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/security.h>
123 #include <linux/slab.h>
124 #include <linux/syscalls.h>
125 #include <linux/time.h>
126 #include <linux/rcupdate.h>
127 #include <linux/pid_namespace.h>
128 #include <linux/hashtable.h>
129 #include <linux/percpu.h>
130 #include <linux/lglock.h>
132 #define CREATE_TRACE_POINTS
133 #include <trace/events/filelock.h>
135 #include <asm/uaccess.h>
137 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
138 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
139 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
140 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
142 static bool lease_breaking(struct file_lock *fl)
144 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
147 static int target_leasetype(struct file_lock *fl)
149 if (fl->fl_flags & FL_UNLOCK_PENDING)
151 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
156 int leases_enable = 1;
157 int lease_break_time = 45;
160 * The global file_lock_list is only used for displaying /proc/locks, so we
161 * keep a list on each CPU, with each list protected by its own spinlock via
162 * the file_lock_lglock. Note that alterations to the list also require that
163 * the relevant flc_lock is held.
165 DEFINE_STATIC_LGLOCK(file_lock_lglock);
166 static DEFINE_PER_CPU(struct hlist_head, file_lock_list);
169 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
170 * It is protected by blocked_lock_lock.
172 * We hash locks by lockowner in order to optimize searching for the lock a
173 * particular lockowner is waiting on.
175 * FIXME: make this value scale via some heuristic? We generally will want more
176 * buckets when we have more lockowners holding locks, but that's a little
177 * difficult to determine without knowing what the workload will look like.
179 #define BLOCKED_HASH_BITS 7
180 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
183 * This lock protects the blocked_hash. Generally, if you're accessing it, you
184 * want to be holding this lock.
186 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
187 * pointer for file_lock structures that are acting as lock requests (in
188 * contrast to those that are acting as records of acquired locks).
190 * Note that when we acquire this lock in order to change the above fields,
191 * we often hold the flc_lock as well. In certain cases, when reading the fields
192 * protected by this lock, we can skip acquiring it iff we already hold the
195 * In particular, adding an entry to the fl_block list requires that you hold
196 * both the flc_lock and the blocked_lock_lock (acquired in that order).
197 * Deleting an entry from the list however only requires the file_lock_lock.
199 static DEFINE_SPINLOCK(blocked_lock_lock);
201 static struct kmem_cache *flctx_cache __read_mostly;
202 static struct kmem_cache *filelock_cache __read_mostly;
204 static struct file_lock_context *
205 locks_get_lock_context(struct inode *inode, int type)
207 struct file_lock_context *ctx;
209 /* paired with cmpxchg() below */
210 ctx = smp_load_acquire(&inode->i_flctx);
211 if (likely(ctx) || type == F_UNLCK)
214 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
218 spin_lock_init(&ctx->flc_lock);
219 INIT_LIST_HEAD(&ctx->flc_flock);
220 INIT_LIST_HEAD(&ctx->flc_posix);
221 INIT_LIST_HEAD(&ctx->flc_lease);
224 * Assign the pointer if it's not already assigned. If it is, then
225 * free the context we just allocated.
227 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
228 kmem_cache_free(flctx_cache, ctx);
229 ctx = smp_load_acquire(&inode->i_flctx);
232 trace_locks_get_lock_context(inode, type, ctx);
237 locks_free_lock_context(struct inode *inode)
239 struct file_lock_context *ctx = inode->i_flctx;
242 WARN_ON_ONCE(!list_empty(&ctx->flc_flock));
243 WARN_ON_ONCE(!list_empty(&ctx->flc_posix));
244 WARN_ON_ONCE(!list_empty(&ctx->flc_lease));
245 kmem_cache_free(flctx_cache, ctx);
249 static void locks_init_lock_heads(struct file_lock *fl)
251 INIT_HLIST_NODE(&fl->fl_link);
252 INIT_LIST_HEAD(&fl->fl_list);
253 INIT_LIST_HEAD(&fl->fl_block);
254 init_waitqueue_head(&fl->fl_wait);
257 /* Allocate an empty lock structure. */
258 struct file_lock *locks_alloc_lock(void)
260 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
263 locks_init_lock_heads(fl);
267 EXPORT_SYMBOL_GPL(locks_alloc_lock);
269 void locks_release_private(struct file_lock *fl)
272 if (fl->fl_ops->fl_release_private)
273 fl->fl_ops->fl_release_private(fl);
278 if (fl->fl_lmops->lm_put_owner) {
279 fl->fl_lmops->lm_put_owner(fl->fl_owner);
285 EXPORT_SYMBOL_GPL(locks_release_private);
287 /* Free a lock which is not in use. */
288 void locks_free_lock(struct file_lock *fl)
290 BUG_ON(waitqueue_active(&fl->fl_wait));
291 BUG_ON(!list_empty(&fl->fl_list));
292 BUG_ON(!list_empty(&fl->fl_block));
293 BUG_ON(!hlist_unhashed(&fl->fl_link));
295 locks_release_private(fl);
296 kmem_cache_free(filelock_cache, fl);
298 EXPORT_SYMBOL(locks_free_lock);
301 locks_dispose_list(struct list_head *dispose)
303 struct file_lock *fl;
305 while (!list_empty(dispose)) {
306 fl = list_first_entry(dispose, struct file_lock, fl_list);
307 list_del_init(&fl->fl_list);
312 void locks_init_lock(struct file_lock *fl)
314 memset(fl, 0, sizeof(struct file_lock));
315 locks_init_lock_heads(fl);
318 EXPORT_SYMBOL(locks_init_lock);
321 * Initialize a new lock from an existing file_lock structure.
323 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
325 new->fl_owner = fl->fl_owner;
326 new->fl_pid = fl->fl_pid;
328 new->fl_flags = fl->fl_flags;
329 new->fl_type = fl->fl_type;
330 new->fl_start = fl->fl_start;
331 new->fl_end = fl->fl_end;
332 new->fl_lmops = fl->fl_lmops;
336 if (fl->fl_lmops->lm_get_owner)
337 fl->fl_lmops->lm_get_owner(fl->fl_owner);
340 EXPORT_SYMBOL(locks_copy_conflock);
342 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
344 /* "new" must be a freshly-initialized lock */
345 WARN_ON_ONCE(new->fl_ops);
347 locks_copy_conflock(new, fl);
349 new->fl_file = fl->fl_file;
350 new->fl_ops = fl->fl_ops;
353 if (fl->fl_ops->fl_copy_lock)
354 fl->fl_ops->fl_copy_lock(new, fl);
358 EXPORT_SYMBOL(locks_copy_lock);
360 static inline int flock_translate_cmd(int cmd) {
362 return cmd & (LOCK_MAND | LOCK_RW);
374 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
375 static struct file_lock *
376 flock_make_lock(struct file *filp, unsigned int cmd)
378 struct file_lock *fl;
379 int type = flock_translate_cmd(cmd);
382 return ERR_PTR(type);
384 fl = locks_alloc_lock();
386 return ERR_PTR(-ENOMEM);
390 fl->fl_pid = current->tgid;
391 fl->fl_flags = FL_FLOCK;
393 fl->fl_end = OFFSET_MAX;
398 static int assign_type(struct file_lock *fl, long type)
412 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
415 switch (l->l_whence) {
420 fl->fl_start = filp->f_pos;
423 fl->fl_start = i_size_read(file_inode(filp));
428 if (l->l_start > OFFSET_MAX - fl->fl_start)
430 fl->fl_start += l->l_start;
431 if (fl->fl_start < 0)
434 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
435 POSIX-2001 defines it. */
437 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
439 fl->fl_end = fl->fl_start + l->l_len - 1;
441 } else if (l->l_len < 0) {
442 if (fl->fl_start + l->l_len < 0)
444 fl->fl_end = fl->fl_start - 1;
445 fl->fl_start += l->l_len;
447 fl->fl_end = OFFSET_MAX;
449 fl->fl_owner = current->files;
450 fl->fl_pid = current->tgid;
452 fl->fl_flags = FL_POSIX;
456 return assign_type(fl, l->l_type);
459 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
462 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
465 struct flock64 ll = {
467 .l_whence = l->l_whence,
468 .l_start = l->l_start,
472 return flock64_to_posix_lock(filp, fl, &ll);
475 /* default lease lock manager operations */
477 lease_break_callback(struct file_lock *fl)
479 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
484 lease_setup(struct file_lock *fl, void **priv)
486 struct file *filp = fl->fl_file;
487 struct fasync_struct *fa = *priv;
490 * fasync_insert_entry() returns the old entry if any. If there was no
491 * old entry, then it used "priv" and inserted it into the fasync list.
