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 bool lease_breaking(struct file_lock *fl)
144 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
147 static int target_leasetype(struct file_lock *fl)
149 if (fl->fl_flags & FL_UNLOCK_PENDING)
151 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
156 int leases_enable = 1;
157 int lease_break_time = 45;
160 * The global file_lock_list is only used for displaying /proc/locks, so we
161 * keep a list on each CPU, with each list protected by its own spinlock.
162 * Global serialization is done using file_rwsem.
164 * Note that alterations to the list also require that the relevant flc_lock is
167 struct file_lock_list_struct {
169 struct hlist_head hlist;
171 static DEFINE_PER_CPU(struct file_lock_list_struct, file_lock_list);
172 DEFINE_STATIC_PERCPU_RWSEM(file_rwsem);
175 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
176 * It is protected by blocked_lock_lock.
178 * We hash locks by lockowner in order to optimize searching for the lock a
179 * particular lockowner is waiting on.
181 * FIXME: make this value scale via some heuristic? We generally will want more
182 * buckets when we have more lockowners holding locks, but that's a little
183 * difficult to determine without knowing what the workload will look like.
185 #define BLOCKED_HASH_BITS 7
186 static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS);
189 * This lock protects the blocked_hash. Generally, if you're accessing it, you
190 * want to be holding this lock.
192 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
193 * pointer for file_lock structures that are acting as lock requests (in
194 * contrast to those that are acting as records of acquired locks).
196 * Note that when we acquire this lock in order to change the above fields,
197 * we often hold the flc_lock as well. In certain cases, when reading the fields
198 * protected by this lock, we can skip acquiring it iff we already hold the
201 * In particular, adding an entry to the fl_block list requires that you hold
202 * both the flc_lock and the blocked_lock_lock (acquired in that order).
203 * Deleting an entry from the list however only requires the file_lock_lock.
205 static DEFINE_SPINLOCK(blocked_lock_lock);
207 static struct kmem_cache *flctx_cache __read_mostly;
208 static struct kmem_cache *filelock_cache __read_mostly;
210 static struct file_lock_context *
211 locks_get_lock_context(struct inode *inode, int type)
213 struct file_lock_context *ctx;
215 /* paired with cmpxchg() below */
216 ctx = smp_load_acquire(&inode->i_flctx);
217 if (likely(ctx) || type == F_UNLCK)
220 ctx = kmem_cache_alloc(flctx_cache, GFP_KERNEL);
224 spin_lock_init(&ctx->flc_lock);
225 INIT_LIST_HEAD(&ctx->flc_flock);
226 INIT_LIST_HEAD(&ctx->flc_posix);
227 INIT_LIST_HEAD(&ctx->flc_lease);
230 * Assign the pointer if it's not already assigned. If it is, then
231 * free the context we just allocated.
233 if (cmpxchg(&inode->i_flctx, NULL, ctx)) {
234 kmem_cache_free(flctx_cache, ctx);
235 ctx = smp_load_acquire(&inode->i_flctx);
238 trace_locks_get_lock_context(inode, type, ctx);
243 locks_dump_ctx_list(struct list_head *list, char *list_type)
245 struct file_lock *fl;
247 list_for_each_entry(fl, list, fl_list) {
248 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);
253 locks_check_ctx_lists(struct inode *inode)
255 struct file_lock_context *ctx = inode->i_flctx;
257 if (unlikely(!list_empty(&ctx->flc_flock) ||
258 !list_empty(&ctx->flc_posix) ||
259 !list_empty(&ctx->flc_lease))) {
260 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
261 MAJOR(inode->i_sb->s_dev), MINOR(inode->i_sb->s_dev),
263 locks_dump_ctx_list(&ctx->flc_flock, "FLOCK");
264 locks_dump_ctx_list(&ctx->flc_posix, "POSIX");
265 locks_dump_ctx_list(&ctx->flc_lease, "LEASE");
270 locks_check_ctx_file_list(struct file *filp, struct list_head *list,
