4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
7 * Deadlock detection added.
8 * FIXME: one thing isn't handled yet:
9 * - mandatory locks (requires lots of changes elsewhere)
10 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
15 * Converted file_lock_table to a linked list from an array, which eliminates
16 * the limits on how many active file locks are open.
17 * Chad Page (pageone@netcom.com), November 27, 1994
19 * Removed dependency on file descriptors. dup()'ed file descriptors now
20 * get the same locks as the original file descriptors, and a close() on
21 * any file descriptor removes ALL the locks on the file for the current
22 * process. Since locks still depend on the process id, locks are inherited
23 * after an exec() but not after a fork(). This agrees with POSIX, and both
24 * BSD and SVR4 practice.
25 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
27 * Scrapped free list which is redundant now that we allocate locks
28 * dynamically with kmalloc()/kfree().
29 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
36 * FL_FLOCK locks are created with calls to flock(), through the flock()
37 * system call, which is new. Old C libraries implement flock() via fcntl()
38 * and will continue to use the old, broken implementation.
40 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
41 * with a file pointer (filp). As a result they can be shared by a parent
42 * process and its children after a fork(). They are removed when the last
43 * file descriptor referring to the file pointer is closed (unless explicitly
46 * FL_FLOCK locks never deadlock, an existing lock is always removed before
47 * upgrading from shared to exclusive (or vice versa). When this happens
48 * any processes blocked by the current lock are woken up and allowed to
49 * run before the new lock is applied.
50 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
52 * Removed some race conditions in flock_lock_file(), marked other possible
53 * races. Just grep for FIXME to see them.
54 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
56 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
57 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
58 * once we've checked for blocking and deadlocking.
59 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
66 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
67 * check if a file has mandatory locks, used by mmap(), open() and creat() to
68 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
75 * Fixed deadlock condition for pathological code that mixes calls to
76 * flock() and fcntl().
77 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
79 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
80 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
81 * guarantee sensible behaviour in the case where file system modules might
82 * be compiled with different options than the kernel itself.
83 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
85 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
86 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
87 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
89 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
90 * locks. Changed process synchronisation to avoid dereferencing locks that
91 * have already been freed.
92 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
97 * Made mandatory locking a mount option. Default is not to allow mandatory
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
104 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
105 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
107 * Use slab allocator instead of kmalloc/kfree.
108 * Use generic list implementation from <linux/list.h>.
109 * Sped up posix_locks_deadlock by only considering blocked locks.
110 * Matthew Wilcox <willy@debian.org>, March, 2000.
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
117 #include <linux/capability.h>
118 #include <linux/file.h>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/module.h>
123 #include <linux/security.h>
124 #include <linux/slab.h>
125 #include <linux/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/rcupdate.h>
128 #include <linux/pid_namespace.h>
129 #include <linux/hashtable.h>
130 #include <linux/percpu.h>
131 #include <linux/lglock.h>
133 #define CREATE_TRACE_POINTS
134 #include <trace/events/filelock.h>
136 #include <asm/uaccess.h>
138 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
139 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
140 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
141 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
143 static bool lease_breaking(struct file_lock *fl)
145 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
148 static int target_leasetype(struct file_lock *fl)
150 if (fl->fl_flags & FL_UNLOCK_PENDING)
152 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
157 int leases_enable = 1;
158 int lease_break_time = 45;
161 * The global file_lock_list is only used for displaying /proc/locks, so we
162 * keep a list on each CPU, with each list protected by its own spinlock via
163 * the file_lock_lglock. Note that alterations to the list also require that
164 * the relevant flc_lock is held.
166 DEFINE_STATIC_LGLOCK(file_lock_lglock);
167 static DEFINE_PER_CPU(struct hlist_head, file_lock_list);
170 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
171 * It is protected by blocked_lock_lock.
173 * We hash locks by lockowner in order to optimize searching for the lock a
174 * particular lockowner is waiting on.
176 * FIXME: make this value scale via some heuristic? We generally will want more
177 * buckets when we have more lockowners holding locks, but that's a little
178 * difficult to determine without knowing what the workload will look like.
180 #define BLOCKED_HASH_BITS 7
181 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
184 * This lock protects the blocked_hash. Generally, if you're accessing it, you
185 * want to be holding this lock.
187 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
188 * pointer for file_lock structures that are acting as lock requests (in
189 * contrast to those that are acting as records of acquired locks).
191 * Note that when we acquire this lock in order to change the above fields,
192 * we often hold the flc_lock as well. In certain cases, when reading the fields
193 * protected by this lock, we can skip acquiring it iff we already hold the
196 * In particular, adding an entry to the fl_block list requires that you hold
197 * both the flc_lock and the blocked_lock_lock (acquired in that order).
198 * Deleting an entry from the list however only requires the file_lock_lock.
200 static DEFINE_SPINLOCK(blocked_lock_lock);
202 static struct kmem_cache *flctx_cache __read_mostly;
203 static struct kmem_cache *filelock_cache __read_mostly;
205 static struct file_lock_context *
206 locks_get_lock_context(struct inode *inode, int type)
208 struct file_lock_context *ctx;
210 /* paired with cmpxchg() below */
211 ctx = smp_load_acquire(&inode->i_flctx);
212 if (likely(ctx) || type == F_UNLCK)
215 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
219 spin_lock_init(&ctx->flc_lock);
220 INIT_LIST_HEAD(&ctx->flc_flock);
221 INIT_LIST_HEAD(&ctx->flc_posix);
222 INIT_LIST_HEAD(&ctx->flc_lease);
225 * Assign the pointer if it's not already assigned. If it is, then
226 * free the context we just allocated.
228 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
229 kmem_cache_free(flctx_cache, ctx);
230 ctx = smp_load_acquire(&inode->i_flctx);
237 locks_free_lock_context(struct file_lock_context *ctx)
240 WARN_ON_ONCE(!list_empty(&ctx->flc_flock));
241 WARN_ON_ONCE(!list_empty(&ctx->flc_posix));
242 WARN_ON_ONCE(!list_empty(&ctx->flc_lease));
243 kmem_cache_free(flctx_cache, ctx);
247 static void locks_init_lock_heads(struct file_lock *fl)
249 INIT_HLIST_NODE(&fl->fl_link);
250 INIT_LIST_HEAD(&fl->fl_list);
251 INIT_LIST_HEAD(&fl->fl_block);
252 init_waitqueue_head(&fl->fl_wait);
255 /* Allocate an empty lock structure. */
256 struct file_lock *locks_alloc_lock(void)
258 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
261 locks_init_lock_heads(fl);
265 EXPORT_SYMBOL_GPL(locks_alloc_lock);
267 void locks_release_private(struct file_lock *fl)
270 if (fl->fl_ops->fl_release_private)
271 fl->fl_ops->fl_release_private(fl);
276 if (fl->fl_lmops->lm_put_owner) {
277 fl->fl_lmops->lm_put_owner(fl->fl_owner);
283 EXPORT_SYMBOL_GPL(locks_release_private);
285 /* Free a lock which is not in use. */
286 void locks_free_lock(struct file_lock *fl)
288 BUG_ON(waitqueue_active(&fl->fl_wait));
289 BUG_ON(!list_empty(&fl->fl_list));
290 BUG_ON(!list_empty(&fl->fl_block));
291 BUG_ON(!hlist_unhashed(&fl->fl_link));
293 locks_release_private(fl);
294 kmem_cache_free(filelock_cache, fl);
296 EXPORT_SYMBOL(locks_free_lock);
299 locks_dispose_list(struct list_head *dispose)
301 struct file_lock *fl;
303 while (!list_empty(dispose)) {
304 fl = list_first_entry(dispose, struct file_lock, fl_list);
305 list_del_init(&fl->fl_list);
310 void locks_init_lock(struct file_lock *fl)
312 memset(fl, 0, sizeof(struct file_lock));
313 locks_init_lock_heads(fl);
316 EXPORT_SYMBOL(locks_init_lock);
319 * Initialize a new lock from an existing file_lock structure.
