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)
141 static inline bool is_remote_lock(struct file *filp)
143 return likely(!(filp->f_path.dentry->d_sb->s_flags & MS_NOREMOTELOCK));
146 static bool lease_breaking(struct file_lock *fl)
148 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
151 static int target_leasetype(struct file_lock *fl)
153 if (fl->fl_flags & FL_UNLOCK_PENDING)
155 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
160 int leases_enable = 1;
161 int lease_break_time = 45;
164 * The global file_lock_list is only used for displaying /proc/locks, so we
165 * keep a list on each CPU, with each list protected by its own spinlock.
166 * Global serialization is done using file_rwsem.
168 * Note that alterations to the list also require that the relevant flc_lock is
171 struct file_lock_list_struct {
173 struct hlist_head hlist;
175 static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
176 DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
179 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
180 * It is protected by blocked_lock_lock.
182 * We hash locks by lockowner in order to optimize searching for the lock a
183 * particular lockowner is waiting on.
185 * FIXME: make this value scale via some heuristic? We generally will want more
186 * buckets when we have more lockowners holding locks, but that's a little
187 * difficult to determine without knowing what the workload will look like.
189 #define BLOCKED_HASH_BITS 7
190 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
193 * This lock protects the blocked_hash. Generally, if you're accessing it, you
194 * want to be holding this lock.
196 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
197 * pointer for file_lock structures that are acting as lock requests (in
198 * contrast to those that are acting as records of acquired locks).
200 * Note that when we acquire this lock in order to change the above fields,
201 * we often hold the flc_lock as well. In certain cases, when reading the fields
202 * protected by this lock, we can skip acquiring it iff we already hold the
205 * In particular, adding an entry to the fl_block list requires that you hold
206 * both the flc_lock and the blocked_lock_lock (acquired in that order).
207 * Deleting an entry from the list however only requires the file_lock_lock.
209 static DEFINE_SPINLOCK(blocked_lock_lock);
211 static struct kmem_cache *flctx_cache __read_mostly;
212 static struct kmem_cache *filelock_cache __read_mostly;
214 static struct file_lock_context *
215 locks_get_lock_context(struct inode *inode, int type)
217 struct file_lock_context *ctx;
219 /* paired with cmpxchg() below */
220 ctx = smp_load_acquire(&inode->i_flctx);
221 if (likely(ctx) || type == F_UNLCK)
224 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
228 spin_lock_init(&ctx->flc_lock);
229 INIT_LIST_HEAD(&ctx->flc_flock);
230 INIT_LIST_HEAD(&ctx->flc_posix);
231 INIT_LIST_HEAD(&ctx->flc_lease);
234 * Assign the pointer if it's not already assigned. If it is, then
235 * free the context we just allocated.
237 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
238 kmem_cache_free(flctx_cache, ctx);
239 ctx = smp_load_acquire(&inode->i_flctx);
242 trace_locks_get_lock_context(inode, type, ctx);
247 locks_dump_ctx_list(struct list_head *list, char *list_type)
249 struct file_lock *fl;
251 list_for_each_entry(fl, list, fl_list) {
252 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);
257 locks_check_ctx_lists(struct inode *inode)
259 struct file_lock_context *ctx = inode->i_flctx;
261 if (unlikely(!list_empty(&ctx->flc_flock) ||
262 !list_empty(&ctx->flc_posix) ||
263 !list_empty(&ctx->flc_lease))) {
264 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
265 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
267 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
268 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
269 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
274 locks_free_lock_context(struct inode *inode)
276 struct file_lock_context *ctx = inode->i_flctx;
279 locks_check_ctx_lists(inode);
280 kmem_cache_free(flctx_cache, ctx);
284 static void locks_init_lock_heads(struct file_lock *fl)
286 INIT_HLIST_NODE(&fl->fl_link);
287 INIT_LIST_HEAD(&fl->fl_list);
288 INIT_LIST_HEAD(&fl->fl_block);
289 init_waitqueue_head(&fl->fl_wait);
292 /* Allocate an empty lock structure. */
293 struct file_lock *locks_alloc_lock(void)
295 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
298 locks_init_lock_heads(fl);
302 EXPORT_SYMBOL_GPL(locks_alloc_lock);
304 void locks_release_private(struct file_lock *fl)
307 if (fl->fl_ops->fl_release_private)
308 fl->fl_ops->fl_release_private(fl);
313 if (fl->fl_lmops->lm_put_owner) {
314 fl->fl_lmops->lm_put_owner(fl->fl_owner);
320 EXPORT_SYMBOL_GPL(locks_release_private);
322 /* Free a lock which is not in use. */
323 void locks_free_lock(struct file_lock *fl)
325 BUG_ON(waitqueue_active(&fl->fl_wait));
326 BUG_ON(!list_empty(&fl->fl_list));
327 BUG_ON(!list_empty(&fl->fl_block));
328 BUG_ON(!hlist_unhashed(&fl->fl_link));
330 locks_release_private(fl);
331 kmem_cache_free(filelock_cache, fl);
333 EXPORT_SYMBOL(locks_free_lock);
336 locks_dispose_list(struct list_head *dispose)
338 struct file_lock *fl;
340 while (!list_empty(dispose)) {
341 fl = list_first_entry(dispose, struct file_lock, fl_list);
342 list_del_init(&fl->fl_list);
347 void locks_init_lock(struct file_lock *fl)
349 memset(fl, 0, sizeof(struct file_lock));
350 locks_init_lock_heads(fl);
353 EXPORT_SYMBOL(locks_init_lock);
356 * Initialize a new lock from an existing file_lock structure.
358 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
360 new->fl_owner = fl->fl_owner;
361 new->fl_pid = fl->fl_pid;
363 new->fl_flags = fl->fl_flags;
364 new->fl_type = fl->fl_type;
365 new->fl_start = fl->fl_start;
366 new->fl_end = fl->fl_end;
367 new->fl_lmops = fl->fl_lmops;
371 if (fl->fl_lmops->lm_get_owner)
372 fl->fl_lmops->lm_get_owner(fl->fl_owner);
375 EXPORT_SYMBOL(locks_copy_conflock);
377 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
379 /* "new" must be a freshly-initialized lock */
380 WARN_ON_ONCE(new->fl_ops);
382 locks_copy_conflock(new, fl);
384 new->fl_file = fl->fl_file;
385 new->fl_ops = fl->fl_ops;
388 if (fl->fl_ops->fl_copy_lock)
389 fl->fl_ops->fl_copy_lock(new, fl);
393 EXPORT_SYMBOL(locks_copy_lock);
395 static inline int flock_translate_cmd(int cmd) {
397 return cmd & (LOCK_MAND | LOCK_RW);
409 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
410 static struct file_lock *
411 flock_make_lock(struct file *filp, unsigned int cmd)
413 struct file_lock *fl;
414 int type = flock_translate_cmd(cmd);
417 return ERR_PTR(type);
419 fl = locks_alloc_lock();
421 return ERR_PTR(-ENOMEM);
425 fl->fl_pid = current->tgid;
426 fl->fl_flags = FL_FLOCK;
428 fl->fl_end = OFFSET_MAX;
433 static int assign_type(struct file_lock *fl, long type)
447 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
450 switch (l->l_whence) {
455 fl->fl_start = filp->f_pos;
458 fl->fl_start = i_size_read(file_inode(filp));
463 if (l->l_start > OFFSET_MAX - fl->fl_start)
465 fl->fl_start += l->l_start;
466 if (fl->fl_start < 0)
469 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
470 POSIX-2001 defines it. */
472 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
474 fl->fl_end = fl->fl_start + l->l_len - 1;
476 } else if (l->l_len < 0) {
477 if (fl->fl_start + l->l_len < 0)
479 fl->fl_end = fl->fl_start - 1;
480 fl->fl_start += l->l_len;
482 fl->fl_end = OFFSET_MAX;
484 fl->fl_owner = current->files;
485 fl->fl_pid = current->tgid;
487 fl->fl_flags = FL_POSIX;
491 return assign_type(fl, l->l_type);
494 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
497 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
500 struct flock64 ll = {
502 .l_whence = l->l_whence,
503 .l_start = l->l_start,
507 return flock64_to_posix_lock(filp, fl, &ll);
510 /* default lease lock manager operations */
512 lease_break_callback(struct file_lock *fl)
514 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
519 lease_setup(struct file_lock *fl, void **priv)
521 struct file *filp = fl->fl_file;
522 struct fasync_struct *fa = *priv;
525 * fasync_insert_entry() returns the old entry if any. If there was no
526 * old entry, then it used "priv" and inserted it into the fasync list.