273 struct file_lock *fl;
274 struct inode *inode = locks_inode(filp);
276 list_for_each_entry(fl, list, fl_list)
277 if (fl->fl_file == filp)
278 pr_warn("Leaked %s lock on dev=0x%x:0x%x ino=0x%lx "
279 " fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n",
280 list_type, MAJOR(inode->i_sb->s_dev),
281 MINOR(inode->i_sb->s_dev), inode->i_ino,
282 fl->fl_owner, fl->fl_flags, fl->fl_type, fl->fl_pid);
286 locks_free_lock_context(struct inode *inode)
288 struct file_lock_context *ctx = inode->i_flctx;
291 locks_check_ctx_lists(inode);
292 kmem_cache_free(flctx_cache, ctx);
296 static void locks_init_lock_heads(struct file_lock *fl)
298 INIT_HLIST_NODE(&fl->fl_link);
299 INIT_LIST_HEAD(&fl->fl_list);
300 INIT_LIST_HEAD(&fl->fl_block);
301 init_waitqueue_head(&fl->fl_wait);
304 /* Allocate an empty lock structure. */
305 struct file_lock *locks_alloc_lock(void)
307 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
310 locks_init_lock_heads(fl);
314 EXPORT_SYMBOL_GPL(locks_alloc_lock);
316 void locks_release_private(struct file_lock *fl)
319 if (fl->fl_ops->fl_release_private)
320 fl->fl_ops->fl_release_private(fl);
325 if (fl->fl_lmops->lm_put_owner) {
326 fl->fl_lmops->lm_put_owner(fl->fl_owner);
332 EXPORT_SYMBOL_GPL(locks_release_private);
334 /* Free a lock which is not in use. */
335 void locks_free_lock(struct file_lock *fl)
337 BUG_ON(waitqueue_active(&fl->fl_wait));
338 BUG_ON(!list_empty(&fl->fl_list));
339 BUG_ON(!list_empty(&fl->fl_block));
340 BUG_ON(!hlist_unhashed(&fl->fl_link));
342 locks_release_private(fl);
343 kmem_cache_free(filelock_cache, fl);
345 EXPORT_SYMBOL(locks_free_lock);
348 locks_dispose_list(struct list_head *dispose)
350 struct file_lock *fl;
352 while (!list_empty(dispose)) {
353 fl = list_first_entry(dispose, struct file_lock, fl_list);
354 list_del_init(&fl->fl_list);
359 void locks_init_lock(struct file_lock *fl)
361 memset(fl, 0, sizeof(struct file_lock));
362 locks_init_lock_heads(fl);
365 EXPORT_SYMBOL(locks_init_lock);
368 * Initialize a new lock from an existing file_lock structure.
370 void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
372 new->fl_owner = fl->fl_owner;
373 new->fl_pid = fl->fl_pid;
375 new->fl_flags = fl->fl_flags;
376 new->fl_type = fl->fl_type;
377 new->fl_start = fl->fl_start;
378 new->fl_end = fl->fl_end;
379 new->fl_lmops = fl->fl_lmops;
383 if (fl->fl_lmops->lm_get_owner)
384 fl->fl_lmops->lm_get_owner(fl->fl_owner);
387 EXPORT_SYMBOL(locks_copy_conflock);
389 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
391 /* "new" must be a freshly-initialized lock */
392 WARN_ON_ONCE(new->fl_ops);
394 locks_copy_conflock(new, fl);
396 new->fl_file = fl->fl_file;
397 new->fl_ops = fl->fl_ops;
400 if (fl->fl_ops->fl_copy_lock)
401 fl->fl_ops->fl_copy_lock(new, fl);
405 EXPORT_SYMBOL(locks_copy_lock);
407 static inline int flock_translate_cmd(int cmd) {
409 return cmd & (LOCK_MAND | LOCK_RW);
421 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
422 static struct file_lock *
423 flock_make_lock(struct file *filp, unsigned int cmd)
425 struct file_lock *fl;
426 int type = flock_translate_cmd(cmd);
429 return ERR_PTR(type);
431 fl = locks_alloc_lock();
433 return ERR_PTR(-ENOMEM);
437 fl->fl_pid = current->tgid;
438 fl->fl_flags = FL_FLOCK;
440 fl->fl_end = OFFSET_MAX;
445 static int assign_type(struct file_lock *fl, long type)
459 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
462 switch (l->l_whence) {
467 fl->fl_start = filp->f_pos;
470 fl->fl_start = i_size_read(file_inode(filp));
475 if (l->l_start > OFFSET_MAX - fl->fl_start)
477 fl->fl_start += l->l_start;
478 if (fl->fl_start < 0)
481 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
482 POSIX-2001 defines it. */
484 if (l->l_len - 1 > OFFSET_MAX - fl->fl_start)
486 fl->fl_end = fl->fl_start + l->l_len - 1;
488 } else if (l->l_len < 0) {
