1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Code which implements an OCFS2 specific interface to our DLM.
8 * Copyright (C) 2003, 2004 Oracle. All rights reserved.
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.
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.
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.
26 #include <linux/types.h>
27 #include <linux/slab.h>
28 #include <linux/highmem.h>
30 #include <linux/kthread.h>
31 #include <linux/pagemap.h>
32 #include <linux/debugfs.h>
33 #include <linux/seq_file.h>
34 #include <linux/time.h>
35 #include <linux/quotaops.h>
36 #include <linux/sched/signal.h>
38 #define MLOG_MASK_PREFIX ML_DLM_GLUE
39 #include <cluster/masklog.h>
42 #include "ocfs2_lockingver.h"
47 #include "extent_map.h"
49 #include "heartbeat.h"
52 #include "stackglue.h"
57 #include "refcounttree.h"
60 #include "buffer_head_io.h"
62 struct ocfs2_mask_waiter {
63 struct list_head mw_item;
65 struct completion mw_complete;
66 unsigned long mw_mask;
67 unsigned long mw_goal;
68 #ifdef CONFIG_OCFS2_FS_STATS
69 ktime_t mw_lock_start;
73 static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres);
74 static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres);
75 static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres);
76 static struct ocfs2_super *ocfs2_get_qinfo_osb(struct ocfs2_lock_res *lockres);
79 * Return value from ->downconvert_worker functions.
81 * These control the precise actions of ocfs2_unblock_lock()
82 * and ocfs2_process_blocked_lock()
85 enum ocfs2_unblock_action {
86 UNBLOCK_CONTINUE = 0, /* Continue downconvert */
87 UNBLOCK_CONTINUE_POST = 1, /* Continue downconvert, fire
88 * ->post_unlock callback */
89 UNBLOCK_STOP_POST = 2, /* Do not downconvert, fire
90 * ->post_unlock() callback. */
93 struct ocfs2_unblock_ctl {
95 enum ocfs2_unblock_action unblock_action;
98 /* Lockdep class keys */
99 #ifdef CONFIG_DEBUG_LOCK_ALLOC
100 static struct lock_class_key lockdep_keys[OCFS2_NUM_LOCK_TYPES];
103 static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres,
105 static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres);
107 static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
110 static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres,
113 static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
114 struct ocfs2_lock_res *lockres);
116 static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres);
118 static int ocfs2_check_refcount_downconvert(struct ocfs2_lock_res *lockres,
120 static int ocfs2_refcount_convert_worker(struct ocfs2_lock_res *lockres,
123 #define mlog_meta_lvb(__level, __lockres) ocfs2_dump_meta_lvb_info(__level, __PRETTY_FUNCTION__, __LINE__, __lockres)
125 /* This aids in debugging situations where a bad LVB might be involved. */
126 static void ocfs2_dump_meta_lvb_info(u64 level,
127 const char *function,
129 struct ocfs2_lock_res *lockres)
131 struct ocfs2_meta_lvb *lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
133 mlog(level, "LVB information for %s (called from %s:%u):\n",
134 lockres->l_name, function, line);
135 mlog(level, "version: %u, clusters: %u, generation: 0x%x\n",
136 lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters),
137 be32_to_cpu(lvb->lvb_igeneration));
138 mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n",
139 (unsigned long long)be64_to_cpu(lvb->lvb_isize),
140 be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid),
141 be16_to_cpu(lvb->lvb_imode));
142 mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, "
143 "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink),
144 (long long)be64_to_cpu(lvb->lvb_iatime_packed),
145 (long long)be64_to_cpu(lvb->lvb_ictime_packed),
146 (long long)be64_to_cpu(lvb->lvb_imtime_packed),
147 be32_to_cpu(lvb->lvb_iattr));
152 * OCFS2 Lock Resource Operations
154 * These fine tune the behavior of the generic dlmglue locking infrastructure.
156 * The most basic of lock types can point ->l_priv to their respective
157 * struct ocfs2_super and allow the default actions to manage things.
159 * Right now, each lock type also needs to implement an init function,
160 * and trivial lock/unlock wrappers. ocfs2_simple_drop_lockres()
161 * should be called when the lock is no longer needed (i.e., object
164 struct ocfs2_lock_res_ops {
166 * Translate an ocfs2_lock_res * into an ocfs2_super *. Define
167 * this callback if ->l_priv is not an ocfs2_super pointer
169 struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *);
172 * Optionally called in the downconvert thread after a
173 * successful downconvert. The lockres will not be referenced
174 * after this callback is called, so it is safe to free
177 * The exact semantics of when this is called are controlled
178 * by ->downconvert_worker()
180 void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *);
183 * Allow a lock type to add checks to determine whether it is
184 * safe to downconvert a lock. Return 0 to re-queue the
185 * downconvert at a later time, nonzero to continue.
187 * For most locks, the default checks that there are no
188 * incompatible holders are sufficient.
190 * Called with the lockres spinlock held.
192 int (*check_downconvert)(struct ocfs2_lock_res *, int);
195 * Allows a lock type to populate the lock value block. This
196 * is called on downconvert, and when we drop a lock.
198 * Locks that want to use this should set LOCK_TYPE_USES_LVB
199 * in the flags field.
201 * Called with the lockres spinlock held.
203 void (*set_lvb)(struct ocfs2_lock_res *);
206 * Called from the downconvert thread when it is determined
207 * that a lock will be downconverted. This is called without
208 * any locks held so the function can do work that might
209 * schedule (syncing out data, etc).
211 * This should return any one of the ocfs2_unblock_action
212 * values, depending on what it wants the thread to do.
214 int (*downconvert_worker)(struct ocfs2_lock_res *, int);
217 * LOCK_TYPE_* flags which describe the specific requirements
218 * of a lock type. Descriptions of each individual flag follow.
224 * Some locks want to "refresh" potentially stale data when a
225 * meaningful (PRMODE or EXMODE) lock level is first obtained. If this
226 * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the
227 * individual lockres l_flags member from the ast function. It is
228 * expected that the locking wrapper will clear the
229 * OCFS2_LOCK_NEEDS_REFRESH flag when done.
231 #define LOCK_TYPE_REQUIRES_REFRESH 0x1
234 * Indicate that a lock type makes use of the lock value block. The
235 * ->set_lvb lock type callback must be defined.
237 #define LOCK_TYPE_USES_LVB 0x2
239 static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = {
240 .get_osb = ocfs2_get_inode_osb,
244 static struct ocfs2_lock_res_ops ocfs2_inode_inode_lops = {
245 .get_osb = ocfs2_get_inode_osb,
246 .check_downconvert = ocfs2_check_meta_downconvert,
247 .set_lvb = ocfs2_set_meta_lvb,
248 .downconvert_worker = ocfs2_data_convert_worker,
249 .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB,
252 static struct ocfs2_lock_res_ops ocfs2_super_lops = {
253 .flags = LOCK_TYPE_REQUIRES_REFRESH,
256 static struct ocfs2_lock_res_ops ocfs2_rename_lops = {
260 static struct ocfs2_lock_res_ops ocfs2_nfs_sync_lops = {
264 static struct ocfs2_lock_res_ops ocfs2_trim_fs_lops = {
265 .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB,
268 static struct ocfs2_lock_res_ops ocfs2_orphan_scan_lops = {
269 .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB,
272 static struct ocfs2_lock_res_ops ocfs2_dentry_lops = {
273 .get_osb = ocfs2_get_dentry_osb,
274 .post_unlock = ocfs2_dentry_post_unlock,
275 .downconvert_worker = ocfs2_dentry_convert_worker,
279 static struct ocfs2_lock_res_ops ocfs2_inode_open_lops = {
280 .get_osb = ocfs2_get_inode_osb,
284 static struct ocfs2_lock_res_ops ocfs2_flock_lops = {
285 .get_osb = ocfs2_get_file_osb,
289 static struct ocfs2_lock_res_ops ocfs2_qinfo_lops = {
290 .set_lvb = ocfs2_set_qinfo_lvb,
291 .get_osb = ocfs2_get_qinfo_osb,
292 .flags = LOCK_TYPE_REQUIRES_REFRESH | LOCK_TYPE_USES_LVB,
295 static struct ocfs2_lock_res_ops ocfs2_refcount_block_lops = {
296 .check_downconvert = ocfs2_check_refcount_downconvert,
297 .downconvert_worker = ocfs2_refcount_convert_worker,
301 static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
303 return lockres->l_type == OCFS2_LOCK_TYPE_META ||
304 lockres->l_type == OCFS2_LOCK_TYPE_RW ||
305 lockres->l_type == OCFS2_LOCK_TYPE_OPEN;
308 static inline struct ocfs2_lock_res *ocfs2_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb)
310 return container_of(lksb, struct ocfs2_lock_res, l_lksb);
313 static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres)
315 BUG_ON(!ocfs2_is_inode_lock(lockres));
317 return (struct inode *) lockres->l_priv;
320 static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres)
322 BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY);
324 return (struct ocfs2_dentry_lock *)lockres->l_priv;
327 static inline struct ocfs2_mem_dqinfo *ocfs2_lock_res_qinfo(struct ocfs2_lock_res *lockres)
329 BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_QINFO);
331 return (struct ocfs2_mem_dqinfo *)lockres->l_priv;
334 static inline struct ocfs2_refcount_tree *
335 ocfs2_lock_res_refcount_tree(struct ocfs2_lock_res *res)
337 return container_of(res, struct ocfs2_refcount_tree, rf_lockres);
340 static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres)
342 if (lockres->l_ops->get_osb)
343 return lockres->l_ops->get_osb(lockres);
345 return (struct ocfs2_super *)lockres->l_priv;
348 static int ocfs2_lock_create(struct ocfs2_super *osb,
349 struct ocfs2_lock_res *lockres,
352 static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
354 static void __ocfs2_cluster_unlock(struct ocfs2_super *osb,
355 struct ocfs2_lock_res *lockres,
356 int level, unsigned long caller_ip);
357 static inline void ocfs2_cluster_unlock(struct ocfs2_super *osb,
358 struct ocfs2_lock_res *lockres,
361 __ocfs2_cluster_unlock(osb, lockres, level, _RET_IP_);
364 static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres);
365 static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres);
366 static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres);
367 static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level);
368 static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
369 struct ocfs2_lock_res *lockres);
370 static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
372 #define ocfs2_log_dlm_error(_func, _err, _lockres) do { \
373 if ((_lockres)->l_type != OCFS2_LOCK_TYPE_DENTRY) \
374 mlog(ML_ERROR, "DLM error %d while calling %s on resource %s\n", \
375 _err, _func, _lockres->l_name); \
377 mlog(ML_ERROR, "DLM error %d while calling %s on resource %.*s%08x\n", \
378 _err, _func, OCFS2_DENTRY_LOCK_INO_START - 1, (_lockres)->l_name, \
379 (unsigned int)ocfs2_get_dentry_lock_ino(_lockres)); \
381 static int ocfs2_downconvert_thread(void *arg);
382 static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb,
383 struct ocfs2_lock_res *lockres);
384 static int ocfs2_inode_lock_update(struct inode *inode,
385 struct buffer_head **bh);
386 static void ocfs2_drop_osb_locks(struct ocfs2_super *osb);
387 static inline int ocfs2_highest_compat_lock_level(int level);
388 static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
390 static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
391 struct ocfs2_lock_res *lockres,
394 unsigned int generation);
395 static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
396 struct ocfs2_lock_res *lockres);
397 static int ocfs2_cancel_convert(struct ocfs2_super *osb,
398 struct ocfs2_lock_res *lockres);
401 static void ocfs2_build_lock_name(enum ocfs2_lock_type type,
408 BUG_ON(type >= OCFS2_NUM_LOCK_TYPES);
410 len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x",
411 ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD,
412 (long long)blkno, generation);
414 BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1));
416 mlog(0, "built lock resource with name: %s\n", name);
419 static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock);
421 static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res,
422 struct ocfs2_dlm_debug *dlm_debug)
424 mlog(0, "Add tracking for lockres %s\n", res->l_name);
426 spin_lock(&ocfs2_dlm_tracking_lock);
427 list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking);
428 spin_unlock(&ocfs2_dlm_tracking_lock);
431 static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res)
433 spin_lock(&ocfs2_dlm_tracking_lock);
434 if (!list_empty(&res->l_debug_list))
435 list_del_init(&res->l_debug_list);
436 spin_unlock(&ocfs2_dlm_tracking_lock);
439 #ifdef CONFIG_OCFS2_FS_STATS
440 static void ocfs2_init_lock_stats(struct ocfs2_lock_res *res)
442 res->l_lock_refresh = 0;
443 memset(&res->l_lock_prmode, 0, sizeof(struct ocfs2_lock_stats));
444 memset(&res->l_lock_exmode, 0, sizeof(struct ocfs2_lock_stats));
447 static void ocfs2_update_lock_stats(struct ocfs2_lock_res *res, int level,
448 struct ocfs2_mask_waiter *mw, int ret)
452 struct ocfs2_lock_stats *stats;
454 if (level == LKM_PRMODE)
455 stats = &res->l_lock_prmode;
456 else if (level == LKM_EXMODE)
457 stats = &res->l_lock_exmode;
461 kt = ktime_sub(ktime_get(), mw->mw_lock_start);
462 usec = ktime_to_us(kt);
465 stats->ls_total += ktime_to_ns(kt);
467 if (unlikely(stats->ls_gets == 0)) {
469 stats->ls_total = ktime_to_ns(kt);
472 if (stats->ls_max < usec)
473 stats->ls_max = usec;
479 static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres)
481 lockres->l_lock_refresh++;
484 static inline void ocfs2_init_start_time(struct ocfs2_mask_waiter *mw)
486 mw->mw_lock_start = ktime_get();
489 static inline void ocfs2_init_lock_stats(struct ocfs2_lock_res *res)
492 static inline void ocfs2_update_lock_stats(struct ocfs2_lock_res *res,
493 int level, struct ocfs2_mask_waiter *mw, int ret)
496 static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres)
499 static inline void ocfs2_init_start_time(struct ocfs2_mask_waiter *mw)
504 static void ocfs2_lock_res_init_common(struct ocfs2_super *osb,
505 struct ocfs2_lock_res *res,
506 enum ocfs2_lock_type type,
507 struct ocfs2_lock_res_ops *ops,
514 res->l_level = DLM_LOCK_IV;
515 res->l_requested = DLM_LOCK_IV;
516 res->l_blocking = DLM_LOCK_IV;
517 res->l_action = OCFS2_AST_INVALID;
518 res->l_unlock_action = OCFS2_UNLOCK_INVALID;
520 res->l_flags = OCFS2_LOCK_INITIALIZED;
522 ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug);
524 ocfs2_init_lock_stats(res);
525 #ifdef CONFIG_DEBUG_LOCK_ALLOC
526 if (type != OCFS2_LOCK_TYPE_OPEN)
527 lockdep_init_map(&res->l_lockdep_map, ocfs2_lock_type_strings[type],
528 &lockdep_keys[type], 0);
530 res->l_lockdep_map.key = NULL;
534 void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res)
536 /* This also clears out the lock status block */
537 memset(res, 0, sizeof(struct ocfs2_lock_res));
538 spin_lock_init(&res->l_lock);
539 init_waitqueue_head(&res->l_event);
540 INIT_LIST_HEAD(&res->l_blocked_list);
541 INIT_LIST_HEAD(&res->l_mask_waiters);
542 INIT_LIST_HEAD(&res->l_holders);
545 void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
546 enum ocfs2_lock_type type,
547 unsigned int generation,
550 struct ocfs2_lock_res_ops *ops;
553 case OCFS2_LOCK_TYPE_RW:
554 ops = &ocfs2_inode_rw_lops;
556 case OCFS2_LOCK_TYPE_META:
557 ops = &ocfs2_inode_inode_lops;
559 case OCFS2_LOCK_TYPE_OPEN:
560 ops = &ocfs2_inode_open_lops;
563 mlog_bug_on_msg(1, "type: %d\n", type);
564 ops = NULL; /* thanks, gcc */
568 ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno,
569 generation, res->l_name);
570 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode);
573 static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres)
575 struct inode *inode = ocfs2_lock_res_inode(lockres);
577 return OCFS2_SB(inode->i_sb);
580 static struct ocfs2_super *ocfs2_get_qinfo_osb(struct ocfs2_lock_res *lockres)
582 struct ocfs2_mem_dqinfo *info = lockres->l_priv;
584 return OCFS2_SB(info->dqi_gi.dqi_sb);
587 static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres)
589 struct ocfs2_file_private *fp = lockres->l_priv;
591 return OCFS2_SB(fp->fp_file->f_mapping->host->i_sb);
594 static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres)
596 __be64 inode_blkno_be;
598 memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START],
601 return be64_to_cpu(inode_blkno_be);
604 static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres)
606 struct ocfs2_dentry_lock *dl = lockres->l_priv;
608 return OCFS2_SB(dl->dl_inode->i_sb);
611 void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl,
612 u64 parent, struct inode *inode)
615 u64 inode_blkno = OCFS2_I(inode)->ip_blkno;
616 __be64 inode_blkno_be = cpu_to_be64(inode_blkno);
617 struct ocfs2_lock_res *lockres = &dl->dl_lockres;
619 ocfs2_lock_res_init_once(lockres);
622 * Unfortunately, the standard lock naming scheme won't work
623 * here because we have two 16 byte values to use. Instead,
624 * we'll stuff the inode number as a binary value. We still
625 * want error prints to show something without garbling the
626 * display, so drop a null byte in there before the inode
627 * number. A future version of OCFS2 will likely use all
628 * binary lock names. The stringified names have been a
629 * tremendous aid in debugging, but now that the debugfs
630 * interface exists, we can mangle things there if need be.