527 * Clear the pointer to indicate that it shouldn't be freed.
529 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
532 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
535 static const struct lock_manager_operations lease_manager_ops = {
536 .lm_break = lease_break_callback,
537 .lm_change = lease_modify,
538 .lm_setup = lease_setup,
542 * Initialize a lease, use the default lock manager operations
544 static int lease_init(struct file *filp, long type, struct file_lock *fl)
546 if (assign_type(fl, type) != 0)
550 fl->fl_pid = current->tgid;
553 fl->fl_flags = FL_LEASE;
555 fl->fl_end = OFFSET_MAX;
557 fl->fl_lmops = &lease_manager_ops;
561 /* Allocate a file_lock initialised to this type of lease */
562 static struct file_lock *lease_alloc(struct file *filp, long type)
564 struct file_lock *fl = locks_alloc_lock();
568 return ERR_PTR(error);
570 error = lease_init(filp, type, fl);
573 return ERR_PTR(error);
578 /* Check if two locks overlap each other.
580 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
582 return ((fl1->fl_end >= fl2->fl_start) &&
583 (fl2->fl_end >= fl1->fl_start));
587 * Check whether two locks have the same owner.
589 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
591 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
592 return fl2->fl_lmops == fl1->fl_lmops &&
593 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
594 return fl1->fl_owner == fl2->fl_owner;
597 /* Must be called with the flc_lock held! */
598 static void locks_insert_global_locks(struct file_lock *fl)
600 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
602 percpu_rwsem_assert_held(&file_rwsem);
604 spin_lock(&fll->lock);
605 fl->fl_link_cpu = smp_processor_id();
606 hlist_add_head(&fl->fl_link, &fll->hlist);
607 spin_unlock(&fll->lock);
610 /* Must be called with the flc_lock held! */
611 static void locks_delete_global_locks(struct file_lock *fl)
613 struct file_lock_list_struct *fll;
615 percpu_rwsem_assert_held(&file_rwsem);
618 * Avoid taking lock if already unhashed. This is safe since this check
619 * is done while holding the flc_lock, and new insertions into the list
620 * also require that it be held.
622 if (hlist_unhashed(&fl->fl_link))
625 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
626 spin_lock(&fll->lock);
627 hlist_del_init(&fl->fl_link);
628 spin_unlock(&fll->lock);
632 posix_owner_key(struct file_lock *fl)
634 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
635 return fl->fl_lmops->lm_owner_key(fl);
636 return (unsigned long)fl->fl_owner;
639 static void locks_insert_global_blocked(struct file_lock *waiter)
641 lockdep_assert_held(&blocked_lock_lock);
643 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
646 static void locks_delete_global_blocked(struct file_lock *waiter)
648 lockdep_assert_held(&blocked_lock_lock);
650 hash_del(&waiter->fl_link);
653 /* Remove waiter from blocker's block list.
654 * When blocker ends up pointing to itself then the list is empty.
656 * Must be called with blocked_lock_lock held.
658 static void __locks_delete_block(struct file_lock *waiter)
660 locks_delete_global_blocked(waiter);
661 list_del_init(&waiter->fl_block);
662 waiter->fl_next = NULL;
665 static void locks_delete_block(struct file_lock *waiter)
667 spin_lock(&blocked_lock_lock);
668 __locks_delete_block(waiter);
669 spin_unlock(&blocked_lock_lock);
672 /* Insert waiter into blocker's block list.
673 * We use a circular list so that processes can be easily woken up in
674 * the order they blocked. The documentation doesn't require this but
675 * it seems like the reasonable thing to do.
677 * Must be called with both the flc_lock and blocked_lock_lock held. The
678 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
679 * that the flc_lock is also held on insertions we can avoid taking the
680 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
682 static void __locks_insert_block(struct file_lock *blocker,
683 struct file_lock *waiter)
685 BUG_ON(!list_empty(&waiter->fl_block));
686 waiter->fl_next = blocker;
687 list_add_tail(&waiter->fl_block, &blocker->fl_block);
688 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
689 locks_insert_global_blocked(waiter);
692 /* Must be called with flc_lock held. */
693 static void locks_insert_block(struct file_lock *blocker,
694 struct file_lock *waiter)
696 spin_lock(&blocked_lock_lock);
697 __locks_insert_block(blocker, waiter);
698 spin_unlock(&blocked_lock_lock);
702 * Wake up processes blocked waiting for blocker.
704 * Must be called with the inode->flc_lock held!
706 static void locks_wake_up_blocks(struct file_lock *blocker)
709 * Avoid taking global lock if list is empty. This is safe since new
710 * blocked requests are only added to the list under the flc_lock, and
711 * the flc_lock is always held here. Note that removal from the fl_block
712 * list does not require the flc_lock, so we must recheck list_empty()
713 * after acquiring the blocked_lock_lock.
715 if (list_empty(&blocker->fl_block))
718 spin_lock(&blocked_lock_lock);
719 while (!list_empty(&blocker->fl_block)) {
720 struct file_lock *waiter;
722 waiter = list_first_entry(&blocker->fl_block,
723 struct file_lock, fl_block);
724 __locks_delete_block(waiter);
725 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
726 waiter->fl_lmops->lm_notify(waiter);
728 wake_up(&waiter->fl_wait);
730 spin_unlock(&blocked_lock_lock);
734 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
736 fl->fl_nspid = get_pid(task_tgid(current));
737 list_add_tail(&fl->fl_list, before);
738 locks_insert_global_locks(fl);
742 locks_unlink_lock_ctx(struct file_lock *fl)
744 locks_delete_global_locks(fl);
745 list_del_init(&fl->fl_list);
747 put_pid(fl->fl_nspid);
750 locks_wake_up_blocks(fl);
754 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
756 locks_unlink_lock_ctx(fl);
758 list_add(&fl->fl_list, dispose);
763 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
764 * checks for shared/exclusive status of overlapping locks.
766 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
768 if (sys_fl->fl_type == F_WRLCK)
770 if (caller_fl->fl_type == F_WRLCK)
775 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
776 * checking before calling the locks_conflict().
778 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
780 /* POSIX locks owned by the same process do not conflict with
783 if (posix_same_owner(caller_fl, sys_fl))
786 /* Check whether they overlap */
787 if (!locks_overlap(caller_fl, sys_fl))
790 return (locks_conflict(caller_fl, sys_fl));
793 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
794 * checking before calling the locks_conflict().
