projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: Add ->check_downconvert callback in dlmglue
[deliverable/linux.git] / fs / ocfs2 / dlmglue.c
CommitLineData
ccd979bd
MF		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * dlmglue.c
5 *
6 * Code which implements an OCFS2 specific interface to our DLM.
7 *
8 * Copyright (C) 2003, 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#include <linux/types.h>
27#include <linux/slab.h>
28#include <linux/highmem.h>
29#include <linux/mm.h>
30#include <linux/smp_lock.h>
31#include <linux/crc32.h>
32#include <linux/kthread.h>
33#include <linux/pagemap.h>
34#include <linux/debugfs.h>
35#include <linux/seq_file.h>
36
37#include <cluster/heartbeat.h>
38#include <cluster/nodemanager.h>
39#include <cluster/tcp.h>
40
41#include <dlm/dlmapi.h>
42
43#define MLOG_MASK_PREFIX ML_DLM_GLUE
44#include <cluster/masklog.h>
45
46#include "ocfs2.h"
47
48#include "alloc.h"
d680efe9 49#include "dcache.h"
ccd979bd
MF		 50#include "dlmglue.h"
51#include "extent_map.h"
52#include "heartbeat.h"
53#include "inode.h"
54#include "journal.h"
55#include "slot_map.h"
56#include "super.h"
57#include "uptodate.h"
58#include "vote.h"
59
60#include "buffer_head_io.h"
61
62struct ocfs2_mask_waiter {
63 struct list_head mw_item;
64 int mw_status;
65 struct completion mw_complete;
66 unsigned long mw_mask;
67 unsigned long mw_goal;
68};
69
54a7e755
MF		 70static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres);
71static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres);
ccd979bd 72
d680efe9
MF		 73/*
74 * Return value from ocfs2_convert_worker_t functions.
75 *
76 * These control the precise actions of ocfs2_generic_unblock_lock()
77 * and ocfs2_process_blocked_lock()
78 *
79 */
80enum ocfs2_unblock_action {
81 UNBLOCK_CONTINUE = 0, /* Continue downconvert */
82 UNBLOCK_CONTINUE_POST = 1, /* Continue downconvert, fire
83 * ->post_unlock callback */
84 UNBLOCK_STOP_POST = 2, /* Do not downconvert, fire
85 * ->post_unlock() callback. */
86};
87
88struct ocfs2_unblock_ctl {
89 int requeue;
90 enum ocfs2_unblock_action unblock_action;
91};
92
ccd979bd 93static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres,
d680efe9 94 struct ocfs2_unblock_ctl *ctl);
ccd979bd 95static int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
d680efe9 96 struct ocfs2_unblock_ctl *ctl);
ccd979bd 97static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
d680efe9
MF		 98 struct ocfs2_unblock_ctl *ctl);
99static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres,
100 struct ocfs2_unblock_ctl *ctl);
ccd979bd 101static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
d680efe9
MF		 102 struct ocfs2_unblock_ctl *ctl);
103
104static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
105 struct ocfs2_lock_res *lockres);
ccd979bd 106
f625c979
MF		 107/*
108 * OCFS2 Lock Resource Operations
109 *
110 * These fine tune the behavior of the generic dlmglue locking infrastructure.
111 */
ccd979bd 112struct ocfs2_lock_res_ops {
54a7e755
MF		 113 /*
114 * Translate an ocfs2_lock_res * into an ocfs2_super *. Define
115 * this callback if ->l_priv is not an ocfs2_super pointer
116 */
117 struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *);
d680efe9
MF		 118 int (*unblock)(struct ocfs2_lock_res *, struct ocfs2_unblock_ctl *);
119 void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *);
f625c979 120
16d5b956
MF		 121 /*
122 * Allow a lock type to add checks to determine whether it is
123 * safe to downconvert a lock. Return 0 to re-queue the
124 * downconvert at a later time, nonzero to continue.
125 *
126 * For most locks, the default checks that there are no
127 * incompatible holders are sufficient.
128 *
129 * Called with the lockres spinlock held.
130 */
131 int (*check_downconvert)(struct ocfs2_lock_res *, int);
132
f625c979
MF		 133 /*
134 * LOCK_TYPE_* flags which describe the specific requirements
135 * of a lock type. Descriptions of each individual flag follow.
136 */
137 int flags;
ccd979bd
MF		 138};
139
f625c979
MF		 140/*
141 * Some locks want to "refresh" potentially stale data when a
142 * meaningful (PRMODE or EXMODE) lock level is first obtained. If this
143 * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the
144 * individual lockres l_flags member from the ast function. It is
145 * expected that the locking wrapper will clear the
146 * OCFS2_LOCK_NEEDS_REFRESH flag when done.
147 */
148#define LOCK_TYPE_REQUIRES_REFRESH 0x1
149
b80fc012
MF		 150/*
151 * Indicate that a lock type makes use of the lock value block.
152 */
153#define LOCK_TYPE_USES_LVB 0x2
154
d680efe9
MF		 155typedef int (ocfs2_convert_worker_t)(struct ocfs2_lock_res *, int);
156static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
157 struct ocfs2_lock_res *lockres,
158 struct ocfs2_unblock_ctl *ctl,
159 ocfs2_convert_worker_t *worker);
160
ccd979bd 161static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = {
54a7e755 162 .get_osb = ocfs2_get_inode_osb,
ccd979bd 163 .unblock = ocfs2_unblock_inode_lock,
f625c979 164 .flags = 0,
ccd979bd
MF		 165};
166
167static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = {
54a7e755 168 .get_osb = ocfs2_get_inode_osb,
ccd979bd 169 .unblock = ocfs2_unblock_meta,
b80fc012 170 .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB,
ccd979bd
MF		 171};
172
ccd979bd 173static struct ocfs2_lock_res_ops ocfs2_inode_data_lops = {
54a7e755 174 .get_osb = ocfs2_get_inode_osb,
ccd979bd 175 .unblock = ocfs2_unblock_data,
f625c979 176 .flags = 0,
ccd979bd
MF		 177};
178
179static struct ocfs2_lock_res_ops ocfs2_super_lops = {
ccd979bd 180 .unblock = ocfs2_unblock_osb_lock,
f625c979 181 .flags = LOCK_TYPE_REQUIRES_REFRESH,
ccd979bd
MF		 182};
183
184static struct ocfs2_lock_res_ops ocfs2_rename_lops = {
ccd979bd 185 .unblock = ocfs2_unblock_osb_lock,
f625c979 186 .flags = 0,
ccd979bd
MF		 187};
188
d680efe9 189static struct ocfs2_lock_res_ops ocfs2_dentry_lops = {
54a7e755 190 .get_osb = ocfs2_get_dentry_osb,
d680efe9
MF		 191 .unblock = ocfs2_unblock_dentry_lock,
192 .post_unlock = ocfs2_dentry_post_unlock,
f625c979 193 .flags = 0,
d680efe9
MF		 194};
195
ccd979bd
MF		 196static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
197{
198 return lockres->l_type == OCFS2_LOCK_TYPE_META ||
199 lockres->l_type == OCFS2_LOCK_TYPE_DATA ||
200 lockres->l_type == OCFS2_LOCK_TYPE_RW;
201}
202
ccd979bd
MF		 203static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres)
204{
205 BUG_ON(!ocfs2_is_inode_lock(lockres));
206
207 return (struct inode *) lockres->l_priv;
208}
209
d680efe9
MF		 210static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres)
211{
212 BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY);
213
214 return (struct ocfs2_dentry_lock *)lockres->l_priv;
215}
216
54a7e755
MF		 217static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres)
218{
219 if (lockres->l_ops->get_osb)
220 return lockres->l_ops->get_osb(lockres);
221
222 return (struct ocfs2_super *)lockres->l_priv;
223}
224
ccd979bd
MF		 225static int ocfs2_lock_create(struct ocfs2_super *osb,
226 struct ocfs2_lock_res *lockres,
227 int level,
228 int dlm_flags);
229static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
230 int wanted);
231static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
232 struct ocfs2_lock_res *lockres,
233 int level);
234static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres);
235static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres);
236static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres);
237static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level);
238static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
239 struct ocfs2_lock_res *lockres);
240static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
241 int convert);
242#define ocfs2_log_dlm_error(_func, _stat, _lockres) do { \
243 mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on " \
244 "resource %s: %s\n", dlm_errname(_stat), _func, \
245 _lockres->l_name, dlm_errmsg(_stat)); \
246} while (0)
247static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
248 struct ocfs2_lock_res *lockres);
249static int ocfs2_meta_lock_update(struct inode *inode,
250 struct buffer_head **bh);
251static void ocfs2_drop_osb_locks(struct ocfs2_super *osb);
252static inline int ocfs2_highest_compat_lock_level(int level);
253static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode,
254 struct ocfs2_lock_res *lockres,
255 int new_level);
256
ccd979bd
MF		 257static void ocfs2_build_lock_name(enum ocfs2_lock_type type,
258 u64 blkno,
259 u32 generation,
260 char *name)
261{
262 int len;
263
264 mlog_entry_void();
265
266 BUG_ON(type >= OCFS2_NUM_LOCK_TYPES);
267
b0697053
MF		 268 len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x",
269 ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD,
270 (long long)blkno, generation);
ccd979bd
MF		 271
272 BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1));
273
274 mlog(0, "built lock resource with name: %s\n", name);
275
276 mlog_exit_void();
277}
278
34af946a 279static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock);
ccd979bd
MF		 280
281static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res,
282 struct ocfs2_dlm_debug *dlm_debug)
283{
284 mlog(0, "Add tracking for lockres %s\n", res->l_name);
285
286 spin_lock(&ocfs2_dlm_tracking_lock);
287 list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking);
288 spin_unlock(&ocfs2_dlm_tracking_lock);
289}
290
291static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res)
292{
293 spin_lock(&ocfs2_dlm_tracking_lock);
294 if (!list_empty(&res->l_debug_list))
295 list_del_init(&res->l_debug_list);
296 spin_unlock(&ocfs2_dlm_tracking_lock);
297}
298
299static void ocfs2_lock_res_init_common(struct ocfs2_super *osb,
300 struct ocfs2_lock_res *res,
301 enum ocfs2_lock_type type,
ccd979bd
MF		 302 struct ocfs2_lock_res_ops *ops,
303 void *priv)
304{
ccd979bd
MF		 305 res->l_type = type;
306 res->l_ops = ops;
307 res->l_priv = priv;
308
309 res->l_level = LKM_IVMODE;
310 res->l_requested = LKM_IVMODE;
311 res->l_blocking = LKM_IVMODE;
312 res->l_action = OCFS2_AST_INVALID;
313 res->l_unlock_action = OCFS2_UNLOCK_INVALID;
314
315 res->l_flags = OCFS2_LOCK_INITIALIZED;
316
317 ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug);
318}
319
320void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res)
321{
322 /* This also clears out the lock status block */
323 memset(res, 0, sizeof(struct ocfs2_lock_res));
324 spin_lock_init(&res->l_lock);
325 init_waitqueue_head(&res->l_event);
326 INIT_LIST_HEAD(&res->l_blocked_list);
327 INIT_LIST_HEAD(&res->l_mask_waiters);
328}
329
330void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
331 enum ocfs2_lock_type type,
24c19ef4 332 unsigned int generation,
ccd979bd
MF		 333 struct inode *inode)
334{
335 struct ocfs2_lock_res_ops *ops;
336
337 switch(type) {
338 case OCFS2_LOCK_TYPE_RW:
339 ops = &ocfs2_inode_rw_lops;
340 break;
341 case OCFS2_LOCK_TYPE_META:
342 ops = &ocfs2_inode_meta_lops;
343 break;
344 case OCFS2_LOCK_TYPE_DATA:
345 ops = &ocfs2_inode_data_lops;
346 break;
347 default:
348 mlog_bug_on_msg(1, "type: %d\n", type);
349 ops = NULL; /* thanks, gcc */
350 break;
351 };
352
d680efe9 353 ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno,
24c19ef4 354 generation, res->l_name);
d680efe9
MF		 355 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode);
356}
357
54a7e755
MF		 358static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres)
359{
360 struct inode *inode = ocfs2_lock_res_inode(lockres);
361
362 return OCFS2_SB(inode->i_sb);
363}
364
d680efe9
MF		 365static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres)
366{
367 __be64 inode_blkno_be;
368
369 memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START],
370 sizeof(__be64));
371
372 return be64_to_cpu(inode_blkno_be);
373}
374
54a7e755
MF		 375static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres)
376{
377 struct ocfs2_dentry_lock *dl = lockres->l_priv;
378
379 return OCFS2_SB(dl->dl_inode->i_sb);
380}
381
d680efe9
MF		 382void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
383 u64 parent, struct inode *inode)
384{
385 int len;
386 u64 inode_blkno = OCFS2_I(inode)->ip_blkno;
387 __be64 inode_blkno_be = cpu_to_be64(inode_blkno);
388 struct ocfs2_lock_res *lockres = &dl->dl_lockres;
389
390 ocfs2_lock_res_init_once(lockres);
391
392 /*
393 * Unfortunately, the standard lock naming scheme won't work
394 * here because we have two 16 byte values to use. Instead,
395 * we'll stuff the inode number as a binary value. We still
396 * want error prints to show something without garbling the
397 * display, so drop a null byte in there before the inode
398 * number. A future version of OCFS2 will likely use all
399 * binary lock names. The stringified names have been a
400 * tremendous aid in debugging, but now that the debugfs
401 * interface exists, we can mangle things there if need be.