632 * NOTE: We also drop the standard "pad" value (the total lock
633 * name size stays the same though - the last part is all
634 * zeros due to the memset in ocfs2_lock_res_init_once()
636 len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START,
638 ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY),
641 BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1));
643 memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be,
646 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
647 OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops,
651 static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res,
652 struct ocfs2_super *osb)
654 /* Superblock lockres doesn't come from a slab so we call init
655 * once on it manually. */
656 ocfs2_lock_res_init_once(res);
657 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO,
659 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER,
660 &ocfs2_super_lops, osb);
663 static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
664 struct ocfs2_super *osb)
666 /* Rename lockres doesn't come from a slab so we call init
667 * once on it manually. */
668 ocfs2_lock_res_init_once(res);
669 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name);
670 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME,
671 &ocfs2_rename_lops, osb);
674 static void ocfs2_nfs_sync_lock_res_init(struct ocfs2_lock_res *res,
675 struct ocfs2_super *osb)
677 /* nfs_sync lockres doesn't come from a slab so we call init
678 * once on it manually. */
679 ocfs2_lock_res_init_once(res);
680 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_NFS_SYNC, 0, 0, res->l_name);
681 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_NFS_SYNC,
682 &ocfs2_nfs_sync_lops, osb);
685 void ocfs2_trim_fs_lock_res_init(struct ocfs2_super *osb)
687 struct ocfs2_lock_res *lockres = &osb->osb_trim_fs_lockres;
689 ocfs2_lock_res_init_once(lockres);
690 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_TRIM_FS, 0, 0, lockres->l_name);
691 ocfs2_lock_res_init_common(osb, lockres, OCFS2_LOCK_TYPE_TRIM_FS,
692 &ocfs2_trim_fs_lops, osb);
695 void ocfs2_trim_fs_lock_res_uninit(struct ocfs2_super *osb)
697 struct ocfs2_lock_res *lockres = &osb->osb_trim_fs_lockres;
699 ocfs2_simple_drop_lockres(osb, lockres);
700 ocfs2_lock_res_free(lockres);
703 static void ocfs2_orphan_scan_lock_res_init(struct ocfs2_lock_res *res,
704 struct ocfs2_super *osb)
706 ocfs2_lock_res_init_once(res);
707 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_ORPHAN_SCAN, 0, 0, res->l_name);
708 ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_ORPHAN_SCAN,
709 &ocfs2_orphan_scan_lops, osb);
712 void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres,
713 struct ocfs2_file_private *fp)
715 struct inode *inode = fp->fp_file->f_mapping->host;
716 struct ocfs2_inode_info *oi = OCFS2_I(inode);
718 ocfs2_lock_res_init_once(lockres);
719 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_FLOCK, oi->ip_blkno,
720 inode->i_generation, lockres->l_name);
721 ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres,
722 OCFS2_LOCK_TYPE_FLOCK, &ocfs2_flock_lops,
724 lockres->l_flags |= OCFS2_LOCK_NOCACHE;
727 void ocfs2_qinfo_lock_res_init(struct ocfs2_lock_res *lockres,
728 struct ocfs2_mem_dqinfo *info)
730 ocfs2_lock_res_init_once(lockres);
731 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_QINFO, info->dqi_gi.dqi_type,
733 ocfs2_lock_res_init_common(OCFS2_SB(info->dqi_gi.dqi_sb), lockres,
734 OCFS2_LOCK_TYPE_QINFO, &ocfs2_qinfo_lops,
738 void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res *lockres,
739 struct ocfs2_super *osb, u64 ref_blkno,
740 unsigned int generation)
742 ocfs2_lock_res_init_once(lockres);
743 ocfs2_build_lock_name(OCFS2_LOCK_TYPE_REFCOUNT, ref_blkno,
744 generation, lockres->l_name);
745 ocfs2_lock_res_init_common(osb, lockres, OCFS2_LOCK_TYPE_REFCOUNT,
746 &ocfs2_refcount_block_lops, osb);
749 void ocfs2_lock_res_free(struct ocfs2_lock_res *res)
751 if (!(res->l_flags & OCFS2_LOCK_INITIALIZED))
754 ocfs2_remove_lockres_tracking(res);
756 mlog_bug_on_msg(!list_empty(&res->l_blocked_list),
757 "Lockres %s is on the blocked list\n",
759 mlog_bug_on_msg(!list_empty(&res->l_mask_waiters),
760 "Lockres %s has mask waiters pending\n",
762 mlog_bug_on_msg(spin_is_locked(&res->l_lock),
763 "Lockres %s is locked\n",
765 mlog_bug_on_msg(res->l_ro_holders,
766 "Lockres %s has %u ro holders\n",
767 res->l_name, res->l_ro_holders);
768 mlog_bug_on_msg(res->l_ex_holders,
769 "Lockres %s has %u ex holders\n",
770 res->l_name, res->l_ex_holders);
772 /* Need to clear out the lock status block for the dlm */
773 memset(&res->l_lksb, 0, sizeof(res->l_lksb));
779 * Keep a list of processes who have interest in a lockres.
780 * Note: this is now only uesed for check recursive cluster locking.
782 static inline void ocfs2_add_holder(struct ocfs2_lock_res *lockres,
783 struct ocfs2_lock_holder *oh)
785 INIT_LIST_HEAD(&oh->oh_list);
786 oh->oh_owner_pid = get_pid(task_pid(current));
788 spin_lock(&lockres->l_lock);
789 list_add_tail(&oh->oh_list, &lockres->l_holders);
790 spin_unlock(&lockres->l_lock);
793 static struct ocfs2_lock_holder *
794 ocfs2_pid_holder(struct ocfs2_lock_res *lockres,
797 struct ocfs2_lock_holder *oh;
799 spin_lock(&lockres->l_lock);
800 list_for_each_entry(oh, &lockres->l_holders, oh_list) {
801 if (oh->oh_owner_pid == pid) {
802 spin_unlock(&lockres->l_lock);
806 spin_unlock(&lockres->l_lock);
810 static inline void ocfs2_remove_holder(struct ocfs2_lock_res *lockres,
811 struct ocfs2_lock_holder *oh)
813 spin_lock(&lockres->l_lock);
814 list_del(&oh->oh_list);
815 spin_unlock(&lockres->l_lock);
817 put_pid(oh->oh_owner_pid);
821 static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres,
828 lockres->l_ex_holders++;
831 lockres->l_ro_holders++;
838 static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres,
845 BUG_ON(!lockres->l_ex_holders);
846 lockres->l_ex_holders--;
849 BUG_ON(!lockres->l_ro_holders);
850 lockres->l_ro_holders--;
857 /* WARNING: This function lives in a world where the only three lock
858 * levels are EX, PR, and NL. It *will* have to be adjusted when more
859 * lock types are added. */
860 static inline int ocfs2_highest_compat_lock_level(int level)
862 int new_level = DLM_LOCK_EX;
864 if (level == DLM_LOCK_EX)
865 new_level = DLM_LOCK_NL;
866 else if (level == DLM_LOCK_PR)
867 new_level = DLM_LOCK_PR;
871 static void lockres_set_flags(struct ocfs2_lock_res *lockres,
872 unsigned long newflags)
874 struct ocfs2_mask_waiter *mw, *tmp;
876 assert_spin_locked(&lockres->l_lock);
878 lockres->l_flags = newflags;
880 list_for_each_entry_safe(mw, tmp, &lockres->l_mask_waiters, mw_item) {
881 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
884 list_del_init(&mw->mw_item);
886 complete(&mw->mw_complete);
889 static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or)
891 lockres_set_flags(lockres, lockres->l_flags | or);
893 static void lockres_clear_flags(struct ocfs2_lock_res *lockres,
896 lockres_set_flags(lockres, lockres->l_flags & ~clear);
899 static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres)
901 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
902 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
903 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
904 BUG_ON(lockres->l_blocking <= DLM_LOCK_NL);
906 lockres->l_level = lockres->l_requested;
907 if (lockres->l_level <=
908 ocfs2_highest_compat_lock_level(lockres->l_blocking)) {
909 lockres->l_blocking = DLM_LOCK_NL;
910 lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
912 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
915 static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres)
917 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
918 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
920 /* Convert from RO to EX doesn't really need anything as our
921 * information is already up to data. Convert from NL to
922 * *anything* however should mark ourselves as needing an
924 if (lockres->l_level == DLM_LOCK_NL &&
925 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
926 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
928 lockres->l_level = lockres->l_requested;
931 * We set the OCFS2_LOCK_UPCONVERT_FINISHING flag before clearing
932 * the OCFS2_LOCK_BUSY flag to prevent the dc thread from
933 * downconverting the lock before the upconvert has fully completed.
934 * Do not prevent the dc thread from downconverting if NONBLOCK lock
935 * had already returned.
937 if (!(lockres->l_flags & OCFS2_LOCK_NONBLOCK_FINISHED))
938 lockres_or_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
940 lockres_clear_flags(lockres, OCFS2_LOCK_NONBLOCK_FINISHED);
942 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
945 static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres)
947 BUG_ON((!(lockres->l_flags & OCFS2_LOCK_BUSY)));
948 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
950 if (lockres->l_requested > DLM_LOCK_NL &&
951 !(lockres->l_flags & OCFS2_LOCK_LOCAL) &&
952 lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
953 lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
955 lockres->l_level = lockres->l_requested;
956 lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED);
957 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
960 static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
963 int needs_downconvert = 0;
965 assert_spin_locked(&lockres->l_lock);
967 if (level > lockres->l_blocking) {
968 /* only schedule a downconvert if we haven't already scheduled
969 * one that goes low enough to satisfy the level we're
970 * blocking. this also catches the case where we get
972 if (ocfs2_highest_compat_lock_level(level) <
973 ocfs2_highest_compat_lock_level(lockres->l_blocking))
974 needs_downconvert = 1;
976 lockres->l_blocking = level;
979 mlog(ML_BASTS, "lockres %s, block %d, level %d, l_block %d, dwn %d\n",
980 lockres->l_name, level, lockres->l_level, lockres->l_blocking,
983 if (needs_downconvert)
984 lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
985 mlog(0, "needs_downconvert = %d\n", needs_downconvert);
986 return needs_downconvert;
990 * OCFS2_LOCK_PENDING and l_pending_gen.
992 * Why does OCFS2_LOCK_PENDING exist? To close a race between setting
993 * OCFS2_LOCK_BUSY and calling ocfs2_dlm_lock(). See ocfs2_unblock_lock()
994 * for more details on the race.
996 * OCFS2_LOCK_PENDING closes the race quite nicely. However, it introduces
997 * a race on itself. In o2dlm, we can get the ast before ocfs2_dlm_lock()
998 * returns. The ast clears OCFS2_LOCK_BUSY, and must therefore clear
999 * OCFS2_LOCK_PENDING at the same time. When ocfs2_dlm_lock() returns,
1000 * the caller is going to try to clear PENDING again. If nothing else is
1001 * happening, __lockres_clear_pending() sees PENDING is unset and does
1004 * But what if another path (eg downconvert thread) has just started a
1005 * new locking action? The other path has re-set PENDING. Our path
1006 * cannot clear PENDING, because that will re-open the original race
1012 * ocfs2_cluster_lock()
1017 * ocfs2_locking_ast() ocfs2_downconvert_thread()
1018 * clear PENDING ocfs2_unblock_lock()
1021 * ocfs2_prepare_downconvert()
1031 * So as you can see, we now have a window where l_lock is not held,
1032 * PENDING is not set, and ocfs2_dlm_lock() has not been called.
1034 * The core problem is that ocfs2_cluster_lock() has cleared the PENDING
1035 * set by ocfs2_prepare_downconvert(). That wasn't nice.
1037 * To solve this we introduce l_pending_gen. A call to
1038 * lockres_clear_pending() will only do so when it is passed a generation
1039 * number that matches the lockres. lockres_set_pending() will return the
1040 * current generation number. When ocfs2_cluster_lock() goes to clear
1041 * PENDING, it passes the generation it got from set_pending(). In our
1042 * example above, the generation numbers will *not* match. Thus,
1043 * ocfs2_cluster_lock() will not clear the PENDING set by
1044 * ocfs2_prepare_downconvert().
1047 /* Unlocked version for ocfs2_locking_ast() */
1048 static void __lockres_clear_pending(struct ocfs2_lock_res *lockres,
1049 unsigned int generation,
1050 struct ocfs2_super *osb)
1052 assert_spin_locked(&lockres->l_lock);
1055 * The ast and locking functions can race us here. The winner
1056 * will clear pending, the loser will not.
1058 if (!(lockres->l_flags & OCFS2_LOCK_PENDING) ||
1059 (lockres->l_pending_gen != generation))
1062 lockres_clear_flags(lockres, OCFS2_LOCK_PENDING);
1063 lockres->l_pending_gen++;
1066 * The downconvert thread may have skipped us because we
1067 * were PENDING. Wake it up.
1069 if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
1070 ocfs2_wake_downconvert_thread(osb);
1073 /* Locked version for callers of ocfs2_dlm_lock() */
1074 static void lockres_clear_pending(struct ocfs2_lock_res *lockres,
1075 unsigned int generation,
1076 struct ocfs2_super *osb)
1078 unsigned long flags;
1080 spin_lock_irqsave(&lockres->l_lock, flags);
1081 __lockres_clear_pending(lockres, generation, osb);
1082 spin_unlock_irqrestore(&lockres->l_lock, flags);
1085 static unsigned int lockres_set_pending(struct ocfs2_lock_res *lockres)
1087 assert_spin_locked(&lockres->l_lock);
1088 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
1090 lockres_or_flags(lockres, OCFS2_LOCK_PENDING);
1092 return lockres->l_pending_gen;
1095 static void ocfs2_blocking_ast(struct ocfs2_dlm_lksb *lksb, int level)
1097 struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
1098 struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
1099 int needs_downconvert;
1100 unsigned long flags;
1102 BUG_ON(level <= DLM_LOCK_NL);
1104 mlog(ML_BASTS, "BAST fired for lockres %s, blocking %d, level %d, "
1105 "type %s\n", lockres->l_name, level, lockres->l_level,
1106 ocfs2_lock_type_string(lockres->l_type));
1109 * We can skip the bast for locks which don't enable caching -
1110 * they'll be dropped at the earliest possible time anyway.