492 * Clear the pointer to indicate that it shouldn't be freed.
494 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
497 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
500 static const struct lock_manager_operations lease_manager_ops = {
501 .lm_break = lease_break_callback,
502 .lm_change = lease_modify,
503 .lm_setup = lease_setup,
507 * Initialize a lease, use the default lock manager operations
509 static int lease_init(struct file *filp, long type, struct file_lock *fl)
511 if (assign_type(fl, type) != 0)
515 fl->fl_pid = current->tgid;
518 fl->fl_flags = FL_LEASE;
520 fl->fl_end = OFFSET_MAX;
522 fl->fl_lmops = &lease_manager_ops;
526 /* Allocate a file_lock initialised to this type of lease */
527 static struct file_lock *lease_alloc(struct file *filp, long type)
529 struct file_lock *fl = locks_alloc_lock();
533 return ERR_PTR(error);
535 error = lease_init(filp, type, fl);
538 return ERR_PTR(error);
543 /* Check if two locks overlap each other.
545 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
547 return ((fl1->fl_end >= fl2->fl_start) &&
548 (fl2->fl_end >= fl1->fl_start));
552 * Check whether two locks have the same owner.
554 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
556 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
557 return fl2->fl_lmops == fl1->fl_lmops &&
558 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
559 return fl1->fl_owner == fl2->fl_owner;
562 /* Must be called with the flc_lock held! */
563 static void locks_insert_global_locks(struct file_lock *fl)
565 lg_local_lock(&file_lock_lglock);
566 fl->fl_link_cpu = smp_processor_id();
567 hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list));
568 lg_local_unlock(&file_lock_lglock);
571 /* Must be called with the flc_lock held! */
572 static void locks_delete_global_locks(struct file_lock *fl)
575 * Avoid taking lock if already unhashed. This is safe since this check
576 * is done while holding the flc_lock, and new insertions into the list
577 * also require that it be held.
579 if (hlist_unhashed(&fl->fl_link))
581 lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu);
582 hlist_del_init(&fl->fl_link);
583 lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu);
587 posix_owner_key(struct file_lock *fl)
589 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
590 return fl->fl_lmops->lm_owner_key(fl);
591 return (unsigned long)fl->fl_owner;
594 static void locks_insert_global_blocked(struct file_lock *waiter)
596 lockdep_assert_held(&blocked_lock_lock);
598 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
601 static void locks_delete_global_blocked(struct file_lock *waiter)
603 lockdep_assert_held(&blocked_lock_lock);
605 hash_del(&waiter->fl_link);
608 /* Remove waiter from blocker's block list.
609 * When blocker ends up pointing to itself then the list is empty.
611 * Must be called with blocked_lock_lock held.
613 static void __locks_delete_block(struct file_lock *waiter)
615 locks_delete_global_blocked(waiter);
616 list_del_init(&waiter->fl_block);
617 waiter->fl_next = NULL;
620 static void locks_delete_block(struct file_lock *waiter)
622 spin_lock(&blocked_lock_lock);
623 __locks_delete_block(waiter);
624 spin_unlock(&blocked_lock_lock);
627 /* Insert waiter into blocker's block list.
628 * We use a circular list so that processes can be easily woken up in
629 * the order they blocked. The documentation doesn't require this but
630 * it seems like the reasonable thing to do.
632 * Must be called with both the flc_lock and blocked_lock_lock held. The
633 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
634 * that the flc_lock is also held on insertions we can avoid taking the
635 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
637 static void __locks_insert_block(struct file_lock *blocker,
638 struct file_lock *waiter)
640 BUG_ON(!list_empty(&waiter->fl_block));
641 waiter->fl_next = blocker;
642 list_add_tail(&waiter->fl_block, &blocker->fl_block);
643 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
644 locks_insert_global_blocked(waiter);
647 /* Must be called with flc_lock held. */
648 static void locks_insert_block(struct file_lock *blocker,
649 struct file_lock *waiter)
651 spin_lock(&blocked_lock_lock);
652 __locks_insert_block(blocker, waiter);
653 spin_unlock(&blocked_lock_lock);
657 * Wake up processes blocked waiting for blocker.
659 * Must be called with the inode->flc_lock held!
661 static void locks_wake_up_blocks(struct file_lock *blocker)
664 * Avoid taking global lock if list is empty. This is safe since new
665 * blocked requests are only added to the list under the flc_lock, and
666 * the flc_lock is always held here. Note that removal from the fl_block
667 * list does not require the flc_lock, so we must recheck list_empty()
668 * after acquiring the blocked_lock_lock.
670 if (list_empty(&blocker->fl_block))
673 spin_lock(&blocked_lock_lock);
674 while (!list_empty(&blocker->fl_block)) {
675 struct file_lock *waiter;
677 waiter = list_first_entry(&blocker->fl_block,
678 struct file_lock, fl_block);
679 __locks_delete_block(waiter);
680 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
681 waiter->fl_lmops->lm_notify(waiter);
683 wake_up(&waiter->fl_wait);
685 spin_unlock(&blocked_lock_lock);
689 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
691 fl->fl_nspid = get_pid(task_tgid(current));
692 list_add_tail(&fl->fl_list, before);
693 locks_insert_global_locks(fl);
697 locks_unlink_lock_ctx(struct file_lock *fl)
699 locks_delete_global_locks(fl);
700 list_del_init(&fl->fl_list);
702 put_pid(fl->fl_nspid);
705 locks_wake_up_blocks(fl);
709 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
711 locks_unlink_lock_ctx(fl);
713 list_add(&fl->fl_list, dispose);
718 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
719 * checks for shared/exclusive status of overlapping locks.
721 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
723 if (sys_fl->fl_type == F_WRLCK)
725 if (caller_fl->fl_type == F_WRLCK)
730 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
731 * checking before calling the locks_conflict().
733 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
735 /* POSIX locks owned by the same process do not conflict with
738 if (posix_same_owner(caller_fl, sys_fl))
741 /* Check whether they overlap */
742 if (!locks_overlap(caller_fl, sys_fl))
745 return (locks_conflict(caller_fl, sys_fl));
748 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
749 * checking before calling the locks_conflict().