796 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
798 /* FLOCK locks referring to the same filp do not conflict with
801 if (caller_fl->fl_file == sys_fl->fl_file)
803 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
806 return (locks_conflict(caller_fl, sys_fl));
810 posix_test_lock(struct file *filp, struct file_lock *fl)
812 struct file_lock *cfl;
813 struct file_lock_context *ctx;
814 struct inode *inode = locks_inode(filp);
816 ctx = smp_load_acquire(&inode->i_flctx);
817 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
818 fl->fl_type = F_UNLCK;
822 spin_lock(&ctx->flc_lock);
823 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
824 if (posix_locks_conflict(fl, cfl)) {
825 locks_copy_conflock(fl, cfl);
827 fl->fl_pid = pid_vnr(cfl->fl_nspid);
831 fl->fl_type = F_UNLCK;
833 spin_unlock(&ctx->flc_lock);
836 EXPORT_SYMBOL(posix_test_lock);
839 * Deadlock detection:
841 * We attempt to detect deadlocks that are due purely to posix file
844 * We assume that a task can be waiting for at most one lock at a time.
845 * So for any acquired lock, the process holding that lock may be
846 * waiting on at most one other lock. That lock in turns may be held by
847 * someone waiting for at most one other lock. Given a requested lock
848 * caller_fl which is about to wait for a conflicting lock block_fl, we
849 * follow this chain of waiters to ensure we are not about to create a
852 * Since we do this before we ever put a process to sleep on a lock, we
853 * are ensured that there is never a cycle; that is what guarantees that
854 * the while() loop in posix_locks_deadlock() eventually completes.
856 * Note: the above assumption may not be true when handling lock
857 * requests from a broken NFS client. It may also fail in the presence
858 * of tasks (such as posix threads) sharing the same open file table.
859 * To handle those cases, we just bail out after a few iterations.
861 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
862 * Because the owner is not even nominally tied to a thread of
863 * execution, the deadlock detection below can't reasonably work well. Just
866 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
867 * locks that just checks for the case where two tasks are attempting to
868 * upgrade from read to write locks on the same inode.
871 #define MAX_DEADLK_ITERATIONS 10
873 /* Find a lock that the owner of the given block_fl is blocking on. */
874 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
876 struct file_lock *fl;
878 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
879 if (posix_same_owner(fl, block_fl))
885 /* Must be called with the blocked_lock_lock held! */
886 static int posix_locks_deadlock(struct file_lock *caller_fl,
887 struct file_lock *block_fl)
891 lockdep_assert_held(&blocked_lock_lock);
894 * This deadlock detector can't reasonably detect deadlocks with
895 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
897 if (IS_OFDLCK(caller_fl))
900 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
901 if (i++ > MAX_DEADLK_ITERATIONS)
903 if (posix_same_owner(caller_fl, block_fl))
909 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
910 * after any leases, but before any posix locks.
912 * Note that if called with an FL_EXISTS argument, the caller may determine
913 * whether or not a lock was successfully freed by testing the return
916 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
918 struct file_lock *new_fl = NULL;
919 struct file_lock *fl;
920 struct file_lock_context *ctx;
925 ctx = locks_get_lock_context(inode, request->fl_type);
927 if (request->fl_type != F_UNLCK)
929 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
932 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
933 new_fl = locks_alloc_lock();
938 percpu_down_read_preempt_disable(&file_rwsem);
939 spin_lock(&ctx->flc_lock);
940 if (request->fl_flags & FL_ACCESS)
943 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
944 if (request->fl_file != fl->fl_file)
946 if (request->fl_type == fl->fl_type)
949 locks_delete_lock_ctx(fl, &dispose);
953 if (request->fl_type == F_UNLCK) {
954 if ((request->fl_flags & FL_EXISTS) && !found)
960 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
961 if (!flock_locks_conflict(request, fl))
964 if (!(request->fl_flags & FL_SLEEP))
966 error = FILE_LOCK_DEFERRED;
967 locks_insert_block(fl, request);
970 if (request->fl_flags & FL_ACCESS)
972 locks_copy_lock(new_fl, request);
973 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
978 spin_unlock(&ctx->flc_lock);
979 percpu_up_read_preempt_enable(&file_rwsem);
981 locks_free_lock(new_fl);
982 locks_dispose_list(&dispose);
986 static int posix_lock_inode(struct inode *inode, struct file_lock *request,
987 struct file_lock *conflock)
989 struct file_lock *fl, *tmp;
990 struct file_lock *new_fl = NULL;
991 struct file_lock *new_fl2 = NULL;
992 struct file_lock *left = NULL;
993 struct file_lock *right = NULL;
994 struct file_lock_context *ctx;
999 ctx = locks_get_lock_context(inode, request->fl_type);
1001 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
1004 * We may need two file_lock structures for this operation,
1005 * so we get them in advance to avoid races.
1007 * In some cases we can be sure, that no new locks will be needed
1009 if (!(request->fl_flags & FL_ACCESS) &&
1010 (request->fl_type != F_UNLCK ||
1011 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1012 new_fl = locks_alloc_lock();
1013 new_fl2 = locks_alloc_lock();
1016 percpu_down_read_preempt_disable(&file_rwsem);
1017 spin_lock(&ctx->flc_lock);
1019 * New lock request. Walk all POSIX locks and look for conflicts. If
1020 * there are any, either return error or put the request on the
1021 * blocker's list of waiters and the global blocked_hash.
1023 if (request->fl_type != F_UNLCK) {
1024 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1025 if (!posix_locks_conflict(request, fl))
1028 locks_copy_conflock(conflock, fl);
1030 if (!(request->fl_flags & FL_SLEEP))
1033 * Deadlock detection and insertion into the blocked
1034 * locks list must be done while holding the same lock!
1037 spin_lock(&blocked_lock_lock);
1038 if (likely(!posix_locks_deadlock(request, fl))) {
1039 error = FILE_LOCK_DEFERRED;
1040 __locks_insert_block(fl, request);
1042 spin_unlock(&blocked_lock_lock);
1047 /* If we're just looking for a conflict, we're done. */
1049 if (request->fl_flags & FL_ACCESS)
1052 /* Find the first old lock with the same owner as the new lock */
1053 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1054 if (posix_same_owner(request, fl))
1058 /* Process locks with this owner. */
1059 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1060 if (!posix_same_owner(request, fl))
1063 /* Detect adjacent or overlapping regions (if same lock type) */
1064 if (request->fl_type == fl->fl_type) {
1065 /* In all comparisons of start vs end, use
1066 * "start - 1" rather than "end + 1". If end
1067 * is OFFSET_MAX, end + 1 will become negative.
1069 if (fl->fl_end < request->fl_start - 1)
1071 /* If the next lock in the list has entirely bigger
1072 * addresses than the new one, insert the lock here.
1074 if (fl->fl_start - 1 > request->fl_end)
1077 /* If we come here, the new and old lock are of the
1078 * same type and adjacent or overlapping. Make one
1079 * lock yielding from the lower start address of both
1080 * locks to the higher end address.
1082 if (fl->fl_start > request->fl_start)
1083 fl->fl_start = request->fl_start;
1085 request->fl_start = fl->fl_start;
1086 if (fl->fl_end < request->fl_end)
1087 fl->fl_end = request->fl_end;
1089 request->fl_end = fl->fl_end;
1091 locks_delete_lock_ctx(fl, &dispose);
1097 /* Processing for different lock types is a bit
1100 if (fl->fl_end < request->fl_start)
1102 if (fl->fl_start > request->fl_end)
1104 if (request->fl_type == F_UNLCK)
1106 if (fl->fl_start < request->fl_start)
1108 /* If the next lock in the list has a higher end
1109 * address than the new one, insert the new one here.
