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>
131 #define CREATE_TRACE_POINTS
132 #include <trace/events/filelock.h>
134 #include <linux/uaccess.h>
136 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
137 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
138 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
139 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
140 #define IS_REMOTELCK(fl) (fl->fl_pid <= 0)
142 static inline bool is_remote_lock(struct file *filp)
144 return likely(!(filp->f_path.dentry->d_sb->s_flags & SB_NOREMOTELOCK));
147 static bool lease_breaking(struct file_lock *fl)
149 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
152 static int target_leasetype(struct file_lock *fl)
154 if (fl->fl_flags & FL_UNLOCK_PENDING)
156 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
161 int leases_enable = 1;
162 int lease_break_time = 45;
165 * The global file_lock_list is only used for displaying /proc/locks, so we
166 * keep a list on each CPU, with each list protected by its own spinlock.
167 * Global serialization is done using file_rwsem.
169 * Note that alterations to the list also require that the relevant flc_lock is
172 struct file_lock_list_struct {
174 struct hlist_head hlist;
176 static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
177 DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
180 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
181 * It is protected by blocked_lock_lock.
183 * We hash locks by lockowner in order to optimize searching for the lock a
184 * particular lockowner is waiting on.
186 * FIXME: make this value scale via some heuristic? We generally will want more
187 * buckets when we have more lockowners holding locks, but that's a little
188 * difficult to determine without knowing what the workload will look like.
190 #define BLOCKED_HASH_BITS 7
191 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
194 * This lock protects the blocked_hash. Generally, if you're accessing it, you
195 * want to be holding this lock.
197 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
198 * pointer for file_lock structures that are acting as lock requests (in
199 * contrast to those that are acting as records of acquired locks).
201 * Note that when we acquire this lock in order to change the above fields,
202 * we often hold the flc_lock as well. In certain cases, when reading the fields
203 * protected by this lock, we can skip acquiring it iff we already hold the
206 * In particular, adding an entry to the fl_block list requires that you hold
207 * both the flc_lock and the blocked_lock_lock (acquired in that order).
208 * Deleting an entry from the list however only requires the file_lock_lock.
210 static DEFINE_SPINLOCK(blocked_lock_lock);
212 static struct kmem_cache *flctx_cache __read_mostly;
213 static struct kmem_cache *filelock_cache __read_mostly;
215 static struct file_lock_context *
216 locks_get_lock_context(struct inode *inode, int type)
218 struct file_lock_context *ctx;
220 /* paired with cmpxchg() below */
221 ctx = smp_load_acquire(&inode->i_flctx);
222 if (likely(ctx) || type == F_UNLCK)
225 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
229 spin_lock_init(&ctx->flc_lock);
230 INIT_LIST_HEAD(&ctx->flc_flock);
231 INIT_LIST_HEAD(&ctx->flc_posix);
232 INIT_LIST_HEAD(&ctx->flc_lease);
235 * Assign the pointer if it's not already assigned. If it is, then
236 * free the context we just allocated.
238 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
239 kmem_cache_free(flctx_cache, ctx);
240 ctx = smp_load_acquire(&inode->i_flctx);
243 trace_locks_get_lock_context(inode, type, ctx);
248 locks_dump_ctx_list(struct list_head *list, char *list_type)
250 struct file_lock *fl;
252 list_for_each_entry(fl, list, fl_list) {
253 pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type, fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
258 locks_check_ctx_lists(struct inode *inode)
260 struct file_lock_context *ctx = inode->i_flctx;
262 if (unlikely(!list_empty(&ctx->flc_flock) ||
263 !list_empty(&ctx->flc_posix) ||
264 !list_empty(&ctx->flc_lease))) {
265 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
266 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
268 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
269 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
270 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
275 locks_check_ctx_file_list(struct file *filp, struct list_head *list,
278 struct file_lock *fl;
279 struct inode *inode = locks_inode(filp);
281 list_for_each_entry(fl, list, fl_list)
282 if (fl->fl_file == filp)
283 pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx "
284 " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n",
285 list_type, MAJOR(inode->i_sb->s_dev),
286 MINOR(inode->i_sb->s_dev), inode->i_ino,
287 fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
291 locks_free_lock_context(struct inode *inode)
293 struct file_lock_context *ctx = inode->i_flctx;
296 locks_check_ctx_lists(inode);
297 kmem_cache_free(flctx_cache, ctx);
301 static void locks_init_lock_heads(struct file_lock *fl)
303 INIT_HLIST_NODE(&fl->fl_link);
304 INIT_LIST_HEAD(&fl->fl_list);
305 INIT_LIST_HEAD(&fl->fl_block);
306 init_waitqueue_head(&fl->fl_wait);
309 /* Allocate an empty lock structure. */
310 struct file_lock *locks_alloc_lock(void)
312 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
315 locks_init_lock_heads(fl);
319 EXPORT_SYMBOL_GPL(locks_alloc_lock);
321 void locks_release_private(struct file_lock *fl)
324 if (fl->fl_ops->fl_release_private)
325 fl->fl_ops->fl_release_private(fl);
330 if (fl->fl_lmops->lm_put_owner) {
331 fl->fl_lmops->lm_put_owner(fl->fl_owner);
337 EXPORT_SYMBOL_GPL(locks_release_private);
339 /* Free a lock which is not in use. */
340 void locks_free_lock(struct file_lock *fl)
342 BUG_ON(waitqueue_active(&fl->fl_wait));
343 BUG_ON(!list_empty(&fl->fl_list));
344 BUG_ON(!list_empty(&fl->fl_block));
345 BUG_ON(!hlist_unhashed(&fl->fl_link));
347 locks_release_private(fl);
348 kmem_cache_free(filelock_cache, fl);
350 EXPORT_SYMBOL(locks_free_lock);
353 locks_dispose_list(struct list_head *dispose)
355 struct file_lock *fl;
357 while (!list_empty(dispose)) {
358 fl = list_first_entry(dispose, struct file_lock, fl_list);
359 list_del_init(&fl->fl_list);
364 void locks_init_lock(struct file_lock *fl)
366 memset(fl, 0, sizeof(struct file_lock));
367 locks_init_lock_heads(fl);
370 EXPORT_SYMBOL(locks_init_lock);
373 * Initialize a new lock from an existing file_lock structure.
