.len, res->lockname.name);
3288 /* this call now finishes out the nodemap
3289 * even if one or more nodes die */
ba2bf218 3290 ret = dlm_do_assert_master(dlm, res, iter.node_map,
6714d8e8
KH		 3291 DLM_ASSERT_MASTER_FINISH_MIGRATION);
3292 if (ret < 0) {
3293 /* no longer need to retry. all living nodes contacted. */
3294 mlog_errno(ret);
3295 ret = 0;
3296 }
3297
3298 memset(iter.node_map, 0, sizeof(iter.node_map));
3299 set_bit(old_master, iter.node_map);
3300 mlog(0, "doing assert master of %.*s back to %u\n",
3301 res->lockname.len, res->lockname.name, old_master);
ba2bf218 3302 ret = dlm_do_assert_master(dlm, res, iter.node_map,
6714d8e8
KH		 3303 DLM_ASSERT_MASTER_FINISH_MIGRATION);
3304 if (ret < 0) {
3305 mlog(0, "assert master to original master failed "
3306 "with %d.\n", ret);
3307 /* the only nonzero status here would be because of
3308 * a dead original node. we're done. */
3309 ret = 0;
3310 }
3311
3312 /* all done, set the owner, clear the flag */
3313 spin_lock(&res->spinlock);
3314 dlm_set_lockres_owner(dlm, res, dlm->node_num);
3315 res->state &= ~DLM_LOCK_RES_MIGRATING;
3316 spin_unlock(&res->spinlock);
3317 /* re-dirty it on the new master */
3318 dlm_kick_thread(dlm, res);
3319 wake_up(&res->wq);
3320leave:
3321 return ret;
3322}
3323
3324/*
3325 * LOCKRES AST REFCOUNT
3326 * this is integral to migration
3327 */
3328
3329/* for future intent to call an ast, reserve one ahead of time.
3330 * this should be called only after waiting on the lockres
3331 * with dlm_wait_on_lockres, and while still holding the
3332 * spinlock after the call. */
3333void __dlm_lockres_reserve_ast(struct dlm_lock_resource *res)
3334{
3335 assert_spin_locked(&res->spinlock);
3336 if (res->state & DLM_LOCK_RES_MIGRATING) {
3337 __dlm_print_one_lock_resource(res);
3338 }
3339 BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
3340
3341 atomic_inc(&res->asts_reserved);
3342}
3343
3344/*
3345 * used to drop the reserved ast, either because it went unused,
3346 * or because the ast/bast was actually called.
3347 *
3348 * also, if there is a pending migration on this lockres,
3349 * and this was the last pending ast on the lockres,
3350 * atomically set the MIGRATING flag before we drop the lock.
3351 * this is how we ensure that migration can proceed with no
3352 * asts in progress. note that it is ok if the state of the
3353 * queues is such that a lock should be granted in the future
3354 * or that a bast should be fired, because the new master will
3355 * shuffle the lists on this lockres as soon as it is migrated.
3356 */
3357void dlm_lockres_release_ast(struct dlm_ctxt *dlm,
3358 struct dlm_lock_resource *res)
3359{
3360 if (!atomic_dec_and_lock(&res->asts_reserved, &res->spinlock))
3361 return;
3362
3363 if (!res->migration_pending) {
3364 spin_unlock(&res->spinlock);
3365 return;
3366 }
3367
3368 BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
3369 res->migration_pending = 0;
3370 res->state |= DLM_LOCK_RES_MIGRATING;
3371 spin_unlock(&res->spinlock);
3372 wake_up(&res->wq);
3373 wake_up(&dlm->migration_wq);
3374}
5dad6c39
SE		 3375
3376void dlm_force_free_mles(struct dlm_ctxt *dlm)
3377{
3378 int i;
3379 struct hlist_head *bucket;
3380 struct dlm_master_list_entry *mle;
3381 struct hlist_node *tmp, *list;
3382
3383 /*
3384 * We notified all other nodes that we are exiting the domain and
3385 * marked the dlm state to DLM_CTXT_LEAVING. If any mles are still
3386 * around we force free them and wake any processes that are waiting
3387 * on the mles
3388 */
3389 spin_lock(&dlm->spinlock);
3390 spin_lock(&dlm->master_lock);
3391
3392 BUG_ON(dlm->dlm_state != DLM_CTXT_LEAVING);
3393 BUG_ON((find_next_bit(dlm->domain_map, O2NM_MAX_NODES, 0) < O2NM_MAX_NODES));
3394
3395 for (i = 0; i < DLM_HASH_BUCKETS; i++) {
3396 bucket = dlm_master_hash(dlm, i);
3397 hlist_for_each_safe(list, tmp, bucket) {
3398 mle = hlist_entry(list, struct dlm_master_list_entry,
3399 master_hash_node);
3400 if (mle->type != DLM_MLE_BLOCK) {
3401 mlog(ML_ERROR, "bad mle: %p\n", mle);
3402 dlm_print_one_mle(mle);
3403 }
3404 atomic_set(&mle->woken, 1);
3405 wake_up(&mle->wq);
3406
3407 __dlm_unlink_mle(dlm, mle);
3408 __dlm_mle_detach_hb_events(dlm, mle);
3409 __dlm_put_mle(mle);
3410 }
3411 }
3412 spin_unlock(&dlm->master_lock);
3413 spin_unlock(&dlm->spinlock);
3414}
Linux kernel - Deliverables
This page took 0.756993 seconds and 5 git commands to generate.