1111 if (fl->fl_end > request->fl_end) {
1115 if (fl->fl_start >= request->fl_start) {
1116 /* The new lock completely replaces an old
1117 * one (This may happen several times).
1120 locks_delete_lock_ctx(fl, &dispose);
1124 * Replace the old lock with new_fl, and
1125 * remove the old one. It's safe to do the
1126 * insert here since we know that we won't be
1127 * using new_fl later, and that the lock is
1128 * just replacing an existing lock.
1133 locks_copy_lock(new_fl, request);
1136 locks_insert_lock_ctx(request, &fl->fl_list);
1137 locks_delete_lock_ctx(fl, &dispose);
1144 * The above code only modifies existing locks in case of merging or
1145 * replacing. If new lock(s) need to be inserted all modifications are
1146 * done below this, so it's safe yet to bail out.
1148 error = -ENOLCK; /* "no luck" */
1149 if (right && left == right && !new_fl2)
1154 if (request->fl_type == F_UNLCK) {
1155 if (request->fl_flags & FL_EXISTS)
1164 locks_copy_lock(new_fl, request);
1165 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1170 if (left == right) {
1171 /* The new lock breaks the old one in two pieces,
1172 * so we have to use the second new lock.
1176 locks_copy_lock(left, right);
1177 locks_insert_lock_ctx(left, &fl->fl_list);
1179 right->fl_start = request->fl_end + 1;
1180 locks_wake_up_blocks(right);
1183 left->fl_end = request->fl_start - 1;
1184 locks_wake_up_blocks(left);
1187 spin_unlock(&ctx->flc_lock);
1188 percpu_up_read_preempt_enable(&file_rwsem);
1190 * Free any unused locks.
1193 locks_free_lock(new_fl);
1195 locks_free_lock(new_fl2);
1196 locks_dispose_list(&dispose);
1197 trace_posix_lock_inode(inode, request, error);
1203 * posix_lock_file - Apply a POSIX-style lock to a file
1204 * @filp: The file to apply the lock to
1205 * @fl: The lock to be applied
1206 * @conflock: Place to return a copy of the conflicting lock, if found.
1208 * Add a POSIX style lock to a file.
1209 * We merge adjacent & overlapping locks whenever possible.
1210 * POSIX locks are sorted by owner task, then by starting address
1212 * Note that if called with an FL_EXISTS argument, the caller may determine
1213 * whether or not a lock was successfully freed by testing the return
1214 * value for -ENOENT.
1216 int posix_lock_file(struct file *filp, struct file_lock *fl,
1217 struct file_lock *conflock)
1219 return posix_lock_inode(locks_inode(filp), fl, conflock);
1221 EXPORT_SYMBOL(posix_lock_file);
1224 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1225 * @inode: inode of file to which lock request should be applied
1226 * @fl: The lock to be applied
1228 * Apply a POSIX style lock request to an inode.
1230 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1235 error = posix_lock_inode(inode, fl, NULL);
1236 if (error != FILE_LOCK_DEFERRED)
1238 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1242 locks_delete_block(fl);
1248 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1250 * locks_mandatory_locked - Check for an active lock
1251 * @file: the file to check
1253 * Searches the inode's list of locks to find any POSIX locks which conflict.
1254 * This function is called from locks_verify_locked() only.
1256 int locks_mandatory_locked(struct file *file)
1259 struct inode *inode = locks_inode(file);
1260 struct file_lock_context *ctx;
1261 struct file_lock *fl;
1263 ctx = smp_load_acquire(&inode->i_flctx);
1264 if (!ctx || list_empty_careful(&ctx->flc_posix))
1268 * Search the lock list for this inode for any POSIX locks.
1270 spin_lock(&ctx->flc_lock);
1272 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1273 if (fl->fl_owner != current->files &&
1274 fl->fl_owner != file) {
1279 spin_unlock(&ctx->flc_lock);
1284 * locks_mandatory_area - Check for a conflicting lock
1285 * @inode: the file to check
1286 * @filp: how the file was opened (if it was)
1287 * @start: first byte in the file to check
1288 * @end: lastbyte in the file to check
1289 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1291 * Searches the inode's list of locks to find any POSIX locks which conflict.
1293 int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start,
1294 loff_t end, unsigned char type)
1296 struct file_lock fl;
1300 locks_init_lock(&fl);
1301 fl.fl_pid = current->tgid;
1303 fl.fl_flags = FL_POSIX | FL_ACCESS;
1304 if (filp && !(filp->f_flags & O_NONBLOCK))
1307 fl.fl_start = start;
1313 fl.fl_flags &= ~FL_SLEEP;
1314 error = posix_lock_inode(inode, &fl, NULL);
1320 fl.fl_flags |= FL_SLEEP;
1321 fl.fl_owner = current->files;
1322 error = posix_lock_inode(inode, &fl, NULL);
1323 if (error != FILE_LOCK_DEFERRED)
1325 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1328 * If we've been sleeping someone might have
1329 * changed the permissions behind our back.
1331 if (__mandatory_lock(inode))
1335 locks_delete_block(&fl);
1342 EXPORT_SYMBOL(locks_mandatory_area);
1343 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1345 static void lease_clear_pending(struct file_lock *fl, int arg)
1349 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1352 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1356 /* We already had a lease on this file; just change its type */
1357 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1359 int error = assign_type(fl, arg);
1363 lease_clear_pending(fl, arg);
1364 locks_wake_up_blocks(fl);
1365 if (arg == F_UNLCK) {
1366 struct file *filp = fl->fl_file;
1369 filp->f_owner.signum = 0;
1370 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1371 if (fl->fl_fasync != NULL) {
1372 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1373 fl->fl_fasync = NULL;
1375 locks_delete_lock_ctx(fl, dispose);
1379 EXPORT_SYMBOL(lease_modify);
1381 static bool past_time(unsigned long then)
1384 /* 0 is a special value meaning "this never expires": */
1386 return time_after(jiffies, then);
1389 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1391 struct file_lock_context *ctx = inode->i_flctx;
1392 struct file_lock *fl, *tmp;
1394 lockdep_assert_held(&ctx->flc_lock);
1396 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1397 trace_time_out_leases(inode, fl);
1398 if (past_time(fl->fl_downgrade_time))
1399 lease_modify(fl, F_RDLCK, dispose);
1400 if (past_time(fl->fl_break_time))
1401 lease_modify(fl, F_UNLCK, dispose);
1405 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1407 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1409 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1411 return locks_conflict(breaker, lease);
1415 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1417 struct file_lock_context *ctx = inode->i_flctx;
1418 struct file_lock *fl;
1420 lockdep_assert_held(&ctx->flc_lock);
1422 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1423 if (leases_conflict(fl, breaker))
1430 * __break_lease - revoke all outstanding leases on file
1431 * @inode: the inode of the file to return
1432 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1434 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1437 * break_lease (inlined for speed) has checked there already is at least
1438 * some kind of lock (maybe a lease) on this file. Leases are broken on
1439 * a call to open() or truncate(). This function can sleep unless you
1440 * specified %O_NONBLOCK to your open().
