projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: Remove unused dlmglue functions
[GitHub/mt8127/android_kernel_alcatel_ttab.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);
810d5aeb
MF		 95static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres,
96 int new_level);
97static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres);
98
ccd979bd 99static int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
d680efe9 100 struct ocfs2_unblock_ctl *ctl);
ccd979bd 101static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
d680efe9
MF		 102 struct ocfs2_unblock_ctl *ctl);
103static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres,
104 struct ocfs2_unblock_ctl *ctl);
ccd979bd 105static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
d680efe9
MF		 106 struct ocfs2_unblock_ctl *ctl);
107
108static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
109 struct ocfs2_lock_res *lockres);
ccd979bd 110
f625c979
MF		 111/*
112 * OCFS2 Lock Resource Operations
113 *
114 * These fine tune the behavior of the generic dlmglue locking infrastructure.
115 */
ccd979bd 116struct ocfs2_lock_res_ops {
54a7e755
MF		 117 /*
118 * Translate an ocfs2_lock_res * into an ocfs2_super *. Define
119 * this callback if ->l_priv is not an ocfs2_super pointer
120 */
121 struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *);
d680efe9
MF		 122 int (*unblock)(struct ocfs2_lock_res *, struct ocfs2_unblock_ctl *);
123 void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *);
f625c979 124
16d5b956
MF		 125 /*
126 * Allow a lock type to add checks to determine whether it is
127 * safe to downconvert a lock. Return 0 to re-queue the
128 * downconvert at a later time, nonzero to continue.
129 *
130 * For most locks, the default checks that there are no
131 * incompatible holders are sufficient.
132 *
133 * Called with the lockres spinlock held.
134 */
135 int (*check_downconvert)(struct ocfs2_lock_res *, int);
136
5ef0d4ea
MF		 137 /*
138 * Allows a lock type to populate the lock value block. This
139 * is called on downconvert, and when we drop a lock.
140 *
141 * Locks that want to use this should set LOCK_TYPE_USES_LVB
142 * in the flags field.
143 *
144 * Called with the lockres spinlock held.
145 */
146 void (*set_lvb)(struct ocfs2_lock_res *);
147
f625c979
MF		 148 /*
149 * LOCK_TYPE_* flags which describe the specific requirements
150 * of a lock type. Descriptions of each individual flag follow.
151 */
152 int flags;
ccd979bd
MF		 153};
154
f625c979
MF		 155/*
156 * Some locks want to "refresh" potentially stale data when a
157 * meaningful (PRMODE or EXMODE) lock level is first obtained. If this
158 * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the
159 * individual lockres l_flags member from the ast function. It is
160 * expected that the locking wrapper will clear the
161 * OCFS2_LOCK_NEEDS_REFRESH flag when done.
162 */
163#define LOCK_TYPE_REQUIRES_REFRESH 0x1
164
b80fc012 165/*
5ef0d4ea
MF		 166 * Indicate that a lock type makes use of the lock value block. The
167 * ->set_lvb lock type callback must be defined.
b80fc012
MF		 168 */
169#define LOCK_TYPE_USES_LVB 0x2
170
d680efe9
MF		 171typedef int (ocfs2_convert_worker_t)(struct ocfs2_lock_res *, int);
172static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
173 struct ocfs2_lock_res *lockres,
174 struct ocfs2_unblock_ctl *ctl,
175 ocfs2_convert_worker_t *worker);
176
ccd979bd 177static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = {
54a7e755 178 .get_osb = ocfs2_get_inode_osb,
ccd979bd 179 .unblock = ocfs2_unblock_inode_lock,
f625c979 180 .flags = 0,
ccd979bd
MF		 181};
182
183static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = {
54a7e755 184 .get_osb = ocfs2_get_inode_osb,
ccd979bd 185 .unblock = ocfs2_unblock_meta,
810d5aeb
MF		 186 .check_downconvert = ocfs2_check_meta_downconvert,
187 .set_lvb = ocfs2_set_meta_lvb,
b80fc012 188 .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB,
ccd979bd
MF		 189};
190
ccd979bd 191static struct ocfs2_lock_res_ops ocfs2_inode_data_lops = {
54a7e755 192 .get_osb = ocfs2_get_inode_osb,
ccd979bd 193 .unblock = ocfs2_unblock_data,
f625c979 194 .flags = 0,
ccd979bd
MF		 195};
196
197static struct ocfs2_lock_res_ops ocfs2_super_lops = {
ccd979bd 198 .unblock = ocfs2_unblock_osb_lock,
f625c979 199 .flags = LOCK_TYPE_REQUIRES_REFRESH,
ccd979bd
MF		 200};
201
202static struct ocfs2_lock_res_ops ocfs2_rename_lops = {
ccd979bd 203 .unblock = ocfs2_unblock_osb_lock,
f625c979 204 .flags = 0,
ccd979bd
MF		 205};
206
d680efe9 207static struct ocfs2_lock_res_ops ocfs2_dentry_lops = {
54a7e755 208 .get_osb = ocfs2_get_dentry_osb,
d680efe9
MF		 209 .unblock = ocfs2_unblock_dentry_lock,
210 .post_unlock = ocfs2_dentry_post_unlock,
f625c979 211 .flags = 0,
d680efe9
MF		 212};
213
ccd979bd
MF		 214static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
215{
216 return lockres->l_type == OCFS2_LOCK_TYPE_META ||
217 lockres->l_type == OCFS2_LOCK_TYPE_DATA ||
218 lockres->l_type == OCFS2_LOCK_TYPE_RW;
219}
220
ccd979bd
MF		 221static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres)
222{
223 BUG_ON(!ocfs2_is_inode_lock(lockres));
224
225 return (struct inode *) lockres->l_priv;
226}
227
d680efe9
MF		 228static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres)
229{
230 BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY);
231
232 return (struct ocfs2_dentry_lock *)lockres->l_priv;
233}
234
54a7e755
MF		 235static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres)
236{
237 if (lockres->l_ops->get_osb)
238 return lockres->l_ops->get_osb(lockres);
239
240 return (struct ocfs2_super *)lockres->l_priv;
241}
242
ccd979bd
MF		 243static int ocfs2_lock_create(struct ocfs2_super *osb,
244 struct ocfs2_lock_res *lockres,
245 int level,
246 int dlm_flags);
247static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
248 int wanted);
249static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
250 struct ocfs2_lock_res *lockres,
251 int level);
252static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres);
253static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres);
254static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres);
255static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level);
256static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
257 struct ocfs2_lock_res *lockres);
258static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
259 int convert);
260#define ocfs2_log_dlm_error(_func, _stat, _lockres) do { \
261 mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on " \
262 "resource %s: %s\n", dlm_errname(_stat), _func, \
263 _lockres->l_name, dlm_errmsg(_stat)); \
264} while (0)
265static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
266 struct ocfs2_lock_res *lockres);
267static int ocfs2_meta_lock_update(struct inode *inode,
268 struct buffer_head **bh);
269static void ocfs2_drop_osb_locks(struct ocfs2_super *osb);
270static inline int ocfs2_highest_compat_lock_level(int level);
ccd979bd 271
ccd979bd
MF		 272static void ocfs2_build_lock_name(enum ocfs2_lock_type type,
273 u64 blkno,
274 u32 generation,
275 char *name)
276{
277 int len;
278
279 mlog_entry_void();
280
281 BUG_ON(type >= OCFS2_NUM_LOCK_TYPES);
282
b0697053
MF		 283 len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x",
284 ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD,
285 (long long)blkno, generation);
ccd979bd
MF		 286
287 BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1));
288
289 mlog(0, "built lock resource with name: %s\n", name);
290
291 mlog_exit_void();
292}
293
34af946a 294static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock);
ccd979bd
MF		 295
296static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res,
297 struct ocfs2_dlm_debug *dlm_debug)
298{
299 mlog(0, "Add tracking for lockres %s\n", res->l_name);
300
301 spin_lock(&ocfs2_dlm_tracking_lock);
302 list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking);
303 spin_unlock(&ocfs2_dlm_tracking_lock);
304}
305
306static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res)
307{
308 spin_lock(&ocfs2_dlm_tracking_lock);
309 if (!list_empty(&res->l_debug_list))
310 list_del_init(&res->l_debug_list);
311 spin_unlock(&ocfs2_dlm_tracking_lock);
312}
313
314static void ocfs2_lock_res_init_common(struct ocfs2_super *osb,
315 struct ocfs2_lock_res *res,
316 enum ocfs2_lock_type type,
ccd979bd
MF		 317 struct ocfs2_lock_res_ops *ops,
318 void *priv)
319{
ccd979bd
MF		 320 res->l_type = type;
321 res->l_ops = ops;
322 res->l_priv = priv;
323
324 res->l_level = LKM_IVMODE;
325 res->l_requested = LKM_IVMODE;
326 res->l_blocking = LKM_IVMODE;
327 res->l_action = OCFS2_AST_INVALID;
328 res->l_unlock_action = OCFS2_UNLOCK_INVALID;
329
330 res->l_flags = OCFS2_LOCK_INITIALIZED;
331
332 ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug);
333}
334
335void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res)
336{
337 /* This also clears out the lock status block */
338 memset(res, 0, sizeof(struct ocfs2_lock_res));
339 spin_lock_init(&res->l_lock);
340 init_waitqueue_head(&res->l_event);
341 INIT_LIST_HEAD(&res->l_blocked_list);
342 INIT_LIST_HEAD(&res->l_mask_waiters);
343}
344
345void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
346 enum ocfs2_lock_type type,
24c19ef4 347 unsigned int generation,
ccd979bd
MF		 348 struct inode *inode)
349{
350 struct ocfs2_lock_res_ops *ops;
351
352 switch(type) {
353 case OCFS2_LOCK_TYPE_RW:
354 ops = &ocfs2_inode_rw_lops;
355 break;
356 case OCFS2_LOCK_TYPE_META:
357 ops = &ocfs2_inode_meta_lops;
358 break;
359 case OCFS2_LOCK_TYPE_DATA:
360 ops = &ocfs2_inode_data_lops;
361 break;
362 default:
363 mlog_bug_on_msg(1, "type: %d\n", type);
364 ops = NULL; /* thanks, gcc */
365 break;
366 };
367
d680efe9 368 ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno,
24c19ef4 369 generation, res->l_name);
d680efe9
MF		 370 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode);
371}
372
54a7e755
MF		 373static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres)
374{
375 struct inode *inode = ocfs2_lock_res_inode(lockres);
376
377 return OCFS2_SB(inode->i_sb);
378}
379
d680efe9
MF		 380static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres)
381{
382 __be64 inode_blkno_be;
383
384 memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START],
385 sizeof(__be64));
386
387 return be64_to_cpu(inode_blkno_be);
388}
389
54a7e755
MF		 390static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres)
391{
392 struct ocfs2_dentry_lock *dl = lockres->l_priv;
393
394 return OCFS2_SB(dl->dl_inode->i_sb);
395}
396
d680efe9
MF		 397void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
398 u64 parent, struct inode *inode)
399{
400 int len;
401 u64 inode_blkno = OCFS2_I(inode)->ip_blkno;
402 __be64 inode_blkno_be = cpu_to_be64(inode_blkno);
403 struct ocfs2_lock_res *lockres = &dl->dl_lockres;
404
405 ocfs2_lock_res_init_once(lockres);
406
407 /*
408 * Unfortunately, the standard lock naming scheme won't work
409 * here because we have two 16 byte values to use. Instead,
410 * we'll stuff the inode number as a binary value. We still
411 * want error prints to show something without garbling the
412 * display, so drop a null byte in there before the inode
413 * number. A future version of OCFS2 will likely use all
414 * binary lock names. The stringified names have been a
415 * tremendous aid in debugging, but now that the debugfs
416 * interface exists, we can mangle things there if need be.