751 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
753 /* FLOCK locks referring to the same filp do not conflict with
756 if (caller_fl->fl_file == sys_fl->fl_file)
758 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
761 return (locks_conflict(caller_fl, sys_fl));
765 posix_test_lock(struct file *filp, struct file_lock *fl)
767 struct file_lock *cfl;
768 struct file_lock_context *ctx;
769 struct inode *inode = file_inode(filp);
771 ctx = smp_load_acquire(&inode->i_flctx);
772 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
773 fl->fl_type = F_UNLCK;
777 spin_lock(&ctx->flc_lock);
778 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
779 if (posix_locks_conflict(fl, cfl)) {
780 locks_copy_conflock(fl, cfl);
782 fl->fl_pid = pid_vnr(cfl->fl_nspid);
786 fl->fl_type = F_UNLCK;
788 spin_unlock(&ctx->flc_lock);
791 EXPORT_SYMBOL(posix_test_lock);
794 * Deadlock detection:
796 * We attempt to detect deadlocks that are due purely to posix file
799 * We assume that a task can be waiting for at most one lock at a time.
800 * So for any acquired lock, the process holding that lock may be
801 * waiting on at most one other lock. That lock in turns may be held by
802 * someone waiting for at most one other lock. Given a requested lock
803 * caller_fl which is about to wait for a conflicting lock block_fl, we
804 * follow this chain of waiters to ensure we are not about to create a
807 * Since we do this before we ever put a process to sleep on a lock, we
808 * are ensured that there is never a cycle; that is what guarantees that
809 * the while() loop in posix_locks_deadlock() eventually completes.
811 * Note: the above assumption may not be true when handling lock
812 * requests from a broken NFS client. It may also fail in the presence
813 * of tasks (such as posix threads) sharing the same open file table.
814 * To handle those cases, we just bail out after a few iterations.
816 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
817 * Because the owner is not even nominally tied to a thread of
818 * execution, the deadlock detection below can't reasonably work well. Just
821 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
822 * locks that just checks for the case where two tasks are attempting to
823 * upgrade from read to write locks on the same inode.
826 #define MAX_DEADLK_ITERATIONS 10
828 /* Find a lock that the owner of the given block_fl is blocking on. */
829 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
831 struct file_lock *fl;
833 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
834 if (posix_same_owner(fl, block_fl))
840 /* Must be called with the blocked_lock_lock held! */
841 static int posix_locks_deadlock(struct file_lock *caller_fl,
842 struct file_lock *block_fl)
846 lockdep_assert_held(&blocked_lock_lock);
849 * This deadlock detector can't reasonably detect deadlocks with
850 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
852 if (IS_OFDLCK(caller_fl))
855 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
856 if (i++ > MAX_DEADLK_ITERATIONS)
858 if (posix_same_owner(caller_fl, block_fl))
864 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
865 * after any leases, but before any posix locks.
867 * Note that if called with an FL_EXISTS argument, the caller may determine
868 * whether or not a lock was successfully freed by testing the return
871 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
873 struct file_lock *new_fl = NULL;
874 struct file_lock *fl;
875 struct file_lock_context *ctx;
880 ctx = locks_get_lock_context(inode, request->fl_type);
882 if (request->fl_type != F_UNLCK)
884 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
887 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
888 new_fl = locks_alloc_lock();
893 spin_lock(&ctx->flc_lock);
894 if (request->fl_flags & FL_ACCESS)
897 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
898 if (request->fl_file != fl->fl_file)
900 if (request->fl_type == fl->fl_type)
903 locks_delete_lock_ctx(fl, &dispose);
907 if (request->fl_type == F_UNLCK) {
908 if ((request->fl_flags & FL_EXISTS) && !found)
914 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
915 if (!flock_locks_conflict(request, fl))
918 if (!(request->fl_flags & FL_SLEEP))
920 error = FILE_LOCK_DEFERRED;
921 locks_insert_block(fl, request);
924 if (request->fl_flags & FL_ACCESS)
926 locks_copy_lock(new_fl, request);
927 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
932 spin_unlock(&ctx->flc_lock);
934 locks_free_lock(new_fl);
935 locks_dispose_list(&dispose);
939 static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
941 struct file_lock *fl, *tmp;
942 struct file_lock *new_fl = NULL;
943 struct file_lock *new_fl2 = NULL;
944 struct file_lock *left = NULL;
945 struct file_lock *right = NULL;
946 struct file_lock_context *ctx;
951 ctx = locks_get_lock_context(inode, request->fl_type);
953 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
956 * We may need two file_lock structures for this operation,
957 * so we get them in advance to avoid races.
959 * In some cases we can be sure, that no new locks will be needed
961 if (!(request->fl_flags & FL_ACCESS) &&
962 (request->fl_type != F_UNLCK ||
963 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
964 new_fl = locks_alloc_lock();
965 new_fl2 = locks_alloc_lock();
968 spin_lock(&ctx->flc_lock);
970 * New lock request. Walk all POSIX locks and look for conflicts. If
971 * there are any, either return error or put the request on the
972 * blocker's list of waiters and the global blocked_hash.
974 if (request->fl_type != F_UNLCK) {
975 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
976 if (!posix_locks_conflict(request, fl))
979 locks_copy_conflock(conflock, fl);
981 if (!(request->fl_flags & FL_SLEEP))
984 * Deadlock detection and insertion into the blocked
985 * locks list must be done while holding the same lock!
988 spin_lock(&blocked_lock_lock);
989 if (likely(!posix_locks_deadlock(request, fl))) {
990 error = FILE_LOCK_DEFERRED;
991 __locks_insert_block(fl, request);
993 spin_unlock(&blocked_lock_lock);
998 /* If we're just looking for a conflict, we're done. */
1000 if (request->fl_flags & FL_ACCESS)
1003 /* Find the first old lock with the same owner as the new lock */
1004 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1005 if (posix_same_owner(request, fl))
1009 /* Process locks with this owner. */
1010 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1011 if (!posix_same_owner(request, fl))
1014 /* Detect adjacent or overlapping regions (if same lock type) */
1015 if (request->fl_type == fl->fl_type) {
1016 /* In all comparisons of start vs end, use
1017 * "start - 1" rather than "end + 1". If end
1018 * is OFFSET_MAX, end + 1 will become negative.
1020 if (fl->fl_end < request->fl_start - 1)
1022 /* If the next lock in the list has entirely bigger
1023 * addresses than the new one, insert the lock here.
1025 if (fl->fl_start - 1 > request->fl_end)
1028 /* If we come here, the new and old lock are of the
1029 * same type and adjacent or overlapping. Make one
1030 * lock yielding from the lower start address of both
1031 * locks to the higher end address.
1033 if (fl->fl_start > request->fl_start)
1034 fl->fl_start = request->fl_start;
1036 request->fl_start = fl->fl_start;
1037 if (fl->fl_end < request->fl_end)
1038 fl->fl_end = request->fl_end;
1040 request->fl_end = fl->fl_end;
1042 locks_delete_lock_ctx(fl, &dispose);
1048 /* Processing for different lock types is a bit
1051 if (fl->fl_end < request->fl_start)
1053 if (fl->fl_start > request->fl_end)
1055 if (request->fl_type == F_UNLCK)
1057 if (fl->fl_start < request->fl_start)
1059 /* If the next lock in the list has a higher end
1060 * address than the new one, insert the new one here.
1062 if (fl->fl_end > request->fl_end) {
1066 if (fl->fl_start >= request->fl_start) {
1067 /* The new lock completely replaces an old
1068 * one (This may happen several times).