489 if (fl->fl_start + l->l_len < 0)
491 fl->fl_end = fl->fl_start - 1;
492 fl->fl_start += l->l_len;
494 fl->fl_end = OFFSET_MAX;
496 fl->fl_owner = current->files;
497 fl->fl_pid = current->tgid;
499 fl->fl_flags = FL_POSIX;
503 return assign_type(fl, l->l_type);
506 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
509 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
512 struct flock64 ll = {
514 .l_whence = l->l_whence,
515 .l_start = l->l_start,
519 return flock64_to_posix_lock(filp, fl, &ll);
522 /* default lease lock manager operations */
524 lease_break_callback(struct file_lock *fl)
526 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
531 lease_setup(struct file_lock *fl, void **priv)
533 struct file *filp = fl->fl_file;
534 struct fasync_struct *fa = *priv;
537 * fasync_insert_entry() returns the old entry if any. If there was no
538 * old entry, then it used "priv" and inserted it into the fasync list.
539 * Clear the pointer to indicate that it shouldn't be freed.
541 if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa))
544 __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
547 static const struct lock_manager_operations lease_manager_ops = {
548 .lm_break = lease_break_callback,
549 .lm_change = lease_modify,
550 .lm_setup = lease_setup,
554 * Initialize a lease, use the default lock manager operations
556 static int lease_init(struct file *filp, long type, struct file_lock *fl)
558 if (assign_type(fl, type) != 0)
562 fl->fl_pid = current->tgid;
565 fl->fl_flags = FL_LEASE;
567 fl->fl_end = OFFSET_MAX;
569 fl->fl_lmops = &lease_manager_ops;
573 /* Allocate a file_lock initialised to this type of lease */
574 static struct file_lock *lease_alloc(struct file *filp, long type)
576 struct file_lock *fl = locks_alloc_lock();
580 return ERR_PTR(error);
582 error = lease_init(filp, type, fl);
585 return ERR_PTR(error);
590 /* Check if two locks overlap each other.
592 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
594 return ((fl1->fl_end >= fl2->fl_start) &&
595 (fl2->fl_end >= fl1->fl_start));
599 * Check whether two locks have the same owner.
601 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
603 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
604 return fl2->fl_lmops == fl1->fl_lmops &&
605 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
606 return fl1->fl_owner == fl2->fl_owner;
609 /* Must be called with the flc_lock held! */
610 static void locks_insert_global_locks(struct file_lock *fl)
612 struct file_lock_list_struct *fll = this_cpu_ptr(&file_lock_list);
614 percpu_rwsem_assert_held(&file_rwsem);
616 spin_lock(&fll->lock);
617 fl->fl_link_cpu = smp_processor_id();
618 hlist_add_head(&fl->fl_link, &fll->hlist);
619 spin_unlock(&fll->lock);
622 /* Must be called with the flc_lock held! */
623 static void locks_delete_global_locks(struct file_lock *fl)
625 struct file_lock_list_struct *fll;
627 percpu_rwsem_assert_held(&file_rwsem);
630 * Avoid taking lock if already unhashed. This is safe since this check
631 * is done while holding the flc_lock, and new insertions into the list
632 * also require that it be held.
634 if (hlist_unhashed(&fl->fl_link))
637 fll = per_cpu_ptr(&file_lock_list, fl->fl_link_cpu);
638 spin_lock(&fll->lock);
639 hlist_del_init(&fl->fl_link);
640 spin_unlock(&fll->lock);
644 posix_owner_key(struct file_lock *fl)
646 if (fl->fl_lmops && fl->fl_lmops->lm_owner_key)
647 return fl->fl_lmops->lm_owner_key(fl);
648 return (unsigned long)fl->fl_owner;
651 static void locks_insert_global_blocked(struct file_lock *waiter)
653 lockdep_assert_held(&blocked_lock_lock);
655 hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter));
658 static void locks_delete_global_blocked(struct file_lock *waiter)
660 lockdep_assert_held(&blocked_lock_lock);
662 hash_del(&waiter->fl_link);
665 /* Remove waiter from blocker's block list.