402 *
403 * NOTE: We also drop the standard "pad" value (the total lock
404 * name size stays the same though - the last part is all
405 * zeros due to the memset in ocfs2_lock_res_init_once()
406 */
407 len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START,
408 "%c%016llx",
409 ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY),
410 (long long)parent);
411
412 BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1));
413
414 memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be,
415 sizeof(__be64));
416
417 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
418 OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops,
419 dl);
ccd979bd
MF		 420}
421
422static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res,
423 struct ocfs2_super *osb)
424{
425 /* Superblock lockres doesn't come from a slab so we call init
426 * once on it manually. */
427 ocfs2_lock_res_init_once(res);
d680efe9
MF		 428 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO,
429 0, res->l_name);
ccd979bd 430 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER,
ccd979bd
MF		 431 &ocfs2_super_lops, osb);
432}
433
434static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
435 struct ocfs2_super *osb)
436{
437 /* Rename lockres doesn't come from a slab so we call init
438 * once on it manually. */
439 ocfs2_lock_res_init_once(res);
d680efe9
MF		 440 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name);
441 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME,
ccd979bd
MF		 442 &ocfs2_rename_lops, osb);
443}
444
445void ocfs2_lock_res_free(struct ocfs2_lock_res *res)
446{
447 mlog_entry_void();
448
449 if (!(res->l_flags & OCFS2_LOCK_INITIALIZED))
450 return;
451
452 ocfs2_remove_lockres_tracking(res);
453
454 mlog_bug_on_msg(!list_empty(&res->l_blocked_list),
455 "Lockres %s is on the blocked list\n",
456 res->l_name);
457 mlog_bug_on_msg(!list_empty(&res->l_mask_waiters),
458 "Lockres %s has mask waiters pending\n",
459 res->l_name);
460 mlog_bug_on_msg(spin_is_locked(&res->l_lock),
461 "Lockres %s is locked\n",
462 res->l_name);
463 mlog_bug_on_msg(res->l_ro_holders,
464 "Lockres %s has %u ro holders\n",
465 res->l_name, res->l_ro_holders);
466 mlog_bug_on_msg(res->l_ex_holders,
467 "Lockres %s has %u ex holders\n",
468 res->l_name, res->l_ex_holders);
469
470 /* Need to clear out the lock status block for the dlm */
471 memset(&res->l_lksb, 0, sizeof(res->l_lksb));
472
473 res->l_flags = 0UL;
474 mlog_exit_void();
475}
476
477static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres,
478 int level)
479{
480 mlog_entry_void();
481
482 BUG_ON(!lockres);
483
484 switch(level) {
485 case LKM_EXMODE:
486 lockres->l_ex_holders++;
487 break;
488 case LKM_PRMODE:
489 lockres->l_ro_holders++;
490 break;
491 default:
492 BUG();
493 }
494
495 mlog_exit_void();
496}
497
498static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres,
499 int level)
500{
501 mlog_entry_void();
502
503 BUG_ON(!lockres);
504
505 switch(level) {
506 case LKM_EXMODE:
507 BUG_ON(!lockres->l_ex_holders);
508 lockres->l_ex_holders--;
509 break;
510 case LKM_PRMODE:
511 BUG_ON(!lockres->l_ro_holders);
512 lockres->l_ro_holders--;
513 break;
514 default:
515 BUG();
516 }
517 mlog_exit_void();
518}
519
520/* WARNING: This function lives in a world where the only three lock
521 * levels are EX, PR, and NL. It *will* have to be adjusted when more
522 * lock types are added. */
523static inline int ocfs2_highest_compat_lock_level(int level)
524{
525 int new_level = LKM_EXMODE;
526
527 if (level == LKM_EXMODE)
528 new_level = LKM_NLMODE;
529 else if (level == LKM_PRMODE)
530 new_level = LKM_PRMODE;
531 return new_level;
532}
533
534static void lockres_set_flags(struct ocfs2_lock_res *lockres,
535 unsigned long newflags)
536{
537 struct list_head *pos, *tmp;
538 struct ocfs2_mask_waiter *mw;
539
540 assert_spin_locked(&lockres->l_lock);
541
542 lockres->l_flags = newflags;
543
544 list_for_each_safe(pos, tmp, &lockres->l_mask_waiters) {
545 mw = list_entry(pos, struct ocfs2_mask_waiter, mw_item);
546 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
547 continue;
548
549 list_del_init(&mw->mw_item);
550 mw->mw_status = 0;
551 complete(&mw->mw_complete);
552 }
553}
554static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or)
555{
556 lockres_set_flags(lockres, lockres->l_flags | or);
557}
558static void lockres_clear_flags(struct ocfs2_lock_res *lockres,
559 unsigned long clear)
560{
561 lockres_set_flags(lockres, lockres->l_flags & ~clear);
562}
563
564static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres)
565{
566 mlog_entry_void();
567
568 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
569 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
570 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
571 BUG_ON(lockres->l_blocking <= LKM_NLMODE);
572
573 lockres->l_level = lockres->l_requested;
574 if (lockres->l_level <=
575 ocfs2_highest_compat_lock_level(lockres->l_blocking)) {
576 lockres->l_blocking = LKM_NLMODE;
577 lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
578 }
579 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
580
581 mlog_exit_void();
582}
583
584static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres)
585{
586 mlog_entry_void();
587
588 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
589 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
590
591 /* Convert from RO to EX doesn't really need anything as our
592 * information is already up to data. Convert from NL to
593 * *anything* however should mark ourselves as needing an
594 * update */
f625c979
MF		 595 if (lockres->l_level == LKM_NLMODE &&
596 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
ccd979bd
MF		 597 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
598
599 lockres->l_level = lockres->l_requested;
600 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
601
602 mlog_exit_void();
603}
604
605static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres)
606{
607 mlog_entry_void();
608
609 BUG_ON((!lockres->l_flags & OCFS2_LOCK_BUSY));
610 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
611
612 if (lockres->l_requested > LKM_NLMODE &&
f625c979
MF		 613 !(lockres->l_flags & OCFS2_LOCK_LOCAL) &&
614 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
ccd979bd
MF		 615 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
616
617 lockres->l_level = lockres->l_requested;
618 lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED);
619 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
620
621 mlog_exit_void();
622}
623
ccd979bd
MF		 624static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
625 int level)
626{
627 int needs_downconvert = 0;
628 mlog_entry_void();
629
630 assert_spin_locked(&lockres->l_lock);
631
632 lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
633
634 if (level > lockres->l_blocking) {
635 /* only schedule a downconvert if we haven't already scheduled
636 * one that goes low enough to satisfy the level we're
637 * blocking. this also catches the case where we get
638 * duplicate BASTs */
639 if (ocfs2_highest_compat_lock_level(level) <
640 ocfs2_highest_compat_lock_level(lockres->l_blocking))
641 needs_downconvert = 1;
642
643 lockres->l_blocking = level;
644 }
645
646 mlog_exit(needs_downconvert);
647 return needs_downconvert;
648}
649
aa2623ad 650static void ocfs2_blocking_ast(void *opaque, int level)
ccd979bd 651{
aa2623ad
MF		 652 struct ocfs2_lock_res *lockres = opaque;
653 struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
ccd979bd
MF		 654 int needs_downconvert;
655 unsigned long flags;
656
ccd979bd
MF		 657 BUG_ON(level <= LKM_NLMODE);
658
aa2623ad
MF		 659 mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n",
660 lockres->l_name, level, lockres->l_level,
661 ocfs2_lock_type_string(lockres->l_type));
662
ccd979bd
MF		 663 spin_lock_irqsave(&lockres->l_lock, flags);
664 needs_downconvert = ocfs2_generic_handle_bast(lockres, level);
665 if (needs_downconvert)
666 ocfs2_schedule_blocked_lock(osb, lockres);
667 spin_unlock_irqrestore(&lockres->l_lock, flags);
668
d680efe9
MF		 669 wake_up(&lockres->l_event);
670
ccd979bd 671 ocfs2_kick_vote_thread(osb);
ccd979bd
MF		 672}
673
e92d57df 674static void ocfs2_locking_ast(void *opaque)
ccd979bd 675{
e92d57df 676 struct ocfs2_lock_res *lockres = opaque;
ccd979bd
MF		 677 struct dlm_lockstatus *lksb = &lockres->l_lksb;
678 unsigned long flags;
679
680 spin_lock_irqsave(&lockres->l_lock, flags);
681
682 if (lksb->status != DLM_NORMAL) {
683 mlog(ML_ERROR, "lockres %s: lksb status value of %u!\n",
684 lockres->l_name, lksb->status);
685 spin_unlock_irqrestore(&lockres->l_lock, flags);
686 return;
687 }
688
689 switch(lockres->l_action) {
690 case OCFS2_AST_ATTACH:
691 ocfs2_generic_handle_attach_action(lockres);
e92d57df 692 lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL);
ccd979bd
MF		 693 break;
694 case OCFS2_AST_CONVERT:
695 ocfs2_generic_handle_convert_action(lockres);
696 break;
697 case OCFS2_AST_DOWNCONVERT:
698 ocfs2_generic_handle_downconvert_action(lockres);
699 break;
700 default:
e92d57df
MF		 701 mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u "
702 "lockres flags = 0x%lx, unlock action: %u\n",
703 lockres->l_name, lockres->l_action, lockres->l_flags,
704 lockres->l_unlock_action);
ccd979bd
MF		 705 BUG();
706 }
707
ccd979bd
MF		 708 /* set it to something invalid so if we get called again we
709 * can catch it. */
710 lockres->l_action = OCFS2_AST_INVALID;
ccd979bd
MF		 711
712 wake_up(&lockres->l_event);
d680efe9 713 spin_unlock_irqrestore(&lockres->l_lock, flags);
ccd979bd
MF		 714}
715
ccd979bd
MF		 716static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
717 int convert)
718{
719 unsigned long flags;
720
721 mlog_entry_void();
722 spin_lock_irqsave(&lockres->l_lock, flags);
723 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
724 if (convert)
725 lockres->l_action = OCFS2_AST_INVALID;
726 else
727 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
728 spin_unlock_irqrestore(&lockres->l_lock, flags);
729
730 wake_up(&lockres->l_event);
731 mlog_exit_void();
732}
733
734/* Note: If we detect another process working on the lock (i.e.,
735 * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller
736 * to do the right thing in that case.