1442 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1445 struct file_lock_context *ctx;
1446 struct file_lock *new_fl, *fl, *tmp;
1447 unsigned long break_time;
1448 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1451 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1453 return PTR_ERR(new_fl);
1454 new_fl->fl_flags = type;
1456 /* typically we will check that ctx is non-NULL before calling */
1457 ctx = smp_load_acquire(&inode->i_flctx);
1463 percpu_down_read_preempt_disable(&file_rwsem);
1464 spin_lock(&ctx->flc_lock);
1466 time_out_leases(inode, &dispose);
1468 if (!any_leases_conflict(inode, new_fl))
1472 if (lease_break_time > 0) {
1473 break_time = jiffies + lease_break_time * HZ;
1474 if (break_time == 0)
1475 break_time++; /* so that 0 means no break time */
1478 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1479 if (!leases_conflict(fl, new_fl))
1482 if (fl->fl_flags & FL_UNLOCK_PENDING)
1484 fl->fl_flags |= FL_UNLOCK_PENDING;
1485 fl->fl_break_time = break_time;
1487 if (lease_breaking(fl))
1489 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1490 fl->fl_downgrade_time = break_time;
1492 if (fl->fl_lmops->lm_break(fl))
1493 locks_delete_lock_ctx(fl, &dispose);
1496 if (list_empty(&ctx->flc_lease))
1499 if (mode & O_NONBLOCK) {
1500 trace_break_lease_noblock(inode, new_fl);
1501 error = -EWOULDBLOCK;
1506 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1507 break_time = fl->fl_break_time;
1508 if (break_time != 0)
1509 break_time -= jiffies;
1510 if (break_time == 0)
1512 locks_insert_block(fl, new_fl);
1513 trace_break_lease_block(inode, new_fl);
1514 spin_unlock(&ctx->flc_lock);
1515 percpu_up_read_preempt_enable(&file_rwsem);
1517 locks_dispose_list(&dispose);
1518 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1519 !new_fl->fl_next, break_time);
1521 percpu_down_read_preempt_disable(&file_rwsem);
1522 spin_lock(&ctx->flc_lock);
1523 trace_break_lease_unblock(inode, new_fl);
1524 locks_delete_block(new_fl);
1527 * Wait for the next conflicting lease that has not been
1531 time_out_leases(inode, &dispose);
1532 if (any_leases_conflict(inode, new_fl))
1537 spin_unlock(&ctx->flc_lock);
1538 percpu_up_read_preempt_enable(&file_rwsem);
1539 locks_dispose_list(&dispose);
1540 locks_free_lock(new_fl);
1544 EXPORT_SYMBOL(__break_lease);
1547 * lease_get_mtime - get the last modified time of an inode
1549 * @time: pointer to a timespec which will contain the last modified time
1551 * This is to force NFS clients to flush their caches for files with
1552 * exclusive leases. The justification is that if someone has an
1553 * exclusive lease, then they could be modifying it.
1555 void lease_get_mtime(struct inode *inode, struct timespec *time)
1557 bool has_lease = false;
1558 struct file_lock_context *ctx;
1559 struct file_lock *fl;
1561 ctx = smp_load_acquire(&inode->i_flctx);
1562 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1563 spin_lock(&ctx->flc_lock);
1564 fl = list_first_entry_or_null(&ctx->flc_lease,
1565 struct file_lock, fl_list);
1566 if (fl && (fl->fl_type == F_WRLCK))
1568 spin_unlock(&ctx->flc_lock);
1572 *time = current_time(inode);
1574 *time = inode->i_mtime;
1577 EXPORT_SYMBOL(lease_get_mtime);
1580 * fcntl_getlease - Enquire what lease is currently active
1583 * The value returned by this function will be one of
1584 * (if no lease break is pending):
1586 * %F_RDLCK to indicate a shared lease is held.
1588 * %F_WRLCK to indicate an exclusive lease is held.
1590 * %F_UNLCK to indicate no lease is held.
1592 * (if a lease break is pending):
1594 * %F_RDLCK to indicate an exclusive lease needs to be
1595 * changed to a shared lease (or removed).
1597 * %F_UNLCK to indicate the lease needs to be removed.
1599 * XXX: sfr & willy disagree over whether F_INPROGRESS
1600 * should be returned to userspace.
1602 int fcntl_getlease(struct file *filp)
1604 struct file_lock *fl;
1605 struct inode *inode = locks_inode(filp);
1606 struct file_lock_context *ctx;
1610 ctx = smp_load_acquire(&inode->i_flctx);
1611 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1612 percpu_down_read_preempt_disable(&file_rwsem);
1613 spin_lock(&ctx->flc_lock);
1614 time_out_leases(inode, &dispose);
1615 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1616 if (fl->fl_file != filp)
1618 type = target_leasetype(fl);
1621 spin_unlock(&ctx->flc_lock);
1622 percpu_up_read_preempt_enable(&file_rwsem);
1624 locks_dispose_list(&dispose);
1630 * check_conflicting_open - see if the given dentry points to a file that has
1631 * an existing open that would conflict with the
1633 * @dentry: dentry to check
1634 * @arg: type of lease that we're trying to acquire
1635 * @flags: current lock flags
1637 * Check to see if there's an existing open fd on this file that would
1638 * conflict with the lease we're trying to set.
1641 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1644 struct inode *inode = dentry->d_inode;
1646 if (flags & FL_LAYOUT)
1649 if ((arg == F_RDLCK) &&
1650 (atomic_read(&d_real_inode(dentry)->i_writecount) > 0))
1653 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1654 (atomic_read(&inode->i_count) > 1)))
1661 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1663 struct file_lock *fl, *my_fl = NULL, *lease;
1664 struct dentry *dentry = filp->f_path.dentry;
1665 struct inode *inode = dentry->d_inode;
1666 struct file_lock_context *ctx;
1667 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1672 trace_generic_add_lease(inode, lease);
1674 /* Note that arg is never F_UNLCK here */
1675 ctx = locks_get_lock_context(inode, arg);
1680 * In the delegation case we need mutual exclusion with
1681 * a number of operations that take the i_mutex. We trylock
1682 * because delegations are an optional optimization, and if
1683 * there's some chance of a conflict--we'd rather not
1684 * bother, maybe that's a sign this just isn't a good file to
1685 * hand out a delegation on.
1687 if (is_deleg && !inode_trylock(inode))
1690 if (is_deleg && arg == F_WRLCK) {
1691 /* Write delegations are not currently supported: */
1692 inode_unlock(inode);
1697 percpu_down_read_preempt_disable(&file_rwsem);
1698 spin_lock(&ctx->flc_lock);
1699 time_out_leases(inode, &dispose);
1700 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1705 * At this point, we know that if there is an exclusive
1706 * lease on this file, then we hold it on this filp
1707 * (otherwise our open of this file would have blocked).
1708 * And if we are trying to acquire an exclusive lease,
1709 * then the file is not open by anyone (including us)
1710 * except for this filp.
1713 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1714 if (fl->fl_file == filp &&
1715 fl->fl_owner == lease->fl_owner) {
1721 * No exclusive leases if someone else has a lease on
1727 * Modifying our existing lease is OK, but no getting a
1728 * new lease if someone else is opening for write:
1730 if (fl->fl_flags & FL_UNLOCK_PENDING)
1734 if (my_fl != NULL) {
1736 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1746 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1748 * The check in break_lease() is lockless. It's possible for another
1749 * open to race in after we did the earlier check for a conflicting
1750 * open but before the lease was inserted. Check again for a
1751 * conflicting open and cancel the lease if there is one.
1753 * We also add a barrier here to ensure that the insertion of the lock
1754 * precedes these checks.