1071 locks_delete_lock_ctx(fl, &dispose);
1075 * Replace the old lock with new_fl, and
1076 * remove the old one. It's safe to do the
1077 * insert here since we know that we won't be
1078 * using new_fl later, and that the lock is
1079 * just replacing an existing lock.
1084 locks_copy_lock(new_fl, request);
1087 locks_insert_lock_ctx(request, &fl->fl_list);
1088 locks_delete_lock_ctx(fl, &dispose);
1095 * The above code only modifies existing locks in case of merging or
1096 * replacing. If new lock(s) need to be inserted all modifications are
1097 * done below this, so it's safe yet to bail out.
1099 error = -ENOLCK; /* "no luck" */
1100 if (right && left == right && !new_fl2)
1105 if (request->fl_type == F_UNLCK) {
1106 if (request->fl_flags & FL_EXISTS)
1115 locks_copy_lock(new_fl, request);
1116 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1121 if (left == right) {
1122 /* The new lock breaks the old one in two pieces,
1123 * so we have to use the second new lock.
1127 locks_copy_lock(left, right);
1128 locks_insert_lock_ctx(left, &fl->fl_list);
1130 right->fl_start = request->fl_end + 1;
1131 locks_wake_up_blocks(right);
1134 left->fl_end = request->fl_start - 1;
1135 locks_wake_up_blocks(left);
1138 spin_unlock(&ctx->flc_lock);
1140 * Free any unused locks.
1143 locks_free_lock(new_fl);
1145 locks_free_lock(new_fl2);
1146 locks_dispose_list(&dispose);
1147 trace_posix_lock_inode(inode, request, error);
1153 * posix_lock_file - Apply a POSIX-style lock to a file
1154 * @filp: The file to apply the lock to
1155 * @fl: The lock to be applied
1156 * @conflock: Place to return a copy of the conflicting lock, if found.
1158 * Add a POSIX style lock to a file.
1159 * We merge adjacent & overlapping locks whenever possible.
1160 * POSIX locks are sorted by owner task, then by starting address
1162 * Note that if called with an FL_EXISTS argument, the caller may determine
1163 * whether or not a lock was successfully freed by testing the return
1164 * value for -ENOENT.
1166 int posix_lock_file(struct file *filp, struct file_lock *fl,
1167 struct file_lock *conflock)
1169 return __posix_lock_file(file_inode(filp), fl, conflock);
1171 EXPORT_SYMBOL(posix_lock_file);
1174 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1175 * @inode: inode of file to which lock request should be applied
1176 * @fl: The lock to be applied
1178 * Apply a POSIX style lock request to an inode.
1180 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1185 error = __posix_lock_file(inode, fl, NULL);
1186 if (error != FILE_LOCK_DEFERRED)
1188 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1192 locks_delete_block(fl);
1198 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1200 * locks_mandatory_locked - Check for an active lock
1201 * @file: the file to check
1203 * Searches the inode's list of locks to find any POSIX locks which conflict.
1204 * This function is called from locks_verify_locked() only.
1206 int locks_mandatory_locked(struct file *file)
1209 struct inode *inode = file_inode(file);
1210 struct file_lock_context *ctx;
1211 struct file_lock *fl;
1213 ctx = smp_load_acquire(&inode->i_flctx);
1214 if (!ctx || list_empty_careful(&ctx->flc_posix))
1218 * Search the lock list for this inode for any POSIX locks.
1220 spin_lock(&ctx->flc_lock);
1222 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1223 if (fl->fl_owner != current->files &&
1224 fl->fl_owner != file) {
1229 spin_unlock(&ctx->flc_lock);
1234 * locks_mandatory_area - Check for a conflicting lock
1235 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1237 * @inode: the file to check
1238 * @filp: how the file was opened (if it was)
1239 * @offset: start of area to check
1240 * @count: length of area to check
1242 * Searches the inode's list of locks to find any POSIX locks which conflict.
1243 * This function is called from rw_verify_area() and
1244 * locks_verify_truncate().
1246 int locks_mandatory_area(int read_write, struct inode *inode,
1247 struct file *filp, loff_t offset,
1250 struct file_lock fl;
1254 locks_init_lock(&fl);
1255 fl.fl_pid = current->tgid;
1257 fl.fl_flags = FL_POSIX | FL_ACCESS;
1258 if (filp && !(filp->f_flags & O_NONBLOCK))
1260 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1261 fl.fl_start = offset;
1262 fl.fl_end = offset + count - 1;
1267 fl.fl_flags &= ~FL_SLEEP;
1268 error = __posix_lock_file(inode, &fl, NULL);
1274 fl.fl_flags |= FL_SLEEP;
1275 fl.fl_owner = current->files;
1276 error = __posix_lock_file(inode, &fl, NULL);
1277 if (error != FILE_LOCK_DEFERRED)
1279 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1282 * If we've been sleeping someone might have
1283 * changed the permissions behind our back.
1285 if (__mandatory_lock(inode))
1289 locks_delete_block(&fl);
1296 EXPORT_SYMBOL(locks_mandatory_area);
1297 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1299 static void lease_clear_pending(struct file_lock *fl, int arg)
1303 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1306 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1310 /* We already had a lease on this file; just change its type */
1311 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1313 int error = assign_type(fl, arg);
1317 lease_clear_pending(fl, arg);
1318 locks_wake_up_blocks(fl);
1319 if (arg == F_UNLCK) {
1320 struct file *filp = fl->fl_file;
1323 filp->f_owner.signum = 0;
1324 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1325 if (fl->fl_fasync != NULL) {
1326 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1327 fl->fl_fasync = NULL;
1329 locks_delete_lock_ctx(fl, dispose);
1333 EXPORT_SYMBOL(lease_modify);
1335 static bool past_time(unsigned long then)
1338 /* 0 is a special value meaning "this never expires": */
1340 return time_after(jiffies, then);
1343 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1345 struct file_lock_context *ctx = inode->i_flctx;
1346 struct file_lock *fl, *tmp;
1348 lockdep_assert_held(&ctx->flc_lock);
1350 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1351 trace_time_out_leases(inode, fl);
1352 if (past_time(fl->fl_downgrade_time))
1353 lease_modify(fl, F_RDLCK, dispose);
1354 if (past_time(fl->fl_break_time))
1355 lease_modify(fl, F_UNLCK, dispose);
1359 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1361 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1363 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1365 return locks_conflict(breaker, lease);
1369 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1371 struct file_lock_context *ctx = inode->i_flctx;
1372 struct file_lock *fl;
1374 lockdep_assert_held(&ctx->flc_lock);
1376 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1377 if (leases_conflict(fl, breaker))
1384 * __break_lease - revoke all outstanding leases on file
1385 * @inode: the inode of the file to return
1386 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1388 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1391 * break_lease (inlined for speed) has checked there already is at least
1392 * some kind of lock (maybe a lease) on this file. Leases are broken on
1393 * a call to open() or truncate(). This function can sleep unless you
1394 * specified %O_NONBLOCK to your open().