737 */
738static int ocfs2_lock_create(struct ocfs2_super *osb,
739 struct ocfs2_lock_res *lockres,
740 int level,
741 int dlm_flags)
742{
743 int ret = 0;
744 enum dlm_status status;
745 unsigned long flags;
746
747 mlog_entry_void();
748
749 mlog(0, "lock %s, level = %d, flags = %d\n", lockres->l_name, level,
750 dlm_flags);
751
752 spin_lock_irqsave(&lockres->l_lock, flags);
753 if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) ||
754 (lockres->l_flags & OCFS2_LOCK_BUSY)) {
755 spin_unlock_irqrestore(&lockres->l_lock, flags);
756 goto bail;
757 }
758
759 lockres->l_action = OCFS2_AST_ATTACH;
760 lockres->l_requested = level;
761 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
762 spin_unlock_irqrestore(&lockres->l_lock, flags);
763
764 status = dlmlock(osb->dlm,
765 level,
766 &lockres->l_lksb,
767 dlm_flags,
768 lockres->l_name,
f0681062 769 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 770 ocfs2_locking_ast,
ccd979bd 771 lockres,
aa2623ad 772 ocfs2_blocking_ast);
ccd979bd
MF		 773 if (status != DLM_NORMAL) {
774 ocfs2_log_dlm_error("dlmlock", status, lockres);
775 ret = -EINVAL;
776 ocfs2_recover_from_dlm_error(lockres, 1);
777 }
778
779 mlog(0, "lock %s, successfull return from dlmlock\n", lockres->l_name);
780
781bail:
782 mlog_exit(ret);
783 return ret;
784}
785
786static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres,
787 int flag)
788{
789 unsigned long flags;
790 int ret;
791
792 spin_lock_irqsave(&lockres->l_lock, flags);
793 ret = lockres->l_flags & flag;
794 spin_unlock_irqrestore(&lockres->l_lock, flags);
795
796 return ret;
797}
798
799static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres)
800
801{
802 wait_event(lockres->l_event,
803 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY));
804}
805
806static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres)
807
808{
809 wait_event(lockres->l_event,
810 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING));
811}
812
813/* predict what lock level we'll be dropping down to on behalf
814 * of another node, and return true if the currently wanted
815 * level will be compatible with it. */
816static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
817 int wanted)
818{
819 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
820
821 return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking);
822}
823
824static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw)
825{
826 INIT_LIST_HEAD(&mw->mw_item);
827 init_completion(&mw->mw_complete);
828}
829
830static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw)
831{
832 wait_for_completion(&mw->mw_complete);
833 /* Re-arm the completion in case we want to wait on it again */
834 INIT_COMPLETION(mw->mw_complete);
835 return mw->mw_status;
836}
837
838static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres,
839 struct ocfs2_mask_waiter *mw,
840 unsigned long mask,
841 unsigned long goal)
842{
843 BUG_ON(!list_empty(&mw->mw_item));
844
845 assert_spin_locked(&lockres->l_lock);
846
847 list_add_tail(&mw->mw_item, &lockres->l_mask_waiters);
848 mw->mw_mask = mask;
849 mw->mw_goal = goal;
850}
851
852/* returns 0 if the mw that was removed was already satisfied, -EBUSY
853 * if the mask still hadn't reached its goal */
854static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
855 struct ocfs2_mask_waiter *mw)
856{
857 unsigned long flags;
858 int ret = 0;
859
860 spin_lock_irqsave(&lockres->l_lock, flags);
861 if (!list_empty(&mw->mw_item)) {
862 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
863 ret = -EBUSY;
864
865 list_del_init(&mw->mw_item);
866 init_completion(&mw->mw_complete);
867 }
868 spin_unlock_irqrestore(&lockres->l_lock, flags);
869
870 return ret;
871
872}
873
874static int ocfs2_cluster_lock(struct ocfs2_super *osb,
875 struct ocfs2_lock_res *lockres,
876 int level,
877 int lkm_flags,
878 int arg_flags)
879{
880 struct ocfs2_mask_waiter mw;
881 enum dlm_status status;
882 int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR);
883 int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */
884 unsigned long flags;
885
886 mlog_entry_void();
887
888 ocfs2_init_mask_waiter(&mw);
889
b80fc012
MF		 890 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
891 lkm_flags |= LKM_VALBLK;
892
ccd979bd
MF		 893again:
894 wait = 0;
895
896 if (catch_signals && signal_pending(current)) {
897 ret = -ERESTARTSYS;
898 goto out;
899 }
900
901 spin_lock_irqsave(&lockres->l_lock, flags);
902
903 mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING,
904 "Cluster lock called on freeing lockres %s! flags "
905 "0x%lx\n", lockres->l_name, lockres->l_flags);
906
907 /* We only compare against the currently granted level
908 * here. If the lock is blocked waiting on a downconvert,
909 * we'll get caught below. */
910 if (lockres->l_flags & OCFS2_LOCK_BUSY &&
911 level > lockres->l_level) {
912 /* is someone sitting in dlm_lock? If so, wait on
913 * them. */
914 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
915 wait = 1;
916 goto unlock;
917 }
918
919 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
920 /* lock has not been created yet. */
921 spin_unlock_irqrestore(&lockres->l_lock, flags);
922
923 ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0);
924 if (ret < 0) {
925 mlog_errno(ret);
926 goto out;
927 }
928 goto again;
929 }
930
931 if (lockres->l_flags & OCFS2_LOCK_BLOCKED &&
932 !ocfs2_may_continue_on_blocked_lock(lockres, level)) {
933 /* is the lock is currently blocked on behalf of
934 * another node */
935 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0);
936 wait = 1;
937 goto unlock;
938 }
939
940 if (level > lockres->l_level) {
941 if (lockres->l_action != OCFS2_AST_INVALID)
942 mlog(ML_ERROR, "lockres %s has action %u pending\n",
943 lockres->l_name, lockres->l_action);
944
945 lockres->l_action = OCFS2_AST_CONVERT;
946 lockres->l_requested = level;
947 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
948 spin_unlock_irqrestore(&lockres->l_lock, flags);
949
950 BUG_ON(level == LKM_IVMODE);
951 BUG_ON(level == LKM_NLMODE);
952
953 mlog(0, "lock %s, convert from %d to level = %d\n",
954 lockres->l_name, lockres->l_level, level);
955
956 /* call dlm_lock to upgrade lock now */
957 status = dlmlock(osb->dlm,
958 level,
959 &lockres->l_lksb,
b80fc012 960 lkm_flags|LKM_CONVERT,
ccd979bd 961 lockres->l_name,
f0681062 962 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 963 ocfs2_locking_ast,
ccd979bd 964 lockres,
aa2623ad 965 ocfs2_blocking_ast);
ccd979bd
MF		 966 if (status != DLM_NORMAL) {
967 if ((lkm_flags & LKM_NOQUEUE) &&
968 (status == DLM_NOTQUEUED))
969 ret = -EAGAIN;
970 else {
971 ocfs2_log_dlm_error("dlmlock", status,
972 lockres);
973 ret = -EINVAL;
974 }
975 ocfs2_recover_from_dlm_error(lockres, 1);
976 goto out;
977 }
978
979 mlog(0, "lock %s, successfull return from dlmlock\n",
980 lockres->l_name);
981
982 /* At this point we've gone inside the dlm and need to
983 * complete our work regardless. */
984 catch_signals = 0;
985
986 /* wait for busy to clear and carry on */
987 goto again;
988 }
989
990 /* Ok, if we get here then we're good to go. */
991 ocfs2_inc_holders(lockres, level);
992
993 ret = 0;
994unlock:
995 spin_unlock_irqrestore(&lockres->l_lock, flags);
996out:
997 /*
998 * This is helping work around a lock inversion between the page lock
999 * and dlm locks. One path holds the page lock while calling aops
1000 * which block acquiring dlm locks. The voting thread holds dlm
1001 * locks while acquiring page locks while down converting data locks.
1002 * This block is helping an aop path notice the inversion and back
1003 * off to unlock its page lock before trying the dlm lock again.
1004 */
1005 if (wait && arg_flags & OCFS2_LOCK_NONBLOCK &&
1006 mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) {
1007 wait = 0;
1008 if (lockres_remove_mask_waiter(lockres, &mw))
1009 ret = -EAGAIN;
1010 else
1011 goto again;
1012 }
1013 if (wait) {
1014 ret = ocfs2_wait_for_mask(&mw);
1015 if (ret == 0)
1016 goto again;
1017 mlog_errno(ret);
1018 }
1019
1020 mlog_exit(ret);
1021 return ret;
1022}
1023
1024static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
1025 struct ocfs2_lock_res *lockres,
1026 int level)
1027{
1028 unsigned long flags;
1029
1030 mlog_entry_void();
1031 spin_lock_irqsave(&lockres->l_lock, flags);
1032 ocfs2_dec_holders(lockres, level);
1033 ocfs2_vote_on_unlock(osb, lockres);
1034 spin_unlock_irqrestore(&lockres->l_lock, flags);
1035 mlog_exit_void();
1036}
1037
d680efe9
MF		 1038int ocfs2_create_new_lock(struct ocfs2_super *osb,
1039 struct ocfs2_lock_res *lockres,
24c19ef4
MF		 1040 int ex,
1041 int local)
ccd979bd 1042{
d680efe9 1043 int level = ex ? LKM_EXMODE : LKM_PRMODE;
ccd979bd 1044 unsigned long flags;
24c19ef4 1045 int lkm_flags = local ? LKM_LOCAL : 0;
ccd979bd
MF		 1046
1047 spin_lock_irqsave(&lockres->l_lock, flags);
1048 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
1049 lockres_or_flags(lockres, OCFS2_LOCK_LOCAL);
1050 spin_unlock_irqrestore(&lockres->l_lock, flags);
1051
24c19ef4 1052 return ocfs2_lock_create(osb, lockres, level, lkm_flags);
ccd979bd
MF		 1053}
1054
1055/* Grants us an EX lock on the data and metadata resources, skipping
1056 * the normal cluster directory lookup. Use this ONLY on newly created
1057 * inodes which other nodes can't possibly see, and which haven't been
1058 * hashed in the inode hash yet. This can give us a good performance
1059 * increase as it'll skip the network broadcast normally associated
1060 * with creating a new lock resource. */
1061int ocfs2_create_new_inode_locks(struct inode *inode)
1062{
1063 int ret;
d680efe9 1064 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
ccd979bd
MF		 1065
1066 BUG_ON(!inode);
1067 BUG_ON(!ocfs2_inode_is_new(inode));
1068
1069 mlog_entry_void();
1070
b0697053 1071 mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd
MF		 1072
1073 /* NOTE: That we don't increment any of the holder counts, nor
1074 * do we add anything to a journal handle. Since this is
1075 * supposed to be a new inode which the cluster doesn't know
1076 * about yet, there is no need to. As far as the LVB handling
1077 * is concerned, this is basically like acquiring an EX lock
1078 * on a resource which has an invalid one -- we'll set it
1079 * valid when we release the EX. */
1080
24c19ef4 1081 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1);
ccd979bd
MF		 1082 if (ret) {
1083 mlog_errno(ret);
1084 goto bail;
1085 }
1086
24c19ef4
MF		 1087 /*
1088 * We don't want to use LKM_LOCAL on a meta data lock as they
1089 * don't use a generation in their lock names.
1090 */
1091 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_meta_lockres, 1, 0);
ccd979bd
MF		 1092 if (ret) {
1093 mlog_errno(ret);
1094 goto bail;
1095 }
1096
24c19ef4 1097 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_data_lockres, 1, 1);
ccd979bd
MF		 1098 if (ret) {
1099 mlog_errno(ret);
1100 goto bail;
1101 }
1102
1103bail:
1104 mlog_exit(ret);
1105 return ret;
1106}
1107
1108int ocfs2_rw_lock(struct inode *inode, int write)
1109{
1110 int status, level;
1111 struct ocfs2_lock_res *lockres;
1112
1113 BUG_ON(!inode);
1114
1115 mlog_entry_void();
1116
b0697053
MF		 1117 mlog(0, "inode %llu take %s RW lock\n",
1118 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1119 write ? "EXMODE" : "PRMODE");
1120
1121 lockres = &OCFS2_I(inode)->ip_rw_lockres;
1122
1123 level = write ? LKM_EXMODE : LKM_PRMODE;
1124
1125 status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0,
1126 0);
1127 if (status < 0)
1128 mlog_errno(status);
1129
1130 mlog_exit(status);
1131 return status;
1132}
1133
1134void ocfs2_rw_unlock(struct inode *inode, int write)
1135{
1136 int level = write ? LKM_EXMODE : LKM_PRMODE;
1137 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres;
1138
1139 mlog_entry_void();
1140
b0697053
MF		 1141 mlog(0, "inode %llu drop %s RW lock\n",
1142 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1143 write ? "EXMODE" : "PRMODE");
1144
1145 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1146
1147 mlog_exit_void();
1148}
1149
1150int ocfs2_data_lock_full(struct inode *inode,
1151 int write,
1152 int arg_flags)
1153{
1154 int status = 0, level;
1155 struct ocfs2_lock_res *lockres;
1156
1157 BUG_ON(!inode);
1158
1159 mlog_entry_void();
1160
b0697053
MF		 1161 mlog(0, "inode %llu take %s DATA lock\n",
1162 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1163 write ? "EXMODE" : "PRMODE");
1164
1165 /* We'll allow faking a readonly data lock for
1166 * rodevices. */
1167 if (ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) {
1168 if (write) {
1169 status = -EROFS;
1170 mlog_errno(status);
1171 }
1172 goto out;
1173 }
1174
1175 lockres = &OCFS2_I(inode)->ip_data_lockres;
1176
1177 level = write ? LKM_EXMODE : LKM_PRMODE;
1178
1179 status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level,
1180 0, arg_flags);
1181 if (status < 0 && status != -EAGAIN)
1182 mlog_errno(status);
1183
1184out:
1185 mlog_exit(status);
1186 return status;
1187}
1188
1189/* see ocfs2_meta_lock_with_page() */
1190int ocfs2_data_lock_with_page(struct inode *inode,
1191 int write,
1192 struct page *page)
1193{
1194 int ret;
1195
1196 ret = ocfs2_data_lock_full(inode, write, OCFS2_LOCK_NONBLOCK);
1197 if (ret == -EAGAIN) {
1198 unlock_page(page);
1199 if (ocfs2_data_lock(inode, write) == 0)
1200 ocfs2_data_unlock(inode, write);
1201 ret = AOP_TRUNCATED_PAGE;
1202 }
1203
1204 return ret;
1205}
1206
1207static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
1208 struct ocfs2_lock_res *lockres)
1209{
1210 int kick = 0;
1211
1212 mlog_entry_void();
1213
1214 /* If we know that another node is waiting on our lock, kick
1215 * the vote thread * pre-emptively when we reach a release
1216 * condition. */
1217 if (lockres->l_flags & OCFS2_LOCK_BLOCKED) {
1218 switch(lockres->l_blocking) {
1219 case LKM_EXMODE:
1220 if (!lockres->l_ex_holders && !lockres->l_ro_holders)
1221 kick = 1;
1222 break;
1223 case LKM_PRMODE:
1224 if (!lockres->l_ex_holders)
1225 kick = 1;
1226 break;
1227 default:
1228 BUG();
1229 }
1230 }
1231
1232 if (kick)
1233 ocfs2_kick_vote_thread(osb);
1234
1235 mlog_exit_void();
1236}
1237
1238void ocfs2_data_unlock(struct inode *inode,
1239 int write)
1240{
1241 int level = write ? LKM_EXMODE : LKM_PRMODE;
1242 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres;
1243
1244 mlog_entry_void();
1245
b0697053
MF		 1246 mlog(0, "inode %llu drop %s DATA lock\n",
1247 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1248 write ? "EXMODE" : "PRMODE");
1249
1250 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
1251 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1252
1253 mlog_exit_void();
1254}
1255
1256#define OCFS2_SEC_BITS 34
1257#define OCFS2_SEC_SHIFT (64 - 34)
1258#define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1)
1259
1260/* LVB only has room for 64 bits of time here so we pack it for
1261 * now. */
1262static u64 ocfs2_pack_timespec(struct timespec *spec)
1263{
1264 u64 res;
1265 u64 sec = spec->tv_sec;
1266 u32 nsec = spec->tv_nsec;
1267
1268 res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK);
1269
1270 return res;
1271}
1272
1273/* Call this with the lockres locked. I am reasonably sure we don't
1274 * need ip_lock in this function as anyone who would be changing those
1275 * values is supposed to be blocked in ocfs2_meta_lock right now. */
1276static void __ocfs2_stuff_meta_lvb(struct inode *inode)
1277{
1278 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1279 struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
1280 struct ocfs2_meta_lvb *lvb;
1281
1282 mlog_entry_void();
1283
1284 lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1285
24c19ef4
MF		 1286 /*
1287 * Invalidate the LVB of a deleted inode - this way other
1288 * nodes are forced to go to disk and discover the new inode
1289 * status.