417 *
418 * NOTE: We also drop the standard "pad" value (the total lock
419 * name size stays the same though - the last part is all
420 * zeros due to the memset in ocfs2_lock_res_init_once()
421 */
422 len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START,
423 "%c%016llx",
424 ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY),
425 (long long)parent);
426
427 BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1));
428
429 memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be,
430 sizeof(__be64));
431
432 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
433 OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops,
434 dl);
ccd979bd
MF		 435}
436
437static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res,
438 struct ocfs2_super *osb)
439{
440 /* Superblock lockres doesn't come from a slab so we call init
441 * once on it manually. */
442 ocfs2_lock_res_init_once(res);
d680efe9
MF		 443 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO,
444 0, res->l_name);
ccd979bd 445 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER,
ccd979bd
MF		 446 &ocfs2_super_lops, osb);
447}
448
449static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
450 struct ocfs2_super *osb)
451{
452 /* Rename lockres doesn't come from a slab so we call init
453 * once on it manually. */
454 ocfs2_lock_res_init_once(res);
d680efe9
MF		 455 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name);
456 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME,
ccd979bd
MF		 457 &ocfs2_rename_lops, osb);
458}
459
460void ocfs2_lock_res_free(struct ocfs2_lock_res *res)
461{
462 mlog_entry_void();
463
464 if (!(res->l_flags & OCFS2_LOCK_INITIALIZED))
465 return;
466
467 ocfs2_remove_lockres_tracking(res);
468
469 mlog_bug_on_msg(!list_empty(&res->l_blocked_list),
470 "Lockres %s is on the blocked list\n",
471 res->l_name);
472 mlog_bug_on_msg(!list_empty(&res->l_mask_waiters),
473 "Lockres %s has mask waiters pending\n",
474 res->l_name);
475 mlog_bug_on_msg(spin_is_locked(&res->l_lock),
476 "Lockres %s is locked\n",
477 res->l_name);
478 mlog_bug_on_msg(res->l_ro_holders,
479 "Lockres %s has %u ro holders\n",
480 res->l_name, res->l_ro_holders);
481 mlog_bug_on_msg(res->l_ex_holders,
482 "Lockres %s has %u ex holders\n",
483 res->l_name, res->l_ex_holders);
484
485 /* Need to clear out the lock status block for the dlm */
486 memset(&res->l_lksb, 0, sizeof(res->l_lksb));
487
488 res->l_flags = 0UL;
489 mlog_exit_void();
490}
491
492static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres,
493 int level)
494{
495 mlog_entry_void();
496
497 BUG_ON(!lockres);
498
499 switch(level) {
500 case LKM_EXMODE:
501 lockres->l_ex_holders++;
502 break;
503 case LKM_PRMODE:
504 lockres->l_ro_holders++;
505 break;
506 default:
507 BUG();
508 }
509
510 mlog_exit_void();
511}
512
513static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres,
514 int level)
515{
516 mlog_entry_void();
517
518 BUG_ON(!lockres);
519
520 switch(level) {
521 case LKM_EXMODE:
522 BUG_ON(!lockres->l_ex_holders);
523 lockres->l_ex_holders--;
524 break;
525 case LKM_PRMODE:
526 BUG_ON(!lockres->l_ro_holders);
527 lockres->l_ro_holders--;
528 break;
529 default:
530 BUG();
531 }
532 mlog_exit_void();
533}
534
535/* WARNING: This function lives in a world where the only three lock
536 * levels are EX, PR, and NL. It *will* have to be adjusted when more
537 * lock types are added. */
538static inline int ocfs2_highest_compat_lock_level(int level)
539{
540 int new_level = LKM_EXMODE;
541
542 if (level == LKM_EXMODE)
543 new_level = LKM_NLMODE;
544 else if (level == LKM_PRMODE)
545 new_level = LKM_PRMODE;
546 return new_level;
547}
548
549static void lockres_set_flags(struct ocfs2_lock_res *lockres,
550 unsigned long newflags)
551{
552 struct list_head *pos, *tmp;
553 struct ocfs2_mask_waiter *mw;
554
555 assert_spin_locked(&lockres->l_lock);
556
557 lockres->l_flags = newflags;
558
559 list_for_each_safe(pos, tmp, &lockres->l_mask_waiters) {
560 mw = list_entry(pos, struct ocfs2_mask_waiter, mw_item);
561 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
562 continue;
563
564 list_del_init(&mw->mw_item);
565 mw->mw_status = 0;
566 complete(&mw->mw_complete);
567 }
568}
569static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or)
570{
571 lockres_set_flags(lockres, lockres->l_flags | or);
572}
573static void lockres_clear_flags(struct ocfs2_lock_res *lockres,
574 unsigned long clear)
575{
576 lockres_set_flags(lockres, lockres->l_flags & ~clear);
577}
578
579static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres)
580{
581 mlog_entry_void();
582
583 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
584 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
585 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
586 BUG_ON(lockres->l_blocking <= LKM_NLMODE);
587
588 lockres->l_level = lockres->l_requested;
589 if (lockres->l_level <=
590 ocfs2_highest_compat_lock_level(lockres->l_blocking)) {
591 lockres->l_blocking = LKM_NLMODE;
592 lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
593 }
594 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
595
596 mlog_exit_void();
597}
598
599static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres)
600{
601 mlog_entry_void();
602
603 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
604 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
605
606 /* Convert from RO to EX doesn't really need anything as our
607 * information is already up to data. Convert from NL to
608 * *anything* however should mark ourselves as needing an
609 * update */
f625c979
MF		 610 if (lockres->l_level == LKM_NLMODE &&
611 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
ccd979bd
MF		 612 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
613
614 lockres->l_level = lockres->l_requested;
615 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
616
617 mlog_exit_void();
618}
619
620static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres)
621{
622 mlog_entry_void();
623
624 BUG_ON((!lockres->l_flags & OCFS2_LOCK_BUSY));
625 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
626
627 if (lockres->l_requested > LKM_NLMODE &&
f625c979
MF		 628 !(lockres->l_flags & OCFS2_LOCK_LOCAL) &&
629 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
ccd979bd
MF		 630 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
631
632 lockres->l_level = lockres->l_requested;
633 lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED);
634 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
635
636 mlog_exit_void();
637}
638
ccd979bd
MF		 639static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
640 int level)
641{
642 int needs_downconvert = 0;
643 mlog_entry_void();
644
645 assert_spin_locked(&lockres->l_lock);
646
647 lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
648
649 if (level > lockres->l_blocking) {
650 /* only schedule a downconvert if we haven't already scheduled
651 * one that goes low enough to satisfy the level we're
652 * blocking. this also catches the case where we get
653 * duplicate BASTs */
654 if (ocfs2_highest_compat_lock_level(level) <
655 ocfs2_highest_compat_lock_level(lockres->l_blocking))
656 needs_downconvert = 1;
657
658 lockres->l_blocking = level;
659 }
660
661 mlog_exit(needs_downconvert);
662 return needs_downconvert;
663}
664
aa2623ad 665static void ocfs2_blocking_ast(void *opaque, int level)
ccd979bd 666{
aa2623ad
MF		 667 struct ocfs2_lock_res *lockres = opaque;
668 struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
ccd979bd
MF		 669 int needs_downconvert;
670 unsigned long flags;
671
ccd979bd
MF		 672 BUG_ON(level <= LKM_NLMODE);
673
aa2623ad
MF		 674 mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n",
675 lockres->l_name, level, lockres->l_level,
676 ocfs2_lock_type_string(lockres->l_type));
677
ccd979bd
MF		 678 spin_lock_irqsave(&lockres->l_lock, flags);
679 needs_downconvert = ocfs2_generic_handle_bast(lockres, level);
680 if (needs_downconvert)
681 ocfs2_schedule_blocked_lock(osb, lockres);
682 spin_unlock_irqrestore(&lockres->l_lock, flags);
683
d680efe9
MF		 684 wake_up(&lockres->l_event);
685
ccd979bd 686 ocfs2_kick_vote_thread(osb);
ccd979bd
MF		 687}
688
e92d57df 689static void ocfs2_locking_ast(void *opaque)
ccd979bd 690{
e92d57df 691 struct ocfs2_lock_res *lockres = opaque;
ccd979bd
MF		 692 struct dlm_lockstatus *lksb = &lockres->l_lksb;
693 unsigned long flags;
694
695 spin_lock_irqsave(&lockres->l_lock, flags);
696
697 if (lksb->status != DLM_NORMAL) {
698 mlog(ML_ERROR, "lockres %s: lksb status value of %u!\n",
699 lockres->l_name, lksb->status);
700 spin_unlock_irqrestore(&lockres->l_lock, flags);
701 return;
702 }
703
704 switch(lockres->l_action) {
705 case OCFS2_AST_ATTACH:
706 ocfs2_generic_handle_attach_action(lockres);
e92d57df 707 lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL);
ccd979bd
MF		 708 break;
709 case OCFS2_AST_CONVERT:
710 ocfs2_generic_handle_convert_action(lockres);
711 break;
712 case OCFS2_AST_DOWNCONVERT:
713 ocfs2_generic_handle_downconvert_action(lockres);
714 break;
715 default:
e92d57df
MF		 716 mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u "
717 "lockres flags = 0x%lx, unlock action: %u\n",
718 lockres->l_name, lockres->l_action, lockres->l_flags,
719 lockres->l_unlock_action);
ccd979bd
MF		 720 BUG();
721 }
722
ccd979bd
MF		 723 /* set it to something invalid so if we get called again we
724 * can catch it. */
725 lockres->l_action = OCFS2_AST_INVALID;
ccd979bd
MF		 726
727 wake_up(&lockres->l_event);
d680efe9 728 spin_unlock_irqrestore(&lockres->l_lock, flags);
ccd979bd
MF		 729}
730
ccd979bd
MF		 731static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
732 int convert)
733{
734 unsigned long flags;
735
736 mlog_entry_void();
737 spin_lock_irqsave(&lockres->l_lock, flags);
738 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
739 if (convert)
740 lockres->l_action = OCFS2_AST_INVALID;
741 else
742 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
743 spin_unlock_irqrestore(&lockres->l_lock, flags);
744
745 wake_up(&lockres->l_event);
746 mlog_exit_void();
747}
748
749/* Note: If we detect another process working on the lock (i.e.,
750 * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller
751 * to do the right thing in that case.