1757 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1759 locks_unlink_lock_ctx(lease);
1764 if (lease->fl_lmops->lm_setup)
1765 lease->fl_lmops->lm_setup(lease, priv);
1767 spin_unlock(&ctx->flc_lock);
1768 percpu_up_read_preempt_enable(&file_rwsem);
1769 locks_dispose_list(&dispose);
1771 inode_unlock(inode);
1772 if (!error && !my_fl)
1777 static int generic_delete_lease(struct file *filp, void *owner)
1779 int error = -EAGAIN;
1780 struct file_lock *fl, *victim = NULL;
1781 struct inode *inode = locks_inode(filp);
1782 struct file_lock_context *ctx;
1785 ctx = smp_load_acquire(&inode->i_flctx);
1787 trace_generic_delete_lease(inode, NULL);
1791 percpu_down_read_preempt_disable(&file_rwsem);
1792 spin_lock(&ctx->flc_lock);
1793 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1794 if (fl->fl_file == filp &&
1795 fl->fl_owner == owner) {
1800 trace_generic_delete_lease(inode, victim);
1802 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1803 spin_unlock(&ctx->flc_lock);
1804 percpu_up_read_preempt_enable(&file_rwsem);
1805 locks_dispose_list(&dispose);
1810 * generic_setlease - sets a lease on an open file
1811 * @filp: file pointer
1812 * @arg: type of lease to obtain
1813 * @flp: input - file_lock to use, output - file_lock inserted
1814 * @priv: private data for lm_setup (may be NULL if lm_setup
1815 * doesn't require it)
1817 * The (input) flp->fl_lmops->lm_break function is required
1820 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1823 struct inode *inode = locks_inode(filp);
1826 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1828 if (!S_ISREG(inode->i_mode))
1830 error = security_file_lock(filp, arg);
1836 return generic_delete_lease(filp, *priv);
1839 if (!(*flp)->fl_lmops->lm_break) {
1844 return generic_add_lease(filp, arg, flp, priv);
1849 EXPORT_SYMBOL(generic_setlease);
1852 * vfs_setlease - sets a lease on an open file
1853 * @filp: file pointer
1854 * @arg: type of lease to obtain
1855 * @lease: file_lock to use when adding a lease
1856 * @priv: private info for lm_setup when adding a lease (may be
1857 * NULL if lm_setup doesn't require it)
1859 * Call this to establish a lease on the file. The "lease" argument is not
1860 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1861 * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be
1862 * set; if not, this function will return -ENOLCK (and generate a scary-looking
1865 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1866 * may be NULL if the lm_setup operation doesn't require it.
1869 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1871 if (filp->f_op->setlease && is_remote_lock(filp))
1872 return filp->f_op->setlease(filp, arg, lease, priv);
1874 return generic_setlease(filp, arg, lease, priv);
1876 EXPORT_SYMBOL_GPL(vfs_setlease);
1878 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1880 struct file_lock *fl;
1881 struct fasync_struct *new;
1884 fl = lease_alloc(filp, arg);
1888 new = fasync_alloc();
1890 locks_free_lock(fl);
1895 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1897 locks_free_lock(fl);
1904 * fcntl_setlease - sets a lease on an open file
1905 * @fd: open file descriptor
1906 * @filp: file pointer
1907 * @arg: type of lease to obtain
1909 * Call this fcntl to establish a lease on the file.
1910 * Note that you also need to call %F_SETSIG to
1911 * receive a signal when the lease is broken.
1913 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1916 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1917 return do_fcntl_add_lease(fd, filp, arg);
1921 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1922 * @inode: inode of the file to apply to
1923 * @fl: The lock to be applied
1925 * Apply a FLOCK style lock request to an inode.
1927 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1932 error = flock_lock_inode(inode, fl);
1933 if (error != FILE_LOCK_DEFERRED)
1935 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1939 locks_delete_block(fl);
1946 * locks_lock_inode_wait - Apply a lock to an inode
1947 * @inode: inode of the file to apply to
1948 * @fl: The lock to be applied
1950 * Apply a POSIX or FLOCK style lock request to an inode.
1952 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1955 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1957 res = posix_lock_inode_wait(inode, fl);
1960 res = flock_lock_inode_wait(inode, fl);
1967 EXPORT_SYMBOL(locks_lock_inode_wait);
1970 * sys_flock: - flock() system call.
1971 * @fd: the file descriptor to lock.
1972 * @cmd: the type of lock to apply.
1974 * Apply a %FL_FLOCK style lock to an open file descriptor.
1975 * The @cmd can be one of:
1977 * - %LOCK_SH -- a shared lock.
1978 * - %LOCK_EX -- an exclusive lock.
1979 * - %LOCK_UN -- remove an existing lock.
1980 * - %LOCK_MAND -- a 'mandatory' flock.
1981 * This exists to emulate Windows Share Modes.
1983 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1984 * processes read and write access respectively.
1986 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1988 struct fd f = fdget(fd);
1989 struct file_lock *lock;
1990 int can_sleep, unlock;
1997 can_sleep = !(cmd & LOCK_NB);
1999 unlock = (cmd == LOCK_UN);
2001 if (!unlock && !(cmd & LOCK_MAND) &&
2002 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
2005 lock = flock_make_lock(f.file, cmd);
2007 error = PTR_ERR(lock);
2012 lock->fl_flags |= FL_SLEEP;
2014 error = security_file_lock(f.file, lock->fl_type);
2018 if (f.file->f_op->flock && is_remote_lock(f.file))
2019 error = f.file->f_op->flock(f.file,
2020 (can_sleep) ? F_SETLKW : F_SETLK,
2023 error = locks_lock_file_wait(f.file, lock);
2026 locks_free_lock(lock);
2035 * vfs_test_lock - test file byte range lock
2036 * @filp: The file to test lock for
2037 * @fl: The lock to test; also used to hold result
2039 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2040 * setting conf->fl_type to something other than F_UNLCK.
2042 int vfs_test_lock(struct file *filp, struct file_lock *fl)
2044 if (filp->f_op->lock && is_remote_lock(filp))
2045 return filp->f_op->lock(filp, F_GETLK, fl);
2046 posix_test_lock(filp, fl);
2049 EXPORT_SYMBOL_GPL(vfs_test_lock);
2051 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2053 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
2054 #if BITS_PER_LONG == 32
2056 * Make sure we can represent the posix lock via
2057 * legacy 32bit flock.
2059 if (fl->fl_start > OFFT_OFFSET_MAX)
2061 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2064 flock->l_start = fl->fl_start;
2065 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2066 fl->fl_end - fl->fl_start + 1;
2067 flock->l_whence = 0;
2068 flock->l_type = fl->fl_type;
2072 #if BITS_PER_LONG == 32
2073 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2075 flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid;
2076 flock->l_start = fl->fl_start;
2077 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2078 fl->fl_end - fl->fl_start + 1;
2079 flock->l_whence = 0;
2080 flock->l_type = fl->fl_type;
2084 /* Report the first existing lock that would conflict with l.
2085 * This implements the F_GETLK command of fcntl().
2087 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock)
2089 struct file_lock file_lock;
2093 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2096 error = flock_to_posix_lock(filp, &file_lock, flock);
2100 if (cmd == F_OFD_GETLK) {
2102 if (flock->l_pid != 0)
2106 file_lock.fl_flags |= FL_OFDLCK;
2107 file_lock.fl_owner = filp;
2110 error = vfs_test_lock(filp, &file_lock);
2114 flock->l_type = file_lock.fl_type;
2115 if (file_lock.fl_type != F_UNLCK) {
2116 error = posix_lock_to_flock(flock, &file_lock);
2121 locks_release_private(&file_lock);
2127 * vfs_lock_file - file byte range lock
2128 * @filp: The file to apply the lock to
2129 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2130 * @fl: The lock to be applied
2131 * @conf: Place to return a copy of the conflicting lock, if found.
2133 * A caller that doesn't care about the conflicting lock may pass NULL
2134 * as the final argument.
2136 * If the filesystem defines a private ->lock() method, then @conf will
2137 * be left unchanged; so a caller that cares should initialize it to
2138 * some acceptable default.