321 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
323 new->fl_owner = fl->fl_owner;
324 new->fl_pid = fl->fl_pid;
326 new->fl_flags = fl->fl_flags;
327 new->fl_type = fl->fl_type;
328 new->fl_start = fl->fl_start;
329 new->fl_end = fl->fl_end;
330 new->fl_lmops = fl->fl_lmops;
334 if (fl->fl_lmops->lm_get_owner)
335 fl->fl_lmops->lm_get_owner(fl->fl_owner);
338 EXPORT_SYMBOL(locks_copy_conflock);
340 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
342 /* "new" must be a freshly-initialized lock */
343 WARN_ON_ONCE(new->fl_ops);
345 locks_copy_conflock(new, fl);
347 new->fl_file = fl->fl_file;
348 new->fl_ops = fl->fl_ops;
351 if (fl->fl_ops->fl_copy_lock)
352 fl->fl_ops->fl_copy_lock(new, fl);
356 EXPORT_SYMBOL(locks_copy_lock);
358 static inline int flock_translate_cmd(int cmd) {
360 return cmd & (LOCK_MAND | LOCK_RW);
372 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
373 static struct file_lock *
374 flock_make_lock(struct file *filp, unsigned int cmd)
376 struct file_lock *fl;
377 int type = flock_translate_cmd(cmd);
380 return ERR_PTR(type);
382 fl = locks_alloc_lock();
384 return ERR_PTR(-ENOMEM);
388 fl->fl_pid = current->tgid;
389 fl->fl_flags = FL_FLOCK;
391 fl->fl_end = OFFSET_MAX;
396 static int assign_type(struct file_lock *fl, long type)
410 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
413 switch (l->l_whence) {
418 fl->fl_start = filp->f_pos;
421 fl->fl_start = i_size_read(file_inode(filp));
426 if (l->l_start > OFFSET_MAX - fl->fl_start)
428 fl->fl_start += l->l_start;
429 if (fl->fl_start < 0)
432 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
433 POSIX-2001 defines it. */
435 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
437 fl->fl_end = fl->fl_start + l->l_len - 1;
439 } else if (l->l_len < 0) {
440 if (fl->fl_start + l->l_len < 0)
442 fl->fl_end = fl->fl_start - 1;
443 fl->fl_start += l->l_len;
445 fl->fl_end = OFFSET_MAX;
447 fl->fl_owner = current->files;
448 fl->fl_pid = current->tgid;
450 fl->fl_flags = FL_POSIX;
454 return assign_type(fl, l->l_type);
457 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
460 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
463 struct flock64 ll = {
465 .l_whence = l->l_whence,
466 .l_start = l->l_start,
470 return flock64_to_posix_lock(filp, fl, &ll);
473 /* default lease lock manager operations */
475 lease_break_callback(struct file_lock *fl)
477 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
482 lease_setup(struct file_lock *fl, void **priv)
484 struct file *filp = fl->fl_file;
485 struct fasync_struct *fa = *priv;
488 * fasync_insert_entry() returns the old entry if any. If there was no
489 * old entry, then it used "priv" and inserted it into the fasync list.
490 * Clear the pointer to indicate that it shouldn't be freed.
492 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
495 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
498 static const struct lock_manager_operations lease_manager_ops = {
499 .lm_break = lease_break_callback,
500 .lm_change = lease_modify,
501 .lm_setup = lease_setup,
505 * Initialize a lease, use the default lock manager operations
507 static int lease_init(struct file *filp, long type, struct file_lock *fl)
509 if (assign_type(fl, type) != 0)
513 fl->fl_pid = current->tgid;
516 fl->fl_flags = FL_LEASE;
518 fl->fl_end = OFFSET_MAX;
520 fl->fl_lmops = &lease_manager_ops;
524 /* Allocate a file_lock initialised to this type of lease */
525 static struct file_lock *lease_alloc(struct file *filp, long type)
527 struct file_lock *fl = locks_alloc_lock();
531 return ERR_PTR(error);
533 error = lease_init(filp, type, fl);
536 return ERR_PTR(error);
541 /* Check if two locks overlap each other.
543 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
545 return ((fl1->fl_end >= fl2->fl_start) &&
546 (fl2->fl_end >= fl1->fl_start));
550 * Check whether two locks have the same owner.
552 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
554 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
555 return fl2->fl_lmops == fl1->fl_lmops &&
556 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
557 return fl1->fl_owner == fl2->fl_owner;
560 /* Must be called with the flc_lock held! */
561 static void locks_insert_global_locks(struct file_lock *fl)
563 lg_local_lock(&file_lock_lglock);
564 fl->fl_link_cpu = smp_processor_id();
565 hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list));
566 lg_local_unlock(&file_lock_lglock);
569 /* Must be called with the flc_lock held! */
570 static void locks_delete_global_locks(struct file_lock *fl)
573 * Avoid taking lock if already unhashed. This is safe since this check
574 * is done while holding the flc_lock, and new insertions into the list
575 * also require that it be held.
577 if (hlist_unhashed(&fl->fl_link))
579 lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu);
580 hlist_del_init(&fl->fl_link);
581 lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu);
585 posix_owner_key(struct file_lock *fl)
587 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
588 return fl->fl_lmops->lm_owner_key(fl);
589 return (unsigned long)fl->fl_owner;
592 static void locks_insert_global_blocked(struct file_lock *waiter)
594 lockdep_assert_held(&blocked_lock_lock);
596 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
599 static void locks_delete_global_blocked(struct file_lock *waiter)
601 lockdep_assert_held(&blocked_lock_lock);
603 hash_del(&waiter->fl_link);
606 /* Remove waiter from blocker's block list.
607 * When blocker ends up pointing to itself then the list is empty.
609 * Must be called with blocked_lock_lock held.
611 static void __locks_delete_block(struct file_lock *waiter)
613 locks_delete_global_blocked(waiter);
614 list_del_init(&waiter->fl_block);
615 waiter->fl_next = NULL;
618 static void locks_delete_block(struct file_lock *waiter)
620 spin_lock(&blocked_lock_lock);
621 __locks_delete_block(waiter);
622 spin_unlock(&blocked_lock_lock);
625 /* Insert waiter into blocker's block list.
626 * We use a circular list so that processes can be easily woken up in
627 * the order they blocked. The documentation doesn't require this but
628 * it seems like the reasonable thing to do.
630 * Must be called with both the flc_lock and blocked_lock_lock held. The
631 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
632 * that the flc_lock is also held on insertions we can avoid taking the
633 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
635 static void __locks_insert_block(struct file_lock *blocker,
636 struct file_lock *waiter)
638 BUG_ON(!list_empty(&waiter->fl_block));
639 waiter->fl_next = blocker;
640 list_add_tail(&waiter->fl_block, &blocker->fl_block);
641 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
642 locks_insert_global_blocked(waiter);
645 /* Must be called with flc_lock held. */
646 static void locks_insert_block(struct file_lock *blocker,
647 struct file_lock *waiter)
649 spin_lock(&blocked_lock_lock);
650 __locks_insert_block(blocker, waiter);
651 spin_unlock(&blocked_lock_lock);
655 * Wake up processes blocked waiting for blocker.
657 * Must be called with the inode->flc_lock held!