375 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
377 new->fl_owner = fl->fl_owner;
378 new->fl_pid = fl->fl_pid;
380 new->fl_flags = fl->fl_flags;
381 new->fl_type = fl->fl_type;
382 new->fl_start = fl->fl_start;
383 new->fl_end = fl->fl_end;
384 new->fl_lmops = fl->fl_lmops;
388 if (fl->fl_lmops->lm_get_owner)
389 fl->fl_lmops->lm_get_owner(fl->fl_owner);
392 EXPORT_SYMBOL(locks_copy_conflock);
394 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
396 /* "new" must be a freshly-initialized lock */
397 WARN_ON_ONCE(new->fl_ops);
399 locks_copy_conflock(new, fl);
401 new->fl_file = fl->fl_file;
402 new->fl_ops = fl->fl_ops;
405 if (fl->fl_ops->fl_copy_lock)
406 fl->fl_ops->fl_copy_lock(new, fl);
410 EXPORT_SYMBOL(locks_copy_lock);
412 static inline int flock_translate_cmd(int cmd) {
414 return cmd & (LOCK_MAND | LOCK_RW);
426 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
427 static struct file_lock *
428 flock_make_lock(struct file *filp, unsigned int cmd)
430 struct file_lock *fl;
431 int type = flock_translate_cmd(cmd);
434 return ERR_PTR(type);
436 fl = locks_alloc_lock();
438 return ERR_PTR(-ENOMEM);
442 fl->fl_pid = current->tgid;
443 fl->fl_flags = FL_FLOCK;
445 fl->fl_end = OFFSET_MAX;
450 static int assign_type(struct file_lock *fl, long type)
464 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
467 switch (l->l_whence) {
472 fl->fl_start = filp->f_pos;
475 fl->fl_start = i_size_read(file_inode(filp));
480 if (l->l_start > OFFSET_MAX - fl->fl_start)
482 fl->fl_start += l->l_start;
483 if (fl->fl_start < 0)
486 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
487 POSIX-2001 defines it. */
489 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
491 fl->fl_end = fl->fl_start + l->l_len - 1;
493 } else if (l->l_len < 0) {
494 if (fl->fl_start + l->l_len < 0)
496 fl->fl_end = fl->fl_start - 1;
497 fl->fl_start += l->l_len;
499 fl->fl_end = OFFSET_MAX;
501 fl->fl_owner = current->files;
502 fl->fl_pid = current->tgid;
504 fl->fl_flags = FL_POSIX;
508 return assign_type(fl, l->l_type);
511 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
514 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
517 struct flock64 ll = {
519 .l_whence = l->l_whence,
520 .l_start = l->l_start,
524 return flock64_to_posix_lock(filp, fl, &ll);
527 /* default lease lock manager operations */
529 lease_break_callback(struct file_lock *fl)
531 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
536 lease_setup(struct file_lock *fl, void **priv)
538 struct file *filp = fl->fl_file;
539 struct fasync_struct *fa = *priv;
542 * fasync_insert_entry() returns the old entry if any. If there was no
543 * old entry, then it used "priv" and inserted it into the fasync list.
544 * Clear the pointer to indicate that it shouldn't be freed.
546 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
549 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
552 static const struct lock_manager_operations lease_manager_ops = {
553 .lm_break = lease_break_callback,
554 .lm_change = lease_modify,
555 .lm_setup = lease_setup,
559 * Initialize a lease, use the default lock manager operations
561 static int lease_init(struct file *filp, long type, struct file_lock *fl)
563 if (assign_type(fl, type) != 0)
567 fl->fl_pid = current->tgid;
570 fl->fl_flags = FL_LEASE;
572 fl->fl_end = OFFSET_MAX;
574 fl->fl_lmops = &lease_manager_ops;
578 /* Allocate a file_lock initialised to this type of lease */
579 static struct file_lock *lease_alloc(struct file *filp, long type)
581 struct file_lock *fl = locks_alloc_lock();
585 return ERR_PTR(error);
587 error = lease_init(filp, type, fl);
590 return ERR_PTR(error);
595 /* Check if two locks overlap each other.
597 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
599 return ((fl1->fl_end >= fl2->fl_start) &&
600 (fl2->fl_end >= fl1->fl_start));
604 * Check whether two locks have the same owner.
606 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
608 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
609 return fl2->fl_lmops == fl1->fl_lmops &&
610 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
611 return fl1->fl_owner == fl2->fl_owner;
614 /* Must be called with the flc_lock held! */
615 static void locks_insert_global_locks(struct file_lock *fl)
617 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
619 percpu_rwsem_assert_held(&file_rwsem);
621 spin_lock(&fll->lock);
622 fl->fl_link_cpu = smp_processor_id();
623 hlist_add_head(&fl->fl_link, &fll->hlist);
624 spin_unlock(&fll->lock);
627 /* Must be called with the flc_lock held! */
628 static void locks_delete_global_locks(struct file_lock *fl)
630 struct file_lock_list_struct *fll;
632 percpu_rwsem_assert_held(&file_rwsem);
635 * Avoid taking lock if already unhashed. This is safe since this check
636 * is done while holding the flc_lock, and new insertions into the list
637 * also require that it be held.
639 if (hlist_unhashed(&fl->fl_link))
642 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
643 spin_lock(&fll->lock);
644 hlist_del_init(&fl->fl_link);
645 spin_unlock(&fll->lock);
649 posix_owner_key(struct file_lock *fl)
651 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
652 return fl->fl_lmops->lm_owner_key(fl);
653 return (unsigned long)fl->fl_owner;
656 static void locks_insert_global_blocked(struct file_lock *waiter)
658 lockdep_assert_held(&blocked_lock_lock);
660 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
663 static void locks_delete_global_blocked(struct file_lock *waiter)
665 lockdep_assert_held(&blocked_lock_lock);
667 hash_del(&waiter->fl_link);
670 /* Remove waiter from blocker's block list.
671 * When blocker ends up pointing to itself then the list is empty.
673 * Must be called with blocked_lock_lock held.
675 static void __locks_delete_block(struct file_lock *waiter)
677 locks_delete_global_blocked(waiter);
678 list_del_init(&waiter->fl_block);
679 waiter->fl_next = NULL;
682 static void locks_delete_block(struct file_lock *waiter)
684 spin_lock(&blocked_lock_lock);
685 __locks_delete_block(waiter);
686 spin_unlock(&blocked_lock_lock);
689 /* Insert waiter into blocker's block list.
690 * We use a circular list so that processes can be easily woken up in
691 * the order they blocked. The documentation doesn't require this but
692 * it seems like the reasonable thing to do.
694 * Must be called with both the flc_lock and blocked_lock_lock held. The
695 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
696 * that the flc_lock is also held on insertions we can avoid taking the
697 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
699 static void __locks_insert_block(struct file_lock *blocker,
700 struct file_lock *waiter)
702 BUG_ON(!list_empty(&waiter->fl_block));
703 waiter->fl_next = blocker;
704 list_add_tail(&waiter->fl_block, &blocker->fl_block);
705 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
706 locks_insert_global_blocked(waiter);
709 /* Must be called with flc_lock held. */
710 static void locks_insert_block(struct file_lock *blocker,
711 struct file_lock *waiter)
713 spin_lock(&blocked_lock_lock);
714 __locks_insert_block(blocker, waiter);
715 spin_unlock(&blocked_lock_lock);
719 * Wake up processes blocked waiting for blocker.
721 * Must be called with the inode->flc_lock held!
723 static void locks_wake_up_blocks(struct file_lock *blocker)
726 * Avoid taking global lock if list is empty. This is safe since new
727 * blocked requests are only added to the list under the flc_lock, and
728 * the flc_lock is always held here. Note that removal from the fl_block
729 * list does not require the flc_lock, so we must recheck list_empty()
730 * after acquiring the blocked_lock_lock.
732 if (list_empty(&blocker->fl_block))
735 spin_lock(&blocked_lock_lock);
736 while (!list_empty(&blocker->fl_block)) {
737 struct file_lock *waiter;
739 waiter = list_first_entry(&blocker->fl_block,
740 struct file_lock, fl_block);
741 __locks_delete_block(waiter);
742 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
743 waiter->fl_lmops->lm_notify(waiter);
745 wake_up(&waiter->fl_wait);
747 spin_unlock(&blocked_lock_lock);
751 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
753 list_add_tail(&fl->fl_list, before);
754 locks_insert_global_locks(fl);
758 locks_unlink_lock_ctx(struct file_lock *fl)
760 locks_delete_global_locks(fl);
761 list_del_init(&fl->fl_list);
762 locks_wake_up_blocks(fl);
766 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
768 locks_unlink_lock_ctx(fl);
770 list_add(&fl->fl_list, dispose);
775 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
776 * checks for shared/exclusive status of overlapping locks.
778 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
780 if (sys_fl->fl_type == F_WRLCK)
782 if (caller_fl->fl_type == F_WRLCK)
787 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
788 * checking before calling the locks_conflict().
790 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
792 /* POSIX locks owned by the same process do not conflict with
795 if (posix_same_owner(caller_fl, sys_fl))
798 /* Check whether they overlap */
799 if (!locks_overlap(caller_fl, sys_fl))
802 return (locks_conflict(caller_fl, sys_fl));
805 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
806 * checking before calling the locks_conflict().