1396 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1399 struct file_lock_context *ctx;
1400 struct file_lock *new_fl, *fl, *tmp;
1401 unsigned long break_time;
1402 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1405 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1407 return PTR_ERR(new_fl);
1408 new_fl->fl_flags = type;
1410 /* typically we will check that ctx is non-NULL before calling */
1411 ctx = smp_load_acquire(&inode->i_flctx);
1417 spin_lock(&ctx->flc_lock);
1419 time_out_leases(inode, &dispose);
1421 if (!any_leases_conflict(inode, new_fl))
1425 if (lease_break_time > 0) {
1426 break_time = jiffies + lease_break_time * HZ;
1427 if (break_time == 0)
1428 break_time++; /* so that 0 means no break time */
1431 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1432 if (!leases_conflict(fl, new_fl))
1435 if (fl->fl_flags & FL_UNLOCK_PENDING)
1437 fl->fl_flags |= FL_UNLOCK_PENDING;
1438 fl->fl_break_time = break_time;
1440 if (lease_breaking(fl))
1442 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1443 fl->fl_downgrade_time = break_time;
1445 if (fl->fl_lmops->lm_break(fl))
1446 locks_delete_lock_ctx(fl, &dispose);
1449 if (list_empty(&ctx->flc_lease))
1452 if (mode & O_NONBLOCK) {
1453 trace_break_lease_noblock(inode, new_fl);
1454 error = -EWOULDBLOCK;
1459 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1460 break_time = fl->fl_break_time;
1461 if (break_time != 0)
1462 break_time -= jiffies;
1463 if (break_time == 0)
1465 locks_insert_block(fl, new_fl);
1466 trace_break_lease_block(inode, new_fl);
1467 spin_unlock(&ctx->flc_lock);
1468 locks_dispose_list(&dispose);
1469 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1470 !new_fl->fl_next, break_time);
1471 spin_lock(&ctx->flc_lock);
1472 trace_break_lease_unblock(inode, new_fl);
1473 locks_delete_block(new_fl);
1476 * Wait for the next conflicting lease that has not been
1480 time_out_leases(inode, &dispose);
1481 if (any_leases_conflict(inode, new_fl))
1486 spin_unlock(&ctx->flc_lock);
1487 locks_dispose_list(&dispose);
1488 locks_free_lock(new_fl);
1492 EXPORT_SYMBOL(__break_lease);
1495 * lease_get_mtime - get the last modified time of an inode
1497 * @time: pointer to a timespec which will contain the last modified time
1499 * This is to force NFS clients to flush their caches for files with
1500 * exclusive leases. The justification is that if someone has an
1501 * exclusive lease, then they could be modifying it.
1503 void lease_get_mtime(struct inode *inode, struct timespec *time)
1505 bool has_lease = false;
1506 struct file_lock_context *ctx;
1507 struct file_lock *fl;
1509 ctx = smp_load_acquire(&inode->i_flctx);
1510 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1511 spin_lock(&ctx->flc_lock);
1512 fl = list_first_entry_or_null(&ctx->flc_lease,
1513 struct file_lock, fl_list);
1514 if (fl && (fl->fl_type == F_WRLCK))
1516 spin_unlock(&ctx->flc_lock);
1520 *time = current_fs_time(inode->i_sb);
1522 *time = inode->i_mtime;
1525 EXPORT_SYMBOL(lease_get_mtime);
1528 * fcntl_getlease - Enquire what lease is currently active
1531 * The value returned by this function will be one of
1532 * (if no lease break is pending):
1534 * %F_RDLCK to indicate a shared lease is held.
1536 * %F_WRLCK to indicate an exclusive lease is held.
1538 * %F_UNLCK to indicate no lease is held.
1540 * (if a lease break is pending):
1542 * %F_RDLCK to indicate an exclusive lease needs to be
1543 * changed to a shared lease (or removed).
1545 * %F_UNLCK to indicate the lease needs to be removed.
1547 * XXX: sfr & willy disagree over whether F_INPROGRESS
1548 * should be returned to userspace.
1550 int fcntl_getlease(struct file *filp)
1552 struct file_lock *fl;
1553 struct inode *inode = file_inode(filp);
1554 struct file_lock_context *ctx;
1558 ctx = smp_load_acquire(&inode->i_flctx);
1559 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1560 spin_lock(&ctx->flc_lock);
1561 time_out_leases(file_inode(filp), &dispose);
1562 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1563 if (fl->fl_file != filp)
1565 type = target_leasetype(fl);
1568 spin_unlock(&ctx->flc_lock);
1569 locks_dispose_list(&dispose);
1575 * check_conflicting_open - see if the given dentry points to a file that has
1576 * an existing open that would conflict with the
1578 * @dentry: dentry to check
1579 * @arg: type of lease that we're trying to acquire
1580 * @flags: current lock flags
1582 * Check to see if there's an existing open fd on this file that would
1583 * conflict with the lease we're trying to set.
1586 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1589 struct inode *inode = dentry->d_inode;
1591 if (flags & FL_LAYOUT)
1594 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1597 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1598 (atomic_read(&inode->i_count) > 1)))
1605 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1607 struct file_lock *fl, *my_fl = NULL, *lease;
1608 struct dentry *dentry = filp->f_path.dentry;
1609 struct inode *inode = dentry->d_inode;
1610 struct file_lock_context *ctx;
1611 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1616 trace_generic_add_lease(inode, lease);
1618 /* Note that arg is never F_UNLCK here */
1619 ctx = locks_get_lock_context(inode, arg);
1624 * In the delegation case we need mutual exclusion with
1625 * a number of operations that take the i_mutex. We trylock
1626 * because delegations are an optional optimization, and if
1627 * there's some chance of a conflict--we'd rather not
1628 * bother, maybe that's a sign this just isn't a good file to
1629 * hand out a delegation on.
1631 if (is_deleg && !mutex_trylock(&inode->i_mutex))
1634 if (is_deleg && arg == F_WRLCK) {
1635 /* Write delegations are not currently supported: */
1636 mutex_unlock(&inode->i_mutex);
1641 spin_lock(&ctx->flc_lock);
1642 time_out_leases(inode, &dispose);
1643 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1648 * At this point, we know that if there is an exclusive
1649 * lease on this file, then we hold it on this filp
1650 * (otherwise our open of this file would have blocked).
1651 * And if we are trying to acquire an exclusive lease,
1652 * then the file is not open by anyone (including us)
1653 * except for this filp.
1656 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1657 if (fl->fl_file == filp &&
1658 fl->fl_owner == lease->fl_owner) {
1664 * No exclusive leases if someone else has a lease on
1670 * Modifying our existing lease is OK, but no getting a
1671 * new lease if someone else is opening for write:
1673 if (fl->fl_flags & FL_UNLOCK_PENDING)
1677 if (my_fl != NULL) {
1679 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1689 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1691 * The check in break_lease() is lockless. It's possible for another
1692 * open to race in after we did the earlier check for a conflicting
1693 * open but before the lease was inserted. Check again for a
1694 * conflicting open and cancel the lease if there is one.
1696 * We also add a barrier here to ensure that the insertion of the lock
1697 * precedes these checks.