1112 if (lockres->l_flags & OCFS2_LOCK_NOCACHE)
1115 spin_lock_irqsave(&lockres->l_lock, flags);
1116 needs_downconvert = ocfs2_generic_handle_bast(lockres, level);
1117 if (needs_downconvert)
1118 ocfs2_schedule_blocked_lock(osb, lockres);
1119 spin_unlock_irqrestore(&lockres->l_lock, flags);
1121 wake_up(&lockres->l_event);
1123 ocfs2_wake_downconvert_thread(osb);
1126 static void ocfs2_locking_ast(struct ocfs2_dlm_lksb *lksb)
1128 struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
1129 struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
1130 unsigned long flags;
1133 spin_lock_irqsave(&lockres->l_lock, flags);
1135 status = ocfs2_dlm_lock_status(&lockres->l_lksb);
1137 if (status == -EAGAIN) {
1138 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
1143 mlog(ML_ERROR, "lockres %s: lksb status value of %d!\n",
1144 lockres->l_name, status);
1145 spin_unlock_irqrestore(&lockres->l_lock, flags);
1149 mlog(ML_BASTS, "AST fired for lockres %s, action %d, unlock %d, "
1150 "level %d => %d\n", lockres->l_name, lockres->l_action,
1151 lockres->l_unlock_action, lockres->l_level, lockres->l_requested);
1153 switch(lockres->l_action) {
1154 case OCFS2_AST_ATTACH:
1155 ocfs2_generic_handle_attach_action(lockres);
1156 lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL);
1158 case OCFS2_AST_CONVERT:
1159 ocfs2_generic_handle_convert_action(lockres);
1161 case OCFS2_AST_DOWNCONVERT:
1162 ocfs2_generic_handle_downconvert_action(lockres);
1165 mlog(ML_ERROR, "lockres %s: AST fired with invalid action: %u, "
1166 "flags 0x%lx, unlock: %u\n",
1167 lockres->l_name, lockres->l_action, lockres->l_flags,
1168 lockres->l_unlock_action);
1172 /* set it to something invalid so if we get called again we
1174 lockres->l_action = OCFS2_AST_INVALID;
1176 /* Did we try to cancel this lock? Clear that state */
1177 if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT)
1178 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
1181 * We may have beaten the locking functions here. We certainly
1182 * know that dlm_lock() has been called :-)
1183 * Because we can't have two lock calls in flight at once, we
1184 * can use lockres->l_pending_gen.
1186 __lockres_clear_pending(lockres, lockres->l_pending_gen, osb);
1188 wake_up(&lockres->l_event);
1189 spin_unlock_irqrestore(&lockres->l_lock, flags);
1192 static void ocfs2_unlock_ast(struct ocfs2_dlm_lksb *lksb, int error)
1194 struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
1195 unsigned long flags;
1197 mlog(ML_BASTS, "UNLOCK AST fired for lockres %s, action = %d\n",
1198 lockres->l_name, lockres->l_unlock_action);
1200 spin_lock_irqsave(&lockres->l_lock, flags);
1202 mlog(ML_ERROR, "Dlm passes error %d for lock %s, "
1203 "unlock_action %d\n", error, lockres->l_name,
1204 lockres->l_unlock_action);
1205 spin_unlock_irqrestore(&lockres->l_lock, flags);
1209 switch(lockres->l_unlock_action) {
1210 case OCFS2_UNLOCK_CANCEL_CONVERT:
1211 mlog(0, "Cancel convert success for %s\n", lockres->l_name);
1212 lockres->l_action = OCFS2_AST_INVALID;
1213 /* Downconvert thread may have requeued this lock, we
1214 * need to wake it. */
1215 if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
1216 ocfs2_wake_downconvert_thread(ocfs2_get_lockres_osb(lockres));
1218 case OCFS2_UNLOCK_DROP_LOCK:
1219 lockres->l_level = DLM_LOCK_IV;
1225 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
1226 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
1227 wake_up(&lockres->l_event);
1228 spin_unlock_irqrestore(&lockres->l_lock, flags);
1232 * This is the filesystem locking protocol. It provides the lock handling
1233 * hooks for the underlying DLM. It has a maximum version number.
1234 * The version number allows interoperability with systems running at
1235 * the same major number and an equal or smaller minor number.
1237 * Whenever the filesystem does new things with locks (adds or removes a
1238 * lock, orders them differently, does different things underneath a lock),
1239 * the version must be changed. The protocol is negotiated when joining
1240 * the dlm domain. A node may join the domain if its major version is
1241 * identical to all other nodes and its minor version is greater than
1242 * or equal to all other nodes. When its minor version is greater than
1243 * the other nodes, it will run at the minor version specified by the
1246 * If a locking change is made that will not be compatible with older
1247 * versions, the major number must be increased and the minor version set
1248 * to zero. If a change merely adds a behavior that can be disabled when
1249 * speaking to older versions, the minor version must be increased. If a
1250 * change adds a fully backwards compatible change (eg, LVB changes that
1251 * are just ignored by older versions), the version does not need to be
1254 static struct ocfs2_locking_protocol lproto = {
1256 .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
1257 .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
1259 .lp_lock_ast = ocfs2_locking_ast,
1260 .lp_blocking_ast = ocfs2_blocking_ast,
1261 .lp_unlock_ast = ocfs2_unlock_ast,
1264 void ocfs2_set_locking_protocol(void)
1266 ocfs2_stack_glue_set_max_proto_version(&lproto.lp_max_version);
1269 static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
1272 unsigned long flags;
1274 spin_lock_irqsave(&lockres->l_lock, flags);
1275 lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
1276 lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
1278 lockres->l_action = OCFS2_AST_INVALID;
1280 lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
1281 spin_unlock_irqrestore(&lockres->l_lock, flags);
1283 wake_up(&lockres->l_event);
1286 /* Note: If we detect another process working on the lock (i.e.,
1287 * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller
1288 * to do the right thing in that case.
1290 static int ocfs2_lock_create(struct ocfs2_super *osb,
1291 struct ocfs2_lock_res *lockres,
1296 unsigned long flags;
1299 mlog(0, "lock %s, level = %d, flags = %u\n", lockres->l_name, level,
1302 spin_lock_irqsave(&lockres->l_lock, flags);
1303 if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) ||
1304 (lockres->l_flags & OCFS2_LOCK_BUSY)) {
1305 spin_unlock_irqrestore(&lockres->l_lock, flags);
1309 lockres->l_action = OCFS2_AST_ATTACH;
1310 lockres->l_requested = level;
1311 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
1312 gen = lockres_set_pending(lockres);
1313 spin_unlock_irqrestore(&lockres->l_lock, flags);
1315 ret = ocfs2_dlm_lock(osb->cconn,
1320 OCFS2_LOCK_ID_MAX_LEN - 1);
1321 lockres_clear_pending(lockres, gen, osb);
1323 ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
1324 ocfs2_recover_from_dlm_error(lockres, 1);
1327 mlog(0, "lock %s, return from ocfs2_dlm_lock\n", lockres->l_name);
1333 static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres,
1336 unsigned long flags;
1339 spin_lock_irqsave(&lockres->l_lock, flags);
1340 ret = lockres->l_flags & flag;
1341 spin_unlock_irqrestore(&lockres->l_lock, flags);
1346 static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres)
1349 wait_event(lockres->l_event,
1350 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY));
1353 static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres)
1356 wait_event(lockres->l_event,
1357 !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING));
1360 /* predict what lock level we'll be dropping down to on behalf
1361 * of another node, and return true if the currently wanted
1362 * level will be compatible with it. */
1363 static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
1366 BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
1368 return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking);
1371 static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw)
1373 INIT_LIST_HEAD(&mw->mw_item);
1374 init_completion(&mw->mw_complete);
1375 ocfs2_init_start_time(mw);
1378 static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw)
1380 wait_for_completion(&mw->mw_complete);
1381 /* Re-arm the completion in case we want to wait on it again */
1382 reinit_completion(&mw->mw_complete);
1383 return mw->mw_status;
1386 static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres,
1387 struct ocfs2_mask_waiter *mw,
1391 BUG_ON(!list_empty(&mw->mw_item));
1393 assert_spin_locked(&lockres->l_lock);
1395 list_add_tail(&mw->mw_item, &lockres->l_mask_waiters);
1400 /* returns 0 if the mw that was removed was already satisfied, -EBUSY
1401 * if the mask still hadn't reached its goal */
1402 static int __lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
1403 struct ocfs2_mask_waiter *mw)
1407 assert_spin_locked(&lockres->l_lock);
1408 if (!list_empty(&mw->mw_item)) {
1409 if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
1412 list_del_init(&mw->mw_item);
1413 init_completion(&mw->mw_complete);
1419 static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
1420 struct ocfs2_mask_waiter *mw)
1422 unsigned long flags;
1425 spin_lock_irqsave(&lockres->l_lock, flags);
1426 ret = __lockres_remove_mask_waiter(lockres, mw);
1427 spin_unlock_irqrestore(&lockres->l_lock, flags);
1433 static int ocfs2_wait_for_mask_interruptible(struct ocfs2_mask_waiter *mw,
1434 struct ocfs2_lock_res *lockres)
1438 ret = wait_for_completion_interruptible(&mw->mw_complete);
1440 lockres_remove_mask_waiter(lockres, mw);
1442 ret = mw->mw_status;
1443 /* Re-arm the completion in case we want to wait on it again */
1444 reinit_completion(&mw->mw_complete);
1448 static int __ocfs2_cluster_lock(struct ocfs2_super *osb,
1449 struct ocfs2_lock_res *lockres,
1454 unsigned long caller_ip)
1456 struct ocfs2_mask_waiter mw;
1457 int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR);
1458 int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */
1459 unsigned long flags;
1461 int noqueue_attempted = 0;
1465 if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED)) {
1466 mlog_errno(-EINVAL);
1470 ocfs2_init_mask_waiter(&mw);
1472 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
1473 lkm_flags |= DLM_LKF_VALBLK;
1478 spin_lock_irqsave(&lockres->l_lock, flags);
1480 if (catch_signals && signal_pending(current)) {
1485 mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING,
1486 "Cluster lock called on freeing lockres %s! flags "
1487 "0x%lx\n", lockres->l_name, lockres->l_flags);
1489 /* We only compare against the currently granted level
1490 * here. If the lock is blocked waiting on a downconvert,
1491 * we'll get caught below. */
1492 if (lockres->l_flags & OCFS2_LOCK_BUSY &&
1493 level > lockres->l_level) {
1494 /* is someone sitting in dlm_lock? If so, wait on
1496 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
1501 if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING) {
1503 * We've upconverted. If the lock now has a level we can
1504 * work with, we take it. If, however, the lock is not at the
1505 * required level, we go thru the full cycle. One way this could
1506 * happen is if a process requesting an upconvert to PR is
1507 * closely followed by another requesting upconvert to an EX.
1508 * If the process requesting EX lands here, we want it to
1509 * continue attempting to upconvert and let the process
1510 * requesting PR take the lock.
1511 * If multiple processes request upconvert to PR, the first one
1512 * here will take the lock. The others will have to go thru the
1513 * OCFS2_LOCK_BLOCKED check to ensure that there is no pending
1514 * downconvert request.
1516 if (level <= lockres->l_level)
1517 goto update_holders;
1520 if (lockres->l_flags & OCFS2_LOCK_BLOCKED &&
1521 !ocfs2_may_continue_on_blocked_lock(lockres, level)) {
1522 /* is the lock is currently blocked on behalf of
1524 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0);
1529 if (level > lockres->l_level) {
1530 if (noqueue_attempted > 0) {
1534 if (lkm_flags & DLM_LKF_NOQUEUE)
1535 noqueue_attempted = 1;
1537 if (lockres->l_action != OCFS2_AST_INVALID)
1538 mlog(ML_ERROR, "lockres %s has action %u pending\n",
1539 lockres->l_name, lockres->l_action);
1541 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
1542 lockres->l_action = OCFS2_AST_ATTACH;
1543 lkm_flags &= ~DLM_LKF_CONVERT;
1545 lockres->l_action = OCFS2_AST_CONVERT;
1546 lkm_flags |= DLM_LKF_CONVERT;
1549 lockres->l_requested = level;
1550 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
1551 gen = lockres_set_pending(lockres);
1552 spin_unlock_irqrestore(&lockres->l_lock, flags);
1554 BUG_ON(level == DLM_LOCK_IV);
1555 BUG_ON(level == DLM_LOCK_NL);
1557 mlog(ML_BASTS, "lockres %s, convert from %d to %d\n",
1558 lockres->l_name, lockres->l_level, level);
1560 /* call dlm_lock to upgrade lock now */
1561 ret = ocfs2_dlm_lock(osb->cconn,
1566 OCFS2_LOCK_ID_MAX_LEN - 1);
1567 lockres_clear_pending(lockres, gen, osb);
1569 if (!(lkm_flags & DLM_LKF_NOQUEUE) ||
1571 ocfs2_log_dlm_error("ocfs2_dlm_lock",
1574 ocfs2_recover_from_dlm_error(lockres, 1);
1579 mlog(0, "lock %s, successful return from ocfs2_dlm_lock\n",
1582 /* At this point we've gone inside the dlm and need to
1583 * complete our work regardless. */
1586 /* wait for busy to clear and carry on */
1591 /* Ok, if we get here then we're good to go. */
1592 ocfs2_inc_holders(lockres, level);
1596 lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
1598 /* ocfs2_unblock_lock reques on seeing OCFS2_LOCK_UPCONVERT_FINISHING */
1599 kick_dc = (lockres->l_flags & OCFS2_LOCK_BLOCKED);
1601 spin_unlock_irqrestore(&lockres->l_lock, flags);
1603 ocfs2_wake_downconvert_thread(osb);
1606 * This is helping work around a lock inversion between the page lock
1607 * and dlm locks. One path holds the page lock while calling aops
1608 * which block acquiring dlm locks. The voting thread holds dlm
1609 * locks while acquiring page locks while down converting data locks.
1610 * This block is helping an aop path notice the inversion and back
1611 * off to unlock its page lock before trying the dlm lock again.