808 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
810 /* FLOCK locks referring to the same filp do not conflict with
813 if (caller_fl->fl_file == sys_fl->fl_file)
815 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
818 return (locks_conflict(caller_fl, sys_fl));
822 posix_test_lock(struct file *filp, struct file_lock *fl)
824 struct file_lock *cfl;
825 struct file_lock_context *ctx;
826 struct inode *inode = locks_inode(filp);
828 ctx = smp_load_acquire(&inode->i_flctx);
829 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
830 fl->fl_type = F_UNLCK;
834 spin_lock(&ctx->flc_lock);
835 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
836 if (posix_locks_conflict(fl, cfl)) {
837 locks_copy_conflock(fl, cfl);
841 fl->fl_type = F_UNLCK;
843 spin_unlock(&ctx->flc_lock);
846 EXPORT_SYMBOL(posix_test_lock);
849 * Deadlock detection:
851 * We attempt to detect deadlocks that are due purely to posix file
854 * We assume that a task can be waiting for at most one lock at a time.
855 * So for any acquired lock, the process holding that lock may be
856 * waiting on at most one other lock. That lock in turns may be held by
857 * someone waiting for at most one other lock. Given a requested lock
858 * caller_fl which is about to wait for a conflicting lock block_fl, we
859 * follow this chain of waiters to ensure we are not about to create a
862 * Since we do this before we ever put a process to sleep on a lock, we
863 * are ensured that there is never a cycle; that is what guarantees that
864 * the while() loop in posix_locks_deadlock() eventually completes.
866 * Note: the above assumption may not be true when handling lock
867 * requests from a broken NFS client. It may also fail in the presence
868 * of tasks (such as posix threads) sharing the same open file table.
869 * To handle those cases, we just bail out after a few iterations.
871 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
872 * Because the owner is not even nominally tied to a thread of
873 * execution, the deadlock detection below can't reasonably work well. Just
876 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
877 * locks that just checks for the case where two tasks are attempting to
878 * upgrade from read to write locks on the same inode.
881 #define MAX_DEADLK_ITERATIONS 10
883 /* Find a lock that the owner of the given block_fl is blocking on. */
884 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
886 struct file_lock *fl;
888 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
889 if (posix_same_owner(fl, block_fl))
895 /* Must be called with the blocked_lock_lock held! */
896 static int posix_locks_deadlock(struct file_lock *caller_fl,
897 struct file_lock *block_fl)
901 lockdep_assert_held(&blocked_lock_lock);
904 * This deadlock detector can't reasonably detect deadlocks with
905 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
907 if (IS_OFDLCK(caller_fl))
910 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
911 if (i++ > MAX_DEADLK_ITERATIONS)
913 if (posix_same_owner(caller_fl, block_fl))
919 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
920 * after any leases, but before any posix locks.
922 * Note that if called with an FL_EXISTS argument, the caller may determine
923 * whether or not a lock was successfully freed by testing the return
926 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
928 struct file_lock *new_fl = NULL;
929 struct file_lock *fl;
930 struct file_lock_context *ctx;
935 ctx = locks_get_lock_context(inode, request->fl_type);
937 if (request->fl_type != F_UNLCK)
939 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
942 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
943 new_fl = locks_alloc_lock();
948 percpu_down_read_preempt_disable(&file_rwsem);
949 spin_lock(&ctx->flc_lock);
950 if (request->fl_flags & FL_ACCESS)
953 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
954 if (request->fl_file != fl->fl_file)
956 if (request->fl_type == fl->fl_type)
959 locks_delete_lock_ctx(fl, &dispose);
963 if (request->fl_type == F_UNLCK) {
964 if ((request->fl_flags & FL_EXISTS) && !found)
970 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
971 if (!flock_locks_conflict(request, fl))
974 if (!(request->fl_flags & FL_SLEEP))
976 error = FILE_LOCK_DEFERRED;
977 locks_insert_block(fl, request);
980 if (request->fl_flags & FL_ACCESS)
982 locks_copy_lock(new_fl, request);
983 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
988 spin_unlock(&ctx->flc_lock);
989 percpu_up_read_preempt_enable(&file_rwsem);
991 locks_free_lock(new_fl);
992 locks_dispose_list(&dispose);
993 trace_flock_lock_inode(inode, request, error);
997 static int posix_lock_inode(struct inode *inode, struct file_lock *request,
998 struct file_lock *conflock)
1000 struct file_lock *fl, *tmp;
1001 struct file_lock *new_fl = NULL;
1002 struct file_lock *new_fl2 = NULL;
1003 struct file_lock *left = NULL;
1004 struct file_lock *right = NULL;
1005 struct file_lock_context *ctx;
1010 ctx = locks_get_lock_context(inode, request->fl_type);
1012 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
1015 * We may need two file_lock structures for this operation,
1016 * so we get them in advance to avoid races.
1018 * In some cases we can be sure, that no new locks will be needed
1020 if (!(request->fl_flags & FL_ACCESS) &&
1021 (request->fl_type != F_UNLCK ||
1022 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1023 new_fl = locks_alloc_lock();
1024 new_fl2 = locks_alloc_lock();
1027 percpu_down_read_preempt_disable(&file_rwsem);
1028 spin_lock(&ctx->flc_lock);
1030 * New lock request. Walk all POSIX locks and look for conflicts. If
1031 * there are any, either return error or put the request on the
1032 * blocker's list of waiters and the global blocked_hash.
1034 if (request->fl_type != F_UNLCK) {
1035 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1036 if (!posix_locks_conflict(request, fl))
1039 locks_copy_conflock(conflock, fl);
1041 if (!(request->fl_flags & FL_SLEEP))
1044 * Deadlock detection and insertion into the blocked
1045 * locks list must be done while holding the same lock!
1048 spin_lock(&blocked_lock_lock);
1049 if (likely(!posix_locks_deadlock(request, fl))) {
1050 error = FILE_LOCK_DEFERRED;
1051 __locks_insert_block(fl, request);
1053 spin_unlock(&blocked_lock_lock);
1058 /* If we're just looking for a conflict, we're done. */
1060 if (request->fl_flags & FL_ACCESS)
1063 /* Find the first old lock with the same owner as the new lock */
1064 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1065 if (posix_same_owner(request, fl))
1069 /* Process locks with this owner. */
1070 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1071 if (!posix_same_owner(request, fl))
1074 /* Detect adjacent or overlapping regions (if same lock type) */
1075 if (request->fl_type == fl->fl_type) {
1076 /* In all comparisons of start vs end, use
1077 * "start - 1" rather than "end + 1". If end
1078 * is OFFSET_MAX, end + 1 will become negative.
1080 if (fl->fl_end < request->fl_start - 1)
1082 /* If the next lock in the list has entirely bigger
1083 * addresses than the new one, insert the lock here.
1085 if (fl->fl_start - 1 > request->fl_end)
1088 /* If we come here, the new and old lock are of the
1089 * same type and adjacent or overlapping. Make one
1090 * lock yielding from the lower start address of both
1091 * locks to the higher end address.
1093 if (fl->fl_start > request->fl_start)
1094 fl->fl_start = request->fl_start;
1096 request->fl_start = fl->fl_start;
1097 if (fl->fl_end < request->fl_end)
1098 fl->fl_end = request->fl_end;
1100 request->fl_end = fl->fl_end;
1102 locks_delete_lock_ctx(fl, &dispose);
1108 /* Processing for different lock types is a bit
1111 if (fl->fl_end < request->fl_start)
1113 if (fl->fl_start > request->fl_end)
1115 if (request->fl_type == F_UNLCK)
1117 if (fl->fl_start < request->fl_start)
1119 /* If the next lock in the list has a higher end
1120 * address than the new one, insert the new one here.