659 static void locks_wake_up_blocks(struct file_lock *blocker)
662 * Avoid taking global lock if list is empty. This is safe since new
663 * blocked requests are only added to the list under the flc_lock, and
664 * the flc_lock is always held here. Note that removal from the fl_block
665 * list does not require the flc_lock, so we must recheck list_empty()
666 * after acquiring the blocked_lock_lock.
668 if (list_empty(&blocker->fl_block))
671 spin_lock(&blocked_lock_lock);
672 while (!list_empty(&blocker->fl_block)) {
673 struct file_lock *waiter;
675 waiter = list_first_entry(&blocker->fl_block,
676 struct file_lock, fl_block);
677 __locks_delete_block(waiter);
678 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
679 waiter->fl_lmops->lm_notify(waiter);
681 wake_up(&waiter->fl_wait);
683 spin_unlock(&blocked_lock_lock);
687 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
689 fl->fl_nspid = get_pid(task_tgid(current));
690 list_add_tail(&fl->fl_list, before);
691 locks_insert_global_locks(fl);
695 locks_unlink_lock_ctx(struct file_lock *fl)
697 locks_delete_global_locks(fl);
698 list_del_init(&fl->fl_list);
700 put_pid(fl->fl_nspid);
703 locks_wake_up_blocks(fl);
707 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
709 locks_unlink_lock_ctx(fl);
711 list_add(&fl->fl_list, dispose);
716 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
717 * checks for shared/exclusive status of overlapping locks.
719 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
721 if (sys_fl->fl_type == F_WRLCK)
723 if (caller_fl->fl_type == F_WRLCK)
728 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
729 * checking before calling the locks_conflict().
731 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
733 /* POSIX locks owned by the same process do not conflict with
736 if (posix_same_owner(caller_fl, sys_fl))
739 /* Check whether they overlap */
740 if (!locks_overlap(caller_fl, sys_fl))
743 return (locks_conflict(caller_fl, sys_fl));
746 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
747 * checking before calling the locks_conflict().
749 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
751 /* FLOCK locks referring to the same filp do not conflict with
754 if (caller_fl->fl_file == sys_fl->fl_file)
756 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
759 return (locks_conflict(caller_fl, sys_fl));
763 posix_test_lock(struct file *filp, struct file_lock *fl)
765 struct file_lock *cfl;
766 struct file_lock_context *ctx;
767 struct inode *inode = file_inode(filp);
769 ctx = smp_load_acquire(&inode->i_flctx);
770 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
771 fl->fl_type = F_UNLCK;
775 spin_lock(&ctx->flc_lock);
776 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
777 if (posix_locks_conflict(fl, cfl)) {
778 locks_copy_conflock(fl, cfl);
780 fl->fl_pid = pid_vnr(cfl->fl_nspid);
784 fl->fl_type = F_UNLCK;
786 spin_unlock(&ctx->flc_lock);
789 EXPORT_SYMBOL(posix_test_lock);
792 * Deadlock detection:
794 * We attempt to detect deadlocks that are due purely to posix file
797 * We assume that a task can be waiting for at most one lock at a time.
798 * So for any acquired lock, the process holding that lock may be
799 * waiting on at most one other lock. That lock in turns may be held by
800 * someone waiting for at most one other lock. Given a requested lock
801 * caller_fl which is about to wait for a conflicting lock block_fl, we
802 * follow this chain of waiters to ensure we are not about to create a
805 * Since we do this before we ever put a process to sleep on a lock, we
806 * are ensured that there is never a cycle; that is what guarantees that
807 * the while() loop in posix_locks_deadlock() eventually completes.
809 * Note: the above assumption may not be true when handling lock
810 * requests from a broken NFS client. It may also fail in the presence
811 * of tasks (such as posix threads) sharing the same open file table.
812 * To handle those cases, we just bail out after a few iterations.
814 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
815 * Because the owner is not even nominally tied to a thread of
816 * execution, the deadlock detection below can't reasonably work well. Just
819 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
820 * locks that just checks for the case where two tasks are attempting to
821 * upgrade from read to write locks on the same inode.
824 #define MAX_DEADLK_ITERATIONS 10
826 /* Find a lock that the owner of the given block_fl is blocking on. */
827 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
829 struct file_lock *fl;
831 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
832 if (posix_same_owner(fl, block_fl))
838 /* Must be called with the blocked_lock_lock held! */
839 static int posix_locks_deadlock(struct file_lock *caller_fl,
840 struct file_lock *block_fl)
844 lockdep_assert_held(&blocked_lock_lock);
847 * This deadlock detector can't reasonably detect deadlocks with
848 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
850 if (IS_OFDLCK(caller_fl))
853 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
854 if (i++ > MAX_DEADLK_ITERATIONS)
856 if (posix_same_owner(caller_fl, block_fl))
862 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
863 * after any leases, but before any posix locks.
865 * Note that if called with an FL_EXISTS argument, the caller may determine
866 * whether or not a lock was successfully freed by testing the return
869 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
871 struct file_lock *new_fl = NULL;
872 struct file_lock *fl;
873 struct file_lock_context *ctx;
878 ctx = locks_get_lock_context(inode, request->fl_type);
880 if (request->fl_type != F_UNLCK)
882 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
885 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
886 new_fl = locks_alloc_lock();
891 spin_lock(&ctx->flc_lock);
892 if (request->fl_flags & FL_ACCESS)
895 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
896 if (request->fl_file != fl->fl_file)
898 if (request->fl_type == fl->fl_type)
901 locks_delete_lock_ctx(fl, &dispose);
905 if (request->fl_type == F_UNLCK) {
906 if ((request->fl_flags & FL_EXISTS) && !found)
912 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
913 if (!flock_locks_conflict(request, fl))
916 if (!(request->fl_flags & FL_SLEEP))
918 error = FILE_LOCK_DEFERRED;
919 locks_insert_block(fl, request);
922 if (request->fl_flags & FL_ACCESS)
924 locks_copy_lock(new_fl, request);
925 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
930 spin_unlock(&ctx->flc_lock);
932 locks_free_lock(new_fl);
933 locks_dispose_list(&dispose);
937 static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
939 struct file_lock *fl, *tmp;
940 struct file_lock *new_fl = NULL;
941 struct file_lock *new_fl2 = NULL;
942 struct file_lock *left = NULL;
943 struct file_lock *right = NULL;
944 struct file_lock_context *ctx;
949 ctx = locks_get_lock_context(inode, request->fl_type);
951 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
954 * We may need two file_lock structures for this operation,
955 * so we get them in advance to avoid races.
957 * In some cases we can be sure, that no new locks will be needed
959 if (!(request->fl_flags & FL_ACCESS) &&
960 (request->fl_type != F_UNLCK ||
961 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
962 new_fl = locks_alloc_lock();
963 new_fl2 = locks_alloc_lock();
966 spin_lock(&ctx->flc_lock);
968 * New lock request. Walk all POSIX locks and look for conflicts. If
969 * there are any, either return error or put the request on the
970 * blocker's list of waiters and the global blocked_hash.
972 if (request->fl_type != F_UNLCK) {
973 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
974 if (!posix_locks_conflict(request, fl))
977 locks_copy_conflock(conflock, fl);
979 if (!(request->fl_flags & FL_SLEEP))
982 * Deadlock detection and insertion into the blocked
983 * locks list must be done while holding the same lock!
986 spin_lock(&blocked_lock_lock);
987 if (likely(!posix_locks_deadlock(request, fl))) {
988 error = FILE_LOCK_DEFERRED;
989 __locks_insert_block(fl, request);
991 spin_unlock(&blocked_lock_lock);
996 /* If we're just looking for a conflict, we're done. */
998 if (request->fl_flags & FL_ACCESS)
1001 /* Find the first old lock with the same owner as the new lock */
1002 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1003 if (posix_same_owner(request, fl))
1007 /* Process locks with this owner. */
1008 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1009 if (!posix_same_owner(request, fl))
1012 /* Detect adjacent or overlapping regions (if same lock type) */
1013 if (request->fl_type == fl->fl_type) {
1014 /* In all comparisons of start vs end, use
1015 * "start - 1" rather than "end + 1". If end
1016 * is OFFSET_MAX, end + 1 will become negative.