752 */
753static int ocfs2_lock_create(struct ocfs2_super *osb,
754 struct ocfs2_lock_res *lockres,
755 int level,
756 int dlm_flags)
757{
758 int ret = 0;
759 enum dlm_status status;
760 unsigned long flags;
761
762 mlog_entry_void();
763
764 mlog(0, "lock %s, level = %d, flags = %d\n", lockres->l_name, level,
765 dlm_flags);
766
767 spin_lock_irqsave(&lockres->l_lock, flags);
768 if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) ||
769 (lockres->l_flags & OCFS2_LOCK_BUSY)) {
770 spin_unlock_irqrestore(&lockres->l_lock, flags);
771 goto bail;
772 }
773
774 lockres->l_action = OCFS2_AST_ATTACH;
775 lockres->l_requested = level;
776 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
777 spin_unlock_irqrestore(&lockres->l_lock, flags);
778
779 status = dlmlock(osb->dlm,
780 level,
781 &lockres->l_lksb,
782 dlm_flags,
783 lockres->l_name,
f0681062 784 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 785 ocfs2_locking_ast,
ccd979bd 786 lockres,
aa2623ad 787 ocfs2_blocking_ast);
ccd979bd
MF		 788 if (status != DLM_NORMAL) {
789 ocfs2_log_dlm_error("dlmlock", status, lockres);
790 ret = -EINVAL;
791 ocfs2_recover_from_dlm_error(lockres, 1);
792 }
793
794 mlog(0, "lock %s, successfull return from dlmlock\n", lockres->l_name);
795
796bail:
797 mlog_exit(ret);
798 return ret;
799}
800
801static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres,
802 int flag)
803{
804 unsigned long flags;
805 int ret;
806
807 spin_lock_irqsave(&lockres->l_lock, flags);
808 ret = lockres->l_flags & flag;
809 spin_unlock_irqrestore(&lockres->l_lock, flags);
810
811 return ret;
812}
813
814static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres)
815
816{
817 wait_event(lockres->l_event,
818 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY));
819}
820
821static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres)
822
823{
824 wait_event(lockres->l_event,
825 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING));
826}
827
828/* predict what lock level we'll be dropping down to on behalf
829 * of another node, and return true if the currently wanted
830 * level will be compatible with it. */
831static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
832 int wanted)
833{
834 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
835
836 return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking);
837}
838
839static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw)
840{
841 INIT_LIST_HEAD(&mw->mw_item);
842 init_completion(&mw->mw_complete);
843}
844
845static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw)
846{
847 wait_for_completion(&mw->mw_complete);
848 /* Re-arm the completion in case we want to wait on it again */
849 INIT_COMPLETION(mw->mw_complete);
850 return mw->mw_status;
851}
852
853static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres,
854 struct ocfs2_mask_waiter *mw,
855 unsigned long mask,
856 unsigned long goal)
857{
858 BUG_ON(!list_empty(&mw->mw_item));
859
860 assert_spin_locked(&lockres->l_lock);
861
862 list_add_tail(&mw->mw_item, &lockres->l_mask_waiters);
863 mw->mw_mask = mask;
864 mw->mw_goal = goal;
865}
866
867/* returns 0 if the mw that was removed was already satisfied, -EBUSY
868 * if the mask still hadn't reached its goal */
869static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
870 struct ocfs2_mask_waiter *mw)
871{
872 unsigned long flags;
873 int ret = 0;
874
875 spin_lock_irqsave(&lockres->l_lock, flags);
876 if (!list_empty(&mw->mw_item)) {
877 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
878 ret = -EBUSY;
879
880 list_del_init(&mw->mw_item);
881 init_completion(&mw->mw_complete);
882 }
883 spin_unlock_irqrestore(&lockres->l_lock, flags);
884
885 return ret;
886
887}
888
889static int ocfs2_cluster_lock(struct ocfs2_super *osb,
890 struct ocfs2_lock_res *lockres,
891 int level,
892 int lkm_flags,
893 int arg_flags)
894{
895 struct ocfs2_mask_waiter mw;
896 enum dlm_status status;
897 int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR);
898 int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */
899 unsigned long flags;
900
901 mlog_entry_void();
902
903 ocfs2_init_mask_waiter(&mw);
904
b80fc012
MF		 905 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
906 lkm_flags |= LKM_VALBLK;
907
ccd979bd
MF		 908again:
909 wait = 0;
910
911 if (catch_signals && signal_pending(current)) {
912 ret = -ERESTARTSYS;
913 goto out;
914 }
915
916 spin_lock_irqsave(&lockres->l_lock, flags);
917
918 mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING,
919 "Cluster lock called on freeing lockres %s! flags "
920 "0x%lx\n", lockres->l_name, lockres->l_flags);
921
922 /* We only compare against the currently granted level
923 * here. If the lock is blocked waiting on a downconvert,
924 * we'll get caught below. */
925 if (lockres->l_flags & OCFS2_LOCK_BUSY &&
926 level > lockres->l_level) {
927 /* is someone sitting in dlm_lock? If so, wait on
928 * them. */
929 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
930 wait = 1;
931 goto unlock;
932 }
933
934 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
935 /* lock has not been created yet. */
936 spin_unlock_irqrestore(&lockres->l_lock, flags);
937
938 ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0);
939 if (ret < 0) {
940 mlog_errno(ret);
941 goto out;
942 }
943 goto again;
944 }
945
946 if (lockres->l_flags & OCFS2_LOCK_BLOCKED &&
947 !ocfs2_may_continue_on_blocked_lock(lockres, level)) {
948 /* is the lock is currently blocked on behalf of
949 * another node */
950 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0);
951 wait = 1;
952 goto unlock;
953 }
954
955 if (level > lockres->l_level) {
956 if (lockres->l_action != OCFS2_AST_INVALID)
957 mlog(ML_ERROR, "lockres %s has action %u pending\n",
958 lockres->l_name, lockres->l_action);
959
960 lockres->l_action = OCFS2_AST_CONVERT;
961 lockres->l_requested = level;
962 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
963 spin_unlock_irqrestore(&lockres->l_lock, flags);
964
965 BUG_ON(level == LKM_IVMODE);
966 BUG_ON(level == LKM_NLMODE);
967
968 mlog(0, "lock %s, convert from %d to level = %d\n",
969 lockres->l_name, lockres->l_level, level);
970
971 /* call dlm_lock to upgrade lock now */
972 status = dlmlock(osb->dlm,
973 level,
974 &lockres->l_lksb,
b80fc012 975 lkm_flags|LKM_CONVERT,
ccd979bd 976 lockres->l_name,
f0681062 977 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 978 ocfs2_locking_ast,
ccd979bd 979 lockres,
aa2623ad 980 ocfs2_blocking_ast);
ccd979bd
MF		 981 if (status != DLM_NORMAL) {
982 if ((lkm_flags & LKM_NOQUEUE) &&
983 (status == DLM_NOTQUEUED))
984 ret = -EAGAIN;
985 else {
986 ocfs2_log_dlm_error("dlmlock", status,
987 lockres);
988 ret = -EINVAL;
989 }
990 ocfs2_recover_from_dlm_error(lockres, 1);
991 goto out;
992 }
993
994 mlog(0, "lock %s, successfull return from dlmlock\n",
995 lockres->l_name);
996
997 /* At this point we've gone inside the dlm and need to
998 * complete our work regardless. */
999 catch_signals = 0;
1000
1001 /* wait for busy to clear and carry on */
1002 goto again;
1003 }
1004
1005 /* Ok, if we get here then we're good to go. */
1006 ocfs2_inc_holders(lockres, level);
1007
1008 ret = 0;
1009unlock:
1010 spin_unlock_irqrestore(&lockres->l_lock, flags);
1011out:
1012 /*
1013 * This is helping work around a lock inversion between the page lock
1014 * and dlm locks. One path holds the page lock while calling aops
1015 * which block acquiring dlm locks. The voting thread holds dlm
1016 * locks while acquiring page locks while down converting data locks.
1017 * This block is helping an aop path notice the inversion and back
1018 * off to unlock its page lock before trying the dlm lock again.
1019 */
1020 if (wait && arg_flags & OCFS2_LOCK_NONBLOCK &&
1021 mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) {
1022 wait = 0;
1023 if (lockres_remove_mask_waiter(lockres, &mw))
1024 ret = -EAGAIN;
1025 else
1026 goto again;
1027 }
1028 if (wait) {
1029 ret = ocfs2_wait_for_mask(&mw);
1030 if (ret == 0)
1031 goto again;
1032 mlog_errno(ret);
1033 }
1034
1035 mlog_exit(ret);
1036 return ret;
1037}
1038
1039static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
1040 struct ocfs2_lock_res *lockres,
1041 int level)
1042{
1043 unsigned long flags;
1044
1045 mlog_entry_void();
1046 spin_lock_irqsave(&lockres->l_lock, flags);
1047 ocfs2_dec_holders(lockres, level);
1048 ocfs2_vote_on_unlock(osb, lockres);
1049 spin_unlock_irqrestore(&lockres->l_lock, flags);
1050 mlog_exit_void();
1051}
1052
d680efe9
MF		 1053int ocfs2_create_new_lock(struct ocfs2_super *osb,
1054 struct ocfs2_lock_res *lockres,
24c19ef4
MF		 1055 int ex,
1056 int local)
ccd979bd 1057{
d680efe9 1058 int level = ex ? LKM_EXMODE : LKM_PRMODE;
ccd979bd 1059 unsigned long flags;
24c19ef4 1060 int lkm_flags = local ? LKM_LOCAL : 0;
ccd979bd
MF		 1061
1062 spin_lock_irqsave(&lockres->l_lock, flags);
1063 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
1064 lockres_or_flags(lockres, OCFS2_LOCK_LOCAL);
1065 spin_unlock_irqrestore(&lockres->l_lock, flags);
1066
24c19ef4 1067 return ocfs2_lock_create(osb, lockres, level, lkm_flags);
ccd979bd
MF		 1068}
1069
1070/* Grants us an EX lock on the data and metadata resources, skipping
1071 * the normal cluster directory lookup. Use this ONLY on newly created
1072 * inodes which other nodes can't possibly see, and which haven't been
1073 * hashed in the inode hash yet. This can give us a good performance
1074 * increase as it'll skip the network broadcast normally associated
1075 * with creating a new lock resource. */
1076int ocfs2_create_new_inode_locks(struct inode *inode)
1077{
1078 int ret;
d680efe9 1079 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
ccd979bd
MF		 1080
1081 BUG_ON(!inode);
1082 BUG_ON(!ocfs2_inode_is_new(inode));
1083
1084 mlog_entry_void();
1085
b0697053 1086 mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd
MF		 1087
1088 /* NOTE: That we don't increment any of the holder counts, nor
1089 * do we add anything to a journal handle. Since this is
1090 * supposed to be a new inode which the cluster doesn't know
1091 * about yet, there is no need to. As far as the LVB handling
1092 * is concerned, this is basically like acquiring an EX lock
1093 * on a resource which has an invalid one -- we'll set it
1094 * valid when we release the EX. */
1095
24c19ef4 1096 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1);
ccd979bd
MF		 1097 if (ret) {
1098 mlog_errno(ret);
1099 goto bail;
1100 }
1101
24c19ef4
MF		 1102 /*
1103 * We don't want to use LKM_LOCAL on a meta data lock as they
1104 * don't use a generation in their lock names.
1105 */
1106 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_meta_lockres, 1, 0);
ccd979bd
MF		 1107 if (ret) {
1108 mlog_errno(ret);
1109 goto bail;
1110 }
1111
24c19ef4 1112 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_data_lockres, 1, 1);
ccd979bd
MF		 1113 if (ret) {
1114 mlog_errno(ret);
1115 goto bail;
1116 }
1117
1118bail:
1119 mlog_exit(ret);
1120 return ret;
1121}
1122
1123int ocfs2_rw_lock(struct inode *inode, int write)
1124{
1125 int status, level;
1126 struct ocfs2_lock_res *lockres;
1127
1128 BUG_ON(!inode);
1129
1130 mlog_entry_void();
1131
b0697053
MF		 1132 mlog(0, "inode %llu take %s RW lock\n",
1133 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1134 write ? "EXMODE" : "PRMODE");
1135
1136 lockres = &OCFS2_I(inode)->ip_rw_lockres;
1137
1138 level = write ? LKM_EXMODE : LKM_PRMODE;
1139
1140 status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0,
1141 0);
1142 if (status < 0)
1143 mlog_errno(status);
1144
1145 mlog_exit(status);
1146 return status;
1147}
1148
1149void ocfs2_rw_unlock(struct inode *inode, int write)
1150{
1151 int level = write ? LKM_EXMODE : LKM_PRMODE;
1152 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres;
1153
1154 mlog_entry_void();
1155
b0697053
MF		 1156 mlog(0, "inode %llu drop %s RW lock\n",
1157 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1158 write ? "EXMODE" : "PRMODE");
1159
1160 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1161
1162 mlog_exit_void();
1163}
1164
1165int ocfs2_data_lock_full(struct inode *inode,
1166 int write,
1167 int arg_flags)
1168{
1169 int status = 0, level;
1170 struct ocfs2_lock_res *lockres;
1171
1172 BUG_ON(!inode);
1173
1174 mlog_entry_void();
1175
b0697053
MF		 1176 mlog(0, "inode %llu take %s DATA lock\n",
1177 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1178 write ? "EXMODE" : "PRMODE");
1179
1180 /* We'll allow faking a readonly data lock for
1181 * rodevices. */
1182 if (ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) {
1183 if (write) {
1184 status = -EROFS;
1185 mlog_errno(status);
1186 }
1187 goto out;
1188 }
1189
1190 lockres = &OCFS2_I(inode)->ip_data_lockres;
1191
1192 level = write ? LKM_EXMODE : LKM_PRMODE;
1193
1194 status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level,
1195 0, arg_flags);
1196 if (status < 0 && status != -EAGAIN)
1197 mlog_errno(status);
1198
1199out:
1200 mlog_exit(status);
1201 return status;
1202}
1203
1204/* see ocfs2_meta_lock_with_page() */
1205int ocfs2_data_lock_with_page(struct inode *inode,
1206 int write,
1207 struct page *page)
1208{
1209 int ret;
1210
1211 ret = ocfs2_data_lock_full(inode, write, OCFS2_LOCK_NONBLOCK);
1212 if (ret == -EAGAIN) {
1213 unlock_page(page);
1214 if (ocfs2_data_lock(inode, write) == 0)
1215 ocfs2_data_unlock(inode, write);
1216 ret = AOP_TRUNCATED_PAGE;
1217 }
1218
1219 return ret;
1220}
1221
1222static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
1223 struct ocfs2_lock_res *lockres)
1224{
1225 int kick = 0;
1226
1227 mlog_entry_void();
1228
1229 /* If we know that another node is waiting on our lock, kick
1230 * the vote thread * pre-emptively when we reach a release
1231 * condition. */
1232 if (lockres->l_flags & OCFS2_LOCK_BLOCKED) {
1233 switch(lockres->l_blocking) {
1234 case LKM_EXMODE:
1235 if (!lockres->l_ex_holders && !lockres->l_ro_holders)
1236 kick = 1;
1237 break;
1238 case LKM_PRMODE:
1239 if (!lockres->l_ex_holders)
1240 kick = 1;
1241 break;
1242 default:
1243 BUG();
1244 }
1245 }
1246
1247 if (kick)
1248 ocfs2_kick_vote_thread(osb);
1249
1250 mlog_exit_void();
1251}
1252
1253void ocfs2_data_unlock(struct inode *inode,
1254 int write)
1255{
1256 int level = write ? LKM_EXMODE : LKM_PRMODE;
1257 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres;
1258
1259 mlog_entry_void();
1260
b0697053
MF		 1261 mlog(0, "inode %llu drop %s DATA lock\n",
1262 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1263 write ? "EXMODE" : "PRMODE");
1264
1265 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
1266 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1267
1268 mlog_exit_void();
1269}
1270
1271#define OCFS2_SEC_BITS 34
1272#define OCFS2_SEC_SHIFT (64 - 34)
1273#define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1)
1274
1275/* LVB only has room for 64 bits of time here so we pack it for
1276 * now. */
1277static u64 ocfs2_pack_timespec(struct timespec *spec)
1278{
1279 u64 res;
1280 u64 sec = spec->tv_sec;
1281 u32 nsec = spec->tv_nsec;
1282
1283 res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK);
1284
1285 return res;
1286}
1287
1288/* Call this with the lockres locked. I am reasonably sure we don't
1289 * need ip_lock in this function as anyone who would be changing those
1290 * values is supposed to be blocked in ocfs2_meta_lock right now. */
1291static void __ocfs2_stuff_meta_lvb(struct inode *inode)
1292{
1293 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1294 struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
1295 struct ocfs2_meta_lvb *lvb;
1296
1297 mlog_entry_void();
1298
1299 lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1300
24c19ef4
MF		 1301 /*
1302 * Invalidate the LVB of a deleted inode - this way other
1303 * nodes are forced to go to disk and discover the new inode
1304 * status.