1122 if (fl->fl_end > request->fl_end) {
1126 if (fl->fl_start >= request->fl_start) {
1127 /* The new lock completely replaces an old
1128 * one (This may happen several times).
1131 locks_delete_lock_ctx(fl, &dispose);
1135 * Replace the old lock with new_fl, and
1136 * remove the old one. It's safe to do the
1137 * insert here since we know that we won't be
1138 * using new_fl later, and that the lock is
1139 * just replacing an existing lock.
1144 locks_copy_lock(new_fl, request);
1147 locks_insert_lock_ctx(request, &fl->fl_list);
1148 locks_delete_lock_ctx(fl, &dispose);
1155 * The above code only modifies existing locks in case of merging or
1156 * replacing. If new lock(s) need to be inserted all modifications are
1157 * done below this, so it's safe yet to bail out.
1159 error = -ENOLCK; /* "no luck" */
1160 if (right && left == right && !new_fl2)
1165 if (request->fl_type == F_UNLCK) {
1166 if (request->fl_flags & FL_EXISTS)
1175 locks_copy_lock(new_fl, request);
1176 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1181 if (left == right) {
1182 /* The new lock breaks the old one in two pieces,
1183 * so we have to use the second new lock.
1187 locks_copy_lock(left, right);
1188 locks_insert_lock_ctx(left, &fl->fl_list);
1190 right->fl_start = request->fl_end + 1;
1191 locks_wake_up_blocks(right);
1194 left->fl_end = request->fl_start - 1;
1195 locks_wake_up_blocks(left);
1198 spin_unlock(&ctx->flc_lock);
1199 percpu_up_read_preempt_enable(&file_rwsem);
1201 * Free any unused locks.
1204 locks_free_lock(new_fl);
1206 locks_free_lock(new_fl2);
1207 locks_dispose_list(&dispose);
1208 trace_posix_lock_inode(inode, request, error);
1214 * posix_lock_file - Apply a POSIX-style lock to a file
1215 * @filp: The file to apply the lock to
1216 * @fl: The lock to be applied
1217 * @conflock: Place to return a copy of the conflicting lock, if found.
1219 * Add a POSIX style lock to a file.
1220 * We merge adjacent & overlapping locks whenever possible.
1221 * POSIX locks are sorted by owner task, then by starting address
1223 * Note that if called with an FL_EXISTS argument, the caller may determine
1224 * whether or not a lock was successfully freed by testing the return
1225 * value for -ENOENT.
1227 int posix_lock_file(struct file *filp, struct file_lock *fl,
1228 struct file_lock *conflock)
1230 return posix_lock_inode(locks_inode(filp), fl, conflock);
1232 EXPORT_SYMBOL(posix_lock_file);
1235 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1236 * @inode: inode of file to which lock request should be applied
1237 * @fl: The lock to be applied
1239 * Apply a POSIX style lock request to an inode.
1241 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1246 error = posix_lock_inode(inode, fl, NULL);
1247 if (error != FILE_LOCK_DEFERRED)
1249 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1253 locks_delete_block(fl);
1259 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1261 * locks_mandatory_locked - Check for an active lock
1262 * @file: the file to check
1264 * Searches the inode's list of locks to find any POSIX locks which conflict.
1265 * This function is called from locks_verify_locked() only.
1267 int locks_mandatory_locked(struct file *file)
1270 struct inode *inode = locks_inode(file);
1271 struct file_lock_context *ctx;
1272 struct file_lock *fl;
1274 ctx = smp_load_acquire(&inode->i_flctx);
1275 if (!ctx || list_empty_careful(&ctx->flc_posix))
1279 * Search the lock list for this inode for any POSIX locks.
1281 spin_lock(&ctx->flc_lock);
1283 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1284 if (fl->fl_owner != current->files &&
1285 fl->fl_owner != file) {
1290 spin_unlock(&ctx->flc_lock);
1295 * locks_mandatory_area - Check for a conflicting lock
1296 * @inode: the file to check
1297 * @filp: how the file was opened (if it was)
1298 * @start: first byte in the file to check
1299 * @end: lastbyte in the file to check
1300 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1302 * Searches the inode's list of locks to find any POSIX locks which conflict.
1304 int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start,
1305 loff_t end, unsigned char type)
1307 struct file_lock fl;
1311 locks_init_lock(&fl);
1312 fl.fl_pid = current->tgid;
1314 fl.fl_flags = FL_POSIX | FL_ACCESS;
1315 if (filp && !(filp->f_flags & O_NONBLOCK))
1318 fl.fl_start = start;
1324 fl.fl_flags &= ~FL_SLEEP;
1325 error = posix_lock_inode(inode, &fl, NULL);
1331 fl.fl_flags |= FL_SLEEP;
1332 fl.fl_owner = current->files;
1333 error = posix_lock_inode(inode, &fl, NULL);
1334 if (error != FILE_LOCK_DEFERRED)
1336 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1339 * If we've been sleeping someone might have
1340 * changed the permissions behind our back.
1342 if (__mandatory_lock(inode))
1346 locks_delete_block(&fl);
1353 EXPORT_SYMBOL(locks_mandatory_area);
1354 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1356 static void lease_clear_pending(struct file_lock *fl, int arg)
1360 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1363 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1367 /* We already had a lease on this file; just change its type */
1368 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1370 int error = assign_type(fl, arg);
1374 lease_clear_pending(fl, arg);
1375 locks_wake_up_blocks(fl);
1376 if (arg == F_UNLCK) {
1377 struct file *filp = fl->fl_file;
1380 filp->f_owner.signum = 0;
1381 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1382 if (fl->fl_fasync != NULL) {
1383 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1384 fl->fl_fasync = NULL;
1386 locks_delete_lock_ctx(fl, dispose);
1390 EXPORT_SYMBOL(lease_modify);
1392 static bool past_time(unsigned long then)
1395 /* 0 is a special value meaning "this never expires": */
1397 return time_after(jiffies, then);
1400 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1402 struct file_lock_context *ctx = inode->i_flctx;
1403 struct file_lock *fl, *tmp;
1405 lockdep_assert_held(&ctx->flc_lock);
1407 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1408 trace_time_out_leases(inode, fl);
1409 if (past_time(fl->fl_downgrade_time))
1410 lease_modify(fl, F_RDLCK, dispose);
1411 if (past_time(fl->fl_break_time))
1412 lease_modify(fl, F_UNLCK, dispose);
1416 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1418 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1420 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1422 return locks_conflict(breaker, lease);
1426 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1428 struct file_lock_context *ctx = inode->i_flctx;
1429 struct file_lock *fl;
1431 lockdep_assert_held(&ctx->flc_lock);
1433 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1434 if (leases_conflict(fl, breaker))
1441 * __break_lease - revoke all outstanding leases on file
1442 * @inode: the inode of the file to return
1443 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1445 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1448 * break_lease (inlined for speed) has checked there already is at least
1449 * some kind of lock (maybe a lease) on this file. Leases are broken on
1450 * a call to open() or truncate(). This function can sleep unless you
1451 * specified %O_NONBLOCK to your open().