1018 if (fl->fl_end < request->fl_start - 1)
1020 /* If the next lock in the list has entirely bigger
1021 * addresses than the new one, insert the lock here.
1023 if (fl->fl_start - 1 > request->fl_end)
1026 /* If we come here, the new and old lock are of the
1027 * same type and adjacent or overlapping. Make one
1028 * lock yielding from the lower start address of both
1029 * locks to the higher end address.
1031 if (fl->fl_start > request->fl_start)
1032 fl->fl_start = request->fl_start;
1034 request->fl_start = fl->fl_start;
1035 if (fl->fl_end < request->fl_end)
1036 fl->fl_end = request->fl_end;
1038 request->fl_end = fl->fl_end;
1040 locks_delete_lock_ctx(fl, &dispose);
1046 /* Processing for different lock types is a bit
1049 if (fl->fl_end < request->fl_start)
1051 if (fl->fl_start > request->fl_end)
1053 if (request->fl_type == F_UNLCK)
1055 if (fl->fl_start < request->fl_start)
1057 /* If the next lock in the list has a higher end
1058 * address than the new one, insert the new one here.
1060 if (fl->fl_end > request->fl_end) {
1064 if (fl->fl_start >= request->fl_start) {
1065 /* The new lock completely replaces an old
1066 * one (This may happen several times).
1069 locks_delete_lock_ctx(fl, &dispose);
1073 * Replace the old lock with new_fl, and
1074 * remove the old one. It's safe to do the
1075 * insert here since we know that we won't be
1076 * using new_fl later, and that the lock is
1077 * just replacing an existing lock.
1082 locks_copy_lock(new_fl, request);
1085 locks_insert_lock_ctx(request, &fl->fl_list);
1086 locks_delete_lock_ctx(fl, &dispose);
1093 * The above code only modifies existing locks in case of merging or
1094 * replacing. If new lock(s) need to be inserted all modifications are
1095 * done below this, so it's safe yet to bail out.
1097 error = -ENOLCK; /* "no luck" */
1098 if (right && left == right && !new_fl2)
1103 if (request->fl_type == F_UNLCK) {
1104 if (request->fl_flags & FL_EXISTS)
1113 locks_copy_lock(new_fl, request);
1114 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1119 if (left == right) {
1120 /* The new lock breaks the old one in two pieces,
1121 * so we have to use the second new lock.
1125 locks_copy_lock(left, right);
1126 locks_insert_lock_ctx(left, &fl->fl_list);
1128 right->fl_start = request->fl_end + 1;
1129 locks_wake_up_blocks(right);
1132 left->fl_end = request->fl_start - 1;
1133 locks_wake_up_blocks(left);
1136 spin_unlock(&ctx->flc_lock);
1138 * Free any unused locks.
1141 locks_free_lock(new_fl);
1143 locks_free_lock(new_fl2);
1144 locks_dispose_list(&dispose);
1149 * posix_lock_file - Apply a POSIX-style lock to a file
1150 * @filp: The file to apply the lock to
1151 * @fl: The lock to be applied
1152 * @conflock: Place to return a copy of the conflicting lock, if found.
1154 * Add a POSIX style lock to a file.
1155 * We merge adjacent & overlapping locks whenever possible.
1156 * POSIX locks are sorted by owner task, then by starting address
1158 * Note that if called with an FL_EXISTS argument, the caller may determine
1159 * whether or not a lock was successfully freed by testing the return
1160 * value for -ENOENT.
1162 int posix_lock_file(struct file *filp, struct file_lock *fl,
1163 struct file_lock *conflock)
1165 return __posix_lock_file(file_inode(filp), fl, conflock);
1167 EXPORT_SYMBOL(posix_lock_file);
1170 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1171 * @inode: inode of file to which lock request should be applied
1172 * @fl: The lock to be applied
1174 * Apply a POSIX style lock request to an inode.
1176 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1181 error = __posix_lock_file(inode, fl, NULL);
1182 if (error != FILE_LOCK_DEFERRED)
1184 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1188 locks_delete_block(fl);
1195 * locks_mandatory_locked - Check for an active lock
1196 * @file: the file to check
1198 * Searches the inode's list of locks to find any POSIX locks which conflict.
1199 * This function is called from locks_verify_locked() only.
1201 int locks_mandatory_locked(struct file *file)
1204 struct inode *inode = file_inode(file);
1205 struct file_lock_context *ctx;
1206 struct file_lock *fl;
1208 ctx = smp_load_acquire(&inode->i_flctx);
1209 if (!ctx || list_empty_careful(&ctx->flc_posix))
1213 * Search the lock list for this inode for any POSIX locks.
1215 spin_lock(&ctx->flc_lock);
1217 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1218 if (fl->fl_owner != current->files &&
1219 fl->fl_owner != file) {
1224 spin_unlock(&ctx->flc_lock);
1229 * locks_mandatory_area - Check for a conflicting lock
1230 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1232 * @inode: the file to check
1233 * @filp: how the file was opened (if it was)
1234 * @offset: start of area to check
1235 * @count: length of area to check
1237 * Searches the inode's list of locks to find any POSIX locks which conflict.
1238 * This function is called from rw_verify_area() and
1239 * locks_verify_truncate().
1241 int locks_mandatory_area(int read_write, struct inode *inode,
1242 struct file *filp, loff_t offset,
1245 struct file_lock fl;
1249 locks_init_lock(&fl);
1250 fl.fl_pid = current->tgid;
1252 fl.fl_flags = FL_POSIX | FL_ACCESS;
1253 if (filp && !(filp->f_flags & O_NONBLOCK))
1255 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1256 fl.fl_start = offset;
1257 fl.fl_end = offset + count - 1;
1262 fl.fl_flags &= ~FL_SLEEP;
1263 error = __posix_lock_file(inode, &fl, NULL);
1269 fl.fl_flags |= FL_SLEEP;
1270 fl.fl_owner = current->files;
1271 error = __posix_lock_file(inode, &fl, NULL);
1272 if (error != FILE_LOCK_DEFERRED)
1274 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1277 * If we've been sleeping someone might have
1278 * changed the permissions behind our back.
1280 if (__mandatory_lock(inode))
1284 locks_delete_block(&fl);
1291 EXPORT_SYMBOL(locks_mandatory_area);
1293 static void lease_clear_pending(struct file_lock *fl, int arg)
1297 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1300 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1304 /* We already had a lease on this file; just change its type */
1305 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1307 int error = assign_type(fl, arg);
1311 lease_clear_pending(fl, arg);
1312 locks_wake_up_blocks(fl);
1313 if (arg == F_UNLCK) {
1314 struct file *filp = fl->fl_file;
1317 filp->f_owner.signum = 0;
1318 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1319 if (fl->fl_fasync != NULL) {
1320 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1321 fl->fl_fasync = NULL;
1323 locks_delete_lock_ctx(fl, dispose);
1327 EXPORT_SYMBOL(lease_modify);
1329 static bool past_time(unsigned long then)
1332 /* 0 is a special value meaning "this never expires": */
1334 return time_after(jiffies, then);
1337 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1339 struct file_lock_context *ctx = inode->i_flctx;
1340 struct file_lock *fl, *tmp;
1342 lockdep_assert_held(&ctx->flc_lock);
1344 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1345 trace_time_out_leases(inode, fl);
1346 if (past_time(fl->fl_downgrade_time))
1347 lease_modify(fl, F_RDLCK, dispose);
1348 if (past_time(fl->fl_break_time))
1349 lease_modify(fl, F_UNLCK, dispose);
1353 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1355 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1357 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1359 return locks_conflict(breaker, lease);
1363 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1365 struct file_lock_context *ctx = inode->i_flctx;
1366 struct file_lock *fl;
1368 lockdep_assert_held(&ctx->flc_lock);
1370 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1371 if (leases_conflict(fl, breaker))
1378 * __break_lease - revoke all outstanding leases on file
1379 * @inode: the inode of the file to return
1380 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1382 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1385 * break_lease (inlined for speed) has checked there already is at least
1386 * some kind of lock (maybe a lease) on this file. Leases are broken on
1387 * a call to open() or truncate(). This function can sleep unless you
1388 * specified %O_NONBLOCK to your open().