1305 */
1306 if (oi->ip_flags & OCFS2_INODE_DELETED) {
1307 lvb->lvb_version = 0;
1308 goto out;
1309 }
1310
4d3b83f7 1311 lvb->lvb_version = OCFS2_LVB_VERSION;
ccd979bd
MF		 1312 lvb->lvb_isize = cpu_to_be64(i_size_read(inode));
1313 lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters);
1314 lvb->lvb_iuid = cpu_to_be32(inode->i_uid);
1315 lvb->lvb_igid = cpu_to_be32(inode->i_gid);
1316 lvb->lvb_imode = cpu_to_be16(inode->i_mode);
1317 lvb->lvb_inlink = cpu_to_be16(inode->i_nlink);
1318 lvb->lvb_iatime_packed =
1319 cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime));
1320 lvb->lvb_ictime_packed =
1321 cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime));
1322 lvb->lvb_imtime_packed =
1323 cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime));
ca4d147e 1324 lvb->lvb_iattr = cpu_to_be32(oi->ip_attr);
f9e2d82e 1325 lvb->lvb_igeneration = cpu_to_be32(inode->i_generation);
ccd979bd 1326
24c19ef4 1327out:
ccd979bd
MF		 1328 mlog_meta_lvb(0, lockres);
1329
1330 mlog_exit_void();
1331}
1332
1333static void ocfs2_unpack_timespec(struct timespec *spec,
1334 u64 packed_time)
1335{
1336 spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT;
1337 spec->tv_nsec = packed_time & OCFS2_NSEC_MASK;
1338}
1339
1340static void ocfs2_refresh_inode_from_lvb(struct inode *inode)
1341{
1342 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1343 struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
1344 struct ocfs2_meta_lvb *lvb;
1345
1346 mlog_entry_void();
1347
1348 mlog_meta_lvb(0, lockres);
1349
1350 lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1351
1352 /* We're safe here without the lockres lock... */
1353 spin_lock(&oi->ip_lock);
1354 oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters);
1355 i_size_write(inode, be64_to_cpu(lvb->lvb_isize));
1356
ca4d147e
HP		 1357 oi->ip_attr = be32_to_cpu(lvb->lvb_iattr);
1358 ocfs2_set_inode_flags(inode);
1359
ccd979bd
MF		 1360 /* fast-symlinks are a special case */
1361 if (S_ISLNK(inode->i_mode) && !oi->ip_clusters)
1362 inode->i_blocks = 0;
1363 else
1364 inode->i_blocks =
1365 ocfs2_align_bytes_to_sectors(i_size_read(inode));
1366
1367 inode->i_uid = be32_to_cpu(lvb->lvb_iuid);
1368 inode->i_gid = be32_to_cpu(lvb->lvb_igid);
1369 inode->i_mode = be16_to_cpu(lvb->lvb_imode);
1370 inode->i_nlink = be16_to_cpu(lvb->lvb_inlink);
1371 ocfs2_unpack_timespec(&inode->i_atime,
1372 be64_to_cpu(lvb->lvb_iatime_packed));
1373 ocfs2_unpack_timespec(&inode->i_mtime,
1374 be64_to_cpu(lvb->lvb_imtime_packed));
1375 ocfs2_unpack_timespec(&inode->i_ctime,
1376 be64_to_cpu(lvb->lvb_ictime_packed));
1377 spin_unlock(&oi->ip_lock);
1378
1379 mlog_exit_void();
1380}
1381
f9e2d82e
MF		 1382static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode,
1383 struct ocfs2_lock_res *lockres)
ccd979bd
MF		 1384{
1385 struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
1386
f9e2d82e
MF		 1387 if (lvb->lvb_version == OCFS2_LVB_VERSION
1388 && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation)
ccd979bd
MF		 1389 return 1;
1390 return 0;
1391}
1392
1393/* Determine whether a lock resource needs to be refreshed, and
1394 * arbitrate who gets to refresh it.
1395 *
1396 * 0 means no refresh needed.
1397 *
1398 * > 0 means you need to refresh this and you MUST call
1399 * ocfs2_complete_lock_res_refresh afterwards. */
1400static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres)
1401{
1402 unsigned long flags;
1403 int status = 0;
1404
1405 mlog_entry_void();
1406
1407refresh_check:
1408 spin_lock_irqsave(&lockres->l_lock, flags);
1409 if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
1410 spin_unlock_irqrestore(&lockres->l_lock, flags);
1411 goto bail;
1412 }
1413
1414 if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
1415 spin_unlock_irqrestore(&lockres->l_lock, flags);
1416
1417 ocfs2_wait_on_refreshing_lock(lockres);
1418 goto refresh_check;
1419 }
1420
1421 /* Ok, I'll be the one to refresh this lock. */
1422 lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING);
1423 spin_unlock_irqrestore(&lockres->l_lock, flags);
1424
1425 status = 1;
1426bail:
1427 mlog_exit(status);
1428 return status;
1429}
1430
1431/* If status is non zero, I'll mark it as not being in refresh
1432 * anymroe, but i won't clear the needs refresh flag. */
1433static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres,
1434 int status)
1435{
1436 unsigned long flags;
1437 mlog_entry_void();
1438
1439 spin_lock_irqsave(&lockres->l_lock, flags);
1440 lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING);
1441 if (!status)
1442 lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
1443 spin_unlock_irqrestore(&lockres->l_lock, flags);
1444
1445 wake_up(&lockres->l_event);
1446
1447 mlog_exit_void();
1448}
1449
1450/* may or may not return a bh if it went to disk. */
1451static int ocfs2_meta_lock_update(struct inode *inode,
1452 struct buffer_head **bh)
1453{
1454 int status = 0;
1455 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1456 struct ocfs2_lock_res *lockres;
1457 struct ocfs2_dinode *fe;
1458
1459 mlog_entry_void();
1460
1461 spin_lock(&oi->ip_lock);
1462 if (oi->ip_flags & OCFS2_INODE_DELETED) {
b0697053 1463 mlog(0, "Orphaned inode %llu was deleted while we "
ccd979bd 1464 "were waiting on a lock. ip_flags = 0x%x\n",
b0697053 1465 (unsigned long long)oi->ip_blkno, oi->ip_flags);
ccd979bd
MF		 1466 spin_unlock(&oi->ip_lock);
1467 status = -ENOENT;
1468 goto bail;
1469 }
1470 spin_unlock(&oi->ip_lock);
1471
1472 lockres = &oi->ip_meta_lockres;
1473
1474 if (!ocfs2_should_refresh_lock_res(lockres))
1475 goto bail;
1476
1477 /* This will discard any caching information we might have had
1478 * for the inode metadata. */
1479 ocfs2_metadata_cache_purge(inode);
1480
1481 /* will do nothing for inode types that don't use the extent
1482 * map (directories, bitmap files, etc) */
1483 ocfs2_extent_map_trunc(inode, 0);
1484
f9e2d82e 1485 if (ocfs2_meta_lvb_is_trustable(inode, lockres)) {
b0697053
MF		 1486 mlog(0, "Trusting LVB on inode %llu\n",
1487 (unsigned long long)oi->ip_blkno);
ccd979bd
MF		 1488 ocfs2_refresh_inode_from_lvb(inode);
1489 } else {
1490 /* Boo, we have to go to disk. */
1491 /* read bh, cast, ocfs2_refresh_inode */
1492 status = ocfs2_read_block(OCFS2_SB(inode->i_sb), oi->ip_blkno,
1493 bh, OCFS2_BH_CACHED, inode);
1494 if (status < 0) {
1495 mlog_errno(status);
1496 goto bail_refresh;
1497 }
1498 fe = (struct ocfs2_dinode *) (*bh)->b_data;
1499
1500 /* This is a good chance to make sure we're not
1501 * locking an invalid object.
1502 *
1503 * We bug on a stale inode here because we checked
1504 * above whether it was wiped from disk. The wiping
1505 * node provides a guarantee that we receive that
1506 * message and can mark the inode before dropping any
1507 * locks associated with it. */
1508 if (!OCFS2_IS_VALID_DINODE(fe)) {
1509 OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
1510 status = -EIO;
1511 goto bail_refresh;
1512 }
1513 mlog_bug_on_msg(inode->i_generation !=
1514 le32_to_cpu(fe->i_generation),
b0697053 1515 "Invalid dinode %llu disk generation: %u "
ccd979bd 1516 "inode->i_generation: %u\n",
b0697053
MF		 1517 (unsigned long long)oi->ip_blkno,
1518 le32_to_cpu(fe->i_generation),
ccd979bd
MF		 1519 inode->i_generation);
1520 mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) ||
1521 !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)),
b0697053
MF		 1522 "Stale dinode %llu dtime: %llu flags: 0x%x\n",
1523 (unsigned long long)oi->ip_blkno,
1524 (unsigned long long)le64_to_cpu(fe->i_dtime),
ccd979bd
MF		 1525 le32_to_cpu(fe->i_flags));
1526
1527 ocfs2_refresh_inode(inode, fe);
1528 }
1529
1530 status = 0;
1531bail_refresh:
1532 ocfs2_complete_lock_res_refresh(lockres, status);
1533bail:
1534 mlog_exit(status);
1535 return status;
1536}
1537
1538static int ocfs2_assign_bh(struct inode *inode,
1539 struct buffer_head **ret_bh,
1540 struct buffer_head *passed_bh)
1541{
1542 int status;
1543
1544 if (passed_bh) {
1545 /* Ok, the update went to disk for us, use the
1546 * returned bh. */
1547 *ret_bh = passed_bh;
1548 get_bh(*ret_bh);
1549
1550 return 0;
1551 }
1552
1553 status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
1554 OCFS2_I(inode)->ip_blkno,
1555 ret_bh,
1556 OCFS2_BH_CACHED,
1557 inode);
1558 if (status < 0)
1559 mlog_errno(status);
1560
1561 return status;
1562}
1563
1564/*
1565 * returns < 0 error if the callback will never be called, otherwise
1566 * the result of the lock will be communicated via the callback.