1453 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1456 struct file_lock_context *ctx;
1457 struct file_lock *new_fl, *fl, *tmp;
1458 unsigned long break_time;
1459 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1462 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1464 return PTR_ERR(new_fl);
1465 new_fl->fl_flags = type;
1467 /* typically we will check that ctx is non-NULL before calling */
1468 ctx = smp_load_acquire(&inode->i_flctx);
1474 percpu_down_read_preempt_disable(&file_rwsem);
1475 spin_lock(&ctx->flc_lock);
1477 time_out_leases(inode, &dispose);
1479 if (!any_leases_conflict(inode, new_fl))
1483 if (lease_break_time > 0) {
1484 break_time = jiffies + lease_break_time * HZ;
1485 if (break_time == 0)
1486 break_time++; /* so that 0 means no break time */
1489 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1490 if (!leases_conflict(fl, new_fl))
1493 if (fl->fl_flags & FL_UNLOCK_PENDING)
1495 fl->fl_flags |= FL_UNLOCK_PENDING;
1496 fl->fl_break_time = break_time;
1498 if (lease_breaking(fl))
1500 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1501 fl->fl_downgrade_time = break_time;
1503 if (fl->fl_lmops->lm_break(fl))
1504 locks_delete_lock_ctx(fl, &dispose);
1507 if (list_empty(&ctx->flc_lease))
1510 if (mode & O_NONBLOCK) {
1511 trace_break_lease_noblock(inode, new_fl);
1512 error = -EWOULDBLOCK;
1517 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1518 break_time = fl->fl_break_time;
1519 if (break_time != 0)
1520 break_time -= jiffies;
1521 if (break_time == 0)
1523 locks_insert_block(fl, new_fl);
1524 trace_break_lease_block(inode, new_fl);
1525 spin_unlock(&ctx->flc_lock);
1526 percpu_up_read_preempt_enable(&file_rwsem);
1528 locks_dispose_list(&dispose);
1529 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1530 !new_fl->fl_next, break_time);
1532 percpu_down_read_preempt_disable(&file_rwsem);
1533 spin_lock(&ctx->flc_lock);
1534 trace_break_lease_unblock(inode, new_fl);
1535 locks_delete_block(new_fl);
1538 * Wait for the next conflicting lease that has not been
1542 time_out_leases(inode, &dispose);
1543 if (any_leases_conflict(inode, new_fl))
1548 spin_unlock(&ctx->flc_lock);
1549 percpu_up_read_preempt_enable(&file_rwsem);
1550 locks_dispose_list(&dispose);
1551 locks_free_lock(new_fl);
1555 EXPORT_SYMBOL(__break_lease);
1558 * lease_get_mtime - update modified time of an inode with exclusive lease
1560 * @time: pointer to a timespec which contains the last modified time
1562 * This is to force NFS clients to flush their caches for files with
1563 * exclusive leases. The justification is that if someone has an
1564 * exclusive lease, then they could be modifying it.
1566 void lease_get_mtime(struct inode *inode, struct timespec *time)
1568 bool has_lease = false;
1569 struct file_lock_context *ctx;
1570 struct file_lock *fl;
1572 ctx = smp_load_acquire(&inode->i_flctx);
1573 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1574 spin_lock(&ctx->flc_lock);
1575 fl = list_first_entry_or_null(&ctx->flc_lease,
1576 struct file_lock, fl_list);
1577 if (fl && (fl->fl_type == F_WRLCK))
1579 spin_unlock(&ctx->flc_lock);
1583 *time = current_time(inode);
1586 EXPORT_SYMBOL(lease_get_mtime);
1589 * fcntl_getlease - Enquire what lease is currently active
1592 * The value returned by this function will be one of
1593 * (if no lease break is pending):
1595 * %F_RDLCK to indicate a shared lease is held.
1597 * %F_WRLCK to indicate an exclusive lease is held.
1599 * %F_UNLCK to indicate no lease is held.
1601 * (if a lease break is pending):
1603 * %F_RDLCK to indicate an exclusive lease needs to be
1604 * changed to a shared lease (or removed).
1606 * %F_UNLCK to indicate the lease needs to be removed.
1608 * XXX: sfr & willy disagree over whether F_INPROGRESS
1609 * should be returned to userspace.
1611 int fcntl_getlease(struct file *filp)
1613 struct file_lock *fl;
1614 struct inode *inode = locks_inode(filp);
1615 struct file_lock_context *ctx;
1619 ctx = smp_load_acquire(&inode->i_flctx);
1620 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1621 percpu_down_read_preempt_disable(&file_rwsem);
1622 spin_lock(&ctx->flc_lock);
1623 time_out_leases(inode, &dispose);
1624 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1625 if (fl->fl_file != filp)
1627 type = target_leasetype(fl);
1630 spin_unlock(&ctx->flc_lock);
1631 percpu_up_read_preempt_enable(&file_rwsem);
1633 locks_dispose_list(&dispose);
1639 * check_conflicting_open - see if the given dentry points to a file that has
1640 * an existing open that would conflict with the
1642 * @dentry: dentry to check
1643 * @arg: type of lease that we're trying to acquire
1644 * @flags: current lock flags
1646 * Check to see if there's an existing open fd on this file that would
1647 * conflict with the lease we're trying to set.
1650 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1653 struct inode *inode = dentry->d_inode;
1655 if (flags & FL_LAYOUT)
1658 if ((arg == F_RDLCK) &&
1659 (atomic_read(&d_real_inode(dentry)->i_writecount) > 0))
1662 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1663 (atomic_read(&inode->i_count) > 1)))
1670 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1672 struct file_lock *fl, *my_fl = NULL, *lease;
1673 struct dentry *dentry = filp->f_path.dentry;
1674 struct inode *inode = dentry->d_inode;
1675 struct file_lock_context *ctx;
1676 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1681 trace_generic_add_lease(inode, lease);
1683 /* Note that arg is never F_UNLCK here */
1684 ctx = locks_get_lock_context(inode, arg);
1689 * In the delegation case we need mutual exclusion with
1690 * a number of operations that take the i_mutex. We trylock
1691 * because delegations are an optional optimization, and if
1692 * there's some chance of a conflict--we'd rather not
1693 * bother, maybe that's a sign this just isn't a good file to
1694 * hand out a delegation on.
1696 if (is_deleg && !inode_trylock(inode))
1699 if (is_deleg && arg == F_WRLCK) {
1700 /* Write delegations are not currently supported: */
1701 inode_unlock(inode);
1706 percpu_down_read_preempt_disable(&file_rwsem);
1707 spin_lock(&ctx->flc_lock);
1708 time_out_leases(inode, &dispose);
1709 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1714 * At this point, we know that if there is an exclusive
1715 * lease on this file, then we hold it on this filp
1716 * (otherwise our open of this file would have blocked).
1717 * And if we are trying to acquire an exclusive lease,
1718 * then the file is not open by anyone (including us)
1719 * except for this filp.
1722 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1723 if (fl->fl_file == filp &&
1724 fl->fl_owner == lease->fl_owner) {
1730 * No exclusive leases if someone else has a lease on
1736 * Modifying our existing lease is OK, but no getting a
1737 * new lease if someone else is opening for write:
1739 if (fl->fl_flags & FL_UNLOCK_PENDING)
1743 if (my_fl != NULL) {
1745 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1755 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1757 * The check in break_lease() is lockless. It's possible for another
1758 * open to race in after we did the earlier check for a conflicting
1759 * open but before the lease was inserted. Check again for a
1760 * conflicting open and cancel the lease if there is one.
1762 * We also add a barrier here to ensure that the insertion of the lock
1763 * precedes these checks.