1390 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1393 struct file_lock_context *ctx;
1394 struct file_lock *new_fl, *fl, *tmp;
1395 unsigned long break_time;
1396 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1399 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1401 return PTR_ERR(new_fl);
1402 new_fl->fl_flags = type;
1404 /* typically we will check that ctx is non-NULL before calling */
1405 ctx = smp_load_acquire(&inode->i_flctx);
1411 spin_lock(&ctx->flc_lock);
1413 time_out_leases(inode, &dispose);
1415 if (!any_leases_conflict(inode, new_fl))
1419 if (lease_break_time > 0) {
1420 break_time = jiffies + lease_break_time * HZ;
1421 if (break_time == 0)
1422 break_time++; /* so that 0 means no break time */
1425 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1426 if (!leases_conflict(fl, new_fl))
1429 if (fl->fl_flags & FL_UNLOCK_PENDING)
1431 fl->fl_flags |= FL_UNLOCK_PENDING;
1432 fl->fl_break_time = break_time;
1434 if (lease_breaking(fl))
1436 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1437 fl->fl_downgrade_time = break_time;
1439 if (fl->fl_lmops->lm_break(fl))
1440 locks_delete_lock_ctx(fl, &dispose);
1443 if (list_empty(&ctx->flc_lease))
1446 if (mode & O_NONBLOCK) {
1447 trace_break_lease_noblock(inode, new_fl);
1448 error = -EWOULDBLOCK;
1453 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1454 break_time = fl->fl_break_time;
1455 if (break_time != 0)
1456 break_time -= jiffies;
1457 if (break_time == 0)
1459 locks_insert_block(fl, new_fl);
1460 trace_break_lease_block(inode, new_fl);
1461 spin_unlock(&ctx->flc_lock);
1462 locks_dispose_list(&dispose);
1463 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1464 !new_fl->fl_next, break_time);
1465 spin_lock(&ctx->flc_lock);
1466 trace_break_lease_unblock(inode, new_fl);
1467 locks_delete_block(new_fl);
1470 * Wait for the next conflicting lease that has not been
1474 time_out_leases(inode, &dispose);
1475 if (any_leases_conflict(inode, new_fl))
1480 spin_unlock(&ctx->flc_lock);
1481 locks_dispose_list(&dispose);
1482 locks_free_lock(new_fl);
1486 EXPORT_SYMBOL(__break_lease);
1489 * lease_get_mtime - get the last modified time of an inode
1491 * @time: pointer to a timespec which will contain the last modified time
1493 * This is to force NFS clients to flush their caches for files with
1494 * exclusive leases. The justification is that if someone has an
1495 * exclusive lease, then they could be modifying it.
1497 void lease_get_mtime(struct inode *inode, struct timespec *time)
1499 bool has_lease = false;
1500 struct file_lock_context *ctx;
1501 struct file_lock *fl;
1503 ctx = smp_load_acquire(&inode->i_flctx);
1504 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1505 spin_lock(&ctx->flc_lock);
1506 if (!list_empty(&ctx->flc_lease)) {
1507 fl = list_first_entry(&ctx->flc_lease,
1508 struct file_lock, fl_list);
1509 if (fl->fl_type == F_WRLCK)
1512 spin_unlock(&ctx->flc_lock);
1516 *time = current_fs_time(inode->i_sb);
1518 *time = inode->i_mtime;
1521 EXPORT_SYMBOL(lease_get_mtime);
1524 * fcntl_getlease - Enquire what lease is currently active
1527 * The value returned by this function will be one of
1528 * (if no lease break is pending):
1530 * %F_RDLCK to indicate a shared lease is held.
1532 * %F_WRLCK to indicate an exclusive lease is held.
1534 * %F_UNLCK to indicate no lease is held.
1536 * (if a lease break is pending):
1538 * %F_RDLCK to indicate an exclusive lease needs to be
1539 * changed to a shared lease (or removed).
1541 * %F_UNLCK to indicate the lease needs to be removed.
1543 * XXX: sfr & willy disagree over whether F_INPROGRESS
1544 * should be returned to userspace.
1546 int fcntl_getlease(struct file *filp)
1548 struct file_lock *fl;
1549 struct inode *inode = file_inode(filp);
1550 struct file_lock_context *ctx;
1554 ctx = smp_load_acquire(&inode->i_flctx);
1555 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1556 spin_lock(&ctx->flc_lock);
1557 time_out_leases(file_inode(filp), &dispose);
1558 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1559 if (fl->fl_file != filp)
1561 type = target_leasetype(fl);
1564 spin_unlock(&ctx->flc_lock);
1565 locks_dispose_list(&dispose);
1571 * check_conflicting_open - see if the given dentry points to a file that has
1572 * an existing open that would conflict with the
1574 * @dentry: dentry to check
1575 * @arg: type of lease that we're trying to acquire
1576 * @flags: current lock flags
1578 * Check to see if there's an existing open fd on this file that would
1579 * conflict with the lease we're trying to set.
1582 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1585 struct inode *inode = dentry->d_inode;
1587 if (flags & FL_LAYOUT)
1590 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1593 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1594 (atomic_read(&inode->i_count) > 1)))
1601 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1603 struct file_lock *fl, *my_fl = NULL, *lease;
1604 struct dentry *dentry = filp->f_path.dentry;
1605 struct inode *inode = file_inode(filp);
1606 struct file_lock_context *ctx;
1607 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1612 trace_generic_add_lease(inode, lease);
1614 /* Note that arg is never F_UNLCK here */
1615 ctx = locks_get_lock_context(inode, arg);
1620 * In the delegation case we need mutual exclusion with
1621 * a number of operations that take the i_mutex. We trylock
1622 * because delegations are an optional optimization, and if
1623 * there's some chance of a conflict--we'd rather not
1624 * bother, maybe that's a sign this just isn't a good file to
1625 * hand out a delegation on.
1627 if (is_deleg && !mutex_trylock(&inode->i_mutex))
1630 if (is_deleg && arg == F_WRLCK) {
1631 /* Write delegations are not currently supported: */
1632 mutex_unlock(&inode->i_mutex);
1637 spin_lock(&ctx->flc_lock);
1638 time_out_leases(inode, &dispose);
1639 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1644 * At this point, we know that if there is an exclusive
1645 * lease on this file, then we hold it on this filp
1646 * (otherwise our open of this file would have blocked).
1647 * And if we are trying to acquire an exclusive lease,
1648 * then the file is not open by anyone (including us)
1649 * except for this filp.