666 * When blocker ends up pointing to itself then the list is empty.
668 * Must be called with blocked_lock_lock held.
670 static void __locks_delete_block(struct file_lock *waiter)
672 locks_delete_global_blocked(waiter);
673 list_del_init(&waiter->fl_block);
674 waiter->fl_next = NULL;
677 static void locks_delete_block(struct file_lock *waiter)
679 spin_lock(&blocked_lock_lock);
680 __locks_delete_block(waiter);
681 spin_unlock(&blocked_lock_lock);
684 /* Insert waiter into blocker's block list.
685 * We use a circular list so that processes can be easily woken up in
686 * the order they blocked. The documentation doesn't require this but
687 * it seems like the reasonable thing to do.
689 * Must be called with both the flc_lock and blocked_lock_lock held. The
690 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
691 * that the flc_lock is also held on insertions we can avoid taking the
692 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
694 static void __locks_insert_block(struct file_lock *blocker,
695 struct file_lock *waiter)
697 BUG_ON(!list_empty(&waiter->fl_block));
698 waiter->fl_next = blocker;
699 list_add_tail(&waiter->fl_block, &blocker->fl_block);
700 if (IS_POSIX(blocker) && !IS_OFDLCK(blocker))
701 locks_insert_global_blocked(waiter);
704 /* Must be called with flc_lock held. */
705 static void locks_insert_block(struct file_lock *blocker,
706 struct file_lock *waiter)
708 spin_lock(&blocked_lock_lock);
709 __locks_insert_block(blocker, waiter);
710 spin_unlock(&blocked_lock_lock);
714 * Wake up processes blocked waiting for blocker.
716 * Must be called with the inode->flc_lock held!
718 static void locks_wake_up_blocks(struct file_lock *blocker)
721 * Avoid taking global lock if list is empty. This is safe since new
722 * blocked requests are only added to the list under the flc_lock, and
723 * the flc_lock is always held here. Note that removal from the fl_block
724 * list does not require the flc_lock, so we must recheck list_empty()
725 * after acquiring the blocked_lock_lock.
727 if (list_empty(&blocker->fl_block))
730 spin_lock(&blocked_lock_lock);
731 while (!list_empty(&blocker->fl_block)) {
732 struct file_lock *waiter;
734 waiter = list_first_entry(&blocker->fl_block,
735 struct file_lock, fl_block);
736 __locks_delete_block(waiter);
737 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
738 waiter->fl_lmops->lm_notify(waiter);
740 wake_up(&waiter->fl_wait);
742 spin_unlock(&blocked_lock_lock);
746 locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before)
748 list_add_tail(&fl->fl_list, before);
749 locks_insert_global_locks(fl);
753 locks_unlink_lock_ctx(struct file_lock *fl)
755 locks_delete_global_locks(fl);
756 list_del_init(&fl->fl_list);
757 locks_wake_up_blocks(fl);
761 locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose)
763 locks_unlink_lock_ctx(fl);
765 list_add(&fl->fl_list, dispose);
770 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
771 * checks for shared/exclusive status of overlapping locks.
773 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
775 if (sys_fl->fl_type == F_WRLCK)
777 if (caller_fl->fl_type == F_WRLCK)
782 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
783 * checking before calling the locks_conflict().
785 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
787 /* POSIX locks owned by the same process do not conflict with
790 if (posix_same_owner(caller_fl, sys_fl))
793 /* Check whether they overlap */
794 if (!locks_overlap(caller_fl, sys_fl))
797 return (locks_conflict(caller_fl, sys_fl));
800 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
801 * checking before calling the locks_conflict().