1567 */
1568int ocfs2_meta_lock_full(struct inode *inode,
1569 struct ocfs2_journal_handle *handle,
1570 struct buffer_head **ret_bh,
1571 int ex,
1572 int arg_flags)
1573{
1574 int status, level, dlm_flags, acquired;
1575 struct ocfs2_lock_res *lockres;
1576 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1577 struct buffer_head *local_bh = NULL;
1578
1579 BUG_ON(!inode);
1580
1581 mlog_entry_void();
1582
b0697053
MF		 1583 mlog(0, "inode %llu, take %s META lock\n",
1584 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1585 ex ? "EXMODE" : "PRMODE");
1586
1587 status = 0;
1588 acquired = 0;
1589 /* We'll allow faking a readonly metadata lock for
1590 * rodevices. */
1591 if (ocfs2_is_hard_readonly(osb)) {
1592 if (ex)
1593 status = -EROFS;
1594 goto bail;
1595 }
1596
1597 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
1598 wait_event(osb->recovery_event,
1599 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
1600
1601 acquired = 0;
1602 lockres = &OCFS2_I(inode)->ip_meta_lockres;
1603 level = ex ? LKM_EXMODE : LKM_PRMODE;
1604 dlm_flags = 0;
1605 if (arg_flags & OCFS2_META_LOCK_NOQUEUE)
1606 dlm_flags |= LKM_NOQUEUE;
1607
1608 status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags);
1609 if (status < 0) {
1610 if (status != -EAGAIN && status != -EIOCBRETRY)
1611 mlog_errno(status);
1612 goto bail;
1613 }
1614
1615 /* Notify the error cleanup path to drop the cluster lock. */
1616 acquired = 1;
1617
1618 /* We wait twice because a node may have died while we were in
1619 * the lower dlm layers. The second time though, we've
1620 * committed to owning this lock so we don't allow signals to
1621 * abort the operation. */
1622 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
1623 wait_event(osb->recovery_event,
1624 ocfs2_node_map_is_empty(osb, &osb->recovery_map));
1625
24c19ef4
MF		 1626 /*
1627 * We only see this flag if we're being called from
1628 * ocfs2_read_locked_inode(). It means we're locking an inode
1629 * which hasn't been populated yet, so clear the refresh flag
1630 * and let the caller handle it.
1631 */
1632 if (inode->i_state & I_NEW) {
1633 status = 0;
1634 ocfs2_complete_lock_res_refresh(lockres, 0);
1635 goto bail;
1636 }
1637
ccd979bd
MF		 1638 /* This is fun. The caller may want a bh back, or it may
1639 * not. ocfs2_meta_lock_update definitely wants one in, but
1640 * may or may not read one, depending on what's in the
1641 * LVB. The result of all of this is that we've *only* gone to
1642 * disk if we have to, so the complexity is worthwhile. */
1643 status = ocfs2_meta_lock_update(inode, &local_bh);
1644 if (status < 0) {
1645 if (status != -ENOENT)
1646 mlog_errno(status);
1647 goto bail;
1648 }
1649
1650 if (ret_bh) {
1651 status = ocfs2_assign_bh(inode, ret_bh, local_bh);
1652 if (status < 0) {
1653 mlog_errno(status);
1654 goto bail;
1655 }
1656 }
1657
1658 if (handle) {
1659 status = ocfs2_handle_add_lock(handle, inode);
1660 if (status < 0)
1661 mlog_errno(status);
1662 }
1663
1664bail:
1665 if (status < 0) {
1666 if (ret_bh && (*ret_bh)) {
1667 brelse(*ret_bh);
1668 *ret_bh = NULL;
1669 }
1670 if (acquired)
1671 ocfs2_meta_unlock(inode, ex);
1672 }
1673
1674 if (local_bh)
1675 brelse(local_bh);
1676
1677 mlog_exit(status);
1678 return status;
1679}
1680
1681/*
1682 * This is working around a lock inversion between tasks acquiring DLM locks
1683 * while holding a page lock and the vote thread which blocks dlm lock acquiry
1684 * while acquiring page locks.
1685 *
1686 * ** These _with_page variantes are only intended to be called from aop
1687 * methods that hold page locks and return a very specific *positive* error
1688 * code that aop methods pass up to the VFS -- test for errors with != 0. **
1689 *
1690 * The DLM is called such that it returns -EAGAIN if it would have blocked
1691 * waiting for the vote thread. In that case we unlock our page so the vote
1692 * thread can make progress. Once we've done this we have to return
1693 * AOP_TRUNCATED_PAGE so the aop method that called us can bubble that back up
1694 * into the VFS who will then immediately retry the aop call.
1695 *
1696 * We do a blocking lock and immediate unlock before returning, though, so that
1697 * the lock has a great chance of being cached on this node by the time the VFS
1698 * calls back to retry the aop. This has a potential to livelock as nodes
1699 * ping locks back and forth, but that's a risk we're willing to take to avoid
1700 * the lock inversion simply.
1701 */
1702int ocfs2_meta_lock_with_page(struct inode *inode,
1703 struct ocfs2_journal_handle *handle,
1704 struct buffer_head **ret_bh,
1705 int ex,
1706 struct page *page)
1707{
1708 int ret;
1709
1710 ret = ocfs2_meta_lock_full(inode, handle, ret_bh, ex,
1711 OCFS2_LOCK_NONBLOCK);
1712 if (ret == -EAGAIN) {
1713 unlock_page(page);
1714 if (ocfs2_meta_lock(inode, handle, ret_bh, ex) == 0)
1715 ocfs2_meta_unlock(inode, ex);
1716 ret = AOP_TRUNCATED_PAGE;
1717 }
1718
1719 return ret;
1720}
1721
1722void ocfs2_meta_unlock(struct inode *inode,
1723 int ex)
1724{
1725 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1726 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
1727
1728 mlog_entry_void();
1729
b0697053
MF		 1730 mlog(0, "inode %llu drop %s META lock\n",
1731 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ccd979bd
MF		 1732 ex ? "EXMODE" : "PRMODE");
1733
1734 if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
1735 ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
1736
1737 mlog_exit_void();
1738}
1739
1740int ocfs2_super_lock(struct ocfs2_super *osb,
1741 int ex)
1742{
1743 int status;
1744 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1745 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1746 struct buffer_head *bh;
1747 struct ocfs2_slot_info *si = osb->slot_info;
1748
1749 mlog_entry_void();
1750
1751 if (ocfs2_is_hard_readonly(osb))
1752 return -EROFS;
1753
1754 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
1755 if (status < 0) {
1756 mlog_errno(status);
1757 goto bail;
1758 }
1759
1760 /* The super block lock path is really in the best position to
1761 * know when resources covered by the lock need to be
1762 * refreshed, so we do it here. Of course, making sense of
1763 * everything is up to the caller :) */
1764 status = ocfs2_should_refresh_lock_res(lockres);
1765 if (status < 0) {
1766 mlog_errno(status);
1767 goto bail;
1768 }
1769 if (status) {
1770 bh = si->si_bh;
1771 status = ocfs2_read_block(osb, bh->b_blocknr, &bh, 0,
1772 si->si_inode);
1773 if (status == 0)
1774 ocfs2_update_slot_info(si);
1775
1776 ocfs2_complete_lock_res_refresh(lockres, status);
1777
1778 if (status < 0)
1779 mlog_errno(status);
1780 }
1781bail:
1782 mlog_exit(status);
1783 return status;
1784}
1785
1786void ocfs2_super_unlock(struct ocfs2_super *osb,
1787 int ex)
1788{
1789 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1790 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
1791
1792 ocfs2_cluster_unlock(osb, lockres, level);
1793}
1794
1795int ocfs2_rename_lock(struct ocfs2_super *osb)
1796{
1797 int status;
1798 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
1799
1800 if (ocfs2_is_hard_readonly(osb))
1801 return -EROFS;
1802
1803 status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0);
1804 if (status < 0)
1805 mlog_errno(status);
1806
1807 return status;
1808}
1809
1810void ocfs2_rename_unlock(struct ocfs2_super *osb)
1811{
1812 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
1813
1814 ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE);
1815}
1816
d680efe9
MF		 1817int ocfs2_dentry_lock(struct dentry *dentry, int ex)
1818{
1819 int ret;
1820 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1821 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
1822 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
1823
1824 BUG_ON(!dl);
1825
1826 if (ocfs2_is_hard_readonly(osb))
1827 return -EROFS;
1828
1829 ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0);
1830 if (ret < 0)
1831 mlog_errno(ret);
1832
1833 return ret;
1834}
1835
1836void ocfs2_dentry_unlock(struct dentry *dentry, int ex)
1837{
1838 int level = ex ? LKM_EXMODE : LKM_PRMODE;
1839 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
1840 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
1841
1842 ocfs2_cluster_unlock(osb, &dl->dl_lockres, level);
1843}
1844
ccd979bd
MF		 1845/* Reference counting of the dlm debug structure. We want this because
1846 * open references on the debug inodes can live on after a mount, so
1847 * we can't rely on the ocfs2_super to always exist. */
1848static void ocfs2_dlm_debug_free(struct kref *kref)
1849{
1850 struct ocfs2_dlm_debug *dlm_debug;
1851
1852 dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt);
1853
1854 kfree(dlm_debug);
1855}
1856
1857void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug)
1858{
1859 if (dlm_debug)
1860 kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free);
1861}
1862
1863static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug)
1864{
1865 kref_get(&debug->d_refcnt);
1866}
1867
1868struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void)
1869{
1870 struct ocfs2_dlm_debug *dlm_debug;
1871
1872 dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL);
1873 if (!dlm_debug) {
1874 mlog_errno(-ENOMEM);
1875 goto out;
1876 }
1877
1878 kref_init(&dlm_debug->d_refcnt);
1879 INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking);
1880 dlm_debug->d_locking_state = NULL;
1881out:
1882 return dlm_debug;
1883}
1884
1885/* Access to this is arbitrated for us via seq_file->sem. */
1886struct ocfs2_dlm_seq_priv {
1887 struct ocfs2_dlm_debug *p_dlm_debug;
1888 struct ocfs2_lock_res p_iter_res;
1889 struct ocfs2_lock_res p_tmp_res;
1890};
1891
1892static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start,
1893 struct ocfs2_dlm_seq_priv *priv)
1894{
1895 struct ocfs2_lock_res *iter, *ret = NULL;
1896 struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug;
1897
1898 assert_spin_locked(&ocfs2_dlm_tracking_lock);
1899
1900 list_for_each_entry(iter, &start->l_debug_list, l_debug_list) {
1901 /* discover the head of the list */
1902 if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) {
1903 mlog(0, "End of list found, %p\n", ret);
1904 break;
1905 }
1906
1907 /* We track our "dummy" iteration lockres' by a NULL
1908 * l_ops field. */
1909 if (iter->l_ops != NULL) {
1910 ret = iter;
1911 break;
1912 }
1913 }
1914
1915 return ret;
1916}
1917
1918static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos)
1919{
1920 struct ocfs2_dlm_seq_priv *priv = m->private;
1921 struct ocfs2_lock_res *iter;
1922
1923 spin_lock(&ocfs2_dlm_tracking_lock);
1924 iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv);
1925 if (iter) {
1926 /* Since lockres' have the lifetime of their container
1927 * (which can be inodes, ocfs2_supers, etc) we want to
1928 * copy this out to a temporary lockres while still
1929 * under the spinlock. Obviously after this we can't
1930 * trust any pointers on the copy returned, but that's
1931 * ok as the information we want isn't typically held
1932 * in them. */
1933 priv->p_tmp_res = *iter;
1934 iter = &priv->p_tmp_res;
1935 }
1936 spin_unlock(&ocfs2_dlm_tracking_lock);
1937