1700 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1702 locks_unlink_lock_ctx(lease);
1707 if (lease->fl_lmops->lm_setup)
1708 lease->fl_lmops->lm_setup(lease, priv);
1710 spin_unlock(&ctx->flc_lock);
1711 locks_dispose_list(&dispose);
1713 mutex_unlock(&inode->i_mutex);
1714 if (!error && !my_fl)
1719 static int generic_delete_lease(struct file *filp, void *owner)
1721 int error = -EAGAIN;
1722 struct file_lock *fl, *victim = NULL;
1723 struct inode *inode = file_inode(filp);
1724 struct file_lock_context *ctx;
1727 ctx = smp_load_acquire(&inode->i_flctx);
1729 trace_generic_delete_lease(inode, NULL);
1733 spin_lock(&ctx->flc_lock);
1734 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1735 if (fl->fl_file == filp &&
1736 fl->fl_owner == owner) {
1741 trace_generic_delete_lease(inode, victim);
1743 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1744 spin_unlock(&ctx->flc_lock);
1745 locks_dispose_list(&dispose);
1750 * generic_setlease - sets a lease on an open file
1751 * @filp: file pointer
1752 * @arg: type of lease to obtain
1753 * @flp: input - file_lock to use, output - file_lock inserted
1754 * @priv: private data for lm_setup (may be NULL if lm_setup
1755 * doesn't require it)
1757 * The (input) flp->fl_lmops->lm_break function is required
1760 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1763 struct inode *inode = file_inode(filp);
1766 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1768 if (!S_ISREG(inode->i_mode))
1770 error = security_file_lock(filp, arg);
1776 return generic_delete_lease(filp, *priv);
1779 if (!(*flp)->fl_lmops->lm_break) {
1784 return generic_add_lease(filp, arg, flp, priv);
1789 EXPORT_SYMBOL(generic_setlease);
1792 * vfs_setlease - sets a lease on an open file
1793 * @filp: file pointer
1794 * @arg: type of lease to obtain
1795 * @lease: file_lock to use when adding a lease
1796 * @priv: private info for lm_setup when adding a lease (may be
1797 * NULL if lm_setup doesn't require it)
1799 * Call this to establish a lease on the file. The "lease" argument is not
1800 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1801 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1802 * if not, this function will return -ENOLCK (and generate a scary-looking
1805 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1806 * may be NULL if the lm_setup operation doesn't require it.
1809 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1811 if (filp->f_op->setlease)
1812 return filp->f_op->setlease(filp, arg, lease, priv);
1814 return generic_setlease(filp, arg, lease, priv);
1816 EXPORT_SYMBOL_GPL(vfs_setlease);
1818 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1820 struct file_lock *fl;
1821 struct fasync_struct *new;
1824 fl = lease_alloc(filp, arg);
1828 new = fasync_alloc();
1830 locks_free_lock(fl);
1835 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1837 locks_free_lock(fl);
1844 * fcntl_setlease - sets a lease on an open file
1845 * @fd: open file descriptor
1846 * @filp: file pointer
1847 * @arg: type of lease to obtain
1849 * Call this fcntl to establish a lease on the file.
1850 * Note that you also need to call %F_SETSIG to
1851 * receive a signal when the lease is broken.
1853 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1856 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1857 return do_fcntl_add_lease(fd, filp, arg);
1861 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1862 * @inode: inode of the file to apply to
1863 * @fl: The lock to be applied
1865 * Apply a FLOCK style lock request to an inode.
1867 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1872 error = flock_lock_inode(inode, fl);
1873 if (error != FILE_LOCK_DEFERRED)
1875 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1879 locks_delete_block(fl);
1886 * locks_lock_inode_wait - Apply a lock to an inode
1887 * @inode: inode of the file to apply to
1888 * @fl: The lock to be applied
1890 * Apply a POSIX or FLOCK style lock request to an inode.
1892 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1895 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1897 res = posix_lock_inode_wait(inode, fl);
1900 res = flock_lock_inode_wait(inode, fl);
1907 EXPORT_SYMBOL(locks_lock_inode_wait);
1910 * sys_flock: - flock() system call.
1911 * @fd: the file descriptor to lock.
1912 * @cmd: the type of lock to apply.
1914 * Apply a %FL_FLOCK style lock to an open file descriptor.
1915 * The @cmd can be one of
1917 * %LOCK_SH -- a shared lock.
1919 * %LOCK_EX -- an exclusive lock.
1921 * %LOCK_UN -- remove an existing lock.
1923 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1925 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1926 * processes read and write access respectively.
1928 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1930 struct fd f = fdget(fd);
1931 struct file_lock *lock;
1932 int can_sleep, unlock;
1939 can_sleep = !(cmd & LOCK_NB);
1941 unlock = (cmd == LOCK_UN);
1943 if (!unlock && !(cmd & LOCK_MAND) &&
1944 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
1947 lock = flock_make_lock(f.file, cmd);
1949 error = PTR_ERR(lock);
1954 lock->fl_flags |= FL_SLEEP;
1956 error = security_file_lock(f.file, lock->fl_type);
1960 if (f.file->f_op->flock)
1961 error = f.file->f_op->flock(f.file,
1962 (can_sleep) ? F_SETLKW : F_SETLK,
1965 error = locks_lock_file_wait(f.file, lock);
1968 locks_free_lock(lock);
1977 * vfs_test_lock - test file byte range lock
1978 * @filp: The file to test lock for
1979 * @fl: The lock to test; also used to hold result
1981 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1982 * setting conf->fl_type to something other than F_UNLCK.
1984 int vfs_test_lock(struct file *filp, struct file_lock *fl)
1986 if (filp->f_op->lock)
1987 return filp->f_op->lock(filp, F_GETLK, fl);
1988 posix_test_lock(filp, fl);
1991 EXPORT_SYMBOL_GPL(vfs_test_lock);
1993 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1995 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
1996 #if BITS_PER_LONG == 32
1998 * Make sure we can represent the posix lock via
1999 * legacy 32bit flock.
2001 if (fl->fl_start > OFFT_OFFSET_MAX)
2003 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2006 flock->l_start = fl->fl_start;
2007 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2008 fl->fl_end - fl->fl_start + 1;
2009 flock->l_whence = 0;
2010 flock->l_type = fl->fl_type;
2014 #if BITS_PER_LONG == 32
2015 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2017 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
2018 flock->l_start = fl->fl_start;
2019 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2020 fl->fl_end - fl->fl_start + 1;
2021 flock->l_whence = 0;
2022 flock->l_type = fl->fl_type;
2026 /* Report the first existing lock that would conflict with l.
2027 * This implements the F_GETLK command of fcntl().
2029 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
2031 struct file_lock file_lock;
2036 if (copy_from_user(&flock, l, sizeof(flock)))
2039 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2042 error = flock_to_posix_lock(filp, &file_lock, &flock);
2046 if (cmd == F_OFD_GETLK) {
2048 if (flock.l_pid != 0)
2052 file_lock.fl_flags |= FL_OFDLCK;
2053 file_lock.fl_owner = filp;
2056 error = vfs_test_lock(filp, &file_lock);
2060 flock.l_type = file_lock.fl_type;
2061 if (file_lock.fl_type != F_UNLCK) {
2062 error = posix_lock_to_flock(&flock, &file_lock);
2067 if (!copy_to_user(l, &flock, sizeof(flock)))
2070 locks_release_private(&file_lock);
2076 * vfs_lock_file - file byte range lock
2077 * @filp: The file to apply the lock to
2078 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2079 * @fl: The lock to be applied
2080 * @conf: Place to return a copy of the conflicting lock, if found.
2082 * A caller that doesn't care about the conflicting lock may pass NULL
2083 * as the final argument.
2085 * If the filesystem defines a private ->lock() method, then @conf will
2086 * be left unchanged; so a caller that cares should initialize it to
2087 * some acceptable default.
2089 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2090 * locks, the ->lock() interface may return asynchronously, before the lock has
2091 * been granted or denied by the underlying filesystem, if (and only if)
2092 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2093 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2094 * the request is for a blocking lock. When ->lock() does return asynchronously,
2095 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2096 * request completes.