1652 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1653 if (fl->fl_file == filp &&
1654 fl->fl_owner == lease->fl_owner) {
1660 * No exclusive leases if someone else has a lease on
1666 * Modifying our existing lease is OK, but no getting a
1667 * new lease if someone else is opening for write:
1669 if (fl->fl_flags & FL_UNLOCK_PENDING)
1673 if (my_fl != NULL) {
1675 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1685 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1687 * The check in break_lease() is lockless. It's possible for another
1688 * open to race in after we did the earlier check for a conflicting
1689 * open but before the lease was inserted. Check again for a
1690 * conflicting open and cancel the lease if there is one.
1692 * We also add a barrier here to ensure that the insertion of the lock
1693 * precedes these checks.
1696 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1698 locks_unlink_lock_ctx(lease);
1703 if (lease->fl_lmops->lm_setup)
1704 lease->fl_lmops->lm_setup(lease, priv);
1706 spin_unlock(&ctx->flc_lock);
1707 locks_dispose_list(&dispose);
1709 mutex_unlock(&inode->i_mutex);
1710 if (!error && !my_fl)
1715 static int generic_delete_lease(struct file *filp, void *owner)
1717 int error = -EAGAIN;
1718 struct file_lock *fl, *victim = NULL;
1719 struct inode *inode = file_inode(filp);
1720 struct file_lock_context *ctx;
1723 ctx = smp_load_acquire(&inode->i_flctx);
1725 trace_generic_delete_lease(inode, NULL);
1729 spin_lock(&ctx->flc_lock);
1730 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1731 if (fl->fl_file == filp &&
1732 fl->fl_owner == owner) {
1737 trace_generic_delete_lease(inode, victim);
1739 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1740 spin_unlock(&ctx->flc_lock);
1741 locks_dispose_list(&dispose);
1746 * generic_setlease - sets a lease on an open file
1747 * @filp: file pointer
1748 * @arg: type of lease to obtain
1749 * @flp: input - file_lock to use, output - file_lock inserted
1750 * @priv: private data for lm_setup (may be NULL if lm_setup
1751 * doesn't require it)
1753 * The (input) flp->fl_lmops->lm_break function is required
1756 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1759 struct inode *inode = file_inode(filp);
1762 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1764 if (!S_ISREG(inode->i_mode))
1766 error = security_file_lock(filp, arg);
1772 return generic_delete_lease(filp, *priv);
1775 if (!(*flp)->fl_lmops->lm_break) {
1780 return generic_add_lease(filp, arg, flp, priv);
1785 EXPORT_SYMBOL(generic_setlease);
1788 * vfs_setlease - sets a lease on an open file
1789 * @filp: file pointer
1790 * @arg: type of lease to obtain
1791 * @lease: file_lock to use when adding a lease
1792 * @priv: private info for lm_setup when adding a lease (may be
1793 * NULL if lm_setup doesn't require it)
1795 * Call this to establish a lease on the file. The "lease" argument is not
1796 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1797 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1798 * if not, this function will return -ENOLCK (and generate a scary-looking
1801 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1802 * may be NULL if the lm_setup operation doesn't require it.
1805 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1807 if (filp->f_op->setlease)
1808 return filp->f_op->setlease(filp, arg, lease, priv);
1810 return generic_setlease(filp, arg, lease, priv);
1812 EXPORT_SYMBOL_GPL(vfs_setlease);
1814 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1816 struct file_lock *fl;
1817 struct fasync_struct *new;
1820 fl = lease_alloc(filp, arg);
1824 new = fasync_alloc();
1826 locks_free_lock(fl);
1831 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1833 locks_free_lock(fl);
1840 * fcntl_setlease - sets a lease on an open file
1841 * @fd: open file descriptor
1842 * @filp: file pointer
1843 * @arg: type of lease to obtain
1845 * Call this fcntl to establish a lease on the file.
1846 * Note that you also need to call %F_SETSIG to
1847 * receive a signal when the lease is broken.
1849 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1852 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1853 return do_fcntl_add_lease(fd, filp, arg);
1857 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1858 * @inode: inode of the file to apply to
1859 * @fl: The lock to be applied
1861 * Apply a FLOCK style lock request to an inode.
1863 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1868 error = flock_lock_inode(inode, fl);
1869 if (error != FILE_LOCK_DEFERRED)
1871 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1875 locks_delete_block(fl);
1882 * locks_lock_inode_wait - Apply a lock to an inode
1883 * @inode: inode of the file to apply to
1884 * @fl: The lock to be applied
1886 * Apply a POSIX or FLOCK style lock request to an inode.
1888 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1891 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1893 res = posix_lock_inode_wait(inode, fl);
1896 res = flock_lock_inode_wait(inode, fl);
1903 EXPORT_SYMBOL(locks_lock_inode_wait);
1906 * sys_flock: - flock() system call.
1907 * @fd: the file descriptor to lock.
1908 * @cmd: the type of lock to apply.
1910 * Apply a %FL_FLOCK style lock to an open file descriptor.
1911 * The @cmd can be one of
1913 * %LOCK_SH -- a shared lock.
1915 * %LOCK_EX -- an exclusive lock.
1917 * %LOCK_UN -- remove an existing lock.
1919 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1921 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1922 * processes read and write access respectively.
1924 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1926 struct fd f = fdget(fd);
1927 struct file_lock *lock;
1928 int can_sleep, unlock;
1935 can_sleep = !(cmd & LOCK_NB);
1937 unlock = (cmd == LOCK_UN);
1939 if (!unlock && !(cmd & LOCK_MAND) &&
1940 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
1943 lock = flock_make_lock(f.file, cmd);
1945 error = PTR_ERR(lock);
1950 lock->fl_flags |= FL_SLEEP;
1952 error = security_file_lock(f.file, lock->fl_type);
1956 if (f.file->f_op->flock)
1957 error = f.file->f_op->flock(f.file,
1958 (can_sleep) ? F_SETLKW : F_SETLK,
1961 error = locks_lock_file_wait(f.file, lock);
1964 locks_free_lock(lock);
1973 * vfs_test_lock - test file byte range lock
1974 * @filp: The file to test lock for
1975 * @fl: The lock to test; also used to hold result
1977 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1978 * setting conf->fl_type to something other than F_UNLCK.
1980 int vfs_test_lock(struct file *filp, struct file_lock *fl)
1982 if (filp->f_op->lock)
1983 return filp->f_op->lock(filp, F_GETLK, fl);
1984 posix_test_lock(filp, fl);
1987 EXPORT_SYMBOL_GPL(vfs_test_lock);
1989 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1991 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
1992 #if BITS_PER_LONG == 32
1994 * Make sure we can represent the posix lock via
1995 * legacy 32bit flock.
1997 if (fl->fl_start > OFFT_OFFSET_MAX)
1999 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2002 flock->l_start = fl->fl_start;
2003 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2004 fl->fl_end - fl->fl_start + 1;
2005 flock->l_whence = 0;
2006 flock->l_type = fl->fl_type;
2010 #if BITS_PER_LONG == 32
2011 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2013 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
2014 flock->l_start = fl->fl_start;
2015 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2016 fl->fl_end - fl->fl_start + 1;
2017 flock->l_whence = 0;
2018 flock->l_type = fl->fl_type;
2022 /* Report the first existing lock that would conflict with l.
2023 * This implements the F_GETLK command of fcntl().
2025 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
2027 struct file_lock file_lock;
2032 if (copy_from_user(&flock, l, sizeof(flock)))
2035 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2038 error = flock_to_posix_lock(filp, &file_lock, &flock);
2042 if (cmd == F_OFD_GETLK) {
2044 if (flock.l_pid != 0)
2048 file_lock.fl_flags |= FL_OFDLCK;
2049 file_lock.fl_owner = filp;
2052 error = vfs_test_lock(filp, &file_lock);
2056 flock.l_type = file_lock.fl_type;
2057 if (file_lock.fl_type != F_UNLCK) {
2058 error = posix_lock_to_flock(&flock, &file_lock);
2063 if (!copy_to_user(l, &flock, sizeof(flock)))
2066 locks_release_private(&file_lock);
2072 * vfs_lock_file - file byte range lock
2073 * @filp: The file to apply the lock to
2074 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2075 * @fl: The lock to be applied
2076 * @conf: Place to return a copy of the conflicting lock, if found.