1766 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1768 locks_unlink_lock_ctx(lease);
1773 if (lease->fl_lmops->lm_setup)
1774 lease->fl_lmops->lm_setup(lease, priv);
1776 spin_unlock(&ctx->flc_lock);
1777 percpu_up_read_preempt_enable(&file_rwsem);
1778 locks_dispose_list(&dispose);
1780 inode_unlock(inode);
1781 if (!error && !my_fl)
1786 static int generic_delete_lease(struct file *filp, void *owner)
1788 int error = -EAGAIN;
1789 struct file_lock *fl, *victim = NULL;
1790 struct inode *inode = locks_inode(filp);
1791 struct file_lock_context *ctx;
1794 ctx = smp_load_acquire(&inode->i_flctx);
1796 trace_generic_delete_lease(inode, NULL);
1800 percpu_down_read_preempt_disable(&file_rwsem);
1801 spin_lock(&ctx->flc_lock);
1802 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1803 if (fl->fl_file == filp &&
1804 fl->fl_owner == owner) {
1809 trace_generic_delete_lease(inode, victim);
1811 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1812 spin_unlock(&ctx->flc_lock);
1813 percpu_up_read_preempt_enable(&file_rwsem);
1814 locks_dispose_list(&dispose);
1819 * generic_setlease - sets a lease on an open file
1820 * @filp: file pointer
1821 * @arg: type of lease to obtain
1822 * @flp: input - file_lock to use, output - file_lock inserted
1823 * @priv: private data for lm_setup (may be NULL if lm_setup
1824 * doesn't require it)
1826 * The (input) flp->fl_lmops->lm_break function is required
1829 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1832 struct inode *inode = locks_inode(filp);
1835 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1837 if (!S_ISREG(inode->i_mode))
1839 error = security_file_lock(filp, arg);
1845 return generic_delete_lease(filp, *priv);
1848 if (!(*flp)->fl_lmops->lm_break) {
1853 return generic_add_lease(filp, arg, flp, priv);
1858 EXPORT_SYMBOL(generic_setlease);
1861 * vfs_setlease - sets a lease on an open file
1862 * @filp: file pointer
1863 * @arg: type of lease to obtain
1864 * @lease: file_lock to use when adding a lease
1865 * @priv: private info for lm_setup when adding a lease (may be
1866 * NULL if lm_setup doesn't require it)
1868 * Call this to establish a lease on the file. The "lease" argument is not
1869 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1870 * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be
1871 * set; if not, this function will return -ENOLCK (and generate a scary-looking
1874 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1875 * may be NULL if the lm_setup operation doesn't require it.
1878 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1880 if (filp->f_op->setlease && is_remote_lock(filp))
1881 return filp->f_op->setlease(filp, arg, lease, priv);
1883 return generic_setlease(filp, arg, lease, priv);
1885 EXPORT_SYMBOL_GPL(vfs_setlease);
1887 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1889 struct file_lock *fl;
1890 struct fasync_struct *new;
1893 fl = lease_alloc(filp, arg);
1897 new = fasync_alloc();
1899 locks_free_lock(fl);
1904 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1906 locks_free_lock(fl);
1913 * fcntl_setlease - sets a lease on an open file
1914 * @fd: open file descriptor
1915 * @filp: file pointer
1916 * @arg: type of lease to obtain
1918 * Call this fcntl to establish a lease on the file.
1919 * Note that you also need to call %F_SETSIG to
1920 * receive a signal when the lease is broken.
1922 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1925 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1926 return do_fcntl_add_lease(fd, filp, arg);
1930 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1931 * @inode: inode of the file to apply to
1932 * @fl: The lock to be applied
1934 * Apply a FLOCK style lock request to an inode.
1936 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1941 error = flock_lock_inode(inode, fl);
1942 if (error != FILE_LOCK_DEFERRED)
1944 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1948 locks_delete_block(fl);
1955 * locks_lock_inode_wait - Apply a lock to an inode
1956 * @inode: inode of the file to apply to
1957 * @fl: The lock to be applied
1959 * Apply a POSIX or FLOCK style lock request to an inode.
1961 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1964 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1966 res = posix_lock_inode_wait(inode, fl);
1969 res = flock_lock_inode_wait(inode, fl);
1976 EXPORT_SYMBOL(locks_lock_inode_wait);
1979 * sys_flock: - flock() system call.
1980 * @fd: the file descriptor to lock.
1981 * @cmd: the type of lock to apply.
1983 * Apply a %FL_FLOCK style lock to an open file descriptor.
1984 * The @cmd can be one of:
1986 * - %LOCK_SH -- a shared lock.
1987 * - %LOCK_EX -- an exclusive lock.
1988 * - %LOCK_UN -- remove an existing lock.
1989 * - %LOCK_MAND -- a 'mandatory' flock.
1990 * This exists to emulate Windows Share Modes.
1992 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1993 * processes read and write access respectively.
1995 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1997 struct fd f = fdget(fd);
1998 struct file_lock *lock;
1999 int can_sleep, unlock;
2006 can_sleep = !(cmd & LOCK_NB);
2008 unlock = (cmd == LOCK_UN);
2010 if (!unlock && !(cmd & LOCK_MAND) &&
2011 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
2014 lock = flock_make_lock(f.file, cmd);
2016 error = PTR_ERR(lock);
2021 lock->fl_flags |= FL_SLEEP;
2023 error = security_file_lock(f.file, lock->fl_type);
2027 if (f.file->f_op->flock && is_remote_lock(f.file))
2028 error = f.file->f_op->flock(f.file,
2029 (can_sleep) ? F_SETLKW : F_SETLK,
2032 error = locks_lock_file_wait(f.file, lock);
2035 locks_free_lock(lock);
2044 * vfs_test_lock - test file byte range lock
2045 * @filp: The file to test lock for
2046 * @fl: The lock to test; also used to hold result
2048 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2049 * setting conf->fl_type to something other than F_UNLCK.
2051 int vfs_test_lock(struct file *filp, struct file_lock *fl)
2053 if (filp->f_op->lock && is_remote_lock(filp))
2054 return filp->f_op->lock(filp, F_GETLK, fl);
2055 posix_test_lock(filp, fl);
2058 EXPORT_SYMBOL_GPL(vfs_test_lock);
2061 * locks_translate_pid - translate a file_lock's fl_pid number into a namespace
2062 * @fl: The file_lock who's fl_pid should be translated
2063 * @ns: The namespace into which the pid should be translated
2065 * Used to tranlate a fl_pid into a namespace virtual pid number
2067 static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns)
2074 if (IS_REMOTELCK(fl))
2077 * If the flock owner process is dead and its pid has been already
2078 * freed, the translation below won't work, but we still want to show
2079 * flock owner pid number in init pidns.
2081 if (ns == &init_pid_ns)
2082 return (pid_t)fl->fl_pid;
2085 pid = find_pid_ns(fl->fl_pid, &init_pid_ns);
2086 vnr = pid_nr_ns(pid, ns);
2091 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2093 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2094 #if BITS_PER_LONG == 32
2096 * Make sure we can represent the posix lock via
2097 * legacy 32bit flock.
2099 if (fl->fl_start > OFFT_OFFSET_MAX)
2101 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2104 flock->l_start = fl->fl_start;
2105 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2106 fl->fl_end - fl->fl_start + 1;
2107 flock->l_whence = 0;
2108 flock->l_type = fl->fl_type;
2112 #if BITS_PER_LONG == 32
2113 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2115 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2116 flock->l_start = fl->fl_start;
2117 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2118 fl->fl_end - fl->fl_start + 1;
2119 flock->l_whence = 0;
2120 flock->l_type = fl->fl_type;
2124 /* Report the first existing lock that would conflict with l.
2125 * This implements the F_GETLK command of fcntl().
2127 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock)
2129 struct file_lock *fl;
2132 fl = locks_alloc_lock();
2136 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2139 error = flock_to_posix_lock(filp, fl, flock);
2143 if (cmd == F_OFD_GETLK) {
2145 if (flock->l_pid != 0)
2149 fl->fl_flags |= FL_OFDLCK;
2150 fl->fl_owner = filp;
2153 error = vfs_test_lock(filp, fl);
2157 flock->l_type = fl->fl_type;
2158 if (fl->fl_type != F_UNLCK) {
2159 error = posix_lock_to_flock(flock, fl);
2164 locks_free_lock(fl);
2169 * vfs_lock_file - file byte range lock
2170 * @filp: The file to apply the lock to
2171 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2172 * @fl: The lock to be applied
2173 * @conf: Place to return a copy of the conflicting lock, if found.