2097 * If the request is for non-blocking lock the file system should return
2098 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2099 * with the result. If the request timed out the callback routine will return a
2100 * nonzero return code and the file system should release the lock. The file
2101 * system is also responsible to keep a corresponding posix lock when it
2102 * grants a lock so the VFS can find out which locks are locally held and do
2103 * the correct lock cleanup when required.
2104 * The underlying filesystem must not drop the kernel lock or call
2105 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2108 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2110 if (filp->f_op->lock)
2111 return filp->f_op->lock(filp, cmd, fl);
2113 return posix_lock_file(filp, fl, conf);
2115 EXPORT_SYMBOL_GPL(vfs_lock_file);
2117 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2118 struct file_lock *fl)
2122 error = security_file_lock(filp, fl->fl_type);
2127 error = vfs_lock_file(filp, cmd, fl, NULL);
2128 if (error != FILE_LOCK_DEFERRED)
2130 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2134 locks_delete_block(fl);
2141 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2143 check_fmode_for_setlk(struct file_lock *fl)
2145 switch (fl->fl_type) {
2147 if (!(fl->fl_file->f_mode & FMODE_READ))
2151 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2157 /* Apply the lock described by l to an open file descriptor.
2158 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2160 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2161 struct flock __user *l)
2163 struct file_lock *file_lock = locks_alloc_lock();
2165 struct inode *inode;
2169 if (file_lock == NULL)
2172 inode = file_inode(filp);
2175 * This might block, so we do it before checking the inode.
2178 if (copy_from_user(&flock, l, sizeof(flock)))
2181 /* Don't allow mandatory locks on files that may be memory mapped
2184 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2189 error = flock_to_posix_lock(filp, file_lock, &flock);
2193 error = check_fmode_for_setlk(file_lock);
2198 * If the cmd is requesting file-private locks, then set the
2199 * FL_OFDLCK flag and override the owner.
2204 if (flock.l_pid != 0)
2208 file_lock->fl_flags |= FL_OFDLCK;
2209 file_lock->fl_owner = filp;
2213 if (flock.l_pid != 0)
2217 file_lock->fl_flags |= FL_OFDLCK;
2218 file_lock->fl_owner = filp;
2221 file_lock->fl_flags |= FL_SLEEP;
2224 error = do_lock_file_wait(filp, cmd, file_lock);
2227 * Attempt to detect a close/fcntl race and recover by releasing the
2228 * lock that was just acquired. There is no need to do that when we're
2229 * unlocking though, or for OFD locks.
2231 if (!error && file_lock->fl_type != F_UNLCK &&
2232 !(file_lock->fl_flags & FL_OFDLCK)) {
2234 * We need that spin_lock here - it prevents reordering between
2235 * update of i_flctx->flc_posix and check for it done in
2236 * close(). rcu_read_lock() wouldn't do.
2238 spin_lock(¤t->files->file_lock);
2240 spin_unlock(¤t->files->file_lock);
2242 file_lock->fl_type = F_UNLCK;
2243 error = do_lock_file_wait(filp, cmd, file_lock);
2244 WARN_ON_ONCE(error);
2249 trace_fcntl_setlk(inode, file_lock, error);
2250 locks_free_lock(file_lock);
2254 #if BITS_PER_LONG == 32
2255 /* Report the first existing lock that would conflict with l.
2256 * This implements the F_GETLK command of fcntl().
2258 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
2260 struct file_lock file_lock;
2261 struct flock64 flock;
2265 if (copy_from_user(&flock, l, sizeof(flock)))
2268 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
2271 error = flock64_to_posix_lock(filp, &file_lock, &flock);
2275 if (cmd == F_OFD_GETLK) {
2277 if (flock.l_pid != 0)
2281 file_lock.fl_flags |= FL_OFDLCK;
2282 file_lock.fl_owner = filp;
2285 error = vfs_test_lock(filp, &file_lock);
2289 flock.l_type = file_lock.fl_type;
2290 if (file_lock.fl_type != F_UNLCK)
2291 posix_lock_to_flock64(&flock, &file_lock);
2294 if (!copy_to_user(l, &flock, sizeof(flock)))
2297 locks_release_private(&file_lock);
2302 /* Apply the lock described by l to an open file descriptor.
2303 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2305 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2306 struct flock64 __user *l)
2308 struct file_lock *file_lock = locks_alloc_lock();
2309 struct flock64 flock;
2310 struct inode *inode;
2314 if (file_lock == NULL)
2318 * This might block, so we do it before checking the inode.
2321 if (copy_from_user(&flock, l, sizeof(flock)))
2324 inode = file_inode(filp);
2326 /* Don't allow mandatory locks on files that may be memory mapped
2329 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2334 error = flock64_to_posix_lock(filp, file_lock, &flock);
2338 error = check_fmode_for_setlk(file_lock);
2343 * If the cmd is requesting file-private locks, then set the
2344 * FL_OFDLCK flag and override the owner.
2349 if (flock.l_pid != 0)
2353 file_lock->fl_flags |= FL_OFDLCK;
2354 file_lock->fl_owner = filp;
2358 if (flock.l_pid != 0)
2362 file_lock->fl_flags |= FL_OFDLCK;
2363 file_lock->fl_owner = filp;
2366 file_lock->fl_flags |= FL_SLEEP;
2369 error = do_lock_file_wait(filp, cmd, file_lock);
2372 * Attempt to detect a close/fcntl race and recover by releasing the
2373 * lock that was just acquired. There is no need to do that when we're
2374 * unlocking though, or for OFD locks.
2376 if (!error && file_lock->fl_type != F_UNLCK &&
2377 !(file_lock->fl_flags & FL_OFDLCK)) {
2379 * We need that spin_lock here - it prevents reordering between
2380 * update of i_flctx->flc_posix and check for it done in
2381 * close(). rcu_read_lock() wouldn't do.
2383 spin_lock(¤t->files->file_lock);
2385 spin_unlock(¤t->files->file_lock);
2387 file_lock->fl_type = F_UNLCK;
2388 error = do_lock_file_wait(filp, cmd, file_lock);
2389 WARN_ON_ONCE(error);
2394 locks_free_lock(file_lock);
2397 #endif /* BITS_PER_LONG == 32 */
2400 * This function is called when the file is being removed
2401 * from the task's fd array. POSIX locks belonging to this task
2402 * are deleted at this time.
2404 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2407 struct file_lock lock;
2408 struct file_lock_context *ctx;
2411 * If there are no locks held on this file, we don't need to call
2412 * posix_lock_file(). Another process could be setting a lock on this
2413 * file at the same time, but we wouldn't remove that lock anyway.