2078 * A caller that doesn't care about the conflicting lock may pass NULL
2079 * as the final argument.
2081 * If the filesystem defines a private ->lock() method, then @conf will
2082 * be left unchanged; so a caller that cares should initialize it to
2083 * some acceptable default.
2085 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2086 * locks, the ->lock() interface may return asynchronously, before the lock has
2087 * been granted or denied by the underlying filesystem, if (and only if)
2088 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2089 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2090 * the request is for a blocking lock. When ->lock() does return asynchronously,
2091 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2092 * request completes.
2093 * If the request is for non-blocking lock the file system should return
2094 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2095 * with the result. If the request timed out the callback routine will return a
2096 * nonzero return code and the file system should release the lock. The file
2097 * system is also responsible to keep a corresponding posix lock when it
2098 * grants a lock so the VFS can find out which locks are locally held and do
2099 * the correct lock cleanup when required.
2100 * The underlying filesystem must not drop the kernel lock or call
2101 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2104 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2106 if (filp->f_op->lock)
2107 return filp->f_op->lock(filp, cmd, fl);
2109 return posix_lock_file(filp, fl, conf);
2111 EXPORT_SYMBOL_GPL(vfs_lock_file);
2113 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2114 struct file_lock *fl)
2118 error = security_file_lock(filp, fl->fl_type);
2123 error = vfs_lock_file(filp, cmd, fl, NULL);
2124 if (error != FILE_LOCK_DEFERRED)
2126 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2130 locks_delete_block(fl);
2137 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2139 check_fmode_for_setlk(struct file_lock *fl)
2141 switch (fl->fl_type) {
2143 if (!(fl->fl_file->f_mode & FMODE_READ))
2147 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2153 /* Apply the lock described by l to an open file descriptor.
2154 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2156 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2157 struct flock __user *l)
2159 struct file_lock *file_lock = locks_alloc_lock();
2161 struct inode *inode;
2165 if (file_lock == NULL)
2169 * This might block, so we do it before checking the inode.
2172 if (copy_from_user(&flock, l, sizeof(flock)))
2175 inode = file_inode(filp);
2177 /* Don't allow mandatory locks on files that may be memory mapped
2180 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2185 error = flock_to_posix_lock(filp, file_lock, &flock);
2189 error = check_fmode_for_setlk(file_lock);
2194 * If the cmd is requesting file-private locks, then set the
2195 * FL_OFDLCK flag and override the owner.
2200 if (flock.l_pid != 0)
2204 file_lock->fl_flags |= FL_OFDLCK;
2205 file_lock->fl_owner = filp;
2209 if (flock.l_pid != 0)
2213 file_lock->fl_flags |= FL_OFDLCK;
2214 file_lock->fl_owner = filp;
2217 file_lock->fl_flags |= FL_SLEEP;
2220 error = do_lock_file_wait(filp, cmd, file_lock);
2223 * Attempt to detect a close/fcntl race and recover by
2224 * releasing the lock that was just acquired.
2226 if (!error && file_lock->fl_type != F_UNLCK) {
2228 * We need that spin_lock here - it prevents reordering between
2229 * update of i_flctx->flc_posix and check for it done in
2230 * close(). rcu_read_lock() wouldn't do.
2232 spin_lock(¤t->files->file_lock);
2234 spin_unlock(¤t->files->file_lock);
2236 file_lock->fl_type = F_UNLCK;
2237 error = do_lock_file_wait(filp, cmd, file_lock);
2238 WARN_ON_ONCE(error);
2243 locks_free_lock(file_lock);
2247 #if BITS_PER_LONG == 32
2248 /* Report the first existing lock that would conflict with l.
2249 * This implements the F_GETLK command of fcntl().
2251 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
2253 struct file_lock file_lock;
2254 struct flock64 flock;
2258 if (copy_from_user(&flock, l, sizeof(flock)))
2261 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2264 error = flock64_to_posix_lock(filp, &file_lock, &flock);
2268 if (cmd == F_OFD_GETLK) {
2270 if (flock.l_pid != 0)
2274 file_lock.fl_flags |= FL_OFDLCK;
2275 file_lock.fl_owner = filp;
2278 error = vfs_test_lock(filp, &file_lock);
2282 flock.l_type = file_lock.fl_type;
2283 if (file_lock.fl_type != F_UNLCK)
2284 posix_lock_to_flock64(&flock, &file_lock);
2287 if (!copy_to_user(l, &flock, sizeof(flock)))
2290 locks_release_private(&file_lock);
2295 /* Apply the lock described by l to an open file descriptor.
2296 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2298 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2299 struct flock64 __user *l)
2301 struct file_lock *file_lock = locks_alloc_lock();
2302 struct flock64 flock;
2303 struct inode *inode;
2307 if (file_lock == NULL)
2311 * This might block, so we do it before checking the inode.
2314 if (copy_from_user(&flock, l, sizeof(flock)))
2317 inode = file_inode(filp);
2319 /* Don't allow mandatory locks on files that may be memory mapped
2322 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2327 error = flock64_to_posix_lock(filp, file_lock, &flock);
2331 error = check_fmode_for_setlk(file_lock);
2336 * If the cmd is requesting file-private locks, then set the
2337 * FL_OFDLCK flag and override the owner.
2342 if (flock.l_pid != 0)
2346 file_lock->fl_flags |= FL_OFDLCK;
2347 file_lock->fl_owner = filp;
2351 if (flock.l_pid != 0)
2355 file_lock->fl_flags |= FL_OFDLCK;
2356 file_lock->fl_owner = filp;
2359 file_lock->fl_flags |= FL_SLEEP;
2362 error = do_lock_file_wait(filp, cmd, file_lock);
2365 * Attempt to detect a close/fcntl race and recover by
2366 * releasing the lock that was just acquired.
2368 if (!error && file_lock->fl_type != F_UNLCK) {
2370 * We need that spin_lock here - it prevents reordering between
2371 * update of i_flctx->flc_posix and check for it done in
2372 * close(). rcu_read_lock() wouldn't do.
2374 spin_lock(¤t->files->file_lock);
2376 spin_unlock(¤t->files->file_lock);
2378 file_lock->fl_type = F_UNLCK;
2379 error = do_lock_file_wait(filp, cmd, file_lock);
2380 WARN_ON_ONCE(error);
2385 locks_free_lock(file_lock);
2388 #endif /* BITS_PER_LONG == 32 */
2391 * This function is called when the file is being removed
2392 * from the task's fd array. POSIX locks belonging to this task
2393 * are deleted at this time.
2395 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2397 struct file_lock lock;
2398 struct file_lock_context *ctx;
2401 * If there are no locks held on this file, we don't need to call
2402 * posix_lock_file(). Another process could be setting a lock on this
2403 * file at the same time, but we wouldn't remove that lock anyway.