1613 if (wait && arg_flags & OCFS2_LOCK_NONBLOCK &&
1614 mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) {
1616 spin_lock_irqsave(&lockres->l_lock, flags);
1617 if (__lockres_remove_mask_waiter(lockres, &mw)) {
1619 lockres_or_flags(lockres,
1620 OCFS2_LOCK_NONBLOCK_FINISHED);
1621 spin_unlock_irqrestore(&lockres->l_lock, flags);
1624 spin_unlock_irqrestore(&lockres->l_lock, flags);
1629 ret = ocfs2_wait_for_mask(&mw);
1634 ocfs2_update_lock_stats(lockres, level, &mw, ret);
1636 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1637 if (!ret && lockres->l_lockdep_map.key != NULL) {
1638 if (level == DLM_LOCK_PR)
1639 rwsem_acquire_read(&lockres->l_lockdep_map, l_subclass,
1640 !!(arg_flags & OCFS2_META_LOCK_NOQUEUE),
1643 rwsem_acquire(&lockres->l_lockdep_map, l_subclass,
1644 !!(arg_flags & OCFS2_META_LOCK_NOQUEUE),
1651 static inline int ocfs2_cluster_lock(struct ocfs2_super *osb,
1652 struct ocfs2_lock_res *lockres,
1657 return __ocfs2_cluster_lock(osb, lockres, level, lkm_flags, arg_flags,
1662 static void __ocfs2_cluster_unlock(struct ocfs2_super *osb,
1663 struct ocfs2_lock_res *lockres,
1665 unsigned long caller_ip)
1667 unsigned long flags;
1669 spin_lock_irqsave(&lockres->l_lock, flags);
1670 ocfs2_dec_holders(lockres, level);
1671 ocfs2_downconvert_on_unlock(osb, lockres);
1672 spin_unlock_irqrestore(&lockres->l_lock, flags);
1673 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1674 if (lockres->l_lockdep_map.key != NULL)
1675 rwsem_release(&lockres->l_lockdep_map, 1, caller_ip);
1679 static int ocfs2_create_new_lock(struct ocfs2_super *osb,
1680 struct ocfs2_lock_res *lockres,
1684 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
1685 unsigned long flags;
1686 u32 lkm_flags = local ? DLM_LKF_LOCAL : 0;
1688 spin_lock_irqsave(&lockres->l_lock, flags);
1689 BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
1690 lockres_or_flags(lockres, OCFS2_LOCK_LOCAL);
1691 spin_unlock_irqrestore(&lockres->l_lock, flags);
1693 return ocfs2_lock_create(osb, lockres, level, lkm_flags);
1696 /* Grants us an EX lock on the data and metadata resources, skipping
1697 * the normal cluster directory lookup. Use this ONLY on newly created
1698 * inodes which other nodes can't possibly see, and which haven't been
1699 * hashed in the inode hash yet. This can give us a good performance
1700 * increase as it'll skip the network broadcast normally associated
1701 * with creating a new lock resource. */
1702 int ocfs2_create_new_inode_locks(struct inode *inode)
1705 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1707 BUG_ON(!ocfs2_inode_is_new(inode));
1709 mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno);
1711 /* NOTE: That we don't increment any of the holder counts, nor
1712 * do we add anything to a journal handle. Since this is
1713 * supposed to be a new inode which the cluster doesn't know
1714 * about yet, there is no need to. As far as the LVB handling
1715 * is concerned, this is basically like acquiring an EX lock
1716 * on a resource which has an invalid one -- we'll set it
1717 * valid when we release the EX. */
1719 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1);
1726 * We don't want to use DLM_LKF_LOCAL on a meta data lock as they
1727 * don't use a generation in their lock names.
1729 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_inode_lockres, 1, 0);
1735 ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_open_lockres, 0, 0);
1743 int ocfs2_rw_lock(struct inode *inode, int write)
1746 struct ocfs2_lock_res *lockres;
1747 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1749 mlog(0, "inode %llu take %s RW lock\n",
1750 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1751 write ? "EXMODE" : "PRMODE");
1753 if (ocfs2_mount_local(osb))
1756 lockres = &OCFS2_I(inode)->ip_rw_lockres;
1758 level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
1760 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
1767 int ocfs2_try_rw_lock(struct inode *inode, int write)
1770 struct ocfs2_lock_res *lockres;
1771 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1773 mlog(0, "inode %llu try to take %s RW lock\n",
1774 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1775 write ? "EXMODE" : "PRMODE");
1777 if (ocfs2_mount_local(osb))
1780 lockres = &OCFS2_I(inode)->ip_rw_lockres;
1782 level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
1784 status = ocfs2_cluster_lock(osb, lockres, level, DLM_LKF_NOQUEUE, 0);
1788 void ocfs2_rw_unlock(struct inode *inode, int write)
1790 int level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
1791 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres;
1792 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1794 mlog(0, "inode %llu drop %s RW lock\n",
1795 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1796 write ? "EXMODE" : "PRMODE");
1798 if (!ocfs2_mount_local(osb))
1799 ocfs2_cluster_unlock(osb, lockres, level);
1803 * ocfs2_open_lock always get PR mode lock.
1805 int ocfs2_open_lock(struct inode *inode)
1808 struct ocfs2_lock_res *lockres;
1809 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1811 mlog(0, "inode %llu take PRMODE open lock\n",
1812 (unsigned long long)OCFS2_I(inode)->ip_blkno);
1814 if (ocfs2_is_hard_readonly(osb) || ocfs2_mount_local(osb))
1817 lockres = &OCFS2_I(inode)->ip_open_lockres;
1819 status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_PR, 0, 0);
1827 int ocfs2_try_open_lock(struct inode *inode, int write)
1829 int status = 0, level;
1830 struct ocfs2_lock_res *lockres;
1831 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1833 mlog(0, "inode %llu try to take %s open lock\n",
1834 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1835 write ? "EXMODE" : "PRMODE");
1837 if (ocfs2_is_hard_readonly(osb)) {
1843 if (ocfs2_mount_local(osb))
1846 lockres = &OCFS2_I(inode)->ip_open_lockres;
1848 level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
1851 * The file system may already holding a PRMODE/EXMODE open lock.
1852 * Since we pass DLM_LKF_NOQUEUE, the request won't block waiting on
1853 * other nodes and the -EAGAIN will indicate to the caller that
1854 * this inode is still in use.
1856 status = ocfs2_cluster_lock(osb, lockres, level, DLM_LKF_NOQUEUE, 0);
1863 * ocfs2_open_unlock unlock PR and EX mode open locks.
1865 void ocfs2_open_unlock(struct inode *inode)
1867 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_open_lockres;
1868 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1870 mlog(0, "inode %llu drop open lock\n",
1871 (unsigned long long)OCFS2_I(inode)->ip_blkno);
1873 if (ocfs2_mount_local(osb))
1876 if(lockres->l_ro_holders)
1877 ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_PR);
1878 if(lockres->l_ex_holders)
1879 ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX);
1885 static int ocfs2_flock_handle_signal(struct ocfs2_lock_res *lockres,
1889 struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
1890 unsigned long flags;
1891 struct ocfs2_mask_waiter mw;
1893 ocfs2_init_mask_waiter(&mw);
1896 spin_lock_irqsave(&lockres->l_lock, flags);
1897 if (lockres->l_flags & OCFS2_LOCK_BUSY) {
1898 ret = ocfs2_prepare_cancel_convert(osb, lockres);
1900 spin_unlock_irqrestore(&lockres->l_lock, flags);
1901 ret = ocfs2_cancel_convert(osb, lockres);
1908 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
1909 spin_unlock_irqrestore(&lockres->l_lock, flags);
1911 ocfs2_wait_for_mask(&mw);
1917 * We may still have gotten the lock, in which case there's no
1918 * point to restarting the syscall.
1920 if (lockres->l_level == level)
1923 mlog(0, "Cancel returning %d. flags: 0x%lx, level: %d, act: %d\n", ret,
1924 lockres->l_flags, lockres->l_level, lockres->l_action);
1926 spin_unlock_irqrestore(&lockres->l_lock, flags);
1933 * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of
1934 * flock() calls. The locking approach this requires is sufficiently
1935 * different from all other cluster lock types that we implement a
1936 * separate path to the "low-level" dlm calls. In particular:
1938 * - No optimization of lock levels is done - we take at exactly
1939 * what's been requested.
1941 * - No lock caching is employed. We immediately downconvert to
1942 * no-lock at unlock time. This also means flock locks never go on
1943 * the blocking list).
1945 * - Since userspace can trivially deadlock itself with flock, we make
1946 * sure to allow cancellation of a misbehaving applications flock()
1949 * - Access to any flock lockres doesn't require concurrency, so we
1950 * can simplify the code by requiring the caller to guarantee
1951 * serialization of dlmglue flock calls.
1953 int ocfs2_file_lock(struct file *file, int ex, int trylock)
1955 int ret, level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
1956 unsigned int lkm_flags = trylock ? DLM_LKF_NOQUEUE : 0;
1957 unsigned long flags;
1958 struct ocfs2_file_private *fp = file->private_data;
1959 struct ocfs2_lock_res *lockres = &fp->fp_flock;
1960 struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb);
1961 struct ocfs2_mask_waiter mw;
1963 ocfs2_init_mask_waiter(&mw);
1965 if ((lockres->l_flags & OCFS2_LOCK_BUSY) ||
1966 (lockres->l_level > DLM_LOCK_NL)) {
1968 "File lock \"%s\" has busy or locked state: flags: 0x%lx, "
1969 "level: %u\n", lockres->l_name, lockres->l_flags,
1974 spin_lock_irqsave(&lockres->l_lock, flags);
1975 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
1976 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
1977 spin_unlock_irqrestore(&lockres->l_lock, flags);
1980 * Get the lock at NLMODE to start - that way we
1981 * can cancel the upconvert request if need be.
1983 ret = ocfs2_lock_create(osb, lockres, DLM_LOCK_NL, 0);
1989 ret = ocfs2_wait_for_mask(&mw);
1994 spin_lock_irqsave(&lockres->l_lock, flags);
1997 lockres->l_action = OCFS2_AST_CONVERT;
1998 lkm_flags |= DLM_LKF_CONVERT;
1999 lockres->l_requested = level;
2000 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
2002 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
2003 spin_unlock_irqrestore(&lockres->l_lock, flags);
2005 ret = ocfs2_dlm_lock(osb->cconn, level, &lockres->l_lksb, lkm_flags,
2006 lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1);
2008 if (!trylock || (ret != -EAGAIN)) {
2009 ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
2013 ocfs2_recover_from_dlm_error(lockres, 1);
2014 lockres_remove_mask_waiter(lockres, &mw);
2018 ret = ocfs2_wait_for_mask_interruptible(&mw, lockres);
2019 if (ret == -ERESTARTSYS) {
2021 * Userspace can cause deadlock itself with
2022 * flock(). Current behavior locally is to allow the
2023 * deadlock, but abort the system call if a signal is
2024 * received. We follow this example, otherwise a
2025 * poorly written program could sit in kernel until
2028 * Handling this is a bit more complicated for Ocfs2
2029 * though. We can't exit this function with an
2030 * outstanding lock request, so a cancel convert is
2031 * required. We intentionally overwrite 'ret' - if the
2032 * cancel fails and the lock was granted, it's easier
2033 * to just bubble success back up to the user.
2035 ret = ocfs2_flock_handle_signal(lockres, level);
2036 } else if (!ret && (level > lockres->l_level)) {
2037 /* Trylock failed asynchronously */
2044 mlog(0, "Lock: \"%s\" ex: %d, trylock: %d, returns: %d\n",
2045 lockres->l_name, ex, trylock, ret);
2049 void ocfs2_file_unlock(struct file *file)
2053 unsigned long flags;
2054 struct ocfs2_file_private *fp = file->private_data;
2055 struct ocfs2_lock_res *lockres = &fp->fp_flock;
2056 struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb);
2057 struct ocfs2_mask_waiter mw;
2059 ocfs2_init_mask_waiter(&mw);
2061 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED))
2064 if (lockres->l_level == DLM_LOCK_NL)
2067 mlog(0, "Unlock: \"%s\" flags: 0x%lx, level: %d, act: %d\n",
2068 lockres->l_name, lockres->l_flags, lockres->l_level,
2071 spin_lock_irqsave(&lockres->l_lock, flags);
2073 * Fake a blocking ast for the downconvert code.
2075 lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
2076 lockres->l_blocking = DLM_LOCK_EX;
2078 gen = ocfs2_prepare_downconvert(lockres, DLM_LOCK_NL);
2079 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
2080 spin_unlock_irqrestore(&lockres->l_lock, flags);
2082 ret = ocfs2_downconvert_lock(osb, lockres, DLM_LOCK_NL, 0, gen);
2088 ret = ocfs2_wait_for_mask(&mw);
2093 static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb,
2094 struct ocfs2_lock_res *lockres)
2098 /* If we know that another node is waiting on our lock, kick
2099 * the downconvert thread * pre-emptively when we reach a release
2101 if (lockres->l_flags & OCFS2_LOCK_BLOCKED) {
2102 switch(lockres->l_blocking) {
2104 if (!lockres->l_ex_holders && !lockres->l_ro_holders)
2108 if (!lockres->l_ex_holders)
2117 ocfs2_wake_downconvert_thread(osb);
2120 #define OCFS2_SEC_BITS 34
2121 #define OCFS2_SEC_SHIFT (64 - 34)
2122 #define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1)
2124 /* LVB only has room for 64 bits of time here so we pack it for
2126 static u64 ocfs2_pack_timespec(struct timespec64 *spec)
2129 u64 sec = clamp_t(time64_t, spec->tv_sec, 0, 0x3ffffffffull);
2130 u32 nsec = spec->tv_nsec;
2132 res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK);
2137 /* Call this with the lockres locked. I am reasonably sure we don't
2138 * need ip_lock in this function as anyone who would be changing those
2139 * values is supposed to be blocked in ocfs2_inode_lock right now. */
2140 static void __ocfs2_stuff_meta_lvb(struct inode *inode)
2142 struct ocfs2_inode_info *oi = OCFS2_I(inode);
2143 struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres;
2144 struct ocfs2_meta_lvb *lvb;
2146 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2149 * Invalidate the LVB of a deleted inode - this way other
2150 * nodes are forced to go to disk and discover the new inode
2153 if (oi->ip_flags & OCFS2_INODE_DELETED) {
2154 lvb->lvb_version = 0;
2158 lvb->lvb_version = OCFS2_LVB_VERSION;
2159 lvb->lvb_isize = cpu_to_be64(i_size_read(inode));
2160 lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters);
2161 lvb->lvb_iuid = cpu_to_be32(i_uid_read(inode));
2162 lvb->lvb_igid = cpu_to_be32(i_gid_read(inode));
2163 lvb->lvb_imode = cpu_to_be16(inode->i_mode);
2164 lvb->lvb_inlink = cpu_to_be16(inode->i_nlink);
2165 lvb->lvb_iatime_packed =
2166 cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime));
2167 lvb->lvb_ictime_packed =
2168 cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime));
2169 lvb->lvb_imtime_packed =
2170 cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime));
2171 lvb->lvb_iattr = cpu_to_be32(oi->ip_attr);
2172 lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features);
2173 lvb->lvb_igeneration = cpu_to_be32(inode->i_generation);
2176 mlog_meta_lvb(0, lockres);
2179 static void ocfs2_unpack_timespec(struct timespec64 *spec,
2182 spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT;
2183 spec->tv_nsec = packed_time & OCFS2_NSEC_MASK;
2186 static void ocfs2_refresh_inode_from_lvb(struct inode *inode)
2188 struct ocfs2_inode_info *oi = OCFS2_I(inode);
2189 struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres;
2190 struct ocfs2_meta_lvb *lvb;
2192 mlog_meta_lvb(0, lockres);
2194 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2196 /* We're safe here without the lockres lock... */
2197 spin_lock(&oi->ip_lock);
2198 oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters);
2199 i_size_write(inode, be64_to_cpu(lvb->lvb_isize));
2201 oi->ip_attr = be32_to_cpu(lvb->lvb_iattr);
2202 oi->ip_dyn_features = be16_to_cpu(lvb->lvb_idynfeatures);
2203 ocfs2_set_inode_flags(inode);
2205 /* fast-symlinks are a special case */
2206 if (S_ISLNK(inode->i_mode) && !oi->ip_clusters)
2207 inode->i_blocks = 0;
2209 inode->i_blocks = ocfs2_inode_sector_count(inode);
2211 i_uid_write(inode, be32_to_cpu(lvb->lvb_iuid));
2212 i_gid_write(inode, be32_to_cpu(lvb->lvb_igid));
2213 inode->i_mode = be16_to_cpu(lvb->lvb_imode);
2214 set_nlink(inode, be16_to_cpu(lvb->lvb_inlink));
2215 ocfs2_unpack_timespec(&inode->i_atime,
2216 be64_to_cpu(lvb->lvb_iatime_packed));
2217 ocfs2_unpack_timespec(&inode->i_mtime,
2218 be64_to_cpu(lvb->lvb_imtime_packed));
2219 ocfs2_unpack_timespec(&inode->i_ctime,
2220 be64_to_cpu(lvb->lvb_ictime_packed));
2221 spin_unlock(&oi->ip_lock);
2224 static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode,
2225 struct ocfs2_lock_res *lockres)
2227 struct ocfs2_meta_lvb *lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2229 if (ocfs2_dlm_lvb_valid(&lockres->l_lksb)
2230 && lvb->lvb_version == OCFS2_LVB_VERSION
2231 && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation)
2236 /* Determine whether a lock resource needs to be refreshed, and
2237 * arbitrate who gets to refresh it.