1290 */
1291 if (oi->ip_flags & OCFS2_INODE_DELETED) {
1292 lvb->lvb_version = 0;
1293 goto out;
1294 }
1295
4d3b83f7 1296 lvb->lvb_version = OCFS2_LVB_VERSION;
ccd979bd
MF		 1297 lvb->lvb_isize = cpu_to_be64(i_size_read(inode));
1298 lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters);
1299 lvb->lvb_iuid = cpu_to_be32(inode->i_uid);
1300 lvb->lvb_igid = cpu_to_be32(inode->i_gid);
1301 lvb->lvb_imode = cpu_to_be16(inode->i_mode);
1302 lvb->lvb_inlink = cpu_to_be16(inode->i_nlink);
1303 lvb->lvb_iatime_packed =
1304 cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime));
1305 lvb->lvb_ictime_packed =
1306 cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime));
1307 lvb->lvb_imtime_packed =
1308 cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime));
ca4d147e 1309 lvb->lvb_iattr = cpu_to_be32(oi->ip_attr);
f9e2d82e 1310 lvb->lvb_igeneration = cpu_to_be32(inode->i_generation);
ccd979bd 1311
24c19ef4 1312out:
ccd979bd
MF		 1313 mlog_meta_lvb(0, lockres);
1314
1315 mlog_exit_void();
1316}
1317
1318static void ocfs2_unpack_timespec(struct timespec *spec,
1319 u64 packed_time)
1320{
1321 spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT;
1322 spec->tv_nsec = packed_time & OCFS2_NSEC_MASK;
1323}
1324
1325static void ocfs2_refresh_inode_from_lvb(struct inode *inode)
1326{
1327 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1328 struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
1329 struct ocfs2_meta_lvb *lvb;
1330
1331 mlog_entry_void();
1332
1333 mlog_meta_lvb(0, lockres);
1334
1335 lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1336
1337 /* We're safe here without the lockres lock... */
1338 spin_lock(&oi->ip_lock);
1339 oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters);
1340 i_size_write(inode, be64_to_cpu(lvb->lvb_isize));
1341
ca4d147e
HP		 1342 oi->ip_attr = be32_to_cpu(lvb->lvb_iattr);
1343 ocfs2_set_inode_flags(inode);
1344
ccd979bd
MF		 1345 /* fast-symlinks are a special case */
1346 if (S_ISLNK(inode->i_mode) && !oi->ip_clusters)
1347 inode->i_blocks = 0;
1348 else
1349 inode->i_blocks =
1350 ocfs2_align_bytes_to_sectors(i_size_read(inode));
1351
1352 inode->i_uid = be32_to_cpu(lvb->lvb_iuid);
1353 inode->i_gid = be32_to_cpu(lvb->lvb_igid);
1354 inode->i_mode = be16_to_cpu(lvb->lvb_imode);
1355 inode->i_nlink = be16_to_cpu(lvb->lvb_inlink);
1356 ocfs2_unpack_timespec(&inode->i_atime,
1357 be64_to_cpu(lvb->lvb_iatime_packed));
1358 ocfs2_unpack_timespec(&inode->i_mtime,
1359 be64_to_cpu(lvb->lvb_imtime_packed));
1360 ocfs2_unpack_timespec(&inode->i_ctime,
1361 be64_to_cpu(lvb->lvb_ictime_packed));
1362 spin_unlock(&oi->ip_lock);
1363
1364 mlog_exit_void();
1365}
1366
f9e2d82e
MF		 1367static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode,
1368 struct ocfs2_lock_res *lockres)
ccd979bd
MF		 1369{
1370 struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1371
f9e2d82e
MF		 1372 if (lvb->lvb_version == OCFS2_LVB_VERSION
1373 && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation)
ccd979bd
MF		 1374 return 1;
1375 return 0;
1376}
1377
1378/* Determine whether a lock resource needs to be refreshed, and
1379 * arbitrate who gets to refresh it.
1380 *
1381 * 0 means no refresh needed.
1382 *
1383 * > 0 means you need to refresh this and you MUST call
1384 * ocfs2_complete_lock_res_refresh afterwards. */
1385static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres)
1386{
1387 unsigned long flags;
1388 int status = 0;
1389
1390 mlog_entry_void();
1391
1392refresh_check:
1393 spin_lock_irqsave(&lockres->l_lock, flags);
1394 if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
1395 spin_unlock_irqrestore(&lockres->l_lock, flags);
1396 goto bail;
1397 }
1398
1399 if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
1400 spin_unlock_irqrestore(&lockres->l_lock, flags);
1401
1402 ocfs2_wait_on_refreshing_lock(lockres);
1403 goto refresh_check;
1404 }
1405
1406 /* Ok, I'll be the one to refresh this lock. */
1407 lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING);
1408 spin_unlock_irqrestore(&lockres->l_lock, flags);
1409
1410 status = 1;
1411bail:
1412 mlog_exit(status);
1413 return status;
1414}
1415
1416/* If status is non zero, I'll mark it as not being in refresh
1417 * anymroe, but i won't clear the needs refresh flag. */
1418static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres,
1419 int status)
1420{
1421 unsigned long flags;
1422 mlog_entry_void();
1423
1424 spin_lock_irqsave(&lockres->l_lock, flags);
1425 lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING);
1426 if (!status)
1427 lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
1428 spin_unlock_irqrestore(&lockres->l_lock, flags);
1429
1430 wake_up(&lockres->l_event);
1431
1432 mlog_exit_void();
1433}
1434
1435/* may or may not return a bh if it went to disk. */
1436static int ocfs2_meta_lock_update(struct inode *inode,
1437 struct buffer_head **bh)
1438{
1439 int status = 0;
1440 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1441 struct ocfs2_lock_res *lockres;
1442 struct ocfs2_dinode *fe;
1443
1444 mlog_entry_void();
1445
1446 spin_lock(&oi->ip_lock);
1447 if (oi->ip_flags & OCFS2_INODE_DELETED) {
b0697053 1448 mlog(0, "Orphaned inode %llu was deleted while we "
ccd979bd 1449 "were waiting on a lock. ip_flags = 0x%x\n",
b0697053 1450 (unsigned long long)oi->ip_blkno, oi->ip_flags);
ccd979bd
MF		 1451 spin_unlock(&oi->ip_lock);
1452 status = -ENOENT;
1453 goto bail;
1454 }
1455 spin_unlock(&oi->ip_lock);
1456
1457 lockres = &oi->ip_meta_lockres;
1458
1459 if (!ocfs2_should_refresh_lock_res(lockres))
1460 goto bail;
1461
1462 /* This will discard any caching information we might have had
1463 * for the inode metadata. */
1464 ocfs2_metadata_cache_purge(inode);
1465
1466 /* will do nothing for inode types that don't use the extent
1467 * map (directories, bitmap files, etc) */
1468 ocfs2_extent_map_trunc(inode, 0);
1469
f9e2d82e 1470 if (ocfs2_meta_lvb_is_trustable(inode, lockres)) {
b0697053
MF		 1471 mlog(0, "Trusting LVB on inode %llu\n",
1472 (unsigned long long)oi->ip_blkno);
ccd979bd
MF		 1473 ocfs2_refresh_inode_from_lvb(inode);
1474 } else {
1475 /* Boo, we have to go to disk. */
1476 /* read bh, cast, ocfs2_refresh_inode */
1477 status = ocfs2_read_block(OCFS2_SB(inode->i_sb), oi->ip_blkno,
1478 bh, OCFS2_BH_CACHED, inode);
1479 if (status < 0) {
1480 mlog_errno(status);
1481 goto bail_refresh;
1482 }
1483 fe = (struct ocfs2_dinode *) (*bh)->b_data;
1484
1485 /* This is a good chance to make sure we're not
1486 * locking an invalid object.
1487 *
1488 * We bug on a stale inode here because we checked
1489 * above whether it was wiped from disk. The wiping
1490 * node provides a guarantee that we receive that
1491 * message and can mark the inode before dropping any
1492 * locks associated with it. */
1493 if (!OCFS2_IS_VALID_DINODE(fe)) {
1494 OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
1495 status = -EIO;
1496 goto bail_refresh;
1497 }
1498 mlog_bug_on_msg(inode->i_generation !=
1499 le32_to_cpu(fe->i_generation),
b0697053 1500 "Invalid dinode %llu disk generation: %u "
ccd979bd 1501 "inode->i_generation: %u\n",
b0697053
MF		 1502 (unsigned long long)oi->ip_blkno,
1503 le32_to_cpu(fe->i_generation),
ccd979bd
MF		 1504 inode->i_generation);
1505 mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) ||
1506 !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)),
b0697053
MF		 1507 "Stale dinode %llu dtime: %llu flags: 0x%x\n",
1508 (unsigned long long)oi->ip_blkno,
1509 (unsigned long long)le64_to_cpu(fe->i_dtime),
ccd979bd
MF		 1510 le32_to_cpu(fe->i_flags));
1511
1512 ocfs2_refresh_inode(inode, fe);
1513 }
1514
1515 status = 0;
1516bail_refresh:
1517 ocfs2_complete_lock_res_refresh(lockres, status);
1518bail:
1519 mlog_exit(status);
1520 return status;
1521}
1522
1523static int ocfs2_assign_bh(struct inode *inode,
1524 struct buffer_head **ret_bh,
1525 struct buffer_head *passed_bh)
1526{
1527 int status;
1528
1529 if (passed_bh) {
1530 /* Ok, the update went to disk for us, use the
1531 * returned bh. */
1532 *ret_bh = passed_bh;
1533 get_bh(*ret_bh);
1534
1535 return 0;
1536 }
1537
1538 status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
1539 OCFS2_I(inode)->ip_blkno,
1540 ret_bh,
1541 OCFS2_BH_CACHED,
1542 inode);
1543 if (status < 0)
1544 mlog_errno(status);
1545
1546 return status;
1547}
1548
1549/*
1550 * returns < 0 error if the callback will never be called, otherwise
1551 * the result of the lock will be communicated via the callback.