2175 * A caller that doesn't care about the conflicting lock may pass NULL
2176 * as the final argument.
2178 * If the filesystem defines a private ->lock() method, then @conf will
2179 * be left unchanged; so a caller that cares should initialize it to
2180 * some acceptable default.
2182 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2183 * locks, the ->lock() interface may return asynchronously, before the lock has
2184 * been granted or denied by the underlying filesystem, if (and only if)
2185 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2186 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2187 * the request is for a blocking lock. When ->lock() does return asynchronously,
2188 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2189 * request completes.
2190 * If the request is for non-blocking lock the file system should return
2191 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2192 * with the result. If the request timed out the callback routine will return a
2193 * nonzero return code and the file system should release the lock. The file
2194 * system is also responsible to keep a corresponding posix lock when it
2195 * grants a lock so the VFS can find out which locks are locally held and do
2196 * the correct lock cleanup when required.
2197 * The underlying filesystem must not drop the kernel lock or call
2198 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2201 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2203 if (filp->f_op->lock && is_remote_lock(filp))
2204 return filp->f_op->lock(filp, cmd, fl);
2206 return posix_lock_file(filp, fl, conf);
2208 EXPORT_SYMBOL_GPL(vfs_lock_file);
2210 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2211 struct file_lock *fl)
2215 error = security_file_lock(filp, fl->fl_type);
2220 error = vfs_lock_file(filp, cmd, fl, NULL);
2221 if (error != FILE_LOCK_DEFERRED)
2223 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2227 locks_delete_block(fl);
2234 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2236 check_fmode_for_setlk(struct file_lock *fl)
2238 switch (fl->fl_type) {
2240 if (!(fl->fl_file->f_mode & FMODE_READ))
2244 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2250 /* Apply the lock described by l to an open file descriptor.
2251 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2253 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2254 struct flock *flock)
2256 struct file_lock *file_lock = locks_alloc_lock();
2257 struct inode *inode = locks_inode(filp);
2261 if (file_lock == NULL)
2264 /* Don't allow mandatory locks on files that may be memory mapped
2267 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2272 error = flock_to_posix_lock(filp, file_lock, flock);
2276 error = check_fmode_for_setlk(file_lock);
2281 * If the cmd is requesting file-private locks, then set the
2282 * FL_OFDLCK flag and override the owner.
2287 if (flock->l_pid != 0)
2291 file_lock->fl_flags |= FL_OFDLCK;
2292 file_lock->fl_owner = filp;
2296 if (flock->l_pid != 0)
2300 file_lock->fl_flags |= FL_OFDLCK;
2301 file_lock->fl_owner = filp;
2304 file_lock->fl_flags |= FL_SLEEP;
2307 error = do_lock_file_wait(filp, cmd, file_lock);
2310 * Attempt to detect a close/fcntl race and recover by releasing the
2311 * lock that was just acquired. There is no need to do that when we're
2312 * unlocking though, or for OFD locks.
2314 if (!error && file_lock->fl_type != F_UNLCK &&
2315 !(file_lock->fl_flags & FL_OFDLCK)) {
2317 * We need that spin_lock here - it prevents reordering between
2318 * update of i_flctx->flc_posix and check for it done in
2319 * close(). rcu_read_lock() wouldn't do.
2321 spin_lock(¤t->files->file_lock);
2323 spin_unlock(¤t->files->file_lock);
2325 file_lock->fl_type = F_UNLCK;
2326 error = do_lock_file_wait(filp, cmd, file_lock);
2327 WARN_ON_ONCE(error);
2332 trace_fcntl_setlk(inode, file_lock, error);
2333 locks_free_lock(file_lock);
2337 #if BITS_PER_LONG == 32
2338 /* Report the first existing lock that would conflict with l.
2339 * This implements the F_GETLK command of fcntl().
2341 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock)
2343 struct file_lock *fl;
2346 fl = locks_alloc_lock();
2351 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2354 error = flock64_to_posix_lock(filp, fl, flock);
2358 if (cmd == F_OFD_GETLK) {
2360 if (flock->l_pid != 0)
2364 fl->fl_flags |= FL_OFDLCK;
2365 fl->fl_owner = filp;
2368 error = vfs_test_lock(filp, fl);
2372 flock->l_type = fl->fl_type;
2373 if (fl->fl_type != F_UNLCK)
2374 posix_lock_to_flock64(flock, fl);
2377 locks_free_lock(fl);
2381 /* Apply the lock described by l to an open file descriptor.
2382 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2384 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2385 struct flock64 *flock)
2387 struct file_lock *file_lock = locks_alloc_lock();
2388 struct inode *inode = locks_inode(filp);
2392 if (file_lock == NULL)
2395 /* Don't allow mandatory locks on files that may be memory mapped
2398 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2403 error = flock64_to_posix_lock(filp, file_lock, flock);
2407 error = check_fmode_for_setlk(file_lock);
2412 * If the cmd is requesting file-private locks, then set the
2413 * FL_OFDLCK flag and override the owner.
2418 if (flock->l_pid != 0)
2422 file_lock->fl_flags |= FL_OFDLCK;
2423 file_lock->fl_owner = filp;
2427 if (flock->l_pid != 0)
2431 file_lock->fl_flags |= FL_OFDLCK;
2432 file_lock->fl_owner = filp;
2435 file_lock->fl_flags |= FL_SLEEP;
2438 error = do_lock_file_wait(filp, cmd, file_lock);
2441 * Attempt to detect a close/fcntl race and recover by releasing the
2442 * lock that was just acquired. There is no need to do that when we're
2443 * unlocking though, or for OFD locks.
2445 if (!error && file_lock->fl_type != F_UNLCK &&
2446 !(file_lock->fl_flags & FL_OFDLCK)) {
2448 * We need that spin_lock here - it prevents reordering between
2449 * update of i_flctx->flc_posix and check for it done in
2450 * close(). rcu_read_lock() wouldn't do.
2452 spin_lock(¤t->files->file_lock);
2454 spin_unlock(¤t->files->file_lock);
2456 file_lock->fl_type = F_UNLCK;
2457 error = do_lock_file_wait(filp, cmd, file_lock);
2458 WARN_ON_ONCE(error);
2463 locks_free_lock(file_lock);
2466 #endif /* BITS_PER_LONG == 32 */
2469 * This function is called when the file is being removed
2470 * from the task's fd array. POSIX locks belonging to this task
2471 * are deleted at this time.
2473 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2476 struct inode *inode = locks_inode(filp);
2477 struct file_lock lock;
2478 struct file_lock_context *ctx;
2481 * If there are no locks held on this file, we don't need to call
2482 * posix_lock_file(). Another process could be setting a lock on this
2483 * file at the same time, but we wouldn't remove that lock anyway.