1938 return iter;
1939}
1940
1941static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v)
1942{
1943}
1944
1945static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos)
1946{
1947 struct ocfs2_dlm_seq_priv *priv = m->private;
1948 struct ocfs2_lock_res *iter = v;
1949 struct ocfs2_lock_res *dummy = &priv->p_iter_res;
1950
1951 spin_lock(&ocfs2_dlm_tracking_lock);
1952 iter = ocfs2_dlm_next_res(iter, priv);
1953 list_del_init(&dummy->l_debug_list);
1954 if (iter) {
1955 list_add(&dummy->l_debug_list, &iter->l_debug_list);
1956 priv->p_tmp_res = *iter;
1957 iter = &priv->p_tmp_res;
1958 }
1959 spin_unlock(&ocfs2_dlm_tracking_lock);
1960
1961 return iter;
1962}
1963
1964/* So that debugfs.ocfs2 can determine which format is being used */
1965#define OCFS2_DLM_DEBUG_STR_VERSION 1
1966static int ocfs2_dlm_seq_show(struct seq_file *m, void *v)
1967{
1968 int i;
1969 char *lvb;
1970 struct ocfs2_lock_res *lockres = v;
1971
1972 if (!lockres)
1973 return -EINVAL;
1974
d680efe9
MF		 1975 seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION);
1976
1977 if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY)
1978 seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1,
1979 lockres->l_name,
1980 (unsigned int)ocfs2_get_dentry_lock_ino(lockres));
1981 else
1982 seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name);
1983
1984 seq_printf(m, "%d\t"
ccd979bd
MF		 1985 "0x%lx\t"
1986 "0x%x\t"
1987 "0x%x\t"
1988 "%u\t"
1989 "%u\t"
1990 "%d\t"
1991 "%d\t",
ccd979bd
MF		 1992 lockres->l_level,
1993 lockres->l_flags,
1994 lockres->l_action,
1995 lockres->l_unlock_action,
1996 lockres->l_ro_holders,
1997 lockres->l_ex_holders,
1998 lockres->l_requested,
1999 lockres->l_blocking);
2000
2001 /* Dump the raw LVB */
2002 lvb = lockres->l_lksb.lvb;
2003 for(i = 0; i < DLM_LVB_LEN; i++)
2004 seq_printf(m, "0x%x\t", lvb[i]);
2005
2006 /* End the line */
2007 seq_printf(m, "\n");
2008 return 0;
2009}
2010
2011static struct seq_operations ocfs2_dlm_seq_ops = {
2012 .start = ocfs2_dlm_seq_start,
2013 .stop = ocfs2_dlm_seq_stop,
2014 .next = ocfs2_dlm_seq_next,
2015 .show = ocfs2_dlm_seq_show,
2016};
2017
2018static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file)
2019{
2020 struct seq_file *seq = (struct seq_file *) file->private_data;
2021 struct ocfs2_dlm_seq_priv *priv = seq->private;
2022 struct ocfs2_lock_res *res = &priv->p_iter_res;
2023
2024 ocfs2_remove_lockres_tracking(res);
2025 ocfs2_put_dlm_debug(priv->p_dlm_debug);
2026 return seq_release_private(inode, file);
2027}
2028
2029static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file)
2030{
2031 int ret;
2032 struct ocfs2_dlm_seq_priv *priv;
2033 struct seq_file *seq;
2034 struct ocfs2_super *osb;
2035
2036 priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL);
2037 if (!priv) {
2038 ret = -ENOMEM;
2039 mlog_errno(ret);
2040 goto out;
2041 }
2042 osb = (struct ocfs2_super *) inode->u.generic_ip;
2043 ocfs2_get_dlm_debug(osb->osb_dlm_debug);
2044 priv->p_dlm_debug = osb->osb_dlm_debug;
2045 INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list);
2046
2047 ret = seq_open(file, &ocfs2_dlm_seq_ops);
2048 if (ret) {
2049 kfree(priv);
2050 mlog_errno(ret);
2051 goto out;
2052 }
2053
2054 seq = (struct seq_file *) file->private_data;
2055 seq->private = priv;
2056
2057 ocfs2_add_lockres_tracking(&priv->p_iter_res,
2058 priv->p_dlm_debug);
2059
2060out:
2061 return ret;
2062}
2063
4b6f5d20 2064static const struct file_operations ocfs2_dlm_debug_fops = {
ccd979bd
MF		 2065 .open = ocfs2_dlm_debug_open,
2066 .release = ocfs2_dlm_debug_release,
2067 .read = seq_read,
2068 .llseek = seq_lseek,
2069};
2070
2071static int ocfs2_dlm_init_debug(struct ocfs2_super *osb)
2072{
2073 int ret = 0;
2074 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
2075
2076 dlm_debug->d_locking_state = debugfs_create_file("locking_state",
2077 S_IFREG|S_IRUSR,
2078 osb->osb_debug_root,
2079 osb,
2080 &ocfs2_dlm_debug_fops);
2081 if (!dlm_debug->d_locking_state) {
2082 ret = -EINVAL;
2083 mlog(ML_ERROR,
2084 "Unable to create locking state debugfs file.\n");
2085 goto out;
2086 }
2087
2088 ocfs2_get_dlm_debug(dlm_debug);
2089out:
2090 return ret;
2091}
2092
2093static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb)
2094{
2095 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
2096
2097 if (dlm_debug) {
2098 debugfs_remove(dlm_debug->d_locking_state);
2099 ocfs2_put_dlm_debug(dlm_debug);
2100 }
2101}
2102
2103int ocfs2_dlm_init(struct ocfs2_super *osb)
2104{
2105 int status;
2106 u32 dlm_key;
2107 struct dlm_ctxt *dlm;
2108
2109 mlog_entry_void();
2110
2111 status = ocfs2_dlm_init_debug(osb);
2112 if (status < 0) {
2113 mlog_errno(status);
2114 goto bail;
2115 }
2116
2117 /* launch vote thread */
78427043 2118 osb->vote_task = kthread_run(ocfs2_vote_thread, osb, "ocfs2vote");
ccd979bd
MF		 2119 if (IS_ERR(osb->vote_task)) {
2120 status = PTR_ERR(osb->vote_task);
2121 osb->vote_task = NULL;
2122 mlog_errno(status);
2123 goto bail;
2124 }
2125
2126 /* used by the dlm code to make message headers unique, each
2127 * node in this domain must agree on this. */
2128 dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str));
2129
2130 /* for now, uuid == domain */
2131 dlm = dlm_register_domain(osb->uuid_str, dlm_key);
2132 if (IS_ERR(dlm)) {
2133 status = PTR_ERR(dlm);
2134 mlog_errno(status);
2135 goto bail;
2136 }
2137
2138 ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
2139 ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
2140
2141 dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb);
2142
2143 osb->dlm = dlm;
2144
2145 status = 0;
2146bail:
2147 if (status < 0) {
2148 ocfs2_dlm_shutdown_debug(osb);
2149 if (osb->vote_task)
2150 kthread_stop(osb->vote_task);
2151 }
2152
2153 mlog_exit(status);
2154 return status;
2155}
2156
2157void ocfs2_dlm_shutdown(struct ocfs2_super *osb)
2158{
2159 mlog_entry_void();
2160
2161 dlm_unregister_eviction_cb(&osb->osb_eviction_cb);
2162
2163 ocfs2_drop_osb_locks(osb);
2164
2165 if (osb->vote_task) {
2166 kthread_stop(osb->vote_task);
2167 osb->vote_task = NULL;
2168 }
2169
2170 ocfs2_lock_res_free(&osb->osb_super_lockres);
2171 ocfs2_lock_res_free(&osb->osb_rename_lockres);
2172
2173 dlm_unregister_domain(osb->dlm);
2174 osb->dlm = NULL;
2175
2176 ocfs2_dlm_shutdown_debug(osb);
2177
2178 mlog_exit_void();
2179}
2180
2a45f2d1 2181static void ocfs2_unlock_ast(void *opaque, enum dlm_status status)
ccd979bd
MF		 2182{
2183 struct ocfs2_lock_res *lockres = opaque;
2184 unsigned long flags;
2185
2186 mlog_entry_void();
2187
2188 mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name,
2189 lockres->l_unlock_action);
2190
2191 spin_lock_irqsave(&lockres->l_lock, flags);
2192 /* We tried to cancel a convert request, but it was already
2193 * granted. All we want to do here is clear our unlock
2194 * state. The wake_up call done at the bottom is redundant
2195 * (ocfs2_prepare_cancel_convert doesn't sleep on this) but doesn't
2196 * hurt anything anyway */
2197 if (status == DLM_CANCELGRANT &&
2198 lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
2199 mlog(0, "Got cancelgrant for %s\n", lockres->l_name);
2200
2201 /* We don't clear the busy flag in this case as it
2202 * should have been cleared by the ast which the dlm
2203 * has called. */
2204 goto complete_unlock;
2205 }
2206
2207 if (status != DLM_NORMAL) {
2208 mlog(ML_ERROR, "Dlm passes status %d for lock %s, "
2209 "unlock_action %d\n", status, lockres->l_name,
2210 lockres->l_unlock_action);
2211 spin_unlock_irqrestore(&lockres->l_lock, flags);
2212 return;
2213 }
2214
2215 switch(lockres->l_unlock_action) {
2216 case OCFS2_UNLOCK_CANCEL_CONVERT:
2217 mlog(0, "Cancel convert success for %s\n", lockres->l_name);
2218 lockres->l_action = OCFS2_AST_INVALID;
2219 break;
2220 case OCFS2_UNLOCK_DROP_LOCK:
2221 lockres->l_level = LKM_IVMODE;
2222 break;
2223 default:
2224 BUG();
2225 }
2226
2227 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
2228complete_unlock:
2229 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
2230 spin_unlock_irqrestore(&lockres->l_lock, flags);
2231
2232 wake_up(&lockres->l_event);
2233
2234 mlog_exit_void();
2235}
2236
2237typedef void (ocfs2_pre_drop_cb_t)(struct ocfs2_lock_res *, void *);
2238
2239struct drop_lock_cb {
2240 ocfs2_pre_drop_cb_t *drop_func;
2241 void *drop_data;
2242};
2243
2244static int ocfs2_drop_lock(struct ocfs2_super *osb,
2245 struct ocfs2_lock_res *lockres,
2246 struct drop_lock_cb *dcb)
2247{
2248 enum dlm_status status;
2249 unsigned long flags;
b80fc012 2250 int lkm_flags = 0;
ccd979bd
MF		 2251
2252 /* We didn't get anywhere near actually using this lockres. */
2253 if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED))
2254 goto out;
2255
b80fc012
MF		 2256 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
2257 lkm_flags |= LKM_VALBLK;
2258
ccd979bd
MF		 2259 spin_lock_irqsave(&lockres->l_lock, flags);
2260
2261 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING),
2262 "lockres %s, flags 0x%lx\n",
2263 lockres->l_name, lockres->l_flags);
2264
2265 while (lockres->l_flags & OCFS2_LOCK_BUSY) {
2266 mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = "
2267 "%u, unlock_action = %u\n",
2268 lockres->l_name, lockres->l_flags, lockres->l_action,
2269 lockres->l_unlock_action);
2270
2271 spin_unlock_irqrestore(&lockres->l_lock, flags);
2272
2273 /* XXX: Today we just wait on any busy
2274 * locks... Perhaps we need to cancel converts in the
2275 * future? */
2276 ocfs2_wait_on_busy_lock(lockres);
2277
2278 spin_lock_irqsave(&lockres->l_lock, flags);
2279 }
2280
2281 if (dcb)
2282 dcb->drop_func(lockres, dcb->drop_data);
2283
2284 if (lockres->l_flags & OCFS2_LOCK_BUSY)
2285 mlog(ML_ERROR, "destroying busy lock: \"%s\"\n",
2286 lockres->l_name);
2287 if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
2288 mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name);
2289
2290 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
2291 spin_unlock_irqrestore(&lockres->l_lock, flags);
2292 goto out;
2293 }
2294
2295 lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED);
2296
2297 /* make sure we never get here while waiting for an ast to
2298 * fire. */
2299 BUG_ON(lockres->l_action != OCFS2_AST_INVALID);
2300
2301 /* is this necessary? */
2302 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
2303 lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK;
2304 spin_unlock_irqrestore(&lockres->l_lock, flags);
2305
2306 mlog(0, "lock %s\n", lockres->l_name);
2307
b80fc012 2308 status = dlmunlock(osb->dlm, &lockres->l_lksb, lkm_flags,
2a45f2d1 2309 ocfs2_unlock_ast, lockres);
ccd979bd
MF		 2310 if (status != DLM_NORMAL) {
2311 ocfs2_log_dlm_error("dlmunlock", status, lockres);
2312 mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags);
2313 dlm_print_one_lock(lockres->l_lksb.lockid);
2314 BUG();
2315 }
2316 mlog(0, "lock %s, successfull return from dlmunlock\n",
2317 lockres->l_name);
2318
2319 ocfs2_wait_on_busy_lock(lockres);
2320out:
2321 mlog_exit(0);
2322 return 0;
2323}
2324
2325/* Mark the lockres as being dropped. It will no longer be
2326 * queued if blocking, but we still may have to wait on it
2327 * being dequeued from the vote thread before we can consider
2328 * it safe to drop.