2239 * 0 means no refresh needed.
2241 * > 0 means you need to refresh this and you MUST call
2242 * ocfs2_complete_lock_res_refresh afterwards. */
2243 static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres)
2245 unsigned long flags;
2249 spin_lock_irqsave(&lockres->l_lock, flags);
2250 if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
2251 spin_unlock_irqrestore(&lockres->l_lock, flags);
2255 if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
2256 spin_unlock_irqrestore(&lockres->l_lock, flags);
2258 ocfs2_wait_on_refreshing_lock(lockres);
2262 /* Ok, I'll be the one to refresh this lock. */
2263 lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING);
2264 spin_unlock_irqrestore(&lockres->l_lock, flags);
2268 mlog(0, "status %d\n", status);
2272 /* If status is non zero, I'll mark it as not being in refresh
2273 * anymroe, but i won't clear the needs refresh flag. */
2274 static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres,
2277 unsigned long flags;
2279 spin_lock_irqsave(&lockres->l_lock, flags);
2280 lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING);
2282 lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
2283 spin_unlock_irqrestore(&lockres->l_lock, flags);
2285 wake_up(&lockres->l_event);
2288 /* may or may not return a bh if it went to disk. */
2289 static int ocfs2_inode_lock_update(struct inode *inode,
2290 struct buffer_head **bh)
2293 struct ocfs2_inode_info *oi = OCFS2_I(inode);
2294 struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres;
2295 struct ocfs2_dinode *fe;
2296 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2298 if (ocfs2_mount_local(osb))
2301 spin_lock(&oi->ip_lock);
2302 if (oi->ip_flags & OCFS2_INODE_DELETED) {
2303 mlog(0, "Orphaned inode %llu was deleted while we "
2304 "were waiting on a lock. ip_flags = 0x%x\n",
2305 (unsigned long long)oi->ip_blkno, oi->ip_flags);
2306 spin_unlock(&oi->ip_lock);
2310 spin_unlock(&oi->ip_lock);
2312 if (!ocfs2_should_refresh_lock_res(lockres))
2315 /* This will discard any caching information we might have had
2316 * for the inode metadata. */
2317 ocfs2_metadata_cache_purge(INODE_CACHE(inode));
2319 ocfs2_extent_map_trunc(inode, 0);
2321 if (ocfs2_meta_lvb_is_trustable(inode, lockres)) {
2322 mlog(0, "Trusting LVB on inode %llu\n",
2323 (unsigned long long)oi->ip_blkno);
2324 ocfs2_refresh_inode_from_lvb(inode);
2326 /* Boo, we have to go to disk. */
2327 /* read bh, cast, ocfs2_refresh_inode */
2328 status = ocfs2_read_inode_block(inode, bh);
2333 fe = (struct ocfs2_dinode *) (*bh)->b_data;
2335 /* This is a good chance to make sure we're not
2336 * locking an invalid object. ocfs2_read_inode_block()
2337 * already checked that the inode block is sane.
2339 * We bug on a stale inode here because we checked
2340 * above whether it was wiped from disk. The wiping
2341 * node provides a guarantee that we receive that
2342 * message and can mark the inode before dropping any
2343 * locks associated with it. */
2344 mlog_bug_on_msg(inode->i_generation !=
2345 le32_to_cpu(fe->i_generation),
2346 "Invalid dinode %llu disk generation: %u "
2347 "inode->i_generation: %u\n",
2348 (unsigned long long)oi->ip_blkno,
2349 le32_to_cpu(fe->i_generation),
2350 inode->i_generation);
2351 mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) ||
2352 !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)),
2353 "Stale dinode %llu dtime: %llu flags: 0x%x\n",
2354 (unsigned long long)oi->ip_blkno,
2355 (unsigned long long)le64_to_cpu(fe->i_dtime),
2356 le32_to_cpu(fe->i_flags));
2358 ocfs2_refresh_inode(inode, fe);
2359 ocfs2_track_lock_refresh(lockres);
2364 ocfs2_complete_lock_res_refresh(lockres, status);
2369 static int ocfs2_assign_bh(struct inode *inode,
2370 struct buffer_head **ret_bh,
2371 struct buffer_head *passed_bh)
2376 /* Ok, the update went to disk for us, use the
2378 *ret_bh = passed_bh;
2384 status = ocfs2_read_inode_block(inode, ret_bh);
2392 * returns < 0 error if the callback will never be called, otherwise
2393 * the result of the lock will be communicated via the callback.
2395 int ocfs2_inode_lock_full_nested(struct inode *inode,
2396 struct buffer_head **ret_bh,
2401 int status, level, acquired;
2403 struct ocfs2_lock_res *lockres = NULL;
2404 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2405 struct buffer_head *local_bh = NULL;
2407 mlog(0, "inode %llu, take %s META lock\n",
2408 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2409 ex ? "EXMODE" : "PRMODE");
2413 /* We'll allow faking a readonly metadata lock for
2415 if (ocfs2_is_hard_readonly(osb)) {
2421 if ((arg_flags & OCFS2_META_LOCK_GETBH) ||
2422 ocfs2_mount_local(osb))
2425 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
2426 ocfs2_wait_for_recovery(osb);
2428 lockres = &OCFS2_I(inode)->ip_inode_lockres;
2429 level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2431 if (arg_flags & OCFS2_META_LOCK_NOQUEUE)
2432 dlm_flags |= DLM_LKF_NOQUEUE;
2434 status = __ocfs2_cluster_lock(osb, lockres, level, dlm_flags,
2435 arg_flags, subclass, _RET_IP_);
2437 if (status != -EAGAIN)
2442 /* Notify the error cleanup path to drop the cluster lock. */
2445 /* We wait twice because a node may have died while we were in
2446 * the lower dlm layers. The second time though, we've
2447 * committed to owning this lock so we don't allow signals to
2448 * abort the operation. */
2449 if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
2450 ocfs2_wait_for_recovery(osb);
2454 * We only see this flag if we're being called from
2455 * ocfs2_read_locked_inode(). It means we're locking an inode
2456 * which hasn't been populated yet, so clear the refresh flag
2457 * and let the caller handle it.
2459 if (inode->i_state & I_NEW) {
2462 ocfs2_complete_lock_res_refresh(lockres, 0);
2466 /* This is fun. The caller may want a bh back, or it may
2467 * not. ocfs2_inode_lock_update definitely wants one in, but
2468 * may or may not read one, depending on what's in the
2469 * LVB. The result of all of this is that we've *only* gone to
2470 * disk if we have to, so the complexity is worthwhile. */
2471 status = ocfs2_inode_lock_update(inode, &local_bh);
2473 if (status != -ENOENT)
2479 status = ocfs2_assign_bh(inode, ret_bh, local_bh);
2488 if (ret_bh && (*ret_bh)) {
2493 ocfs2_inode_unlock(inode, ex);
2503 * This is working around a lock inversion between tasks acquiring DLM
2504 * locks while holding a page lock and the downconvert thread which
2505 * blocks dlm lock acquiry while acquiring page locks.
2507 * ** These _with_page variantes are only intended to be called from aop
2508 * methods that hold page locks and return a very specific *positive* error
2509 * code that aop methods pass up to the VFS -- test for errors with != 0. **
2511 * The DLM is called such that it returns -EAGAIN if it would have
2512 * blocked waiting for the downconvert thread. In that case we unlock
2513 * our page so the downconvert thread can make progress. Once we've
2514 * done this we have to return AOP_TRUNCATED_PAGE so the aop method
2515 * that called us can bubble that back up into the VFS who will then
2516 * immediately retry the aop call.
2518 int ocfs2_inode_lock_with_page(struct inode *inode,
2519 struct buffer_head **ret_bh,
2525 ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK);
2526 if (ret == -EAGAIN) {
2529 * If we can't get inode lock immediately, we should not return
2530 * directly here, since this will lead to a softlockup problem.
2531 * The method is to get a blocking lock and immediately unlock
2532 * before returning, this can avoid CPU resource waste due to
2533 * lots of retries, and benefits fairness in getting lock.
2535 if (ocfs2_inode_lock(inode, ret_bh, ex) == 0)
2536 ocfs2_inode_unlock(inode, ex);
2537 ret = AOP_TRUNCATED_PAGE;
2543 int ocfs2_inode_lock_atime(struct inode *inode,
2544 struct vfsmount *vfsmnt,
2545 int *level, int wait)
2550 ret = ocfs2_inode_lock(inode, NULL, 0);
2552 ret = ocfs2_try_inode_lock(inode, NULL, 0);
2561 * If we should update atime, we will get EX lock,
2562 * otherwise we just get PR lock.
2564 if (ocfs2_should_update_atime(inode, vfsmnt)) {
2565 struct buffer_head *bh = NULL;
2567 ocfs2_inode_unlock(inode, 0);
2569 ret = ocfs2_inode_lock(inode, &bh, 1);
2571 ret = ocfs2_try_inode_lock(inode, &bh, 1);
2579 if (ocfs2_should_update_atime(inode, vfsmnt))
2580 ocfs2_update_inode_atime(inode, bh);
2589 void ocfs2_inode_unlock(struct inode *inode,
2592 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2593 struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_inode_lockres;
2594 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2596 mlog(0, "inode %llu drop %s META lock\n",
2597 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2598 ex ? "EXMODE" : "PRMODE");
2600 if (!ocfs2_is_hard_readonly(osb) &&
2601 !ocfs2_mount_local(osb))
2602 ocfs2_cluster_unlock(osb, lockres, level);
2606 * This _tracker variantes are introduced to deal with the recursive cluster
2607 * locking issue. The idea is to keep track of a lock holder on the stack of
2608 * the current process. If there's a lock holder on the stack, we know the
2609 * task context is already protected by cluster locking. Currently, they're
2610 * used in some VFS entry routines.
2612 * return < 0 on error, return == 0 if there's no lock holder on the stack
2613 * before this call, return == 1 if this call would be a recursive locking.
2614 * return == -1 if this lock attempt will cause an upgrade which is forbidden.
2616 * When taking lock levels into account,we face some different situations.
2618 * 1. no lock is held
2619 * In this case, just lock the inode as requested and return 0
2621 * 2. We are holding a lock
2622 * For this situation, things diverges into several cases
2624 * wanted holding what to do
2625 * ex ex see 2.1 below
2626 * ex pr see 2.2 below
2627 * pr ex see 2.1 below
2628 * pr pr see 2.1 below
2630 * 2.1 lock level that is been held is compatible
2631 * with the wanted level, so no lock action will be tacken.
2633 * 2.2 Otherwise, an upgrade is needed, but it is forbidden.
2635 * Reason why upgrade within a process is forbidden is that
2636 * lock upgrade may cause dead lock. The following illustrates
2639 * thread on node1 thread on node2
2640 * ocfs2_inode_lock_tracker(ex=0)
2642 * <====== ocfs2_inode_lock_tracker(ex=1)
2644 * ocfs2_inode_lock_tracker(ex=1)
2646 int ocfs2_inode_lock_tracker(struct inode *inode,
2647 struct buffer_head **ret_bh,
2649 struct ocfs2_lock_holder *oh)
2652 struct ocfs2_lock_res *lockres;
2653 struct ocfs2_lock_holder *tmp_oh;
2654 struct pid *pid = task_pid(current);
2657 lockres = &OCFS2_I(inode)->ip_inode_lockres;
2658 tmp_oh = ocfs2_pid_holder(lockres, pid);
2662 * This corresponds to the case 1.
2663 * We haven't got any lock before.
2665 status = ocfs2_inode_lock_full(inode, ret_bh, ex, 0);
2667 if (status != -ENOENT)
2673 ocfs2_add_holder(lockres, oh);
2677 if (unlikely(ex && !tmp_oh->oh_ex)) {
2679 * case 2.2 upgrade may cause dead lock, forbid it.
2681 mlog(ML_ERROR, "Recursive locking is not permitted to "
2682 "upgrade to EX level from PR level.\n");
2688 * case 2.1 OCFS2_META_LOCK_GETBH flag make ocfs2_inode_lock_full.
2689 * ignore the lock level and just update it.
2692 status = ocfs2_inode_lock_full(inode, ret_bh, ex,
2693 OCFS2_META_LOCK_GETBH);
2695 if (status != -ENOENT)
2700 return tmp_oh ? 1 : 0;
2703 void ocfs2_inode_unlock_tracker(struct inode *inode,
2705 struct ocfs2_lock_holder *oh,
2708 struct ocfs2_lock_res *lockres;
2710 lockres = &OCFS2_I(inode)->ip_inode_lockres;
2711 /* had_lock means that the currect process already takes the cluster
2713 * If had_lock is 1, we have nothing to do here.
2714 * If had_lock is 0, we will release the lock.