803 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
805 /* FLOCK locks referring to the same filp do not conflict with
808 if (caller_fl->fl_file == sys_fl->fl_file)
810 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
813 return (locks_conflict(caller_fl, sys_fl));
817 posix_test_lock(struct file *filp, struct file_lock *fl)
819 struct file_lock *cfl;
820 struct file_lock_context *ctx;
821 struct inode *inode = locks_inode(filp);
823 ctx = smp_load_acquire(&inode->i_flctx);
824 if (!ctx || list_empty_careful(&ctx->flc_posix)) {
825 fl->fl_type = F_UNLCK;
829 spin_lock(&ctx->flc_lock);
830 list_for_each_entry(cfl, &ctx->flc_posix, fl_list) {
831 if (posix_locks_conflict(fl, cfl)) {
832 locks_copy_conflock(fl, cfl);
836 fl->fl_type = F_UNLCK;
838 spin_unlock(&ctx->flc_lock);
841 EXPORT_SYMBOL(posix_test_lock);
844 * Deadlock detection:
846 * We attempt to detect deadlocks that are due purely to posix file
849 * We assume that a task can be waiting for at most one lock at a time.
850 * So for any acquired lock, the process holding that lock may be
851 * waiting on at most one other lock. That lock in turns may be held by
852 * someone waiting for at most one other lock. Given a requested lock
853 * caller_fl which is about to wait for a conflicting lock block_fl, we
854 * follow this chain of waiters to ensure we are not about to create a
857 * Since we do this before we ever put a process to sleep on a lock, we
858 * are ensured that there is never a cycle; that is what guarantees that
859 * the while() loop in posix_locks_deadlock() eventually completes.
861 * Note: the above assumption may not be true when handling lock
862 * requests from a broken NFS client. It may also fail in the presence
863 * of tasks (such as posix threads) sharing the same open file table.
864 * To handle those cases, we just bail out after a few iterations.
866 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
867 * Because the owner is not even nominally tied to a thread of
868 * execution, the deadlock detection below can't reasonably work well. Just
871 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
872 * locks that just checks for the case where two tasks are attempting to
873 * upgrade from read to write locks on the same inode.
876 #define MAX_DEADLK_ITERATIONS 10
878 /* Find a lock that the owner of the given block_fl is blocking on. */
879 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
881 struct file_lock *fl;
883 hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) {
884 if (posix_same_owner(fl, block_fl))
890 /* Must be called with the blocked_lock_lock held! */
891 static int posix_locks_deadlock(struct file_lock *caller_fl,
892 struct file_lock *block_fl)
896 lockdep_assert_held(&blocked_lock_lock);
899 * This deadlock detector can't reasonably detect deadlocks with
900 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
902 if (IS_OFDLCK(caller_fl))
905 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
906 if (i++ > MAX_DEADLK_ITERATIONS)
908 if (posix_same_owner(caller_fl, block_fl))
914 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
915 * after any leases, but before any posix locks.
917 * Note that if called with an FL_EXISTS argument, the caller may determine
918 * whether or not a lock was successfully freed by testing the return
921 static int flock_lock_inode(struct inode *inode, struct file_lock *request)
923 struct file_lock *new_fl = NULL;
924 struct file_lock *fl;
925 struct file_lock_context *ctx;
930 ctx = locks_get_lock_context(inode, request->fl_type);
932 if (request->fl_type != F_UNLCK)
934 return (request->fl_flags & FL_EXISTS) ? -ENOENT : 0;
937 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
938 new_fl = locks_alloc_lock();
943 percpu_down_read_preempt_disable(&file_rwsem);
944 spin_lock(&ctx->flc_lock);
945 if (request->fl_flags & FL_ACCESS)
948 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
949 if (request->fl_file != fl->fl_file)
951 if (request->fl_type == fl->fl_type)
954 locks_delete_lock_ctx(fl, &dispose);
958 if (request->fl_type == F_UNLCK) {
959 if ((request->fl_flags & FL_EXISTS) && !found)
965 list_for_each_entry(fl, &ctx->flc_flock, fl_list) {
966 if (!flock_locks_conflict(request, fl))
969 if (!(request->fl_flags & FL_SLEEP))
971 error = FILE_LOCK_DEFERRED;
972 locks_insert_block(fl, request);
975 if (request->fl_flags & FL_ACCESS)
977 locks_copy_lock(new_fl, request);
978 locks_insert_lock_ctx(new_fl, &ctx->flc_flock);
983 spin_unlock(&ctx->flc_lock);
984 percpu_up_read_preempt_enable(&file_rwsem);
986 locks_free_lock(new_fl);
987 locks_dispose_list(&dispose);
991 static int posix_lock_inode(struct inode *inode, struct file_lock *request,
992 struct file_lock *conflock)
994 struct file_lock *fl, *tmp;
995 struct file_lock *new_fl = NULL;
996 struct file_lock *new_fl2 = NULL;
997 struct file_lock *left = NULL;
998 struct file_lock *right = NULL;
999 struct file_lock_context *ctx;
1004 ctx = locks_get_lock_context(inode, request->fl_type);
1006 return (request->fl_type == F_UNLCK) ? 0 : -ENOMEM;
1009 * We may need two file_lock structures for this operation,
1010 * so we get them in advance to avoid races.