1552 */
1553int ocfs2_meta_lock_full(struct inode *inode,
1554 struct ocfs2_journal_handle *handle,
1555 struct buffer_head **ret_bh,
1556 int ex,
1557 int arg_flags)
1558{
1559 int status, level, dlm_flags, acquired;
1560 struct ocfs2_lock_res *lockres;
1561 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1562 struct buffer_head *local_bh = NULL;
1563
1564 BUG_ON(!inode);
1565
1566 mlog_entry_void();
1567
b0697053
MF		 1568 mlog(0, "inode %llu, take %s META lock\n",
1569 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1570 ex ? "EXMODE" : "PRMODE");
1571
1572 status = 0;
1573 acquired = 0;
1574 /* We'll allow faking a readonly metadata lock for
1575 * rodevices. */
1576 if (ocfs2_is_hard_readonly(osb)) {
1577 if (ex)
1578 status = -EROFS;
1579 goto bail;
1580 }
1581
1582 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
1583 wait_event(osb->recovery_event,
1584 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
1585
1586 acquired = 0;
1587 lockres = &OCFS2_I(inode)->ip_meta_lockres;
1588 level = ex ? LKM_EXMODE : LKM_PRMODE;
1589 dlm_flags = 0;
1590 if (arg_flags & OCFS2_META_LOCK_NOQUEUE)
1591 dlm_flags |= LKM_NOQUEUE;
1592
1593 status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags);
1594 if (status < 0) {
1595 if (status != -EAGAIN && status != -EIOCBRETRY)
1596 mlog_errno(status);
1597 goto bail;
1598 }
1599
1600 /* Notify the error cleanup path to drop the cluster lock. */
1601 acquired = 1;
1602
1603 /* We wait twice because a node may have died while we were in
1604 * the lower dlm layers. The second time though, we've
1605 * committed to owning this lock so we don't allow signals to
1606 * abort the operation. */
1607 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
1608 wait_event(osb->recovery_event,
1609 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
1610
24c19ef4
MF		 1611 /*
1612 * We only see this flag if we're being called from
1613 * ocfs2_read_locked_inode(). It means we're locking an inode
1614 * which hasn't been populated yet, so clear the refresh flag
1615 * and let the caller handle it.
1616 */
1617 if (inode->i_state & I_NEW) {
1618 status = 0;
1619 ocfs2_complete_lock_res_refresh(lockres, 0);
1620 goto bail;
1621 }
1622
ccd979bd
MF		 1623 /* This is fun. The caller may want a bh back, or it may
1624 * not. ocfs2_meta_lock_update definitely wants one in, but
1625 * may or may not read one, depending on what's in the
1626 * LVB. The result of all of this is that we've *only* gone to
1627 * disk if we have to, so the complexity is worthwhile. */
1628 status = ocfs2_meta_lock_update(inode, &local_bh);
1629 if (status < 0) {
1630 if (status != -ENOENT)
1631 mlog_errno(status);
1632 goto bail;
1633 }
1634
1635 if (ret_bh) {
1636 status = ocfs2_assign_bh(inode, ret_bh, local_bh);
1637 if (status < 0) {
1638 mlog_errno(status);
1639 goto bail;
1640 }
1641 }
1642
1643 if (handle) {
1644 status = ocfs2_handle_add_lock(handle, inode);
1645 if (status < 0)
1646 mlog_errno(status);
1647 }
1648
1649bail:
1650 if (status < 0) {
1651 if (ret_bh && (*ret_bh)) {
1652 brelse(*ret_bh);
1653 *ret_bh = NULL;
1654 }
1655 if (acquired)
1656 ocfs2_meta_unlock(inode, ex);
1657 }
1658
1659 if (local_bh)
1660 brelse(local_bh);
1661
1662 mlog_exit(status);
1663 return status;
1664}
1665
1666/*
1667 * This is working around a lock inversion between tasks acquiring DLM locks
1668 * while holding a page lock and the vote thread which blocks dlm lock acquiry
1669 * while acquiring page locks.
1670 *
1671 * ** These _with_page variantes are only intended to be called from aop
1672 * methods that hold page locks and return a very specific *positive* error
1673 * code that aop methods pass up to the VFS -- test for errors with != 0. **
1674 *
1675 * The DLM is called such that it returns -EAGAIN if it would have blocked
1676 * waiting for the vote thread. In that case we unlock our page so the vote
1677 * thread can make progress. Once we've done this we have to return
1678 * AOP_TRUNCATED_PAGE so the aop method that called us can bubble that back up
1679 * into the VFS who will then immediately retry the aop call.
1680 *
1681 * We do a blocking lock and immediate unlock before returning, though, so that
1682 * the lock has a great chance of being cached on this node by the time the VFS
1683 * calls back to retry the aop. This has a potential to livelock as nodes
1684 * ping locks back and forth, but that's a risk we're willing to take to avoid
1685 * the lock inversion simply.
1686 */
1687int ocfs2_meta_lock_with_page(struct inode *inode,
1688 struct ocfs2_journal_handle *handle,
1689 struct buffer_head **ret_bh,
1690 int ex,
1691 struct page *page)
1692{
1693 int ret;
1694
1695 ret = ocfs2_meta_lock_full(inode, handle, ret_bh, ex,
1696 OCFS2_LOCK_NONBLOCK);
1697 if (ret == -EAGAIN) {
1698 unlock_page(page);
1699 if (ocfs2_meta_lock(inode, handle, ret_bh, ex) == 0)
1700 ocfs2_meta_unlock(inode, ex);
1701 ret = AOP_TRUNCATED_PAGE;
1702 }
1703
1704 return ret;
1705}
1706
1707void ocfs2_meta_unlock(struct inode *inode,
1708 int ex)
1709{
1710 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1711 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
1712
1713 mlog_entry_void();
1714
b0697053
MF		 1715 mlog(0, "inode %llu drop %s META lock\n",
1716 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1717 ex ? "EXMODE" : "PRMODE");
1718
1719 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
1720 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1721
1722 mlog_exit_void();
1723}
1724
1725int ocfs2_super_lock(struct ocfs2_super *osb,
1726 int ex)
1727{
1728 int status;
1729 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1730 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1731 struct buffer_head *bh;
1732 struct ocfs2_slot_info *si = osb->slot_info;
1733
1734 mlog_entry_void();
1735
1736 if (ocfs2_is_hard_readonly(osb))
1737 return -EROFS;
1738
1739 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
1740 if (status < 0) {
1741 mlog_errno(status);
1742 goto bail;
1743 }
1744
1745 /* The super block lock path is really in the best position to
1746 * know when resources covered by the lock need to be
1747 * refreshed, so we do it here. Of course, making sense of
1748 * everything is up to the caller :) */
1749 status = ocfs2_should_refresh_lock_res(lockres);
1750 if (status < 0) {
1751 mlog_errno(status);
1752 goto bail;
1753 }
1754 if (status) {
1755 bh = si->si_bh;
1756 status = ocfs2_read_block(osb, bh->b_blocknr, &bh, 0,
1757 si->si_inode);
1758 if (status == 0)
1759 ocfs2_update_slot_info(si);
1760
1761 ocfs2_complete_lock_res_refresh(lockres, status);
1762
1763 if (status < 0)
1764 mlog_errno(status);
1765 }
1766bail:
1767 mlog_exit(status);
1768 return status;
1769}
1770
1771void ocfs2_super_unlock(struct ocfs2_super *osb,
1772 int ex)
1773{
1774 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1775 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1776
1777 ocfs2_cluster_unlock(osb, lockres, level);
1778}
1779
1780int ocfs2_rename_lock(struct ocfs2_super *osb)
1781{
1782 int status;
1783 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
1784
1785 if (ocfs2_is_hard_readonly(osb))
1786 return -EROFS;
1787
1788 status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0);
1789 if (status < 0)
1790 mlog_errno(status);
1791
1792 return status;
1793}
1794
1795void ocfs2_rename_unlock(struct ocfs2_super *osb)
1796{
1797 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
1798
1799 ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE);
1800}
1801
d680efe9
MF		 1802int ocfs2_dentry_lock(struct dentry *dentry, int ex)
1803{
1804 int ret;
1805 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1806 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
1807 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
1808
1809 BUG_ON(!dl);
1810
1811 if (ocfs2_is_hard_readonly(osb))
1812 return -EROFS;
1813
1814 ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0);
1815 if (ret < 0)
1816 mlog_errno(ret);
1817
1818 return ret;
1819}
1820
1821void ocfs2_dentry_unlock(struct dentry *dentry, int ex)
1822{
1823 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1824 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
1825 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
1826
1827 ocfs2_cluster_unlock(osb, &dl->dl_lockres, level);
1828}
1829
ccd979bd
MF		 1830/* Reference counting of the dlm debug structure. We want this because
1831 * open references on the debug inodes can live on after a mount, so
1832 * we can't rely on the ocfs2_super to always exist. */
1833static void ocfs2_dlm_debug_free(struct kref *kref)
1834{
1835 struct ocfs2_dlm_debug *dlm_debug;
1836
1837 dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt);
1838
1839 kfree(dlm_debug);
1840}
1841
1842void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug)
1843{
1844 if (dlm_debug)
1845 kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free);
1846}
1847
1848static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug)
1849{
1850 kref_get(&debug->d_refcnt);
1851}
1852
1853struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void)
1854{
1855 struct ocfs2_dlm_debug *dlm_debug;
1856
1857 dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL);
1858 if (!dlm_debug) {
1859 mlog_errno(-ENOMEM);
1860 goto out;
1861 }
1862
1863 kref_init(&dlm_debug->d_refcnt);
1864 INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking);
1865 dlm_debug->d_locking_state = NULL;
1866out:
1867 return dlm_debug;
1868}
1869
1870/* Access to this is arbitrated for us via seq_file->sem. */
1871struct ocfs2_dlm_seq_priv {
1872 struct ocfs2_dlm_debug *p_dlm_debug;
1873 struct ocfs2_lock_res p_iter_res;
1874 struct ocfs2_lock_res p_tmp_res;
1875};
1876
1877static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start,
1878 struct ocfs2_dlm_seq_priv *priv)
1879{
1880 struct ocfs2_lock_res *iter, *ret = NULL;
1881 struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug;
1882
1883 assert_spin_locked(&ocfs2_dlm_tracking_lock);
1884
1885 list_for_each_entry(iter, &start->l_debug_list, l_debug_list) {
1886 /* discover the head of the list */
1887 if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) {
1888 mlog(0, "End of list found, %p\n", ret);
1889 break;
1890 }
1891
1892 /* We track our "dummy" iteration lockres' by a NULL
1893 * l_ops field. */
1894 if (iter->l_ops != NULL) {
1895 ret = iter;
1896 break;
1897 }
1898 }
1899
1900 return ret;
1901}
1902
1903static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos)
1904{
1905 struct ocfs2_dlm_seq_priv *priv = m->private;
1906 struct ocfs2_lock_res *iter;
1907
1908 spin_lock(&ocfs2_dlm_tracking_lock);
1909 iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv);
1910 if (iter) {
1911 /* Since lockres' have the lifetime of their container
1912 * (which can be inodes, ocfs2_supers, etc) we want to
1913 * copy this out to a temporary lockres while still
1914 * under the spinlock. Obviously after this we can't
1915 * trust any pointers on the copy returned, but that's
1916 * ok as the information we want isn't typically held
1917 * in them. */
1918 priv->p_tmp_res = *iter;
1919 iter = &priv->p_tmp_res;
1920 }
1921 spin_unlock(&ocfs2_dlm_tracking_lock);
1922
1923 return iter;