2329 *
2330 * You can *not* attempt to call cluster_lock on this lockres anymore. */
2331void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres)
2332{
2333 int status;
2334 struct ocfs2_mask_waiter mw;
2335 unsigned long flags;
2336
2337 ocfs2_init_mask_waiter(&mw);
2338
2339 spin_lock_irqsave(&lockres->l_lock, flags);
2340 lockres->l_flags |= OCFS2_LOCK_FREEING;
2341 while (lockres->l_flags & OCFS2_LOCK_QUEUED) {
2342 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0);
2343 spin_unlock_irqrestore(&lockres->l_lock, flags);
2344
2345 mlog(0, "Waiting on lockres %s\n", lockres->l_name);
2346
2347 status = ocfs2_wait_for_mask(&mw);
2348 if (status)
2349 mlog_errno(status);
2350
2351 spin_lock_irqsave(&lockres->l_lock, flags);
2352 }
2353 spin_unlock_irqrestore(&lockres->l_lock, flags);
2354}
2355
d680efe9
MF		 2356void ocfs2_simple_drop_lockres(struct ocfs2_super *osb,
2357 struct ocfs2_lock_res *lockres)
ccd979bd 2358{
d680efe9 2359 int ret;
ccd979bd 2360
d680efe9
MF		 2361 ocfs2_mark_lockres_freeing(lockres);
2362 ret = ocfs2_drop_lock(osb, lockres, NULL);
2363 if (ret)
2364 mlog_errno(ret);
2365}
ccd979bd 2366
d680efe9
MF		 2367static void ocfs2_drop_osb_locks(struct ocfs2_super *osb)
2368{
2369 ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres);
2370 ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres);
ccd979bd
MF		 2371}
2372
2373static void ocfs2_meta_pre_drop(struct ocfs2_lock_res *lockres, void *data)
2374{
2375 struct inode *inode = data;
2376
2377 /* the metadata lock requires a bit more work as we have an
2378 * LVB to worry about. */
2379 if (lockres->l_flags & OCFS2_LOCK_ATTACHED &&
2380 lockres->l_level == LKM_EXMODE &&
2381 !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
2382 __ocfs2_stuff_meta_lvb(inode);
2383}
2384
2385int ocfs2_drop_inode_locks(struct inode *inode)
2386{
2387 int status, err;
2388 struct drop_lock_cb meta_dcb = { ocfs2_meta_pre_drop, inode, };
2389
2390 mlog_entry_void();
2391
2392 /* No need to call ocfs2_mark_lockres_freeing here -
2393 * ocfs2_clear_inode has done it for us. */
2394
2395 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2396 &OCFS2_I(inode)->ip_data_lockres,
2397 NULL);
2398 if (err < 0)
2399 mlog_errno(err);
2400
2401 status = err;
2402
2403 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2404 &OCFS2_I(inode)->ip_meta_lockres,
2405 &meta_dcb);
2406 if (err < 0)
2407 mlog_errno(err);
2408 if (err < 0 && !status)
2409 status = err;
2410
2411 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
2412 &OCFS2_I(inode)->ip_rw_lockres,
2413 NULL);
2414 if (err < 0)
2415 mlog_errno(err);
2416 if (err < 0 && !status)
2417 status = err;
2418
2419 mlog_exit(status);
2420 return status;
2421}
2422
2423static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
2424 int new_level)
2425{
2426 assert_spin_locked(&lockres->l_lock);
2427
2428 BUG_ON(lockres->l_blocking <= LKM_NLMODE);
2429
2430 if (lockres->l_level <= new_level) {
2431 mlog(ML_ERROR, "lockres->l_level (%u) <= new_level (%u)\n",
2432 lockres->l_level, new_level);
2433 BUG();
2434 }
2435
2436 mlog(0, "lock %s, new_level = %d, l_blocking = %d\n",
2437 lockres->l_name, new_level, lockres->l_blocking);
2438
2439 lockres->l_action = OCFS2_AST_DOWNCONVERT;
2440 lockres->l_requested = new_level;
2441 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
2442}
2443
2444static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
2445 struct ocfs2_lock_res *lockres,
2446 int new_level,
2447 int lvb)
2448{
2449 int ret, dlm_flags = LKM_CONVERT;
2450 enum dlm_status status;
2451
2452 mlog_entry_void();
2453
2454 if (lvb)
2455 dlm_flags |= LKM_VALBLK;
2456
2457 status = dlmlock(osb->dlm,
2458 new_level,
2459 &lockres->l_lksb,
2460 dlm_flags,
2461 lockres->l_name,
f0681062 2462 OCFS2_LOCK_ID_MAX_LEN - 1,
e92d57df 2463 ocfs2_locking_ast,
ccd979bd 2464 lockres,
aa2623ad 2465 ocfs2_blocking_ast);
ccd979bd
MF		 2466 if (status != DLM_NORMAL) {
2467 ocfs2_log_dlm_error("dlmlock", status, lockres);
2468 ret = -EINVAL;
2469 ocfs2_recover_from_dlm_error(lockres, 1);
2470 goto bail;
2471 }
2472
2473 ret = 0;
2474bail:
2475 mlog_exit(ret);
2476 return ret;
2477}
2478
2479/* returns 1 when the caller should unlock and call dlmunlock */
2480static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
2481 struct ocfs2_lock_res *lockres)
2482{
2483 assert_spin_locked(&lockres->l_lock);
2484
2485 mlog_entry_void();
2486 mlog(0, "lock %s\n", lockres->l_name);
2487
2488 if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
2489 /* If we're already trying to cancel a lock conversion
2490 * then just drop the spinlock and allow the caller to
2491 * requeue this lock. */
2492
2493 mlog(0, "Lockres %s, skip convert\n", lockres->l_name);
2494 return 0;
2495 }
2496
2497 /* were we in a convert when we got the bast fire? */
2498 BUG_ON(lockres->l_action != OCFS2_AST_CONVERT &&
2499 lockres->l_action != OCFS2_AST_DOWNCONVERT);
2500 /* set things up for the unlockast to know to just
2501 * clear out the ast_action and unset busy, etc. */
2502 lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT;
2503
2504 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY),
2505 "lock %s, invalid flags: 0x%lx\n",
2506 lockres->l_name, lockres->l_flags);
2507
2508 return 1;
2509}
2510
2511static int ocfs2_cancel_convert(struct ocfs2_super *osb,
2512 struct ocfs2_lock_res *lockres)
2513{
2514 int ret;
2515 enum dlm_status status;
2516
2517 mlog_entry_void();
2518 mlog(0, "lock %s\n", lockres->l_name);
2519
2520 ret = 0;
2521 status = dlmunlock(osb->dlm,
2522 &lockres->l_lksb,
2523 LKM_CANCEL,
2a45f2d1 2524 ocfs2_unlock_ast,
ccd979bd
MF		 2525 lockres);
2526 if (status != DLM_NORMAL) {
2527 ocfs2_log_dlm_error("dlmunlock", status, lockres);
2528 ret = -EINVAL;
2529 ocfs2_recover_from_dlm_error(lockres, 0);
2530 }
2531
2532 mlog(0, "lock %s return from dlmunlock\n", lockres->l_name);
2533
ccd979bd
MF		 2534 mlog_exit(ret);
2535 return ret;
2536}
2537
2538static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
2539 struct ocfs2_lock_res *lockres,
d680efe9 2540 struct ocfs2_unblock_ctl *ctl,
ccd979bd
MF		 2541 ocfs2_convert_worker_t *worker)
2542{
2543 unsigned long flags;
2544 int blocking;
2545 int new_level;
2546 int ret = 0;
5ef0d4ea 2547 int set_lvb = 0;
ccd979bd
MF		 2548
2549 mlog_entry_void();
2550
2551 spin_lock_irqsave(&lockres->l_lock, flags);
2552
2553 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
2554
2555recheck:
2556 if (lockres->l_flags & OCFS2_LOCK_BUSY) {
d680efe9 2557 ctl->requeue = 1;
ccd979bd
MF		 2558 ret = ocfs2_prepare_cancel_convert(osb, lockres);
2559 spin_unlock_irqrestore(&lockres->l_lock, flags);
2560 if (ret) {
2561 ret = ocfs2_cancel_convert(osb, lockres);
2562 if (ret < 0)
2563 mlog_errno(ret);
2564 }
2565 goto leave;
2566 }
2567
2568 /* if we're blocking an exclusive and we have *any* holders,
2569 * then requeue. */
2570 if ((lockres->l_blocking == LKM_EXMODE)
f7fbfdd1
MF		 2571 && (lockres->l_ex_holders || lockres->l_ro_holders))
2572 goto leave_requeue;
ccd979bd
MF		 2573
2574 /* If it's a PR we're blocking, then only
2575 * requeue if we've got any EX holders */
2576 if (lockres->l_blocking == LKM_PRMODE &&
f7fbfdd1
MF		 2577 lockres->l_ex_holders)
2578 goto leave_requeue;
2579
2580 /*
2581 * Can we get a lock in this state if the holder counts are
2582 * zero? The meta data unblock code used to check this.
2583 */
2584 if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
2585 && (lockres->l_flags & OCFS2_LOCK_REFRESHING))
2586 goto leave_requeue;
ccd979bd 2587
16d5b956
MF		 2588 new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
2589
2590 if (lockres->l_ops->check_downconvert
2591 && !lockres->l_ops->check_downconvert(lockres, new_level))
2592 goto leave_requeue;
2593
ccd979bd
MF		 2594 /* If we get here, then we know that there are no more
2595 * incompatible holders (and anyone asking for an incompatible
2596 * lock is blocked). We can now downconvert the lock */
2597 if (!worker)
2598 goto downconvert;
2599
2600 /* Some lockres types want to do a bit of work before
2601 * downconverting a lock. Allow that here. The worker function
2602 * may sleep, so we save off a copy of what we're blocking as
2603 * it may change while we're not holding the spin lock. */
2604 blocking = lockres->l_blocking;
2605 spin_unlock_irqrestore(&lockres->l_lock, flags);
2606
d680efe9
MF		 2607 ctl->unblock_action = worker(lockres, blocking);
2608
2609 if (ctl->unblock_action == UNBLOCK_STOP_POST)
2610 goto leave;
ccd979bd
MF		 2611
2612 spin_lock_irqsave(&lockres->l_lock, flags);
2613 if (blocking != lockres->l_blocking) {
2614 /* If this changed underneath us, then we can't drop
2615 * it just yet. */
2616 goto recheck;
2617 }
2618
2619downconvert:
d680efe9 2620 ctl->requeue = 0;
ccd979bd 2621
5ef0d4ea
MF		 2622 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) {
2623 if (lockres->l_level == LKM_EXMODE)
2624 set_lvb = 1;
2625
2626 /*
2627 * We only set the lvb if the lock has been fully
2628 * refreshed - otherwise we risk setting stale
2629 * data. Otherwise, there's no need to actually clear
2630 * out the lvb here as it's value is still valid.
2631 */
2632 if (set_lvb && !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
2633 lockres->l_ops->set_lvb(lockres);
2634 }
2635
ccd979bd
MF		 2636 ocfs2_prepare_downconvert(lockres, new_level);
2637 spin_unlock_irqrestore(&lockres->l_lock, flags);
5ef0d4ea 2638 ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb);
ccd979bd
MF		 2639leave:
2640 mlog_exit(ret);
2641 return ret;
f7fbfdd1
MF		 2642
2643leave_requeue:
2644 spin_unlock_irqrestore(&lockres->l_lock, flags);
2645 ctl->requeue = 1;
2646
2647 mlog_exit(0);
2648 return 0;
ccd979bd
MF		 2649}
2650
d680efe9
MF		 2651static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
2652 int blocking)
ccd979bd
MF		 2653{
2654 struct inode *inode;
2655 struct address_space *mapping;
2656
ccd979bd
MF		 2657 inode = ocfs2_lock_res_inode(lockres);
2658 mapping = inode->i_mapping;
2659
2660 if (filemap_fdatawrite(mapping)) {
b0697053
MF		 2661 mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",
2662 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd
MF		 2663 }
2664 sync_mapping_buffers(mapping);
2665 if (blocking == LKM_EXMODE) {
2666 truncate_inode_pages(mapping, 0);
2667 unmap_mapping_range(mapping, 0, 0, 0);
2668 } else {
2669 /* We only need to wait on the I/O if we're not also
2670 * truncating pages because truncate_inode_pages waits
2671 * for us above. We don't truncate pages if we're
2672 * blocking anything < EXMODE because we want to keep
2673 * them around in that case. */
2674 filemap_fdatawait(mapping);
2675 }
2676
d680efe9 2677 return UNBLOCK_CONTINUE;
ccd979bd
MF		 2678}
2679
2680int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
d680efe9 2681 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2682{
2683 int status;
2684 struct inode *inode;
2685 struct ocfs2_super *osb;
2686
2687 mlog_entry_void();
2688
2689 inode = ocfs2_lock_res_inode(lockres);
2690 osb = OCFS2_SB(inode->i_sb);
2691
b0697053
MF		 2692 mlog(0, "unblock inode %llu\n",
2693 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd 2694
d680efe9 2695 status = ocfs2_generic_unblock_lock(osb, lockres, ctl,
ccd979bd
MF		 2696 ocfs2_data_convert_worker);
2697 if (status < 0)
2698 mlog_errno(status);
2699
b0697053 2700 mlog(0, "inode %llu, requeue = %d\n",
d680efe9 2701 (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue);
ccd979bd
MF		 2702
2703 mlog_exit(status);
2704 return status;
2705}
2706
2707static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
d680efe9 2708 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2709{
2710 int status;
2711 struct inode *inode;
2712
2713 mlog_entry_void();
2714
2715 mlog(0, "Unblock lockres %s\n", lockres->l_name);
2716
2717 inode = ocfs2_lock_res_inode(lockres);
2718
2719 status = ocfs2_generic_unblock_lock(OCFS2_SB(inode->i_sb),
d680efe9 2720 lockres, ctl, NULL);
ccd979bd
MF		 2721 if (status < 0)
2722 mlog_errno(status);
2723
2724 mlog_exit(status);
2725 return status;
2726}
2727
810d5aeb
MF		 2728static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres,
2729 int new_level)
2730{
2731 struct inode *inode = ocfs2_lock_res_inode(lockres);
2732 int checkpointed = ocfs2_inode_fully_checkpointed(inode);
2733
2734 BUG_ON(new_level != LKM_NLMODE && new_level != LKM_PRMODE);
2735 BUG_ON(lockres->l_level != LKM_EXMODE && !checkpointed);
2736
2737 if (checkpointed)
2738 return 1;
2739
2740 ocfs2_start_checkpoint(OCFS2_SB(inode->i_sb));
2741 return 0;
2742}
2743
2744static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres)
2745{
2746 struct inode *inode = ocfs2_lock_res_inode(lockres);
2747
2748 __ocfs2_stuff_meta_lvb(inode);
2749}
2750
d680efe9
MF		 2751static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres,
2752 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2753{
2754 int status;
2755 struct inode *inode;
2756
2757 mlog_entry_void();
2758
2759 inode = ocfs2_lock_res_inode(lockres);
2760
b0697053
MF		 2761 mlog(0, "unblock inode %llu\n",
2762 (unsigned long long)OCFS2_I(inode)->ip_blkno);
ccd979bd 2763
810d5aeb
MF		 2764 status = ocfs2_generic_unblock_lock(OCFS2_SB(inode->i_sb),
2765 lockres, ctl, NULL);
ccd979bd
MF		 2766 if (status < 0)
2767 mlog_errno(status);
2768
b0697053 2769 mlog(0, "inode %llu, requeue = %d\n",
d680efe9 2770 (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue);
ccd979bd
MF		 2771
2772 mlog_exit(status);
2773 return status;
2774}
2775
d680efe9
MF		 2776/*
2777 * Does the final reference drop on our dentry lock. Right now this
2778 * happens in the vote thread, but we could choose to simplify the
2779 * dlmglue API and push these off to the ocfs2_wq in the future.