2405 ctx = smp_load_acquire(&file_inode(filp)->i_flctx);
2406 if (!ctx || list_empty(&ctx->flc_posix))
2409 lock.fl_type = F_UNLCK;
2410 lock.fl_flags = FL_POSIX | FL_CLOSE;
2412 lock.fl_end = OFFSET_MAX;
2413 lock.fl_owner = owner;
2414 lock.fl_pid = current->tgid;
2415 lock.fl_file = filp;
2417 lock.fl_lmops = NULL;
2419 vfs_lock_file(filp, F_SETLK, &lock, NULL);
2421 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2422 lock.fl_ops->fl_release_private(&lock);
2425 EXPORT_SYMBOL(locks_remove_posix);
2427 /* The i_flctx must be valid when calling into here */
2429 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2431 struct file_lock fl = {
2433 .fl_pid = current->tgid,
2435 .fl_flags = FL_FLOCK,
2437 .fl_end = OFFSET_MAX,
2439 struct inode *inode = file_inode(filp);
2441 if (list_empty(&flctx->flc_flock))
2444 if (filp->f_op->flock)
2445 filp->f_op->flock(filp, F_SETLKW, &fl);
2447 flock_lock_inode(inode, &fl);
2449 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2450 fl.fl_ops->fl_release_private(&fl);
2453 /* The i_flctx must be valid when calling into here */
2455 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2457 struct file_lock *fl, *tmp;
2460 if (list_empty(&ctx->flc_lease))
2463 spin_lock(&ctx->flc_lock);
2464 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2465 if (filp == fl->fl_file)
2466 lease_modify(fl, F_UNLCK, &dispose);
2467 spin_unlock(&ctx->flc_lock);
2468 locks_dispose_list(&dispose);
2472 * This function is called on the last close of an open file.
2474 void locks_remove_file(struct file *filp)
2476 struct file_lock_context *ctx;
2478 ctx = smp_load_acquire(&file_inode(filp)->i_flctx);
2482 /* remove any OFD locks */
2483 locks_remove_posix(filp, filp);
2485 /* remove flock locks */
2486 locks_remove_flock(filp, ctx);
2488 /* remove any leases */
2489 locks_remove_lease(filp, ctx);
2493 * posix_unblock_lock - stop waiting for a file lock
2494 * @waiter: the lock which was waiting
2496 * lockd needs to block waiting for locks.
2499 posix_unblock_lock(struct file_lock *waiter)
2503 spin_lock(&blocked_lock_lock);
2504 if (waiter->fl_next)
2505 __locks_delete_block(waiter);
2508 spin_unlock(&blocked_lock_lock);
2511 EXPORT_SYMBOL(posix_unblock_lock);
2514 * vfs_cancel_lock - file byte range unblock lock
2515 * @filp: The file to apply the unblock to
2516 * @fl: The lock to be unblocked
2518 * Used by lock managers to cancel blocked requests
2520 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2522 if (filp->f_op->lock)
2523 return filp->f_op->lock(filp, F_CANCELLK, fl);
2527 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2529 #ifdef CONFIG_PROC_FS
2530 #include <linux/proc_fs.h>
2531 #include <linux/seq_file.h>
2533 struct locks_iterator {
2538 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2539 loff_t id, char *pfx)
2541 struct inode *inode = NULL;
2542 unsigned int fl_pid;
2545 fl_pid = pid_vnr(fl->fl_nspid);
2547 fl_pid = fl->fl_pid;
2549 if (fl->fl_file != NULL)
2550 inode = file_inode(fl->fl_file);
2552 seq_printf(f, "%lld:%s ", id, pfx);
2554 if (fl->fl_flags & FL_ACCESS)
2555 seq_puts(f, "ACCESS");
2556 else if (IS_OFDLCK(fl))
2557 seq_puts(f, "OFDLCK");
2559 seq_puts(f, "POSIX ");
2561 seq_printf(f, " %s ",
2562 (inode == NULL) ? "*NOINODE*" :
2563 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2564 } else if (IS_FLOCK(fl)) {
2565 if (fl->fl_type & LOCK_MAND) {
2566 seq_puts(f, "FLOCK MSNFS ");
2568 seq_puts(f, "FLOCK ADVISORY ");
2570 } else if (IS_LEASE(fl)) {
2571 if (fl->fl_flags & FL_DELEG)
2572 seq_puts(f, "DELEG ");
2574 seq_puts(f, "LEASE ");
2576 if (lease_breaking(fl))
2577 seq_puts(f, "BREAKING ");
2578 else if (fl->fl_file)
2579 seq_puts(f, "ACTIVE ");
2581 seq_puts(f, "BREAKER ");
2583 seq_puts(f, "UNKNOWN UNKNOWN ");
2585 if (fl->fl_type & LOCK_MAND) {
2586 seq_printf(f, "%s ",
2587 (fl->fl_type & LOCK_READ)
2588 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2589 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2591 seq_printf(f, "%s ",
2592 (lease_breaking(fl))
2593 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2594 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2597 /* userspace relies on this representation of dev_t */
2598 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2599 MAJOR(inode->i_sb->s_dev),
2600 MINOR(inode->i_sb->s_dev), inode->i_ino);
2602 seq_printf(f, "%d <none>:0 ", fl_pid);
2605 if (fl->fl_end == OFFSET_MAX)
2606 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2608 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2610 seq_puts(f, "0 EOF\n");
2614 static int locks_show(struct seq_file *f, void *v)
2616 struct locks_iterator *iter = f->private;
2617 struct file_lock *fl, *bfl;
2619 fl = hlist_entry(v, struct file_lock, fl_link);
2621 lock_get_status(f, fl, iter->li_pos, "");
2623 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2624 lock_get_status(f, bfl, iter->li_pos, " ->");
2629 static void __show_fd_locks(struct seq_file *f,
2630 struct list_head *head, int *id,
2631 struct file *filp, struct files_struct *files)
2633 struct file_lock *fl;
2635 list_for_each_entry(fl, head, fl_list) {
2637 if (filp != fl->fl_file)
2639 if (fl->fl_owner != files &&
2640 fl->fl_owner != filp)
2644 seq_puts(f, "lock:\t");
2645 lock_get_status(f, fl, *id, "");
2649 void show_fd_locks(struct seq_file *f,
2650 struct file *filp, struct files_struct *files)
2652 struct inode *inode = file_inode(filp);
2653 struct file_lock_context *ctx;
2656 ctx = smp_load_acquire(&inode->i_flctx);
2660 spin_lock(&ctx->flc_lock);
2661 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2662 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2663 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2664 spin_unlock(&ctx->flc_lock);
2667 static void *locks_start(struct seq_file *f, loff_t *pos)
2668 __acquires(&blocked_lock_lock)
2670 struct locks_iterator *iter = f->private;
2672 iter->li_pos = *pos + 1;
2673 lg_global_lock(&file_lock_lglock);
2674 spin_lock(&blocked_lock_lock);
2675 return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos);
2678 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2680 struct locks_iterator *iter = f->private;
2683 return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos);
2686 static void locks_stop(struct seq_file *f, void *v)
2687 __releases(&blocked_lock_lock)
2689 spin_unlock(&blocked_lock_lock);
2690 lg_global_unlock(&file_lock_lglock);
2693 static const struct seq_operations locks_seq_operations = {
2694 .start = locks_start,
2700 static int locks_open(struct inode *inode, struct file *filp)
2702 return seq_open_private(filp, &locks_seq_operations,
2703 sizeof(struct locks_iterator));
2706 static const struct file_operations proc_locks_operations = {
2709 .llseek = seq_lseek,
2710 .release = seq_release_private,
2713 static int __init proc_locks_init(void)
2715 proc_create("locks", 0, NULL, &proc_locks_operations);
2718 module_init(proc_locks_init);
2721 static int __init filelock_init(void)
2725 flctx_cache = kmem_cache_create("file_lock_ctx",
2726 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2728 filelock_cache = kmem_cache_create("file_lock_cache",
2729 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2731 lg_lock_init(&file_lock_lglock, "file_lock_lglock");
2733 for_each_possible_cpu(i)
2734 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i));
2739 core_initcall(filelock_init);