1924}
1925
1926static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v)
1927{
1928}
1929
1930static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos)
1931{
1932 struct ocfs2_dlm_seq_priv *priv = m->private;
1933 struct ocfs2_lock_res *iter = v;
1934 struct ocfs2_lock_res *dummy = &priv->p_iter_res;
1935
1936 spin_lock(&ocfs2_dlm_tracking_lock);
1937 iter = ocfs2_dlm_next_res(iter, priv);
1938 list_del_init(&dummy->l_debug_list);
1939 if (iter) {
1940 list_add(&dummy->l_debug_list, &iter->l_debug_list);
1941 priv->p_tmp_res = *iter;
1942 iter = &priv->p_tmp_res;
1943 }
1944 spin_unlock(&ocfs2_dlm_tracking_lock);
1945
1946 return iter;
1947}
1948
1949/* So that debugfs.ocfs2 can determine which format is being used */
1950#define OCFS2_DLM_DEBUG_STR_VERSION 1
1951static int ocfs2_dlm_seq_show(struct seq_file *m, void *v)
1952{
1953 int i;
1954 char *lvb;
1955 struct ocfs2_lock_res *lockres = v;
1956
1957 if (!lockres)
1958 return -EINVAL;
1959
d680efe9
MF		 1960 seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION);
1961
1962 if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY)
1963 seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1,
1964 lockres->l_name,
1965 (unsigned int)ocfs2_get_dentry_lock_ino(lockres));
1966 else
1967 seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name);
1968
1969 seq_printf(m, "%d\t"
ccd979bd
MF		 1970 "0x%lx\t"
1971 "0x%x\t"
1972 "0x%x\t"
1973 "%u\t"
1974 "%u\t"
1975 "%d\t"
1976 "%d\t",
ccd979bd
MF		 1977 lockres->l_level,
1978 lockres->l_flags,
1979 lockres->l_action,
1980 lockres->l_unlock_action,
1981 lockres->l_ro_holders,
1982 lockres->l_ex_holders,
1983 lockres->l_requested,
1984 lockres->l_blocking);
1985
1986 /* Dump the raw LVB */
1987 lvb = lockres->l_lksb.lvb;
1988 for(i = 0; i < DLM_LVB_LEN; i++)
1989 seq_printf(m, "0x%x\t", lvb[i]);
1990
1991 /* End the line */
1992 seq_printf(m, "\n");
1993 return 0;
1994}
1995
1996static struct seq_operations ocfs2_dlm_seq_ops = {
1997 .start = ocfs2_dlm_seq_start,
1998 .stop = ocfs2_dlm_seq_stop,
1999 .next = ocfs2_dlm_seq_next,
2000 .show = ocfs2_dlm_seq_show,
2001};
2002
2003static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file)
2004{
2005 struct seq_file *seq = (struct seq_file *) file->private_data;
2006 struct ocfs2_dlm_seq_priv *priv = seq->private;
2007 struct ocfs2_lock_res *res = &priv->p_iter_res;
2008
2009 ocfs2_remove_lockres_tracking(res);
2010 ocfs2_put_dlm_debug(priv->p_dlm_debug);
2011 return seq_release_private(inode, file);
2012}
2013
2014static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file)
2015{
2016 int ret;
2017 struct ocfs2_dlm_seq_priv *priv;
2018 struct seq_file *seq;
2019 struct ocfs2_super *osb;
2020
2021 priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL);
2022 if (!priv) {
2023 ret = -ENOMEM;
2024 mlog_errno(ret);
2025 goto out;
2026 }
2027 osb = (struct ocfs2_super *) inode->u.generic_ip;
2028 ocfs2_get_dlm_debug(osb->osb_dlm_debug);
2029 priv->p_dlm_debug = osb->osb_dlm_debug;
2030 INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list);
2031
2032 ret = seq_open(file, &ocfs2_dlm_seq_ops);
2033 if (ret) {
2034 kfree(priv);
2035 mlog_errno(ret);
2036 goto out;
2037 }
2038
2039 seq = (struct seq_file *) file->private_data;
2040 seq->private = priv;
2041
2042 ocfs2_add_lockres_tracking(&priv->p_iter_res,
2043 priv->p_dlm_debug);
2044
2045out:
2046 return ret;
2047}
2048
4b6f5d20 2049static const struct file_operations ocfs2_dlm_debug_fops = {
ccd979bd
MF		 2050 .open = ocfs2_dlm_debug_open,
2051 .release = ocfs2_dlm_debug_release,
2052 .read = seq_read,
2053 .llseek = seq_lseek,
2054};
2055
2056static int ocfs2_dlm_init_debug(struct ocfs2_super *osb)
2057{
2058 int ret = 0;
2059 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
2060
2061 dlm_debug->d_locking_state = debugfs_create_file("locking_state",
2062 S_IFREG|S_IRUSR,
2063 osb->osb_debug_root,
2064 osb,
2065 &ocfs2_dlm_debug_fops);
2066 if (!dlm_debug->d_locking_state) {
2067 ret = -EINVAL;
2068 mlog(ML_ERROR,
2069 "Unable to create locking state debugfs file.\n");
2070 goto out;
2071 }
2072
2073 ocfs2_get_dlm_debug(dlm_debug);
2074out:
2075 return ret;
2076}
2077
2078static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb)
2079{
2080 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
2081
2082 if (dlm_debug) {
2083 debugfs_remove(dlm_debug->d_locking_state);
2084 ocfs2_put_dlm_debug(dlm_debug);
2085 }
2086}
2087
2088int ocfs2_dlm_init(struct ocfs2_super *osb)
2089{
2090 int status;
2091 u32 dlm_key;
2092 struct dlm_ctxt *dlm;
2093
2094 mlog_entry_void();
2095
2096 status = ocfs2_dlm_init_debug(osb);
2097 if (status < 0) {
2098 mlog_errno(status);
2099 goto bail;
2100 }
2101
2102 /* launch vote thread */
78427043 2103 osb->vote_task = kthread_run(ocfs2_vote_thread, osb, "ocfs2vote");
ccd979bd
MF		 2104 if (IS_ERR(osb->vote_task)) {
2105 status = PTR_ERR(osb->vote_task);
2106 osb->vote_task = NULL;
2107 mlog_errno(status);
2108 goto bail;
2109 }
2110
2111 /* used by the dlm code to make message headers unique, each
2112 * node in this domain must agree on this. */
2113 dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str));
2114
2115 /* for now, uuid == domain */
2116 dlm = dlm_register_domain(osb->uuid_str, dlm_key);
2117 if (IS_ERR(dlm)) {
2118 status = PTR_ERR(dlm);
2119 mlog_errno(status);
2120 goto bail;
2121 }
2122
2123 ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
2124 ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
2125
2126 dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb);
2127
2128 osb->dlm = dlm;
2129
2130 status = 0;
2131bail:
2132 if (status < 0) {
2133 ocfs2_dlm_shutdown_debug(osb);
2134 if (osb->vote_task)
2135 kthread_stop(osb->vote_task);
2136 }
2137
2138 mlog_exit(status);
2139 return status;
2140}
2141
2142void ocfs2_dlm_shutdown(struct ocfs2_super *osb)
2143{
2144 mlog_entry_void();
2145
2146 dlm_unregister_eviction_cb(&osb->osb_eviction_cb);
2147
2148 ocfs2_drop_osb_locks(osb);
2149
2150 if (osb->vote_task) {
2151 kthread_stop(osb->vote_task);
2152 osb->vote_task = NULL;
2153 }
2154
2155 ocfs2_lock_res_free(&osb->osb_super_lockres);
2156 ocfs2_lock_res_free(&osb->osb_rename_lockres);
2157
2158 dlm_unregister_domain(osb->dlm);
2159 osb->dlm = NULL;
2160
2161 ocfs2_dlm_shutdown_debug(osb);
2162
2163 mlog_exit_void();
2164}
2165
2a45f2d1 2166static void ocfs2_unlock_ast(void *opaque, enum dlm_status status)
ccd979bd
MF		 2167{
2168 struct ocfs2_lock_res *lockres = opaque;
2169 unsigned long flags;
2170
2171 mlog_entry_void();
2172
2173 mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name,
2174 lockres->l_unlock_action);
2175
2176 spin_lock_irqsave(&lockres->l_lock, flags);
2177 /* We tried to cancel a convert request, but it was already
2178 * granted. All we want to do here is clear our unlock
2179 * state. The wake_up call done at the bottom is redundant
2180 * (ocfs2_prepare_cancel_convert doesn't sleep on this) but doesn't
2181 * hurt anything anyway */
2182 if (status == DLM_CANCELGRANT &&
2183 lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
2184 mlog(0, "Got cancelgrant for %s\n", lockres->l_name);
2185
2186 /* We don't clear the busy flag in this case as it
2187 * should have been cleared by the ast which the dlm
2188 * has called. */
2189 goto complete_unlock;
2190 }
2191
2192 if (status != DLM_NORMAL) {
2193 mlog(ML_ERROR, "Dlm passes status %d for lock %s, "
2194 "unlock_action %d\n", status, lockres->l_name,
2195 lockres->l_unlock_action);
2196 spin_unlock_irqrestore(&lockres->l_lock, flags);
2197 return;
2198 }
2199
2200 switch(lockres->l_unlock_action) {
2201 case OCFS2_UNLOCK_CANCEL_CONVERT:
2202 mlog(0, "Cancel convert success for %s\n", lockres->l_name);
2203 lockres->l_action = OCFS2_AST_INVALID;
2204 break;
2205 case OCFS2_UNLOCK_DROP_LOCK:
2206 lockres->l_level = LKM_IVMODE;
2207 break;
2208 default:
2209 BUG();
2210 }
2211
2212 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
2213complete_unlock:
2214 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
2215 spin_unlock_irqrestore(&lockres->l_lock, flags);
2216
2217 wake_up(&lockres->l_event);
2218
2219 mlog_exit_void();
2220}
2221
2222typedef void (ocfs2_pre_drop_cb_t)(struct ocfs2_lock_res *, void *);
2223
2224struct drop_lock_cb {
2225 ocfs2_pre_drop_cb_t *drop_func;
2226 void *drop_data;
2227};
2228
2229static int ocfs2_drop_lock(struct ocfs2_super *osb,
2230 struct ocfs2_lock_res *lockres,
2231 struct drop_lock_cb *dcb)
2232{
2233 enum dlm_status status;
2234 unsigned long flags;
b80fc012 2235 int lkm_flags = 0;
ccd979bd
MF		 2236
2237 /* We didn't get anywhere near actually using this lockres. */
2238 if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED))
2239 goto out;
2240
b80fc012
MF		 2241 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
2242 lkm_flags |= LKM_VALBLK;
2243
ccd979bd
MF		 2244 spin_lock_irqsave(&lockres->l_lock, flags);
2245
2246 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING),
2247 "lockres %s, flags 0x%lx\n",
2248 lockres->l_name, lockres->l_flags);
2249
2250 while (lockres->l_flags & OCFS2_LOCK_BUSY) {
2251 mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = "
2252 "%u, unlock_action = %u\n",
2253 lockres->l_name, lockres->l_flags, lockres->l_action,
2254 lockres->l_unlock_action);
2255
2256 spin_unlock_irqrestore(&lockres->l_lock, flags);
2257
2258 /* XXX: Today we just wait on any busy
2259 * locks... Perhaps we need to cancel converts in the
2260 * future? */
2261 ocfs2_wait_on_busy_lock(lockres);
2262
2263 spin_lock_irqsave(&lockres->l_lock, flags);
2264 }
2265
2266 if (dcb)
2267 dcb->drop_func(lockres, dcb->drop_data);
2268
2269 if (lockres->l_flags & OCFS2_LOCK_BUSY)
2270 mlog(ML_ERROR, "destroying busy lock: \"%s\"\n",
2271 lockres->l_name);
2272 if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
2273 mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name);
2274
2275 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
2276 spin_unlock_irqrestore(&lockres->l_lock, flags);
2277 goto out;
2278 }
2279
2280 lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED);
2281
2282 /* make sure we never get here while waiting for an ast to
2283 * fire. */
2284 BUG_ON(lockres->l_action != OCFS2_AST_INVALID);
2285
2286 /* is this necessary? */
2287 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
2288 lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK;
2289 spin_unlock_irqrestore(&lockres->l_lock, flags);
2290
2291 mlog(0, "lock %s\n", lockres->l_name);
2292
b80fc012 2293 status = dlmunlock(osb->dlm, &lockres->l_lksb, lkm_flags,
2a45f2d1 2294 ocfs2_unlock_ast, lockres);
ccd979bd
MF		 2295 if (status != DLM_NORMAL) {
2296 ocfs2_log_dlm_error("dlmunlock", status, lockres);
2297 mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags);
2298 dlm_print_one_lock(lockres->l_lksb.lockid);
2299 BUG();
2300 }
2301 mlog(0, "lock %s, successfull return from dlmunlock\n",
2302 lockres->l_name);
2303
2304 ocfs2_wait_on_busy_lock(lockres);
2305out:
2306 mlog_exit(0);
2307 return 0;
2308}
2309
2310/* Mark the lockres as being dropped. It will no longer be
2311 * queued if blocking, but we still may have to wait on it
2312 * being dequeued from the vote thread before we can consider
2313 * it safe to drop.