2780 */
2781static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
2782 struct ocfs2_lock_res *lockres)
2783{
2784 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2785 ocfs2_dentry_lock_put(osb, dl);
2786}
2787
2788/*
2789 * d_delete() matching dentries before the lock downconvert.
2790 *
2791 * At this point, any process waiting to destroy the
2792 * dentry_lock due to last ref count is stopped by the
2793 * OCFS2_LOCK_QUEUED flag.
2794 *
2795 * We have two potential problems
2796 *
2797 * 1) If we do the last reference drop on our dentry_lock (via dput)
2798 * we'll wind up in ocfs2_release_dentry_lock(), waiting on
2799 * the downconvert to finish. Instead we take an elevated
2800 * reference and push the drop until after we've completed our
2801 * unblock processing.
2802 *
2803 * 2) There might be another process with a final reference,
2804 * waiting on us to finish processing. If this is the case, we
2805 * detect it and exit out - there's no more dentries anyway.
2806 */
2807static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres,
2808 int blocking)
2809{
2810 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2811 struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode);
2812 struct dentry *dentry;
2813 unsigned long flags;
2814 int extra_ref = 0;
2815
2816 /*
2817 * This node is blocking another node from getting a read
2818 * lock. This happens when we've renamed within a
2819 * directory. We've forced the other nodes to d_delete(), but
2820 * we never actually dropped our lock because it's still
2821 * valid. The downconvert code will retain a PR for this node,
2822 * so there's no further work to do.
2823 */
2824 if (blocking == LKM_PRMODE)
2825 return UNBLOCK_CONTINUE;
2826
2827 /*
2828 * Mark this inode as potentially orphaned. The code in
2829 * ocfs2_delete_inode() will figure out whether it actually
2830 * needs to be freed or not.
2831 */
2832 spin_lock(&oi->ip_lock);
2833 oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
2834 spin_unlock(&oi->ip_lock);
2835
2836 /*
2837 * Yuck. We need to make sure however that the check of
2838 * OCFS2_LOCK_FREEING and the extra reference are atomic with
2839 * respect to a reference decrement or the setting of that
2840 * flag.
2841 */
2842 spin_lock_irqsave(&lockres->l_lock, flags);
2843 spin_lock(&dentry_attach_lock);
2844 if (!(lockres->l_flags & OCFS2_LOCK_FREEING)
2845 && dl->dl_count) {
2846 dl->dl_count++;
2847 extra_ref = 1;
2848 }
2849 spin_unlock(&dentry_attach_lock);
2850 spin_unlock_irqrestore(&lockres->l_lock, flags);
2851
2852 mlog(0, "extra_ref = %d\n", extra_ref);
2853
2854 /*
2855 * We have a process waiting on us in ocfs2_dentry_iput(),
2856 * which means we can't have any more outstanding
2857 * aliases. There's no need to do any more work.
2858 */
2859 if (!extra_ref)
2860 return UNBLOCK_CONTINUE;
2861
2862 spin_lock(&dentry_attach_lock);
2863 while (1) {
2864 dentry = ocfs2_find_local_alias(dl->dl_inode,
2865 dl->dl_parent_blkno, 1);
2866 if (!dentry)
2867 break;
2868 spin_unlock(&dentry_attach_lock);
2869
2870 mlog(0, "d_delete(%.*s);\n", dentry->d_name.len,
2871 dentry->d_name.name);
2872
2873 /*
2874 * The following dcache calls may do an
2875 * iput(). Normally we don't want that from the
2876 * downconverting thread, but in this case it's ok
2877 * because the requesting node already has an
2878 * exclusive lock on the inode, so it can't be queued
2879 * for a downconvert.
2880 */
2881 d_delete(dentry);
2882 dput(dentry);
2883
2884 spin_lock(&dentry_attach_lock);
2885 }
2886 spin_unlock(&dentry_attach_lock);
2887
2888 /*
2889 * If we are the last holder of this dentry lock, there is no
2890 * reason to downconvert so skip straight to the unlock.
2891 */
2892 if (dl->dl_count == 1)
2893 return UNBLOCK_STOP_POST;
2894
2895 return UNBLOCK_CONTINUE_POST;
2896}
2897
2898static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres,
2899 struct ocfs2_unblock_ctl *ctl)
2900{
2901 int ret;
2902 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
2903 struct ocfs2_super *osb = OCFS2_SB(dl->dl_inode->i_sb);
2904
2905 mlog(0, "unblock dentry lock: %llu\n",
2906 (unsigned long long)OCFS2_I(dl->dl_inode)->ip_blkno);
2907
2908 ret = ocfs2_generic_unblock_lock(osb,
2909 lockres,
2910 ctl,
2911 ocfs2_dentry_convert_worker);
2912 if (ret < 0)
2913 mlog_errno(ret);
2914
2915 mlog(0, "requeue = %d, post = %d\n", ctl->requeue, ctl->unblock_action);
2916
2917 return ret;
2918}
2919
ccd979bd
MF		 2920/* Generic unblock function for any lockres whose private data is an
2921 * ocfs2_super pointer. */
2922static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
d680efe9 2923 struct ocfs2_unblock_ctl *ctl)
ccd979bd
MF		 2924{
2925 int status;
2926 struct ocfs2_super *osb;
2927
2928 mlog_entry_void();
2929
2930 mlog(0, "Unblock lockres %s\n", lockres->l_name);
2931
aa2623ad 2932 osb = ocfs2_get_lockres_osb(lockres);
ccd979bd
MF		 2933
2934 status = ocfs2_generic_unblock_lock(osb,
2935 lockres,
d680efe9 2936 ctl,
ccd979bd
MF		 2937 NULL);
2938 if (status < 0)
2939 mlog_errno(status);
2940
2941 mlog_exit(status);
2942 return status;
2943}
2944
2945void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
2946 struct ocfs2_lock_res *lockres)
2947{
2948 int status;
d680efe9 2949 struct ocfs2_unblock_ctl ctl = {0, 0,};
ccd979bd
MF		 2950 unsigned long flags;
2951
2952 /* Our reference to the lockres in this function can be
2953 * considered valid until we remove the OCFS2_LOCK_QUEUED
2954 * flag. */
2955
2956 mlog_entry_void();
2957
2958 BUG_ON(!lockres);
2959 BUG_ON(!lockres->l_ops);
2960 BUG_ON(!lockres->l_ops->unblock);
2961
2962 mlog(0, "lockres %s blocked.\n", lockres->l_name);
2963
2964 /* Detect whether a lock has been marked as going away while
2965 * the vote thread was processing other things. A lock can
2966 * still be marked with OCFS2_LOCK_FREEING after this check,
2967 * but short circuiting here will still save us some
2968 * performance. */
2969 spin_lock_irqsave(&lockres->l_lock, flags);
2970 if (lockres->l_flags & OCFS2_LOCK_FREEING)
2971 goto unqueue;
2972 spin_unlock_irqrestore(&lockres->l_lock, flags);
2973
d680efe9 2974 status = lockres->l_ops->unblock(lockres, &ctl);
ccd979bd
MF		 2975 if (status < 0)
2976 mlog_errno(status);
2977
2978 spin_lock_irqsave(&lockres->l_lock, flags);
2979unqueue:
d680efe9 2980 if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) {
ccd979bd
MF		 2981 lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED);
2982 } else
2983 ocfs2_schedule_blocked_lock(osb, lockres);
2984
2985 mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name,
d680efe9 2986 ctl.requeue ? "yes" : "no");
ccd979bd
MF		 2987 spin_unlock_irqrestore(&lockres->l_lock, flags);
2988
d680efe9
MF		 2989 if (ctl.unblock_action != UNBLOCK_CONTINUE
2990 && lockres->l_ops->post_unlock)
2991 lockres->l_ops->post_unlock(osb, lockres);
2992
ccd979bd
MF		 2993 mlog_exit_void();
2994}
2995
2996static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
2997 struct ocfs2_lock_res *lockres)
2998{
2999 mlog_entry_void();
3000
3001 assert_spin_locked(&lockres->l_lock);
3002
3003 if (lockres->l_flags & OCFS2_LOCK_FREEING) {
3004 /* Do not schedule a lock for downconvert when it's on
3005 * the way to destruction - any nodes wanting access
3006 * to the resource will get it soon. */
3007 mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n",
3008 lockres->l_name, lockres->l_flags);
3009 return;
3010 }
3011
3012 lockres_or_flags(lockres, OCFS2_LOCK_QUEUED);
3013
3014 spin_lock(&osb->vote_task_lock);
3015 if (list_empty(&lockres->l_blocked_list)) {
3016 list_add_tail(&lockres->l_blocked_list,
3017 &osb->blocked_lock_list);
3018 osb->blocked_lock_count++;
3019 }
3020 spin_unlock(&osb->vote_task_lock);
3021
3022 mlog_exit_void();
3023}
3024
3025/* This aids in debugging situations where a bad LVB might be involved. */
3026void ocfs2_dump_meta_lvb_info(u64 level,
3027 const char *function,
3028 unsigned int line,
3029 struct ocfs2_lock_res *lockres)
3030{
3031 struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
3032
3033 mlog(level, "LVB information for %s (called from %s:%u):\n",
3034 lockres->l_name, function, line);
f9e2d82e
MF		 3035 mlog(level, "version: %u, clusters: %u, generation: 0x%x\n",
3036 lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters),
3037 be32_to_cpu(lvb->lvb_igeneration));
b0697053
MF		 3038 mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n",
3039 (unsigned long long)be64_to_cpu(lvb->lvb_isize),
3040 be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid),
3041 be16_to_cpu(lvb->lvb_imode));
3042 mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, "
ca4d147e 3043 "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink),
b0697053
MF		 3044 (long long)be64_to_cpu(lvb->lvb_iatime_packed),
3045 (long long)be64_to_cpu(lvb->lvb_ictime_packed),
ca4d147e
HP		 3046 (long long)be64_to_cpu(lvb->lvb_imtime_packed),
3047 be32_to_cpu(lvb->lvb_iattr));
ccd979bd 3048}
kernel source for telekom puls (alcatel/mediatek)