2140 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2141 * locks, the ->lock() interface may return asynchronously, before the lock has
2142 * been granted or denied by the underlying filesystem, if (and only if)
2143 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2144 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2145 * the request is for a blocking lock. When ->lock() does return asynchronously,
2146 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2147 * request completes.
2148 * If the request is for non-blocking lock the file system should return
2149 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2150 * with the result. If the request timed out the callback routine will return a
2151 * nonzero return code and the file system should release the lock. The file
2152 * system is also responsible to keep a corresponding posix lock when it
2153 * grants a lock so the VFS can find out which locks are locally held and do
2154 * the correct lock cleanup when required.
2155 * The underlying filesystem must not drop the kernel lock or call
2156 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2159 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2161 if (filp->f_op->lock && is_remote_lock(filp))
2162 return filp->f_op->lock(filp, cmd, fl);
2164 return posix_lock_file(filp, fl, conf);
2166 EXPORT_SYMBOL_GPL(vfs_lock_file);
2168 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2169 struct file_lock *fl)
2173 error = security_file_lock(filp, fl->fl_type);
2178 error = vfs_lock_file(filp, cmd, fl, NULL);
2179 if (error != FILE_LOCK_DEFERRED)
2181 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2185 locks_delete_block(fl);
2192 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2194 check_fmode_for_setlk(struct file_lock *fl)
2196 switch (fl->fl_type) {
2198 if (!(fl->fl_file->f_mode & FMODE_READ))
2202 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2208 /* Apply the lock described by l to an open file descriptor.
2209 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2211 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2212 struct flock *flock)
2214 struct file_lock *file_lock = locks_alloc_lock();
2215 struct inode *inode = locks_inode(filp);
2219 if (file_lock == NULL)
2222 /* Don't allow mandatory locks on files that may be memory mapped
2225 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2230 error = flock_to_posix_lock(filp, file_lock, flock);
2234 error = check_fmode_for_setlk(file_lock);
2239 * If the cmd is requesting file-private locks, then set the
2240 * FL_OFDLCK flag and override the owner.
2245 if (flock->l_pid != 0)
2249 file_lock->fl_flags |= FL_OFDLCK;
2250 file_lock->fl_owner = filp;
2254 if (flock->l_pid != 0)
2258 file_lock->fl_flags |= FL_OFDLCK;
2259 file_lock->fl_owner = filp;
2262 file_lock->fl_flags |= FL_SLEEP;
2265 error = do_lock_file_wait(filp, cmd, file_lock);
2268 * Attempt to detect a close/fcntl race and recover by releasing the
2269 * lock that was just acquired. There is no need to do that when we're
2270 * unlocking though, or for OFD locks.
2272 if (!error && file_lock->fl_type != F_UNLCK &&
2273 !(file_lock->fl_flags & FL_OFDLCK)) {
2275 * We need that spin_lock here - it prevents reordering between
2276 * update of i_flctx->flc_posix and check for it done in
2277 * close(). rcu_read_lock() wouldn't do.
2279 spin_lock(¤t->files->file_lock);
2281 spin_unlock(¤t->files->file_lock);
2283 file_lock->fl_type = F_UNLCK;
2284 error = do_lock_file_wait(filp, cmd, file_lock);
2285 WARN_ON_ONCE(error);
2290 trace_fcntl_setlk(inode, file_lock, error);
2291 locks_free_lock(file_lock);
2295 #if BITS_PER_LONG == 32
2296 /* Report the first existing lock that would conflict with l.
2297 * This implements the F_GETLK command of fcntl().
2299 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock)
2301 struct file_lock file_lock;
2305 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2308 error = flock64_to_posix_lock(filp, &file_lock, flock);
2312 if (cmd == F_OFD_GETLK) {
2314 if (flock->l_pid != 0)
2318 file_lock.fl_flags |= FL_OFDLCK;
2319 file_lock.fl_owner = filp;
2322 error = vfs_test_lock(filp, &file_lock);
2326 flock->l_type = file_lock.fl_type;
2327 if (file_lock.fl_type != F_UNLCK)
2328 posix_lock_to_flock64(flock, &file_lock);
2330 locks_release_private(&file_lock);
2335 /* Apply the lock described by l to an open file descriptor.
2336 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2338 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2339 struct flock64 *flock)
2341 struct file_lock *file_lock = locks_alloc_lock();
2342 struct inode *inode = locks_inode(filp);
2346 if (file_lock == NULL)
2349 /* Don't allow mandatory locks on files that may be memory mapped
2352 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2357 error = flock64_to_posix_lock(filp, file_lock, flock);
2361 error = check_fmode_for_setlk(file_lock);
2366 * If the cmd is requesting file-private locks, then set the
2367 * FL_OFDLCK flag and override the owner.
2372 if (flock->l_pid != 0)
2376 file_lock->fl_flags |= FL_OFDLCK;
2377 file_lock->fl_owner = filp;
2381 if (flock->l_pid != 0)
2385 file_lock->fl_flags |= FL_OFDLCK;
2386 file_lock->fl_owner = filp;
2389 file_lock->fl_flags |= FL_SLEEP;
2392 error = do_lock_file_wait(filp, cmd, file_lock);
2395 * Attempt to detect a close/fcntl race and recover by releasing the
2396 * lock that was just acquired. There is no need to do that when we're
2397 * unlocking though, or for OFD locks.
2399 if (!error && file_lock->fl_type != F_UNLCK &&
2400 !(file_lock->fl_flags & FL_OFDLCK)) {
2402 * We need that spin_lock here - it prevents reordering between
2403 * update of i_flctx->flc_posix and check for it done in
2404 * close(). rcu_read_lock() wouldn't do.
2406 spin_lock(¤t->files->file_lock);
2408 spin_unlock(¤t->files->file_lock);
2410 file_lock->fl_type = F_UNLCK;
2411 error = do_lock_file_wait(filp, cmd, file_lock);
2412 WARN_ON_ONCE(error);
2417 locks_free_lock(file_lock);
2420 #endif /* BITS_PER_LONG == 32 */
2423 * This function is called when the file is being removed
2424 * from the task's fd array. POSIX locks belonging to this task
2425 * are deleted at this time.
2427 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2430 struct inode *inode = locks_inode(filp);
2431 struct file_lock lock;
2432 struct file_lock_context *ctx;
2435 * If there are no locks held on this file, we don't need to call
2436 * posix_lock_file(). Another process could be setting a lock on this
2437 * file at the same time, but we wouldn't remove that lock anyway.
2439 ctx = smp_load_acquire(&inode->i_flctx);
2440 if (!ctx || list_empty(&ctx->flc_posix))
2443 lock.fl_type = F_UNLCK;
2444 lock.fl_flags = FL_POSIX | FL_CLOSE;
2446 lock.fl_end = OFFSET_MAX;
2447 lock.fl_owner = owner;
2448 lock.fl_pid = current->tgid;
2449 lock.fl_file = filp;
2451 lock.fl_lmops = NULL;
2453 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
2455 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2456 lock.fl_ops->fl_release_private(&lock);
2457 trace_locks_remove_posix(inode, &lock, error);
2460 EXPORT_SYMBOL(locks_remove_posix);
2462 /* The i_flctx must be valid when calling into here */
2464 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2466 struct file_lock fl = {
2468 .fl_pid = current->tgid,
2470 .fl_flags = FL_FLOCK | FL_CLOSE,
2472 .fl_end = OFFSET_MAX,
2474 struct inode *inode = locks_inode(filp);
2476 if (list_empty(&flctx->flc_flock))
2479 if (filp->f_op->flock && is_remote_lock(filp))
2480 filp->f_op->flock(filp, F_SETLKW, &fl);
2482 flock_lock_inode(inode, &fl);
2484 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2485 fl.fl_ops->fl_release_private(&fl);
2488 /* The i_flctx must be valid when calling into here */
2490 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2492 struct file_lock *fl, *tmp;
2495 if (list_empty(&ctx->flc_lease))
2498 percpu_down_read_preempt_disable(&file_rwsem);
2499 spin_lock(&ctx->flc_lock);
2500 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2501 if (filp == fl->fl_file)
2502 lease_modify(fl, F_UNLCK, &dispose);
2503 spin_unlock(&ctx->flc_lock);
2504 percpu_up_read_preempt_enable(&file_rwsem);
2506 locks_dispose_list(&dispose);
2510 * This function is called on the last close of an open file.