2314 *
2315 * You can *not* attempt to call cluster_lock on this lockres anymore. */
2316void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres)
2317{
2318 int status;
2319 struct ocfs2_mask_waiter mw;
2320 unsigned long flags;
2321
2322 ocfs2_init_mask_waiter(&mw);
2323
2324 spin_lock_irqsave(&lockres->l_lock, flags);
2325 lockres->l_flags |= OCFS2_LOCK_FREEING;
2326 while (lockres->l_flags & OCFS2_LOCK_QUEUED) {
2327 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0);
2328 spin_unlock_irqrestore(&lockres->l_lock, flags);
2329
2330 mlog(0, "Waiting on lockres %s\n", lockres->l_name);
2331
2332 status = ocfs2_wait_for_mask(&mw);
2333 if (status)
2334 mlog_errno(status);
2335
2336 spin_lock_irqsave(&lockres->l_lock, flags);
2337 }
2338 spin_unlock_irqrestore(&lockres->l_lock, flags);
2339}
2340
d680efe9
MF		 2341void ocfs2_simple_drop_lockres(struct ocfs2_super *osb,
2342 struct ocfs2_lock_res *lockres)
ccd979bd 2343{
d680efe9 2344 int ret;
ccd979bd 2345
d680efe9
MF		 2346 ocfs2_mark_lockres_freeing(lockres);
2347 ret = ocfs2_drop_lock(osb, lockres, NULL);
2348 if (ret)
2349 mlog_errno(ret);
2350}
ccd979bd 2351
d680efe9
MF		 2352static void ocfs2_drop_osb_locks(struct ocfs2_super *osb)
2353{
2354 ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres);
2355 ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres);
ccd979bd
MF		 2356}
2357
2358static void ocfs2_meta_pre_drop(struct ocfs2_lock_res *lockres, void *data)
2359{
2360 struct inode *inode = data;
2361
2362 /* the metadata lock requires a bit more work as we have an
2363 * LVB to worry about. */
2364 if (lockres->l_flags & OCFS2_LOCK_ATTACHED &&
2365 lockres->l_level == LKM_EXMODE &&
2366 !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
2367 __ocfs2_stuff_meta_lvb(inode);
2368}
2369
2370int ocfs2_drop_inode_locks(struct inode *inode)
2371{
2372 int status, err;
2373 struct drop_lock_cb meta_dcb = { ocfs2_meta_pre_drop, inode, };
2374
2375 mlog_entry_void();
2376
2377 /* No need to call ocfs2_mark_lockres_freeing here -
2378 * ocfs2_clear_inode has done it for us. */
2379
2380 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2381 &OCFS2_I(inode)->ip_data_lockres,
2382 NULL);
2383 if (err < 0)
2384 mlog_errno(err);
2385
2386 status = err;
2387
2388 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2389 &OCFS2_I(inode)->ip_meta_lockres,
2390 &meta_dcb);
2391 if (err < 0)
2392 mlog_errno(err);
2393 if (err < 0 && !status)
2394 status = err;
2395
2396 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2397 &OCFS2_I(inode)->ip_rw_lockres,
2398 NULL);
2399 if (err < 0)
2400 mlog_errno(err);
2401 if (err < 0 && !status)
2402 status = err;
2403
2404 mlog_exit(status);
2405 return status;
2406}
2407
2408static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
2409 int new_level)
2410{
2411 assert_spin_locked(&lockres->l_lock);
2412
2413 BUG_ON(lockres->l_blocking <= LKM_NLMODE);
2414
2415 if (lockres->l_level <= new_level) {
2416 mlog(ML_ERROR, "lockres->l_level (%u) <= new_level (%u)\n",
2417 lockres->l_level, new_level);
2418 BUG();
2419 }
2420
2421 mlog(0, "lock %s, new_level = %d, l_blocking = %d\n",
2422 lockres->l_name, new_level, lockres->l_blocking);
2423
2424 lockres->l_action = OCFS2_AST_DOWNCONVERT;
2425 lockres->l_requested = new_level;
2426 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
2427}
2428
2429static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
2430 struct ocfs2_lock_res *lockres,
2431 int new_level,
2432 int lvb)
2433{
2434 int ret, dlm_flags = LKM_CONVERT;
2435 enum dlm_status status;
2436
2437 mlog_entry_void();
2438
2439 if (lvb)
2440 dlm_flags |= LKM_VALBLK;
2441
2442 status = dlmlock(osb->dlm,
2443 new_level,
2444 &lockres->l_lksb,
2445 dlm_flags,
2446 lockres->l_name,
f0681062 2447 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 2448 ocfs2_locking_ast,
ccd979bd 2449 lockres,
aa2623ad 2450 ocfs2_blocking_ast);
ccd979bd
MF		 2451 if (status != DLM_NORMAL) {
2452 ocfs2_log_dlm_error("dlmlock", status, lockres);
2453 ret = -EINVAL;
2454 ocfs2_recover_from_dlm_error(lockres, 1);
2455 goto bail;
2456 }
2457
2458 ret = 0;
2459bail:
2460 mlog_exit(ret);
2461 return ret;
2462}
2463
2464/* returns 1 when the caller should unlock and call dlmunlock */
2465static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
2466 struct ocfs2_lock_res *lockres)
2467{
2468 assert_spin_locked(&lockres->l_lock);
2469
2470 mlog_entry_void();
2471 mlog(0, "lock %s\n", lockres->l_name);
2472
2473 if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
2474 /* If we're already trying to cancel a lock conversion
2475 * then just drop the spinlock and allow the caller to
2476 * requeue this lock. */
2477
2478 mlog(0, "Lockres %s, skip convert\n", lockres->l_name);
2479 return 0;
2480 }
2481
2482 /* were we in a convert when we got the bast fire? */
2483 BUG_ON(lockres->l_action != OCFS2_AST_CONVERT &&
2484 lockres->l_action != OCFS2_AST_DOWNCONVERT);
2485 /* set things up for the unlockast to know to just
2486 * clear out the ast_action and unset busy, etc. */
2487 lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT;
2488
2489 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY),
2490 "lock %s, invalid flags: 0x%lx\n",
2491 lockres->l_name, lockres->l_flags);
2492
2493 return 1;
2494}
2495
2496static int ocfs2_cancel_convert(struct ocfs2_super *osb,
2497 struct ocfs2_lock_res *lockres)
2498{
2499 int ret;
2500 enum dlm_status status;
2501
2502 mlog_entry_void();
2503 mlog(0, "lock %s\n", lockres->l_name);
2504
2505 ret = 0;
2506 status = dlmunlock(osb->dlm,
2507 &lockres->l_lksb,
2508 LKM_CANCEL,
2a45f2d1 2509 ocfs2_unlock_ast,
ccd979bd
MF		 2510 lockres);
2511 if (status != DLM_NORMAL) {
2512 ocfs2_log_dlm_error("dlmunlock", status, lockres);
2513 ret = -EINVAL;
2514 ocfs2_recover_from_dlm_error(lockres, 0);
2515 }
2516
2517 mlog(0, "lock %s return from dlmunlock\n", lockres->l_name);
2518
2519 mlog_exit(ret);
2520 return ret;
2521}
2522
2523static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode,
2524 struct ocfs2_lock_res *lockres,
2525 int new_level)
2526{
2527 int ret;
2528
2529 mlog_entry_void();
2530
2531 BUG_ON(new_level != LKM_NLMODE && new_level != LKM_PRMODE);
2532
2533 if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
2534 ret = 0;
2535 mlog(0, "lockres %s currently being refreshed -- backing "
2536 "off!\n", lockres->l_name);
2537 } else if (new_level == LKM_PRMODE)
2538 ret = !lockres->l_ex_holders &&
2539 ocfs2_inode_fully_checkpointed(inode);
2540 else /* Must be NLMODE we're converting to. */
2541 ret = !lockres->l_ro_holders && !lockres->l_ex_holders &&
2542 ocfs2_inode_fully_checkpointed(inode);
2543
2544 mlog_exit(ret);
2545 return ret;
2546}
2547
2548static int ocfs2_do_unblock_meta(struct inode *inode,
2549 int *requeue)
2550{
2551 int new_level;
2552 int set_lvb = 0;
2553 int ret = 0;
2554 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
2555 unsigned long flags;
2556
2557 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2558
2559 mlog_entry_void();
2560
2561 spin_lock_irqsave(&lockres->l_lock, flags);
2562
2563 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
2564
2565 mlog(0, "l_level=%d, l_blocking=%d\n", lockres->l_level,
2566 lockres->l_blocking);
2567
2568 BUG_ON(lockres->l_level != LKM_EXMODE &&
2569 lockres->l_level != LKM_PRMODE);
2570
2571 if (lockres->l_flags & OCFS2_LOCK_BUSY) {
2572 *requeue = 1;
2573 ret = ocfs2_prepare_cancel_convert(osb, lockres);
2574 spin_unlock_irqrestore(&lockres->l_lock, flags);
2575 if (ret) {
2576 ret = ocfs2_cancel_convert(osb, lockres);
2577 if (ret < 0)
2578 mlog_errno(ret);
2579 }
2580 goto leave;
2581 }
2582
2583 new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
2584
2585 mlog(0, "l_level=%d, l_blocking=%d, new_level=%d\n",
2586 lockres->l_level, lockres->l_blocking, new_level);
2587
2588 if (ocfs2_can_downconvert_meta_lock(inode, lockres, new_level)) {
2589 if (lockres->l_level == LKM_EXMODE)
2590 set_lvb = 1;
2591
2592 /* If the lock hasn't been refreshed yet (rare), then
2593 * our memory inode values are old and we skip
2594 * stuffing the lvb. There's no need to actually clear
2595 * out the lvb here as it's value is still valid. */
2596 if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
2597 if (set_lvb)
2598 __ocfs2_stuff_meta_lvb(inode);
2599 } else
2600 mlog(0, "lockres %s: downconverting stale lock!\n",
2601 lockres->l_name);
2602
2603 mlog(0, "calling ocfs2_downconvert_lock with l_level=%d, "
2604 "l_blocking=%d, new_level=%d\n",
2605 lockres->l_level, lockres->l_blocking, new_level);
2606
2607 ocfs2_prepare_downconvert(lockres, new_level);
2608 spin_unlock_irqrestore(&lockres->l_lock, flags);
2609 ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb);
2610 goto leave;
2611 }
2612 if (!ocfs2_inode_fully_checkpointed(inode))
2613 ocfs2_start_checkpoint(osb);
2614
2615 *requeue = 1;
2616 spin_unlock_irqrestore(&lockres->l_lock, flags);
2617 ret = 0;
2618leave:
2619 mlog_exit(ret);
2620 return ret;
2621}
2622
2623static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
2624 struct ocfs2_lock_res *lockres,
d680efe9 2625 struct ocfs2_unblock_ctl *ctl,
ccd979bd
MF		 2626 ocfs2_convert_worker_t *worker)
2627{
2628 unsigned long flags;
2629 int blocking;
2630 int new_level;
2631 int ret = 0;
2632
2633 mlog_entry_void();
2634
2635 spin_lock_irqsave(&lockres->l_lock, flags);
2636
2637 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
2638
2639recheck:
2640 if (lockres->l_flags & OCFS2_LOCK_BUSY) {
d680efe9 2641 ctl->requeue = 1;
ccd979bd
MF		 2642 ret = ocfs2_prepare_cancel_convert(osb, lockres);
2643 spin_unlock_irqrestore(&lockres->l_lock, flags);
2644 if (ret) {
2645 ret = ocfs2_cancel_convert(osb, lockres);
2646 if (ret < 0)
2647 mlog_errno(ret);
2648 }
2649 goto leave;
2650 }
2651
2652 /* if we're blocking an exclusive and we have *any* holders,
2653 * then requeue. */
2654 if ((lockres->l_blocking == LKM_EXMODE)
f7fbfdd1
MF		 2655 && (lockres->l_ex_holders || lockres->l_ro_holders))
2656 goto leave_requeue;
ccd979bd
MF		 2657
2658 /* If it's a PR we're blocking, then only
2659 * requeue if we've got any EX holders */
2660 if (lockres->l_blocking == LKM_PRMODE &&
f7fbfdd1
MF		 2661 lockres->l_ex_holders)
2662 goto leave_requeue;
2663
2664 /*
2665 * Can we get a lock in this state if the holder counts are
2666 * zero? The meta data unblock code used to check this.