2485 ctx = smp_load_acquire(&inode->i_flctx);
2486 if (!ctx || list_empty(&ctx->flc_posix))
2489 lock.fl_type = F_UNLCK;
2490 lock.fl_flags = FL_POSIX | FL_CLOSE;
2492 lock.fl_end = OFFSET_MAX;
2493 lock.fl_owner = owner;
2494 lock.fl_pid = current->tgid;
2495 lock.fl_file = filp;
2497 lock.fl_lmops = NULL;
2499 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
2501 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2502 lock.fl_ops->fl_release_private(&lock);
2503 trace_locks_remove_posix(inode, &lock, error);
2506 EXPORT_SYMBOL(locks_remove_posix);
2508 /* The i_flctx must be valid when calling into here */
2510 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2512 struct file_lock fl = {
2514 .fl_pid = current->tgid,
2516 .fl_flags = FL_FLOCK | FL_CLOSE,
2518 .fl_end = OFFSET_MAX,
2520 struct inode *inode = locks_inode(filp);
2522 if (list_empty(&flctx->flc_flock))
2525 if (filp->f_op->flock && is_remote_lock(filp))
2526 filp->f_op->flock(filp, F_SETLKW, &fl);
2528 flock_lock_inode(inode, &fl);
2530 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2531 fl.fl_ops->fl_release_private(&fl);
2534 /* The i_flctx must be valid when calling into here */
2536 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2538 struct file_lock *fl, *tmp;
2541 if (list_empty(&ctx->flc_lease))
2544 percpu_down_read_preempt_disable(&file_rwsem);
2545 spin_lock(&ctx->flc_lock);
2546 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2547 if (filp == fl->fl_file)
2548 lease_modify(fl, F_UNLCK, &dispose);
2549 spin_unlock(&ctx->flc_lock);
2550 percpu_up_read_preempt_enable(&file_rwsem);
2552 locks_dispose_list(&dispose);
2556 * This function is called on the last close of an open file.
2558 void locks_remove_file(struct file *filp)
2560 struct file_lock_context *ctx;
2562 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
2566 /* remove any OFD locks */
2567 locks_remove_posix(filp, filp);
2569 /* remove flock locks */
2570 locks_remove_flock(filp, ctx);
2572 /* remove any leases */
2573 locks_remove_lease(filp, ctx);
2575 spin_lock(&ctx->flc_lock);
2576 locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX");
2577 locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK");
2578 locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE");
2579 spin_unlock(&ctx->flc_lock);
2583 * posix_unblock_lock - stop waiting for a file lock
2584 * @waiter: the lock which was waiting
2586 * lockd needs to block waiting for locks.
2589 posix_unblock_lock(struct file_lock *waiter)
2593 spin_lock(&blocked_lock_lock);
2594 if (waiter->fl_next)
2595 __locks_delete_block(waiter);
2598 spin_unlock(&blocked_lock_lock);
2601 EXPORT_SYMBOL(posix_unblock_lock);
2604 * vfs_cancel_lock - file byte range unblock lock
2605 * @filp: The file to apply the unblock to
2606 * @fl: The lock to be unblocked
2608 * Used by lock managers to cancel blocked requests
2610 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2612 if (filp->f_op->lock && is_remote_lock(filp))
2613 return filp->f_op->lock(filp, F_CANCELLK, fl);
2617 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2619 #ifdef CONFIG_PROC_FS
2620 #include <linux/proc_fs.h>
2621 #include <linux/seq_file.h>
2623 struct locks_iterator {
2628 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2629 loff_t id, char *pfx)
2631 struct inode *inode = NULL;
2632 unsigned int fl_pid;
2633 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2635 fl_pid = locks_translate_pid(fl, proc_pidns);
2637 * If lock owner is dead (and pid is freed) or not visible in current
2638 * pidns, zero is shown as a pid value. Check lock info from
2639 * init_pid_ns to get saved lock pid value.
2642 if (fl->fl_file != NULL)
2643 inode = locks_inode(fl->fl_file);
2645 seq_printf(f, "%lld:%s ", id, pfx);
2647 if (fl->fl_flags & FL_ACCESS)
2648 seq_puts(f, "ACCESS");
2649 else if (IS_OFDLCK(fl))
2650 seq_puts(f, "OFDLCK");
2652 seq_puts(f, "POSIX ");
2654 seq_printf(f, " %s ",
2655 (inode == NULL) ? "*NOINODE*" :
2656 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2657 } else if (IS_FLOCK(fl)) {
2658 if (fl->fl_type & LOCK_MAND) {
2659 seq_puts(f, "FLOCK MSNFS ");
2661 seq_puts(f, "FLOCK ADVISORY ");
2663 } else if (IS_LEASE(fl)) {
2664 if (fl->fl_flags & FL_DELEG)
2665 seq_puts(f, "DELEG ");
2667 seq_puts(f, "LEASE ");
2669 if (lease_breaking(fl))
2670 seq_puts(f, "BREAKING ");
2671 else if (fl->fl_file)
2672 seq_puts(f, "ACTIVE ");
2674 seq_puts(f, "BREAKER ");
2676 seq_puts(f, "UNKNOWN UNKNOWN ");
2678 if (fl->fl_type & LOCK_MAND) {
2679 seq_printf(f, "%s ",
2680 (fl->fl_type & LOCK_READ)
2681 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2682 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2684 seq_printf(f, "%s ",
2685 (lease_breaking(fl))
2686 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2687 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2690 /* userspace relies on this representation of dev_t */
2691 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2692 MAJOR(inode->i_sb->s_dev),
2693 MINOR(inode->i_sb->s_dev), inode->i_ino);
2695 seq_printf(f, "%d <none>:0 ", fl_pid);
2698 if (fl->fl_end == OFFSET_MAX)
2699 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2701 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2703 seq_puts(f, "0 EOF\n");
2707 static int locks_show(struct seq_file *f, void *v)
2709 struct locks_iterator *iter = f->private;
2710 struct file_lock *fl, *bfl;
2711 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2713 fl = hlist_entry(v, struct file_lock, fl_link);
2715 if (locks_translate_pid(fl, proc_pidns) == 0)
2718 lock_get_status(f, fl, iter->li_pos, "");
2720 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2721 lock_get_status(f, bfl, iter->li_pos, " ->");
2726 static void __show_fd_locks(struct seq_file *f,
2727 struct list_head *head, int *id,
2728 struct file *filp, struct files_struct *files)
2730 struct file_lock *fl;
2732 list_for_each_entry(fl, head, fl_list) {
2734 if (filp != fl->fl_file)
2736 if (fl->fl_owner != files &&
2737 fl->fl_owner != filp)
2741 seq_puts(f, "lock:\t");
2742 lock_get_status(f, fl, *id, "");
2746 void show_fd_locks(struct seq_file *f,
2747 struct file *filp, struct files_struct *files)
2749 struct inode *inode = locks_inode(filp);
2750 struct file_lock_context *ctx;
2753 ctx = smp_load_acquire(&inode->i_flctx);
2757 spin_lock(&ctx->flc_lock);
2758 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2759 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2760 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2761 spin_unlock(&ctx->flc_lock);
2764 static void *locks_start(struct seq_file *f, loff_t *pos)
2765 __acquires(&blocked_lock_lock)
2767 struct locks_iterator *iter = f->private;
2769 iter->li_pos = *pos + 1;
2770 percpu_down_write(&file_rwsem);
2771 spin_lock(&blocked_lock_lock);
2772 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
2775 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2777 struct locks_iterator *iter = f->private;
2780 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
2783 static void locks_stop(struct seq_file *f, void *v)
2784 __releases(&blocked_lock_lock)
2786 spin_unlock(&blocked_lock_lock);
2787 percpu_up_write(&file_rwsem);
2790 static const struct seq_operations locks_seq_operations = {
2791 .start = locks_start,
2797 static int __init proc_locks_init(void)
2799 proc_create_seq_private("locks", 0, NULL, &locks_seq_operations,
2800 sizeof(struct locks_iterator), NULL);
2803 fs_initcall(proc_locks_init);
2806 static int __init filelock_init(void)
2810 flctx_cache = kmem_cache_create("file_lock_ctx",
2811 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2813 filelock_cache = kmem_cache_create("file_lock_cache",
2814 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2817 for_each_possible_cpu(i) {
2818 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2820 spin_lock_init(&fll->lock);
2821 INIT_HLIST_HEAD(&fll->hlist);
2827 core_initcall(filelock_init);