2512 void locks_remove_file(struct file *filp)
2514 struct file_lock_context *ctx;
2516 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
2520 /* remove any OFD locks */
2521 locks_remove_posix(filp, filp);
2523 /* remove flock locks */
2524 locks_remove_flock(filp, ctx);
2526 /* remove any leases */
2527 locks_remove_lease(filp, ctx);
2531 * posix_unblock_lock - stop waiting for a file lock
2532 * @waiter: the lock which was waiting
2534 * lockd needs to block waiting for locks.
2537 posix_unblock_lock(struct file_lock *waiter)
2541 spin_lock(&blocked_lock_lock);
2542 if (waiter->fl_next)
2543 __locks_delete_block(waiter);
2546 spin_unlock(&blocked_lock_lock);
2549 EXPORT_SYMBOL(posix_unblock_lock);
2552 * vfs_cancel_lock - file byte range unblock lock
2553 * @filp: The file to apply the unblock to
2554 * @fl: The lock to be unblocked
2556 * Used by lock managers to cancel blocked requests
2558 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2560 if (filp->f_op->lock && is_remote_lock(filp))
2561 return filp->f_op->lock(filp, F_CANCELLK, fl);
2565 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2567 #ifdef CONFIG_PROC_FS
2568 #include <linux/proc_fs.h>
2569 #include <linux/seq_file.h>
2571 struct locks_iterator {
2576 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2577 loff_t id, char *pfx)
2579 struct inode *inode = NULL;
2580 unsigned int fl_pid;
2583 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2585 /* Don't let fl_pid change based on who is reading the file */
2586 fl_pid = pid_nr_ns(fl->fl_nspid, proc_pidns);
2589 * If there isn't a fl_pid don't display who is waiting on
2590 * the lock if we are called from locks_show, or if we are
2591 * called from __show_fd_info - skip lock entirely
2596 fl_pid = fl->fl_pid;
2598 if (fl->fl_file != NULL)
2599 inode = locks_inode(fl->fl_file);
2601 seq_printf(f, "%lld:%s ", id, pfx);
2603 if (fl->fl_flags & FL_ACCESS)
2604 seq_puts(f, "ACCESS");
2605 else if (IS_OFDLCK(fl))
2606 seq_puts(f, "OFDLCK");
2608 seq_puts(f, "POSIX ");
2610 seq_printf(f, " %s ",
2611 (inode == NULL) ? "*NOINODE*" :
2612 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2613 } else if (IS_FLOCK(fl)) {
2614 if (fl->fl_type & LOCK_MAND) {
2615 seq_puts(f, "FLOCK MSNFS ");
2617 seq_puts(f, "FLOCK ADVISORY ");
2619 } else if (IS_LEASE(fl)) {
2620 if (fl->fl_flags & FL_DELEG)
2621 seq_puts(f, "DELEG ");
2623 seq_puts(f, "LEASE ");
2625 if (lease_breaking(fl))
2626 seq_puts(f, "BREAKING ");
2627 else if (fl->fl_file)
2628 seq_puts(f, "ACTIVE ");
2630 seq_puts(f, "BREAKER ");
2632 seq_puts(f, "UNKNOWN UNKNOWN ");
2634 if (fl->fl_type & LOCK_MAND) {
2635 seq_printf(f, "%s ",
2636 (fl->fl_type & LOCK_READ)
2637 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2638 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2640 seq_printf(f, "%s ",
2641 (lease_breaking(fl))
2642 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2643 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2646 /* userspace relies on this representation of dev_t */
2647 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2648 MAJOR(inode->i_sb->s_dev),
2649 MINOR(inode->i_sb->s_dev), inode->i_ino);
2651 seq_printf(f, "%d <none>:0 ", fl_pid);
2654 if (fl->fl_end == OFFSET_MAX)
2655 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2657 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2659 seq_puts(f, "0 EOF\n");
2663 static int locks_show(struct seq_file *f, void *v)
2665 struct locks_iterator *iter = f->private;
2666 struct file_lock *fl, *bfl;
2667 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2669 fl = hlist_entry(v, struct file_lock, fl_link);
2671 if (fl->fl_nspid && !pid_nr_ns(fl->fl_nspid, proc_pidns))
2674 lock_get_status(f, fl, iter->li_pos, "");
2676 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2677 lock_get_status(f, bfl, iter->li_pos, " ->");
2682 static void __show_fd_locks(struct seq_file *f,
2683 struct list_head *head, int *id,
2684 struct file *filp, struct files_struct *files)
2686 struct file_lock *fl;
2688 list_for_each_entry(fl, head, fl_list) {
2690 if (filp != fl->fl_file)
2692 if (fl->fl_owner != files &&
2693 fl->fl_owner != filp)
2697 seq_puts(f, "lock:\t");
2698 lock_get_status(f, fl, *id, "");
2702 void show_fd_locks(struct seq_file *f,
2703 struct file *filp, struct files_struct *files)
2705 struct inode *inode = locks_inode(filp);
2706 struct file_lock_context *ctx;
2709 ctx = smp_load_acquire(&inode->i_flctx);
2713 spin_lock(&ctx->flc_lock);
2714 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2715 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2716 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2717 spin_unlock(&ctx->flc_lock);
2720 static void *locks_start(struct seq_file *f, loff_t *pos)
2721 __acquires(&blocked_lock_lock)
2723 struct locks_iterator *iter = f->private;
2725 iter->li_pos = *pos + 1;
2726 percpu_down_write(&file_rwsem);
2727 spin_lock(&blocked_lock_lock);
2728 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
2731 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2733 struct locks_iterator *iter = f->private;
2736 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
2739 static void locks_stop(struct seq_file *f, void *v)
2740 __releases(&blocked_lock_lock)
2742 spin_unlock(&blocked_lock_lock);
2743 percpu_up_write(&file_rwsem);
2746 static const struct seq_operations locks_seq_operations = {
2747 .start = locks_start,
2753 static int locks_open(struct inode *inode, struct file *filp)
2755 return seq_open_private(filp, &locks_seq_operations,
2756 sizeof(struct locks_iterator));
2759 static const struct file_operations proc_locks_operations = {
2762 .llseek = seq_lseek,
2763 .release = seq_release_private,
2766 static int __init proc_locks_init(void)
2768 proc_create("locks", 0, NULL, &proc_locks_operations);
2771 fs_initcall(proc_locks_init);
2774 static int __init filelock_init(void)
2778 flctx_cache = kmem_cache_create("file_lock_ctx",
2779 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2781 filelock_cache = kmem_cache_create("file_lock_cache",
2782 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2785 for_each_possible_cpu(i) {
2786 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2788 spin_lock_init(&fll->lock);
2789 INIT_HLIST_HEAD(&fll->hlist);
2795 core_initcall(filelock_init);