2717 ocfs2_inode_unlock(inode, oh->oh_ex);
2718 ocfs2_remove_holder(lockres, oh);
2722 int ocfs2_orphan_scan_lock(struct ocfs2_super *osb, u32 *seqno)
2724 struct ocfs2_lock_res *lockres;
2725 struct ocfs2_orphan_scan_lvb *lvb;
2728 if (ocfs2_is_hard_readonly(osb))
2731 if (ocfs2_mount_local(osb))
2734 lockres = &osb->osb_orphan_scan.os_lockres;
2735 status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_EX, 0, 0);
2739 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2740 if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) &&
2741 lvb->lvb_version == OCFS2_ORPHAN_LVB_VERSION)
2742 *seqno = be32_to_cpu(lvb->lvb_os_seqno);
2744 *seqno = osb->osb_orphan_scan.os_seqno + 1;
2749 void ocfs2_orphan_scan_unlock(struct ocfs2_super *osb, u32 seqno)
2751 struct ocfs2_lock_res *lockres;
2752 struct ocfs2_orphan_scan_lvb *lvb;
2754 if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb)) {
2755 lockres = &osb->osb_orphan_scan.os_lockres;
2756 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2757 lvb->lvb_version = OCFS2_ORPHAN_LVB_VERSION;
2758 lvb->lvb_os_seqno = cpu_to_be32(seqno);
2759 ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX);
2763 int ocfs2_super_lock(struct ocfs2_super *osb,
2767 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2768 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
2770 if (ocfs2_is_hard_readonly(osb))
2773 if (ocfs2_mount_local(osb))
2776 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
2782 /* The super block lock path is really in the best position to
2783 * know when resources covered by the lock need to be
2784 * refreshed, so we do it here. Of course, making sense of
2785 * everything is up to the caller :) */
2786 status = ocfs2_should_refresh_lock_res(lockres);
2788 status = ocfs2_refresh_slot_info(osb);
2790 ocfs2_complete_lock_res_refresh(lockres, status);
2793 ocfs2_cluster_unlock(osb, lockres, level);
2796 ocfs2_track_lock_refresh(lockres);
2802 void ocfs2_super_unlock(struct ocfs2_super *osb,
2805 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2806 struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
2808 if (!ocfs2_mount_local(osb))
2809 ocfs2_cluster_unlock(osb, lockres, level);
2812 int ocfs2_rename_lock(struct ocfs2_super *osb)
2815 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
2817 if (ocfs2_is_hard_readonly(osb))
2820 if (ocfs2_mount_local(osb))
2823 status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_EX, 0, 0);
2830 void ocfs2_rename_unlock(struct ocfs2_super *osb)
2832 struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
2834 if (!ocfs2_mount_local(osb))
2835 ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX);
2838 int ocfs2_nfs_sync_lock(struct ocfs2_super *osb, int ex)
2841 struct ocfs2_lock_res *lockres = &osb->osb_nfs_sync_lockres;
2843 if (ocfs2_is_hard_readonly(osb))
2846 if (ocfs2_mount_local(osb))
2849 status = ocfs2_cluster_lock(osb, lockres, ex ? LKM_EXMODE : LKM_PRMODE,
2852 mlog(ML_ERROR, "lock on nfs sync lock failed %d\n", status);
2857 void ocfs2_nfs_sync_unlock(struct ocfs2_super *osb, int ex)
2859 struct ocfs2_lock_res *lockres = &osb->osb_nfs_sync_lockres;
2861 if (!ocfs2_mount_local(osb))
2862 ocfs2_cluster_unlock(osb, lockres,
2863 ex ? LKM_EXMODE : LKM_PRMODE);
2866 int ocfs2_trim_fs_lock(struct ocfs2_super *osb,
2867 struct ocfs2_trim_fs_info *info, int trylock)
2870 struct ocfs2_trim_fs_lvb *lvb;
2871 struct ocfs2_lock_res *lockres = &osb->osb_trim_fs_lockres;
2876 if (ocfs2_is_hard_readonly(osb))
2879 if (ocfs2_mount_local(osb))
2882 status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_EX,
2883 trylock ? DLM_LKF_NOQUEUE : 0, 0);
2885 if (status != -EAGAIN)
2891 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2892 if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) &&
2893 lvb->lvb_version == OCFS2_TRIMFS_LVB_VERSION) {
2895 info->tf_success = lvb->lvb_success;
2896 info->tf_nodenum = be32_to_cpu(lvb->lvb_nodenum);
2897 info->tf_start = be64_to_cpu(lvb->lvb_start);
2898 info->tf_len = be64_to_cpu(lvb->lvb_len);
2899 info->tf_minlen = be64_to_cpu(lvb->lvb_minlen);
2900 info->tf_trimlen = be64_to_cpu(lvb->lvb_trimlen);
2907 void ocfs2_trim_fs_unlock(struct ocfs2_super *osb,
2908 struct ocfs2_trim_fs_info *info)
2910 struct ocfs2_trim_fs_lvb *lvb;
2911 struct ocfs2_lock_res *lockres = &osb->osb_trim_fs_lockres;
2913 if (ocfs2_mount_local(osb))
2917 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
2918 lvb->lvb_version = OCFS2_TRIMFS_LVB_VERSION;
2919 lvb->lvb_success = info->tf_success;
2920 lvb->lvb_nodenum = cpu_to_be32(info->tf_nodenum);
2921 lvb->lvb_start = cpu_to_be64(info->tf_start);
2922 lvb->lvb_len = cpu_to_be64(info->tf_len);
2923 lvb->lvb_minlen = cpu_to_be64(info->tf_minlen);
2924 lvb->lvb_trimlen = cpu_to_be64(info->tf_trimlen);
2927 ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX);
2930 int ocfs2_dentry_lock(struct dentry *dentry, int ex)
2933 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2934 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
2935 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
2939 if (ocfs2_is_hard_readonly(osb)) {
2945 if (ocfs2_mount_local(osb))
2948 ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0);
2955 void ocfs2_dentry_unlock(struct dentry *dentry, int ex)
2957 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
2958 struct ocfs2_dentry_lock *dl = dentry->d_fsdata;
2959 struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb);
2961 if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb))
2962 ocfs2_cluster_unlock(osb, &dl->dl_lockres, level);
2965 /* Reference counting of the dlm debug structure. We want this because
2966 * open references on the debug inodes can live on after a mount, so
2967 * we can't rely on the ocfs2_super to always exist. */
2968 static void ocfs2_dlm_debug_free(struct kref *kref)
2970 struct ocfs2_dlm_debug *dlm_debug;
2972 dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt);
2977 void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug)
2980 kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free);
2983 static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug)
2985 kref_get(&debug->d_refcnt);
2988 struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void)
2990 struct ocfs2_dlm_debug *dlm_debug;
2992 dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL);
2994 mlog_errno(-ENOMEM);
2998 kref_init(&dlm_debug->d_refcnt);
2999 INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking);
3000 dlm_debug->d_locking_state = NULL;
3005 /* Access to this is arbitrated for us via seq_file->sem. */
3006 struct ocfs2_dlm_seq_priv {
3007 struct ocfs2_dlm_debug *p_dlm_debug;
3008 struct ocfs2_lock_res p_iter_res;
3009 struct ocfs2_lock_res p_tmp_res;
3012 static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start,
3013 struct ocfs2_dlm_seq_priv *priv)
3015 struct ocfs2_lock_res *iter, *ret = NULL;
3016 struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug;
3018 assert_spin_locked(&ocfs2_dlm_tracking_lock);
3020 list_for_each_entry(iter, &start->l_debug_list, l_debug_list) {
3021 /* discover the head of the list */
3022 if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) {
3023 mlog(0, "End of list found, %p\n", ret);
3027 /* We track our "dummy" iteration lockres' by a NULL
3029 if (iter->l_ops != NULL) {
3038 static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos)
3040 struct ocfs2_dlm_seq_priv *priv = m->private;
3041 struct ocfs2_lock_res *iter;
3043 spin_lock(&ocfs2_dlm_tracking_lock);
3044 iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv);
3046 /* Since lockres' have the lifetime of their container
3047 * (which can be inodes, ocfs2_supers, etc) we want to
3048 * copy this out to a temporary lockres while still
3049 * under the spinlock. Obviously after this we can't
3050 * trust any pointers on the copy returned, but that's
3051 * ok as the information we want isn't typically held
3053 priv->p_tmp_res = *iter;
3054 iter = &priv->p_tmp_res;
3056 spin_unlock(&ocfs2_dlm_tracking_lock);
3061 static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v)
3065 static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos)
3067 struct ocfs2_dlm_seq_priv *priv = m->private;
3068 struct ocfs2_lock_res *iter = v;
3069 struct ocfs2_lock_res *dummy = &priv->p_iter_res;
3071 spin_lock(&ocfs2_dlm_tracking_lock);
3072 iter = ocfs2_dlm_next_res(iter, priv);
3073 list_del_init(&dummy->l_debug_list);
3075 list_add(&dummy->l_debug_list, &iter->l_debug_list);
3076 priv->p_tmp_res = *iter;
3077 iter = &priv->p_tmp_res;
3079 spin_unlock(&ocfs2_dlm_tracking_lock);
3085 * Version is used by debugfs.ocfs2 to determine the format being used
3088 * - Lock stats printed
3090 * - Max time in lock stats is in usecs (instead of nsecs)
3092 #define OCFS2_DLM_DEBUG_STR_VERSION 3
3093 static int ocfs2_dlm_seq_show(struct seq_file *m, void *v)
3097 struct ocfs2_lock_res *lockres = v;
3102 seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION);
3104 if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY)
3105 seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1,
3107 (unsigned int)ocfs2_get_dentry_lock_ino(lockres));
3109 seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name);
3111 seq_printf(m, "%d\t"
3122 lockres->l_unlock_action,
3123 lockres->l_ro_holders,
3124 lockres->l_ex_holders,
3125 lockres->l_requested,
3126 lockres->l_blocking);
3128 /* Dump the raw LVB */
3129 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
3130 for(i = 0; i < DLM_LVB_LEN; i++)
3131 seq_printf(m, "0x%x\t", lvb[i]);
3133 #ifdef CONFIG_OCFS2_FS_STATS
3134 # define lock_num_prmode(_l) ((_l)->l_lock_prmode.ls_gets)
3135 # define lock_num_exmode(_l) ((_l)->l_lock_exmode.ls_gets)
3136 # define lock_num_prmode_failed(_l) ((_l)->l_lock_prmode.ls_fail)
3137 # define lock_num_exmode_failed(_l) ((_l)->l_lock_exmode.ls_fail)
3138 # define lock_total_prmode(_l) ((_l)->l_lock_prmode.ls_total)
3139 # define lock_total_exmode(_l) ((_l)->l_lock_exmode.ls_total)
3140 # define lock_max_prmode(_l) ((_l)->l_lock_prmode.ls_max)
3141 # define lock_max_exmode(_l) ((_l)->l_lock_exmode.ls_max)
3142 # define lock_refresh(_l) ((_l)->l_lock_refresh)
3144 # define lock_num_prmode(_l) (0)
3145 # define lock_num_exmode(_l) (0)
3146 # define lock_num_prmode_failed(_l) (0)
3147 # define lock_num_exmode_failed(_l) (0)
3148 # define lock_total_prmode(_l) (0ULL)
3149 # define lock_total_exmode(_l) (0ULL)
3150 # define lock_max_prmode(_l) (0)
3151 # define lock_max_exmode(_l) (0)
3152 # define lock_refresh(_l) (0)
3154 /* The following seq_print was added in version 2 of this output */
3155 seq_printf(m, "%u\t"
3164 lock_num_prmode(lockres),
3165 lock_num_exmode(lockres),
3166 lock_num_prmode_failed(lockres),
3167 lock_num_exmode_failed(lockres),
3168 lock_total_prmode(lockres),
3169 lock_total_exmode(lockres),
3170 lock_max_prmode(lockres),
3171 lock_max_exmode(lockres),
3172 lock_refresh(lockres));
3175 seq_printf(m, "\n");
3179 static const struct seq_operations ocfs2_dlm_seq_ops = {
3180 .start = ocfs2_dlm_seq_start,
3181 .stop = ocfs2_dlm_seq_stop,
3182 .next = ocfs2_dlm_seq_next,
3183 .show = ocfs2_dlm_seq_show,
3186 static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file)
3188 struct seq_file *seq = file->private_data;
3189 struct ocfs2_dlm_seq_priv *priv = seq->private;
3190 struct ocfs2_lock_res *res = &priv->p_iter_res;
3192 ocfs2_remove_lockres_tracking(res);
3193 ocfs2_put_dlm_debug(priv->p_dlm_debug);
3194 return seq_release_private(inode, file);
3197 static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file)
3199 struct ocfs2_dlm_seq_priv *priv;
3200 struct ocfs2_super *osb;
3202 priv = __seq_open_private(file, &ocfs2_dlm_seq_ops, sizeof(*priv));
3204 mlog_errno(-ENOMEM);
3208 osb = inode->i_private;
3209 ocfs2_get_dlm_debug(osb->osb_dlm_debug);
3210 priv->p_dlm_debug = osb->osb_dlm_debug;
3211 INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list);
3213 ocfs2_add_lockres_tracking(&priv->p_iter_res,
3219 static const struct file_operations ocfs2_dlm_debug_fops = {
3220 .open = ocfs2_dlm_debug_open,
3221 .release = ocfs2_dlm_debug_release,
3223 .llseek = seq_lseek,
3226 static int ocfs2_dlm_init_debug(struct ocfs2_super *osb)
3229 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
3231 dlm_debug->d_locking_state = debugfs_create_file("locking_state",
3233 osb->osb_debug_root,
3235 &ocfs2_dlm_debug_fops);
3236 if (!dlm_debug->d_locking_state) {
3239 "Unable to create locking state debugfs file.\n");
3243 ocfs2_get_dlm_debug(dlm_debug);
3248 static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb)
3250 struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
3253 debugfs_remove(dlm_debug->d_locking_state);
3254 ocfs2_put_dlm_debug(dlm_debug);
3258 int ocfs2_dlm_init(struct ocfs2_super *osb)
3261 struct ocfs2_cluster_connection *conn = NULL;
3263 if (ocfs2_mount_local(osb)) {
3268 status = ocfs2_dlm_init_debug(osb);
3274 /* launch downconvert thread */
3275 osb->dc_task = kthread_run(ocfs2_downconvert_thread, osb, "ocfs2dc-%s",
3277 if (IS_ERR(osb->dc_task)) {
3278 status = PTR_ERR(osb->dc_task);
3279 osb->dc_task = NULL;
3284 /* for now, uuid == domain */
3285 status = ocfs2_cluster_connect(osb->osb_cluster_stack,
3286 osb->osb_cluster_name,
3287 strlen(osb->osb_cluster_name),
3289 strlen(osb->uuid_str),
3290 &lproto, ocfs2_do_node_down, osb,
3297 status = ocfs2_cluster_this_node(conn, &osb->node_num);
3301 "could not find this host's node number\n");
3302 ocfs2_cluster_disconnect(conn, 0);
3307 ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
3308 ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
3309 ocfs2_nfs_sync_lock_res_init(&osb->osb_nfs_sync_lockres, osb);
3310 ocfs2_orphan_scan_lock_res_init(&osb->osb_orphan_scan.os_lockres, osb);
3315 ocfs2_dlm_shutdown_debug(osb);
3317 kthread_stop(osb->dc_task);
3323 void ocfs2_dlm_shutdown(struct ocfs2_super *osb,
3326 ocfs2_drop_osb_locks(osb);
3329 * Now that we have dropped all locks and ocfs2_dismount_volume()
3330 * has disabled recovery, the DLM won't be talking to us. It's
3331 * safe to tear things down before disconnecting the cluster.
3335 kthread_stop(osb->dc_task);
3336 osb->dc_task = NULL;
3339 ocfs2_lock_res_free(&osb->osb_super_lockres);
3340 ocfs2_lock_res_free(&osb->osb_rename_lockres);
3341 ocfs2_lock_res_free(&osb->osb_nfs_sync_lockres);
3342 ocfs2_lock_res_free(&osb->osb_orphan_scan.os_lockres);
3344 ocfs2_cluster_disconnect(osb->cconn, hangup_pending);
3347 ocfs2_dlm_shutdown_debug(osb);
3350 static int ocfs2_drop_lock(struct ocfs2_super *osb,
3351 struct ocfs2_lock_res *lockres)
3354 unsigned long flags;
3357 /* We didn't get anywhere near actually using this lockres. */
3358 if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED))
3361 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
3362 lkm_flags |= DLM_LKF_VALBLK;
3364 spin_lock_irqsave(&lockres->l_lock, flags);
3366 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING),
3367 "lockres %s, flags 0x%lx\n",
3368 lockres->l_name, lockres->l_flags);
3370 while (lockres->l_flags & OCFS2_LOCK_BUSY) {
3371 mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = "
3372 "%u, unlock_action = %u\n",
3373 lockres->l_name, lockres->l_flags, lockres->l_action,
3374 lockres->l_unlock_action);
3376 spin_unlock_irqrestore(&lockres->l_lock, flags);
3378 /* XXX: Today we just wait on any busy
3379 * locks... Perhaps we need to cancel converts in the
3381 ocfs2_wait_on_busy_lock(lockres);
3383 spin_lock_irqsave(&lockres->l_lock, flags);
3386 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) {
3387 if (lockres->l_flags & OCFS2_LOCK_ATTACHED &&
3388 lockres->l_level == DLM_LOCK_EX &&
3389 !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
3390 lockres->l_ops->set_lvb(lockres);
3393 if (lockres->l_flags & OCFS2_LOCK_BUSY)
3394 mlog(ML_ERROR, "destroying busy lock: \"%s\"\n",
3396 if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
3397 mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name);
3399 if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
3400 spin_unlock_irqrestore(&lockres->l_lock, flags);
3404 lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED);
3406 /* make sure we never get here while waiting for an ast to
3408 BUG_ON(lockres->l_action != OCFS2_AST_INVALID);
3410 /* is this necessary? */
3411 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
3412 lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK;
3413 spin_unlock_irqrestore(&lockres->l_lock, flags);
3415 mlog(0, "lock %s\n", lockres->l_name);
3417 ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags);
3419 ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres);
3420 mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags);
3421 ocfs2_dlm_dump_lksb(&lockres->l_lksb);
3424 mlog(0, "lock %s, successful return from ocfs2_dlm_unlock\n",
3427 ocfs2_wait_on_busy_lock(lockres);
3432 static void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
3433 struct ocfs2_lock_res *lockres);
3435 /* Mark the lockres as being dropped. It will no longer be
3436 * queued if blocking, but we still may have to wait on it
3437 * being dequeued from the downconvert thread before we can consider
3440 * You can *not* attempt to call cluster_lock on this lockres anymore. */
3441 void ocfs2_mark_lockres_freeing(struct ocfs2_super *osb,
3442 struct ocfs2_lock_res *lockres)
3445 struct ocfs2_mask_waiter mw;
3446 unsigned long flags, flags2;
3448 ocfs2_init_mask_waiter(&mw);
3450 spin_lock_irqsave(&lockres->l_lock, flags);
3451 lockres->l_flags |= OCFS2_LOCK_FREEING;
3452 if (lockres->l_flags & OCFS2_LOCK_QUEUED && current == osb->dc_task) {
3454 * We know the downconvert is queued but not in progress
3455 * because we are the downconvert thread and processing
3456 * different lock. So we can just remove the lock from the
3457 * queue. This is not only an optimization but also a way
3458 * to avoid the following deadlock:
3459 * ocfs2_dentry_post_unlock()
3460 * ocfs2_dentry_lock_put()
3461 * ocfs2_drop_dentry_lock()
3463 * ocfs2_evict_inode()
3464 * ocfs2_clear_inode()
3465 * ocfs2_mark_lockres_freeing()
3466 * ... blocks waiting for OCFS2_LOCK_QUEUED
3467 * since we are the downconvert thread which
3468 * should clear the flag.