2415 ctx = smp_load_acquire(&file_inode(filp)->i_flctx);
2416 if (!ctx || list_empty(&ctx->flc_posix))
2419 lock.fl_type = F_UNLCK;
2420 lock.fl_flags = FL_POSIX | FL_CLOSE;
2422 lock.fl_end = OFFSET_MAX;
2423 lock.fl_owner = owner;
2424 lock.fl_pid = current->tgid;
2425 lock.fl_file = filp;
2427 lock.fl_lmops = NULL;
2429 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
2431 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2432 lock.fl_ops->fl_release_private(&lock);
2433 trace_locks_remove_posix(file_inode(filp), &lock, error);
2436 EXPORT_SYMBOL(locks_remove_posix);
2438 /* The i_flctx must be valid when calling into here */
2440 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2442 struct file_lock fl = {
2444 .fl_pid = current->tgid,
2446 .fl_flags = FL_FLOCK,
2448 .fl_end = OFFSET_MAX,
2450 struct inode *inode = file_inode(filp);
2452 if (list_empty(&flctx->flc_flock))
2455 if (filp->f_op->flock)
2456 filp->f_op->flock(filp, F_SETLKW, &fl);
2458 flock_lock_inode(inode, &fl);
2460 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2461 fl.fl_ops->fl_release_private(&fl);
2464 /* The i_flctx must be valid when calling into here */
2466 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2468 struct file_lock *fl, *tmp;
2471 if (list_empty(&ctx->flc_lease))
2474 spin_lock(&ctx->flc_lock);
2475 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2476 if (filp == fl->fl_file)
2477 lease_modify(fl, F_UNLCK, &dispose);
2478 spin_unlock(&ctx->flc_lock);
2479 locks_dispose_list(&dispose);
2483 * This function is called on the last close of an open file.
2485 void locks_remove_file(struct file *filp)
2487 struct file_lock_context *ctx;
2489 ctx = smp_load_acquire(&file_inode(filp)->i_flctx);
2493 /* remove any OFD locks */
2494 locks_remove_posix(filp, filp);
2496 /* remove flock locks */
2497 locks_remove_flock(filp, ctx);
2499 /* remove any leases */
2500 locks_remove_lease(filp, ctx);
2504 * posix_unblock_lock - stop waiting for a file lock
2505 * @waiter: the lock which was waiting
2507 * lockd needs to block waiting for locks.
2510 posix_unblock_lock(struct file_lock *waiter)
2514 spin_lock(&blocked_lock_lock);
2515 if (waiter->fl_next)
2516 __locks_delete_block(waiter);
2519 spin_unlock(&blocked_lock_lock);
2522 EXPORT_SYMBOL(posix_unblock_lock);
2525 * vfs_cancel_lock - file byte range unblock lock
2526 * @filp: The file to apply the unblock to
2527 * @fl: The lock to be unblocked
2529 * Used by lock managers to cancel blocked requests
2531 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2533 if (filp->f_op->lock)
2534 return filp->f_op->lock(filp, F_CANCELLK, fl);
2538 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2540 #ifdef CONFIG_PROC_FS
2541 #include <linux/proc_fs.h>
2542 #include <linux/seq_file.h>
2544 struct locks_iterator {
2549 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2550 loff_t id, char *pfx)
2552 struct inode *inode = NULL;
2553 unsigned int fl_pid;
2556 fl_pid = pid_vnr(fl->fl_nspid);
2558 fl_pid = fl->fl_pid;
2560 if (fl->fl_file != NULL)
2561 inode = file_inode(fl->fl_file);
2563 seq_printf(f, "%lld:%s ", id, pfx);
2565 if (fl->fl_flags & FL_ACCESS)
2566 seq_puts(f, "ACCESS");
2567 else if (IS_OFDLCK(fl))
2568 seq_puts(f, "OFDLCK");
2570 seq_puts(f, "POSIX ");
2572 seq_printf(f, " %s ",
2573 (inode == NULL) ? "*NOINODE*" :
2574 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2575 } else if (IS_FLOCK(fl)) {
2576 if (fl->fl_type & LOCK_MAND) {
2577 seq_puts(f, "FLOCK MSNFS ");
2579 seq_puts(f, "FLOCK ADVISORY ");
2581 } else if (IS_LEASE(fl)) {
2582 if (fl->fl_flags & FL_DELEG)
2583 seq_puts(f, "DELEG ");
2585 seq_puts(f, "LEASE ");
2587 if (lease_breaking(fl))
2588 seq_puts(f, "BREAKING ");
2589 else if (fl->fl_file)
2590 seq_puts(f, "ACTIVE ");
2592 seq_puts(f, "BREAKER ");
2594 seq_puts(f, "UNKNOWN UNKNOWN ");
2596 if (fl->fl_type & LOCK_MAND) {
2597 seq_printf(f, "%s ",
2598 (fl->fl_type & LOCK_READ)
2599 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2600 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2602 seq_printf(f, "%s ",
2603 (lease_breaking(fl))
2604 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2605 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2608 /* userspace relies on this representation of dev_t */
2609 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2610 MAJOR(inode->i_sb->s_dev),
2611 MINOR(inode->i_sb->s_dev), inode->i_ino);
2613 seq_printf(f, "%d <none>:0 ", fl_pid);
2616 if (fl->fl_end == OFFSET_MAX)
2617 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2619 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2621 seq_puts(f, "0 EOF\n");
2625 static int locks_show(struct seq_file *f, void *v)
2627 struct locks_iterator *iter = f->private;
2628 struct file_lock *fl, *bfl;
2630 fl = hlist_entry(v, struct file_lock, fl_link);
2632 lock_get_status(f, fl, iter->li_pos, "");
2634 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2635 lock_get_status(f, bfl, iter->li_pos, " ->");
2640 static void __show_fd_locks(struct seq_file *f,
2641 struct list_head *head, int *id,
2642 struct file *filp, struct files_struct *files)
2644 struct file_lock *fl;
2646 list_for_each_entry(fl, head, fl_list) {
2648 if (filp != fl->fl_file)
2650 if (fl->fl_owner != files &&
2651 fl->fl_owner != filp)
2655 seq_puts(f, "lock:\t");
2656 lock_get_status(f, fl, *id, "");
2660 void show_fd_locks(struct seq_file *f,
2661 struct file *filp, struct files_struct *files)
2663 struct inode *inode = file_inode(filp);
2664 struct file_lock_context *ctx;
2667 ctx = smp_load_acquire(&inode->i_flctx);
2671 spin_lock(&ctx->flc_lock);
2672 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2673 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2674 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2675 spin_unlock(&ctx->flc_lock);
2678 static void *locks_start(struct seq_file *f, loff_t *pos)
2679 __acquires(&blocked_lock_lock)
2681 struct locks_iterator *iter = f->private;
2683 iter->li_pos = *pos + 1;
2684 lg_global_lock(&file_lock_lglock);
2685 spin_lock(&blocked_lock_lock);
2686 return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos);
2689 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2691 struct locks_iterator *iter = f->private;
2694 return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos);
2697 static void locks_stop(struct seq_file *f, void *v)
2698 __releases(&blocked_lock_lock)
2700 spin_unlock(&blocked_lock_lock);
2701 lg_global_unlock(&file_lock_lglock);
2704 static const struct seq_operations locks_seq_operations = {
2705 .start = locks_start,
2711 static int locks_open(struct inode *inode, struct file *filp)
2713 return seq_open_private(filp, &locks_seq_operations,
2714 sizeof(struct locks_iterator));
2717 static const struct file_operations proc_locks_operations = {
2720 .llseek = seq_lseek,
2721 .release = seq_release_private,
2724 static int __init proc_locks_init(void)
2726 proc_create("locks", 0, NULL, &proc_locks_operations);
2729 fs_initcall(proc_locks_init);
2732 static int __init filelock_init(void)
2736 flctx_cache = kmem_cache_create("file_lock_ctx",
2737 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2739 filelock_cache = kmem_cache_create("file_lock_cache",
2740 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2742 lg_lock_init(&file_lock_lglock, "file_lock_lglock");
2744 for_each_possible_cpu(i)
2745 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i));
2750 core_initcall(filelock_init);