2667 */
2668 if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
2669 && (lockres->l_flags & OCFS2_LOCK_REFRESHING))
2670 goto leave_requeue;
ccd979bd 2671
16d5b956
MF		 2672 new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
2673
2674 if (lockres->l_ops->check_downconvert
2675 && !lockres->l_ops->check_downconvert(lockres, new_level))
2676 goto leave_requeue;
2677
ccd979bd
MF		 2678 /* If we get here, then we know that there are no more
2679 * incompatible holders (and anyone asking for an incompatible
2680 * lock is blocked). We can now downconvert the lock */
2681 if (!worker)
2682 goto downconvert;
2683
2684 /* Some lockres types want to do a bit of work before
2685 * downconverting a lock. Allow that here. The worker function
2686 * may sleep, so we save off a copy of what we're blocking as
2687 * it may change while we're not holding the spin lock. */
2688 blocking = lockres->l_blocking;
2689 spin_unlock_irqrestore(&lockres->l_lock, flags);
2690
d680efe9
MF		 2691 ctl->unblock_action = worker(lockres, blocking);
2692
2693 if (ctl->unblock_action == UNBLOCK_STOP_POST)
2694 goto leave;
ccd979bd
MF		 2695
2696 spin_lock_irqsave(&lockres->l_lock, flags);
2697 if (blocking != lockres->l_blocking) {
2698 /* If this changed underneath us, then we can't drop
2699 * it just yet. */
2700 goto recheck;
2701 }
2702
2703downconvert:
d680efe9 2704 ctl->requeue = 0;
ccd979bd
MF		 2705
2706 ocfs2_prepare_downconvert(lockres, new_level);
2707 spin_unlock_irqrestore(&lockres->l_lock, flags);
2708 ret = ocfs2_downconvert_lock(osb, lockres, new_level, 0);
2709leave:
2710 mlog_exit(ret);
2711 return ret;
f7fbfdd1
MF		 2712
2713leave_requeue:
2714 spin_unlock_irqrestore(&lockres->l_lock, flags);
2715 ctl->requeue = 1;
2716
2717 mlog_exit(0);
2718 return 0;
ccd979bd
MF		 2719}
2720
d680efe9
MF		 2721static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
2722 int blocking)
ccd979bd
MF		 2723{
2724 struct inode *inode;
2725 struct address_space *mapping;
2726
ccd979bd
MF		 2727 inode = ocfs2_lock_res_inode(lockres);
2728 mapping = inode->i_mapping;
2729
2730 if (filemap_fdatawrite(mapping)) {
b0697053
MF		 2731 mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",
2732 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd
MF		 2733 }
2734 sync_mapping_buffers(mapping);
2735 if (blocking == LKM_EXMODE) {
2736 truncate_inode_pages(mapping, 0);
2737 unmap_mapping_range(mapping, 0, 0, 0);
2738 } else {
2739 /* We only need to wait on the I/O if we're not also
2740 * truncating pages because truncate_inode_pages waits
2741 * for us above. We don't truncate pages if we're
2742 * blocking anything < EXMODE because we want to keep
2743 * them around in that case. */
2744 filemap_fdatawait(mapping);
2745 }
2746
d680efe9 2747 return UNBLOCK_CONTINUE;
ccd979bd
MF		 2748}
2749
2750int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
d680efe9 2751 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2752{
2753 int status;
2754 struct inode *inode;
2755 struct ocfs2_super *osb;
2756
2757 mlog_entry_void();
2758
2759 inode = ocfs2_lock_res_inode(lockres);
2760 osb = OCFS2_SB(inode->i_sb);
2761
b0697053
MF		 2762 mlog(0, "unblock inode %llu\n",
2763 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd 2764
d680efe9 2765 status = ocfs2_generic_unblock_lock(osb, lockres, ctl,
ccd979bd
MF		 2766 ocfs2_data_convert_worker);
2767 if (status < 0)
2768 mlog_errno(status);
2769
b0697053 2770 mlog(0, "inode %llu, requeue = %d\n",
d680efe9 2771 (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue);
ccd979bd
MF		 2772
2773 mlog_exit(status);
2774 return status;
2775}
2776
2777static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
d680efe9 2778 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2779{
2780 int status;
2781 struct inode *inode;
2782
2783 mlog_entry_void();
2784
2785 mlog(0, "Unblock lockres %s\n", lockres->l_name);
2786
2787 inode = ocfs2_lock_res_inode(lockres);
2788
2789 status = ocfs2_generic_unblock_lock(OCFS2_SB(inode->i_sb),
d680efe9 2790 lockres, ctl, NULL);
ccd979bd
MF		 2791 if (status < 0)
2792 mlog_errno(status);
2793
2794 mlog_exit(status);
2795 return status;
2796}
2797
d680efe9
MF		 2798static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres,
2799 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2800{
2801 int status;
2802 struct inode *inode;
2803
2804 mlog_entry_void();
2805
2806 inode = ocfs2_lock_res_inode(lockres);
2807
b0697053
MF		 2808 mlog(0, "unblock inode %llu\n",
2809 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd 2810
d680efe9 2811 status = ocfs2_do_unblock_meta(inode, &ctl->requeue);
ccd979bd
MF		 2812 if (status < 0)
2813 mlog_errno(status);
2814
b0697053 2815 mlog(0, "inode %llu, requeue = %d\n",
d680efe9 2816 (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue);
ccd979bd
MF		 2817
2818 mlog_exit(status);
2819 return status;
2820}
2821
d680efe9
MF		 2822/*
2823 * Does the final reference drop on our dentry lock. Right now this
2824 * happens in the vote thread, but we could choose to simplify the
2825 * dlmglue API and push these off to the ocfs2_wq in the future.
2826 */
2827static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
2828 struct ocfs2_lock_res *lockres)
2829{
2830 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2831 ocfs2_dentry_lock_put(osb, dl);
2832}
2833
2834/*
2835 * d_delete() matching dentries before the lock downconvert.
2836 *
2837 * At this point, any process waiting to destroy the
2838 * dentry_lock due to last ref count is stopped by the
2839 * OCFS2_LOCK_QUEUED flag.
2840 *
2841 * We have two potential problems
2842 *
2843 * 1) If we do the last reference drop on our dentry_lock (via dput)
2844 * we'll wind up in ocfs2_release_dentry_lock(), waiting on
2845 * the downconvert to finish. Instead we take an elevated
2846 * reference and push the drop until after we've completed our
2847 * unblock processing.
2848 *
2849 * 2) There might be another process with a final reference,
2850 * waiting on us to finish processing. If this is the case, we
2851 * detect it and exit out - there's no more dentries anyway.
2852 */
2853static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres,
2854 int blocking)
2855{
2856 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2857 struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode);
2858 struct dentry *dentry;
2859 unsigned long flags;
2860 int extra_ref = 0;
2861
2862 /*
2863 * This node is blocking another node from getting a read
2864 * lock. This happens when we've renamed within a
2865 * directory. We've forced the other nodes to d_delete(), but
2866 * we never actually dropped our lock because it's still
2867 * valid. The downconvert code will retain a PR for this node,
2868 * so there's no further work to do.
2869 */
2870 if (blocking == LKM_PRMODE)
2871 return UNBLOCK_CONTINUE;
2872
2873 /*
2874 * Mark this inode as potentially orphaned. The code in
2875 * ocfs2_delete_inode() will figure out whether it actually
2876 * needs to be freed or not.
2877 */
2878 spin_lock(&oi->ip_lock);
2879 oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
2880 spin_unlock(&oi->ip_lock);
2881
2882 /*
2883 * Yuck. We need to make sure however that the check of
2884 * OCFS2_LOCK_FREEING and the extra reference are atomic with
2885 * respect to a reference decrement or the setting of that
2886 * flag.
2887 */
2888 spin_lock_irqsave(&lockres->l_lock, flags);
2889 spin_lock(&dentry_attach_lock);
2890 if (!(lockres->l_flags & OCFS2_LOCK_FREEING)
2891 && dl->dl_count) {
2892 dl->dl_count++;
2893 extra_ref = 1;
2894 }
2895 spin_unlock(&dentry_attach_lock);
2896 spin_unlock_irqrestore(&lockres->l_lock, flags);
2897
2898 mlog(0, "extra_ref = %d\n", extra_ref);
2899
2900 /*
2901 * We have a process waiting on us in ocfs2_dentry_iput(),
2902 * which means we can't have any more outstanding
2903 * aliases. There's no need to do any more work.
2904 */
2905 if (!extra_ref)
2906 return UNBLOCK_CONTINUE;
2907
2908 spin_lock(&dentry_attach_lock);
2909 while (1) {
2910 dentry = ocfs2_find_local_alias(dl->dl_inode,
2911 dl->dl_parent_blkno, 1);
2912 if (!dentry)
2913 break;
2914 spin_unlock(&dentry_attach_lock);
2915
2916 mlog(0, "d_delete(%.*s);\n", dentry->d_name.len,
2917 dentry->d_name.name);
2918
2919 /*
2920 * The following dcache calls may do an
2921 * iput(). Normally we don't want that from the
2922 * downconverting thread, but in this case it's ok
2923 * because the requesting node already has an
2924 * exclusive lock on the inode, so it can't be queued
2925 * for a downconvert.
2926 */
2927 d_delete(dentry);
2928 dput(dentry);
2929
2930 spin_lock(&dentry_attach_lock);
2931 }
2932 spin_unlock(&dentry_attach_lock);
2933
2934 /*
2935 * If we are the last holder of this dentry lock, there is no
2936 * reason to downconvert so skip straight to the unlock.
2937 */
2938 if (dl->dl_count == 1)
2939 return UNBLOCK_STOP_POST;
2940
2941 return UNBLOCK_CONTINUE_POST;
2942}
2943
2944static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres,
2945 struct ocfs2_unblock_ctl *ctl)
2946{
2947 int ret;
2948 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2949 struct ocfs2_super *osb = OCFS2_SB(dl->dl_inode->i_sb);
2950
2951 mlog(0, "unblock dentry lock: %llu\n",
2952 (unsigned long long)OCFS2_I(dl->dl_inode)->ip_blkno);
2953
2954 ret = ocfs2_generic_unblock_lock(osb,
2955 lockres,
2956 ctl,
2957 ocfs2_dentry_convert_worker);
2958 if (ret < 0)
2959 mlog_errno(ret);
2960
2961 mlog(0, "requeue = %d, post = %d\n", ctl->requeue, ctl->unblock_action);
2962
2963 return ret;
2964}
2965
ccd979bd
MF		 2966/* Generic unblock function for any lockres whose private data is an
2967 * ocfs2_super pointer. */
2968static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
d680efe9 2969 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2970{
2971 int status;
2972 struct ocfs2_super *osb;
2973
2974 mlog_entry_void();
2975
2976 mlog(0, "Unblock lockres %s\n", lockres->l_name);
2977
aa2623ad 2978 osb = ocfs2_get_lockres_osb(lockres);
ccd979bd
MF		 2979
2980 status = ocfs2_generic_unblock_lock(osb,
2981 lockres,
d680efe9 2982 ctl,
ccd979bd
MF		 2983 NULL);
2984 if (status < 0)
2985 mlog_errno(status);
2986
2987 mlog_exit(status);
2988 return status;
2989}
2990
2991void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
2992 struct ocfs2_lock_res *lockres)
2993{
2994 int status;
d680efe9 2995 struct ocfs2_unblock_ctl ctl = {0, 0,};
ccd979bd
MF		 2996 unsigned long flags;
2997
2998 /* Our reference to the lockres in this function can be
2999 * considered valid until we remove the OCFS2_LOCK_QUEUED
3000 * flag. */
3001
3002 mlog_entry_void();
3003
3004 BUG_ON(!lockres);
3005 BUG_ON(!lockres->l_ops);
3006 BUG_ON(!lockres->l_ops->unblock);
3007
3008 mlog(0, "lockres %s blocked.\n", lockres->l_name);
3009
3010 /* Detect whether a lock has been marked as going away while
3011 * the vote thread was processing other things. A lock can
3012 * still be marked with OCFS2_LOCK_FREEING after this check,
3013 * but short circuiting here will still save us some
3014 * performance. */
3015 spin_lock_irqsave(&lockres->l_lock, flags);
3016 if (lockres->l_flags & OCFS2_LOCK_FREEING)
3017 goto unqueue;
3018 spin_unlock_irqrestore(&lockres->l_lock, flags);
3019
d680efe9 3020 status = lockres->l_ops->unblock(lockres, &ctl);
ccd979bd
MF		 3021 if (status < 0)
3022 mlog_errno(status);
3023
3024 spin_lock_irqsave(&lockres->l_lock, flags);
3025unqueue:
d680efe9 3026 if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) {
ccd979bd
MF		 3027 lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED);
3028 } else
3029 ocfs2_schedule_blocked_lock(osb, lockres);
3030
3031 mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name,
d680efe9 3032 ctl.requeue ? "yes" : "no");
ccd979bd
MF		 3033 spin_unlock_irqrestore(&lockres->l_lock, flags);
3034
d680efe9
MF		 3035 if (ctl.unblock_action != UNBLOCK_CONTINUE
3036 && lockres->l_ops->post_unlock)
3037 lockres->l_ops->post_unlock(osb, lockres);
3038
ccd979bd
MF		 3039 mlog_exit_void();
3040}
3041
3042static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
3043 struct ocfs2_lock_res *lockres)
3044{
3045 mlog_entry_void();
3046
3047 assert_spin_locked(&lockres->l_lock);
3048
3049 if (lockres->l_flags & OCFS2_LOCK_FREEING) {
3050 /* Do not schedule a lock for downconvert when it's on
3051 * the way to destruction - any nodes wanting access
3052 * to the resource will get it soon. */
3053 mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n",
3054 lockres->l_name, lockres->l_flags);
3055 return;
3056 }
3057
3058 lockres_or_flags(lockres, OCFS2_LOCK_QUEUED);
3059
3060 spin_lock(&osb->vote_task_lock);
3061 if (list_empty(&lockres->l_blocked_list)) {
3062 list_add_tail(&lockres->l_blocked_list,
3063 &osb->blocked_lock_list);
3064 osb->blocked_lock_count++;
3065 }
3066 spin_unlock(&osb->vote_task_lock);
3067
3068 mlog_exit_void();
3069}
3070
3071/* This aids in debugging situations where a bad LVB might be involved. */
3072void ocfs2_dump_meta_lvb_info(u64 level,
3073 const char *function,
3074 unsigned int line,
3075 struct ocfs2_lock_res *lockres)
3076{
3077 struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
3078
3079 mlog(level, "LVB information for %s (called from %s:%u):\n",
3080 lockres->l_name, function, line);
f9e2d82e
MF		 3081 mlog(level, "version: %u, clusters: %u, generation: 0x%x\n",
3082 lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters),
3083 be32_to_cpu(lvb->lvb_igeneration));
b0697053
MF		 3084 mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n",
3085 (unsigned long long)be64_to_cpu(lvb->lvb_isize),
3086 be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid),
3087 be16_to_cpu(lvb->lvb_imode));
3088 mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, "
ca4d147e 3089 "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink),
b0697053
MF		 3090 (long long)be64_to_cpu(lvb->lvb_iatime_packed),
3091 (long long)be64_to_cpu(lvb->lvb_ictime_packed),
ca4d147e
HP		 3092 (long long)be64_to_cpu(lvb->lvb_imtime_packed),
3093 be32_to_cpu(lvb->lvb_iattr));
ccd979bd 3094}
Linux kernel - Deliverables
This page took 0.22634 seconds and 5 git commands to generate.