3470 spin_unlock_irqrestore(&lockres->l_lock, flags);
3471 spin_lock_irqsave(&osb->dc_task_lock, flags2);
3472 list_del_init(&lockres->l_blocked_list);
3473 osb->blocked_lock_count--;
3474 spin_unlock_irqrestore(&osb->dc_task_lock, flags2);
3476 * Warn if we recurse into another post_unlock call. Strictly
3477 * speaking it isn't a problem but we need to be careful if
3478 * that happens (stack overflow, deadlocks, ...) so warn if
3479 * ocfs2 grows a path for which this can happen.
3481 WARN_ON_ONCE(lockres->l_ops->post_unlock);
3482 /* Since the lock is freeing we don't do much in the fn below */
3483 ocfs2_process_blocked_lock(osb, lockres);
3486 while (lockres->l_flags & OCFS2_LOCK_QUEUED) {
3487 lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0);
3488 spin_unlock_irqrestore(&lockres->l_lock, flags);
3490 mlog(0, "Waiting on lockres %s\n", lockres->l_name);
3492 status = ocfs2_wait_for_mask(&mw);
3496 spin_lock_irqsave(&lockres->l_lock, flags);
3498 spin_unlock_irqrestore(&lockres->l_lock, flags);
3501 void ocfs2_simple_drop_lockres(struct ocfs2_super *osb,
3502 struct ocfs2_lock_res *lockres)
3506 ocfs2_mark_lockres_freeing(osb, lockres);
3507 ret = ocfs2_drop_lock(osb, lockres);
3512 static void ocfs2_drop_osb_locks(struct ocfs2_super *osb)
3514 ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres);
3515 ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres);
3516 ocfs2_simple_drop_lockres(osb, &osb->osb_nfs_sync_lockres);
3517 ocfs2_simple_drop_lockres(osb, &osb->osb_orphan_scan.os_lockres);
3520 int ocfs2_drop_inode_locks(struct inode *inode)
3524 /* No need to call ocfs2_mark_lockres_freeing here -
3525 * ocfs2_clear_inode has done it for us. */
3527 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
3528 &OCFS2_I(inode)->ip_open_lockres);
3534 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
3535 &OCFS2_I(inode)->ip_inode_lockres);
3538 if (err < 0 && !status)
3541 err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
3542 &OCFS2_I(inode)->ip_rw_lockres);
3545 if (err < 0 && !status)
3551 static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
3554 assert_spin_locked(&lockres->l_lock);
3556 BUG_ON(lockres->l_blocking <= DLM_LOCK_NL);
3558 if (lockres->l_level <= new_level) {
3559 mlog(ML_ERROR, "lockres %s, lvl %d <= %d, blcklst %d, mask %d, "
3560 "type %d, flags 0x%lx, hold %d %d, act %d %d, req %d, "
3561 "block %d, pgen %d\n", lockres->l_name, lockres->l_level,
3562 new_level, list_empty(&lockres->l_blocked_list),
3563 list_empty(&lockres->l_mask_waiters), lockres->l_type,
3564 lockres->l_flags, lockres->l_ro_holders,
3565 lockres->l_ex_holders, lockres->l_action,
3566 lockres->l_unlock_action, lockres->l_requested,
3567 lockres->l_blocking, lockres->l_pending_gen);
3571 mlog(ML_BASTS, "lockres %s, level %d => %d, blocking %d\n",
3572 lockres->l_name, lockres->l_level, new_level, lockres->l_blocking);
3574 lockres->l_action = OCFS2_AST_DOWNCONVERT;
3575 lockres->l_requested = new_level;
3576 lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
3577 return lockres_set_pending(lockres);
3580 static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
3581 struct ocfs2_lock_res *lockres,
3584 unsigned int generation)
3587 u32 dlm_flags = DLM_LKF_CONVERT;
3589 mlog(ML_BASTS, "lockres %s, level %d => %d\n", lockres->l_name,
3590 lockres->l_level, new_level);
3593 * On DLM_LKF_VALBLK, fsdlm behaves differently with o2cb. It always
3594 * expects DLM_LKF_VALBLK being set if the LKB has LVB, so that
3595 * we can recover correctly from node failure. Otherwise, we may get
3596 * invalid LVB in LKB, but without DLM_SBF_VALNOTVALID being set.
3598 if (ocfs2_userspace_stack(osb) &&
3599 lockres->l_ops->flags & LOCK_TYPE_USES_LVB)
3603 dlm_flags |= DLM_LKF_VALBLK;
3605 ret = ocfs2_dlm_lock(osb->cconn,
3610 OCFS2_LOCK_ID_MAX_LEN - 1);
3611 lockres_clear_pending(lockres, generation, osb);
3613 ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
3614 ocfs2_recover_from_dlm_error(lockres, 1);
3623 /* returns 1 when the caller should unlock and call ocfs2_dlm_unlock */
3624 static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
3625 struct ocfs2_lock_res *lockres)
3627 assert_spin_locked(&lockres->l_lock);
3629 if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
3630 /* If we're already trying to cancel a lock conversion
3631 * then just drop the spinlock and allow the caller to
3632 * requeue this lock. */
3633 mlog(ML_BASTS, "lockres %s, skip convert\n", lockres->l_name);
3637 /* were we in a convert when we got the bast fire? */
3638 BUG_ON(lockres->l_action != OCFS2_AST_CONVERT &&
3639 lockres->l_action != OCFS2_AST_DOWNCONVERT);
3640 /* set things up for the unlockast to know to just
3641 * clear out the ast_action and unset busy, etc. */
3642 lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT;
3644 mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY),
3645 "lock %s, invalid flags: 0x%lx\n",
3646 lockres->l_name, lockres->l_flags);
3648 mlog(ML_BASTS, "lockres %s\n", lockres->l_name);
3653 static int ocfs2_cancel_convert(struct ocfs2_super *osb,
3654 struct ocfs2_lock_res *lockres)
3658 ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb,
3661 ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres);
3662 ocfs2_recover_from_dlm_error(lockres, 0);
3665 mlog(ML_BASTS, "lockres %s\n", lockres->l_name);
3670 static int ocfs2_unblock_lock(struct ocfs2_super *osb,
3671 struct ocfs2_lock_res *lockres,
3672 struct ocfs2_unblock_ctl *ctl)
3674 unsigned long flags;
3682 spin_lock_irqsave(&lockres->l_lock, flags);
3686 * Is it still blocking? If not, we have no more work to do.
3688 if (!(lockres->l_flags & OCFS2_LOCK_BLOCKED)) {
3689 BUG_ON(lockres->l_blocking != DLM_LOCK_NL);
3690 spin_unlock_irqrestore(&lockres->l_lock, flags);
3695 if (lockres->l_flags & OCFS2_LOCK_BUSY) {
3697 * This is a *big* race. The OCFS2_LOCK_PENDING flag
3698 * exists entirely for one reason - another thread has set
3699 * OCFS2_LOCK_BUSY, but has *NOT* yet called dlm_lock().
3701 * If we do ocfs2_cancel_convert() before the other thread
3702 * calls dlm_lock(), our cancel will do nothing. We will
3703 * get no ast, and we will have no way of knowing the
3704 * cancel failed. Meanwhile, the other thread will call
3705 * into dlm_lock() and wait...forever.
3707 * Why forever? Because another node has asked for the
3708 * lock first; that's why we're here in unblock_lock().
3710 * The solution is OCFS2_LOCK_PENDING. When PENDING is
3711 * set, we just requeue the unblock. Only when the other
3712 * thread has called dlm_lock() and cleared PENDING will
3713 * we then cancel their request.
3715 * All callers of dlm_lock() must set OCFS2_DLM_PENDING
3716 * at the same time they set OCFS2_DLM_BUSY. They must
3717 * clear OCFS2_DLM_PENDING after dlm_lock() returns.
3719 if (lockres->l_flags & OCFS2_LOCK_PENDING) {
3720 mlog(ML_BASTS, "lockres %s, ReQ: Pending\n",
3726 ret = ocfs2_prepare_cancel_convert(osb, lockres);
3727 spin_unlock_irqrestore(&lockres->l_lock, flags);
3729 ret = ocfs2_cancel_convert(osb, lockres);
3737 * This prevents livelocks. OCFS2_LOCK_UPCONVERT_FINISHING flag is
3738 * set when the ast is received for an upconvert just before the
3739 * OCFS2_LOCK_BUSY flag is cleared. Now if the fs received a bast
3740 * on the heels of the ast, we want to delay the downconvert just
3741 * enough to allow the up requestor to do its task. Because this
3742 * lock is in the blocked queue, the lock will be downconverted
3743 * as soon as the requestor is done with the lock.
3745 if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING)
3749 * How can we block and yet be at NL? We were trying to upconvert
3750 * from NL and got canceled. The code comes back here, and now
3751 * we notice and clear BLOCKING.
3753 if (lockres->l_level == DLM_LOCK_NL) {
3754 BUG_ON(lockres->l_ex_holders || lockres->l_ro_holders);
3755 mlog(ML_BASTS, "lockres %s, Aborting dc\n", lockres->l_name);
3756 lockres->l_blocking = DLM_LOCK_NL;
3757 lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
3758 spin_unlock_irqrestore(&lockres->l_lock, flags);
3762 /* if we're blocking an exclusive and we have *any* holders,
3764 if ((lockres->l_blocking == DLM_LOCK_EX)
3765 && (lockres->l_ex_holders || lockres->l_ro_holders)) {
3766 mlog(ML_BASTS, "lockres %s, ReQ: EX/PR Holders %u,%u\n",
3767 lockres->l_name, lockres->l_ex_holders,
3768 lockres->l_ro_holders);
3772 /* If it's a PR we're blocking, then only
3773 * requeue if we've got any EX holders */
3774 if (lockres->l_blocking == DLM_LOCK_PR &&
3775 lockres->l_ex_holders) {
3776 mlog(ML_BASTS, "lockres %s, ReQ: EX Holders %u\n",
3777 lockres->l_name, lockres->l_ex_holders);
3782 * Can we get a lock in this state if the holder counts are
3783 * zero? The meta data unblock code used to check this.
3785 if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
3786 && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) {
3787 mlog(ML_BASTS, "lockres %s, ReQ: Lock Refreshing\n",
3792 new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
3794 if (lockres->l_ops->check_downconvert
3795 && !lockres->l_ops->check_downconvert(lockres, new_level)) {
3796 mlog(ML_BASTS, "lockres %s, ReQ: Checkpointing\n",
3801 /* If we get here, then we know that there are no more
3802 * incompatible holders (and anyone asking for an incompatible
3803 * lock is blocked). We can now downconvert the lock */
3804 if (!lockres->l_ops->downconvert_worker)
3807 /* Some lockres types want to do a bit of work before
3808 * downconverting a lock. Allow that here. The worker function
3809 * may sleep, so we save off a copy of what we're blocking as
3810 * it may change while we're not holding the spin lock. */
3811 blocking = lockres->l_blocking;
3812 level = lockres->l_level;
3813 spin_unlock_irqrestore(&lockres->l_lock, flags);
3815 ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking);
3817 if (ctl->unblock_action == UNBLOCK_STOP_POST) {
3818 mlog(ML_BASTS, "lockres %s, UNBLOCK_STOP_POST\n",
3823 spin_lock_irqsave(&lockres->l_lock, flags);
3824 if ((blocking != lockres->l_blocking) || (level != lockres->l_level)) {
3825 /* If this changed underneath us, then we can't drop
3827 mlog(ML_BASTS, "lockres %s, block=%d:%d, level=%d:%d, "
3828 "Recheck\n", lockres->l_name, blocking,
3829 lockres->l_blocking, level, lockres->l_level);
3836 if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) {
3837 if (lockres->l_level == DLM_LOCK_EX)
3841 * We only set the lvb if the lock has been fully
3842 * refreshed - otherwise we risk setting stale
3843 * data. Otherwise, there's no need to actually clear
3844 * out the lvb here as it's value is still valid.
3846 if (set_lvb && !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
3847 lockres->l_ops->set_lvb(lockres);
3850 gen = ocfs2_prepare_downconvert(lockres, new_level);
3851 spin_unlock_irqrestore(&lockres->l_lock, flags);
3852 ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb,
3861 spin_unlock_irqrestore(&lockres->l_lock, flags);
3867 static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
3870 struct inode *inode;
3871 struct address_space *mapping;
3872 struct ocfs2_inode_info *oi;
3874 inode = ocfs2_lock_res_inode(lockres);
3875 mapping = inode->i_mapping;
3877 if (S_ISDIR(inode->i_mode)) {
3878 oi = OCFS2_I(inode);
3879 oi->ip_dir_lock_gen++;
3880 mlog(0, "generation: %u\n", oi->ip_dir_lock_gen);
3884 if (!S_ISREG(inode->i_mode))
3888 * We need this before the filemap_fdatawrite() so that it can
3889 * transfer the dirty bit from the PTE to the
3890 * page. Unfortunately this means that even for EX->PR
3891 * downconverts, we'll lose our mappings and have to build
3894 unmap_mapping_range(mapping, 0, 0, 0);
3896 if (filemap_fdatawrite(mapping)) {
3897 mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",
3898 (unsigned long long)OCFS2_I(inode)->ip_blkno);
3900 sync_mapping_buffers(mapping);
3901 if (blocking == DLM_LOCK_EX) {
3902 truncate_inode_pages(mapping, 0);
3904 /* We only need to wait on the I/O if we're not also
3905 * truncating pages because truncate_inode_pages waits
3906 * for us above. We don't truncate pages if we're
3907 * blocking anything < EXMODE because we want to keep
3908 * them around in that case. */
3909 filemap_fdatawait(mapping);
3912 forget_all_cached_acls(inode);
3915 return UNBLOCK_CONTINUE;
3918 static int ocfs2_ci_checkpointed(struct ocfs2_caching_info *ci,
3919 struct ocfs2_lock_res *lockres,
3922 int checkpointed = ocfs2_ci_fully_checkpointed(ci);
3924 BUG_ON(new_level != DLM_LOCK_NL && new_level != DLM_LOCK_PR);
3925 BUG_ON(lockres->l_level != DLM_LOCK_EX && !checkpointed);
3930 ocfs2_start_checkpoint(OCFS2_SB(ocfs2_metadata_cache_get_super(ci)));
3934 static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres,
3937 struct inode *inode = ocfs2_lock_res_inode(lockres);
3939 return ocfs2_ci_checkpointed(INODE_CACHE(inode), lockres, new_level);
3942 static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres)
3944 struct inode *inode = ocfs2_lock_res_inode(lockres);
3946 __ocfs2_stuff_meta_lvb(inode);
3950 * Does the final reference drop on our dentry lock. Right now this
3951 * happens in the downconvert thread, but we could choose to simplify the
3952 * dlmglue API and push these off to the ocfs2_wq in the future.
3954 static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb,
3955 struct ocfs2_lock_res *lockres)
3957 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
3958 ocfs2_dentry_lock_put(osb, dl);
3962 * d_delete() matching dentries before the lock downconvert.
3964 * At this point, any process waiting to destroy the
3965 * dentry_lock due to last ref count is stopped by the
3966 * OCFS2_LOCK_QUEUED flag.
3968 * We have two potential problems
3970 * 1) If we do the last reference drop on our dentry_lock (via dput)
3971 * we'll wind up in ocfs2_release_dentry_lock(), waiting on
3972 * the downconvert to finish. Instead we take an elevated
3973 * reference and push the drop until after we've completed our
3974 * unblock processing.
3976 * 2) There might be another process with a final reference,
3977 * waiting on us to finish processing. If this is the case, we
3978 * detect it and exit out - there's no more dentries anyway.
3980 static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres,
3983 struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres);
3984 struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode);
3985 struct dentry *dentry;
3986 unsigned long flags;
3990 * This node is blocking another node from getting a read
3991 * lock. This happens when we've renamed within a
3992 * directory. We've forced the other nodes to d_delete(), but
3993 * we never actually dropped our lock because it's still
3994 * valid. The downconvert code will retain a PR for this node,
3995 * so there's no further work to do.
3997 if (blocking == DLM_LOCK_PR)
3998 return UNBLOCK_CONTINUE;
4001 * Mark this inode as potentially orphaned. The code in
4002 * ocfs2_delete_inode() will figure out whether it actually
4003 * needs to be freed or not.
4005 spin_lock(&oi->ip_lock);
4006 oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
4007 spin_unlock(&oi->ip_lock);
4010 * Yuck. We need to make sure however that the check of
4011 * OCFS2_LOCK_FREEING and the extra reference are atomic with
4012 * respect to a reference decrement or the setting of that
4015 spin_lock_irqsave(&lockres->l_lock, flags);
4016 spin_lock(&dentry_attach_lock);
4017 if (!(lockres->l_flags & OCFS2_LOCK_FREEING)
4022 spin_unlock(&dentry_attach_lock);
4023 spin_unlock_irqrestore(&lockres->l_lock, flags);
4025 mlog(0, "extra_ref = %d\n", extra_ref);
4028 * We have a process waiting on us in ocfs2_dentry_iput(),
4029 * which means we can't have any more outstanding
4030 * aliases. There's no need to do any more work.
4033 return UNBLOCK_CONTINUE;
4035 spin_lock(&dentry_attach_lock);
4037 dentry = ocfs2_find_local_alias(dl->dl_inode,
4038 dl->dl_parent_blkno, 1);
4041 spin_unlock(&dentry_attach_lock);
4043 if (S_ISDIR(dl->dl_inode->i_mode))
4044 shrink_dcache_parent(dentry);
4046 mlog(0, "d_delete(%pd);\n", dentry);
4049 * The following dcache calls may do an
4050 * iput(). Normally we don't want that from the
4051 * downconverting thread, but in this case it's ok
4052 * because the requesting node already has an
4053 * exclusive lock on the inode, so it can't be queued
4054 * for a downconvert.
4059 spin_lock(&dentry_attach_lock);
4061 spin_unlock(&dentry_attach_lock);
4064 * If we are the last holder of this dentry lock, there is no
4065 * reason to downconvert so skip straight to the unlock.
4067 if (dl->dl_count == 1)
4068 return UNBLOCK_STOP_POST;
4070 return UNBLOCK_CONTINUE_POST;
4073 static int ocfs2_check_refcount_downconvert(struct ocfs2_lock_res *lockres,
4076 struct ocfs2_refcount_tree *tree =
4077 ocfs2_lock_res_refcount_tree(lockres);
4079 return ocfs2_ci_checkpointed(&tree->rf_ci, lockres, new_level);
4082 static int ocfs2_refcount_convert_worker(struct ocfs2_lock_res *lockres,
4085 struct ocfs2_refcount_tree *tree =
4086 ocfs2_lock_res_refcount_tree(lockres);
4088 ocfs2_metadata_cache_purge(&tree->rf_ci);
4090 return UNBLOCK_CONTINUE;
4093 static void ocfs2_set_qinfo_lvb(struct ocfs2_lock_res *lockres)
4095 struct ocfs2_qinfo_lvb *lvb;
4096 struct ocfs2_mem_dqinfo *oinfo = ocfs2_lock_res_qinfo(lockres);
4097 struct mem_dqinfo *info = sb_dqinfo(oinfo->dqi_gi.dqi_sb,
4098 oinfo->dqi_gi.dqi_type);
4100 lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
4101 lvb->lvb_version = OCFS2_QINFO_LVB_VERSION;
4102 lvb->lvb_bgrace = cpu_to_be32(info->dqi_bgrace);
4103 lvb->lvb_igrace = cpu_to_be32(info->dqi_igrace);
4104 lvb->lvb_syncms = cpu_to_be32(oinfo->dqi_syncms);
4105 lvb->lvb_blocks = cpu_to_be32(oinfo->dqi_gi.dqi_blocks);
4106 lvb->lvb_free_blk = cpu_to_be32(oinfo->dqi_gi.dqi_free_blk);
4107 lvb->lvb_free_entry = cpu_to_be32(oinfo->dqi_gi.dqi_free_entry);
4110 void ocfs2_qinfo_unlock(struct ocfs2_mem_dqinfo *oinfo, int ex)
4112 struct ocfs2_lock_res *lockres = &oinfo->dqi_gqlock;
4113 struct ocfs2_super *osb = OCFS2_SB(oinfo->dqi_gi.dqi_sb);
4114 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
4116 if (!ocfs2_is_hard_readonly(osb) && !ocfs2_mount_local(osb))
4117 ocfs2_cluster_unlock(osb, lockres, level);
4120 static int ocfs2_refresh_qinfo(struct ocfs2_mem_dqinfo *oinfo)
4122 struct mem_dqinfo *info = sb_dqinfo(oinfo->dqi_gi.dqi_sb,
4123 oinfo->dqi_gi.dqi_type);
4124 struct ocfs2_lock_res *lockres = &oinfo->dqi_gqlock;
4125 struct ocfs2_qinfo_lvb *lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
4126 struct buffer_head *bh = NULL;
4127 struct ocfs2_global_disk_dqinfo *gdinfo;
4130 if (ocfs2_dlm_lvb_valid(&lockres->l_lksb) &&
4131 lvb->lvb_version == OCFS2_QINFO_LVB_VERSION) {
4132 info->dqi_bgrace = be32_to_cpu(lvb->lvb_bgrace);
4133 info->dqi_igrace = be32_to_cpu(lvb->lvb_igrace);
4134 oinfo->dqi_syncms = be32_to_cpu(lvb->lvb_syncms);
4135 oinfo->dqi_gi.dqi_blocks = be32_to_cpu(lvb->lvb_blocks);
4136 oinfo->dqi_gi.dqi_free_blk = be32_to_cpu(lvb->lvb_free_blk);
4137 oinfo->dqi_gi.dqi_free_entry =
4138 be32_to_cpu(lvb->lvb_free_entry);
4140 status = ocfs2_read_quota_phys_block(oinfo->dqi_gqinode,
4141 oinfo->dqi_giblk, &bh);
4146 gdinfo = (struct ocfs2_global_disk_dqinfo *)
4147 (bh->b_data + OCFS2_GLOBAL_INFO_OFF);
4148 info->dqi_bgrace = le32_to_cpu(gdinfo->dqi_bgrace);
4149 info->dqi_igrace = le32_to_cpu(gdinfo->dqi_igrace);
4150 oinfo->dqi_syncms = le32_to_cpu(gdinfo->dqi_syncms);
4151 oinfo->dqi_gi.dqi_blocks = le32_to_cpu(gdinfo->dqi_blocks);
4152 oinfo->dqi_gi.dqi_free_blk = le32_to_cpu(gdinfo->dqi_free_blk);
4153 oinfo->dqi_gi.dqi_free_entry =
4154 le32_to_cpu(gdinfo->dqi_free_entry);
4156 ocfs2_track_lock_refresh(lockres);
4163 /* Lock quota info, this function expects at least shared lock on the quota file
4164 * so that we can safely refresh quota info from disk. */
4165 int ocfs2_qinfo_lock(struct ocfs2_mem_dqinfo *oinfo, int ex)
4167 struct ocfs2_lock_res *lockres = &oinfo->dqi_gqlock;
4168 struct ocfs2_super *osb = OCFS2_SB(oinfo->dqi_gi.dqi_sb);
4169 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
4172 /* On RO devices, locking really isn't needed... */
4173 if (ocfs2_is_hard_readonly(osb)) {
4178 if (ocfs2_mount_local(osb))
4181 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
4186 if (!ocfs2_should_refresh_lock_res(lockres))
4188 /* OK, we have the lock but we need to refresh the quota info */
4189 status = ocfs2_refresh_qinfo(oinfo);
4191 ocfs2_qinfo_unlock(oinfo, ex);
4192 ocfs2_complete_lock_res_refresh(lockres, status);
4197 int ocfs2_refcount_lock(struct ocfs2_refcount_tree *ref_tree, int ex)
4200 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
4201 struct ocfs2_lock_res *lockres = &ref_tree->rf_lockres;
4202 struct ocfs2_super *osb = lockres->l_priv;
4205 if (ocfs2_is_hard_readonly(osb))
4208 if (ocfs2_mount_local(osb))
4211 status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
4218 void ocfs2_refcount_unlock(struct ocfs2_refcount_tree *ref_tree, int ex)
4220 int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR;
4221 struct ocfs2_lock_res *lockres = &ref_tree->rf_lockres;
4222 struct ocfs2_super *osb = lockres->l_priv;
4224 if (!ocfs2_mount_local(osb))
4225 ocfs2_cluster_unlock(osb, lockres, level);
4228 static void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
4229 struct ocfs2_lock_res *lockres)
4232 struct ocfs2_unblock_ctl ctl = {0, 0,};
4233 unsigned long flags;
4235 /* Our reference to the lockres in this function can be
4236 * considered valid until we remove the OCFS2_LOCK_QUEUED
4240 BUG_ON(!lockres->l_ops);
4242 mlog(ML_BASTS, "lockres %s blocked\n", lockres->l_name);
4244 /* Detect whether a lock has been marked as going away while
4245 * the downconvert thread was processing other things. A lock can
4246 * still be marked with OCFS2_LOCK_FREEING after this check,
4247 * but short circuiting here will still save us some
4249 spin_lock_irqsave(&lockres->l_lock, flags);
4250 if (lockres->l_flags & OCFS2_LOCK_FREEING)
4252 spin_unlock_irqrestore(&lockres->l_lock, flags);
4254 status = ocfs2_unblock_lock(osb, lockres, &ctl);
4258 spin_lock_irqsave(&lockres->l_lock, flags);
4260 if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) {
4261 lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED);
4263 ocfs2_schedule_blocked_lock(osb, lockres);
4265 mlog(ML_BASTS, "lockres %s, requeue = %s.\n", lockres->l_name,
4266 ctl.requeue ? "yes" : "no");
4267 spin_unlock_irqrestore(&lockres->l_lock, flags);
4269 if (ctl.unblock_action != UNBLOCK_CONTINUE
4270 && lockres->l_ops->post_unlock)
4271 lockres->l_ops->post_unlock(osb, lockres);
4274 static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
4275 struct ocfs2_lock_res *lockres)
4277 unsigned long flags;
4279 assert_spin_locked(&lockres->l_lock);
4281 if (lockres->l_flags & OCFS2_LOCK_FREEING) {
4282 /* Do not schedule a lock for downconvert when it's on
4283 * the way to destruction - any nodes wanting access
4284 * to the resource will get it soon. */
4285 mlog(ML_BASTS, "lockres %s won't be scheduled: flags 0x%lx\n",
4286 lockres->l_name, lockres->l_flags);
4290 lockres_or_flags(lockres, OCFS2_LOCK_QUEUED);
4292 spin_lock_irqsave(&osb->dc_task_lock, flags);
4293 if (list_empty(&lockres->l_blocked_list)) {
4294 list_add_tail(&lockres->l_blocked_list,
4295 &osb->blocked_lock_list);
4296 osb->blocked_lock_count++;
4298 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4301 static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb)
4303 unsigned long processed;
4304 unsigned long flags;
4305 struct ocfs2_lock_res *lockres;
4307 spin_lock_irqsave(&osb->dc_task_lock, flags);
4308 /* grab this early so we know to try again if a state change and
4309 * wake happens part-way through our work */
4310 osb->dc_work_sequence = osb->dc_wake_sequence;
4312 processed = osb->blocked_lock_count;
4314 * blocked lock processing in this loop might call iput which can
4315 * remove items off osb->blocked_lock_list. Downconvert up to
4316 * 'processed' number of locks, but stop short if we had some
4317 * removed in ocfs2_mark_lockres_freeing when downconverting.
4319 while (processed && !list_empty(&osb->blocked_lock_list)) {
4320 lockres = list_entry(osb->blocked_lock_list.next,
4321 struct ocfs2_lock_res, l_blocked_list);
4322 list_del_init(&lockres->l_blocked_list);
4323 osb->blocked_lock_count--;
4324 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4329 ocfs2_process_blocked_lock(osb, lockres);
4331 spin_lock_irqsave(&osb->dc_task_lock, flags);
4333 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4336 static int ocfs2_downconvert_thread_lists_empty(struct ocfs2_super *osb)
4339 unsigned long flags;
4341 spin_lock_irqsave(&osb->dc_task_lock, flags);
4342 if (list_empty(&osb->blocked_lock_list))
4345 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4349 static int ocfs2_downconvert_thread_should_wake(struct ocfs2_super *osb)
4351 int should_wake = 0;
4352 unsigned long flags;
4354 spin_lock_irqsave(&osb->dc_task_lock, flags);
4355 if (osb->dc_work_sequence != osb->dc_wake_sequence)
4357 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4362 static int ocfs2_downconvert_thread(void *arg)
4365 struct ocfs2_super *osb = arg;
4367 /* only quit once we've been asked to stop and there is no more
4369 while (!(kthread_should_stop() &&
4370 ocfs2_downconvert_thread_lists_empty(osb))) {
4372 wait_event_interruptible(osb->dc_event,
4373 ocfs2_downconvert_thread_should_wake(osb) ||
4374 kthread_should_stop());
4376 mlog(0, "downconvert_thread: awoken\n");
4378 ocfs2_downconvert_thread_do_work(osb);
4381 osb->dc_task = NULL;
4385 void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb)
4387 unsigned long flags;
4389 spin_lock_irqsave(&osb->dc_task_lock, flags);
4390 /* make sure the voting thread gets a swipe at whatever changes
4391 * the caller may have made to the voting state */
4392 osb->dc_wake_sequence++;
4393 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
4394 wake_up(&osb->dc_event);