1012 * In some cases we can be sure, that no new locks will be needed
1014 if (!(request->fl_flags & FL_ACCESS) &&
1015 (request->fl_type != F_UNLCK ||
1016 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
1017 new_fl = locks_alloc_lock();
1018 new_fl2 = locks_alloc_lock();
1021 percpu_down_read_preempt_disable(&file_rwsem);
1022 spin_lock(&ctx->flc_lock);
1024 * New lock request. Walk all POSIX locks and look for conflicts. If
1025 * there are any, either return error or put the request on the
1026 * blocker's list of waiters and the global blocked_hash.
1028 if (request->fl_type != F_UNLCK) {
1029 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1030 if (!posix_locks_conflict(request, fl))
1033 locks_copy_conflock(conflock, fl);
1035 if (!(request->fl_flags & FL_SLEEP))
1038 * Deadlock detection and insertion into the blocked
1039 * locks list must be done while holding the same lock!
1042 spin_lock(&blocked_lock_lock);
1043 if (likely(!posix_locks_deadlock(request, fl))) {
1044 error = FILE_LOCK_DEFERRED;
1045 __locks_insert_block(fl, request);
1047 spin_unlock(&blocked_lock_lock);
1052 /* If we're just looking for a conflict, we're done. */
1054 if (request->fl_flags & FL_ACCESS)
1057 /* Find the first old lock with the same owner as the new lock */
1058 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1059 if (posix_same_owner(request, fl))
1063 /* Process locks with this owner. */
1064 list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) {
1065 if (!posix_same_owner(request, fl))
1068 /* Detect adjacent or overlapping regions (if same lock type) */
1069 if (request->fl_type == fl->fl_type) {
1070 /* In all comparisons of start vs end, use
1071 * "start - 1" rather than "end + 1". If end
1072 * is OFFSET_MAX, end + 1 will become negative.
1074 if (fl->fl_end < request->fl_start - 1)
1076 /* If the next lock in the list has entirely bigger
1077 * addresses than the new one, insert the lock here.
1079 if (fl->fl_start - 1 > request->fl_end)
1082 /* If we come here, the new and old lock are of the
1083 * same type and adjacent or overlapping. Make one
1084 * lock yielding from the lower start address of both
1085 * locks to the higher end address.
1087 if (fl->fl_start > request->fl_start)
1088 fl->fl_start = request->fl_start;
1090 request->fl_start = fl->fl_start;
1091 if (fl->fl_end < request->fl_end)
1092 fl->fl_end = request->fl_end;
1094 request->fl_end = fl->fl_end;
1096 locks_delete_lock_ctx(fl, &dispose);
1102 /* Processing for different lock types is a bit
1105 if (fl->fl_end < request->fl_start)
1107 if (fl->fl_start > request->fl_end)
1109 if (request->fl_type == F_UNLCK)
1111 if (fl->fl_start < request->fl_start)
1113 /* If the next lock in the list has a higher end
1114 * address than the new one, insert the new one here.
1116 if (fl->fl_end > request->fl_end) {
1120 if (fl->fl_start >= request->fl_start) {
1121 /* The new lock completely replaces an old
1122 * one (This may happen several times).
1125 locks_delete_lock_ctx(fl, &dispose);
1129 * Replace the old lock with new_fl, and
1130 * remove the old one. It's safe to do the
1131 * insert here since we know that we won't be
1132 * using new_fl later, and that the lock is
1133 * just replacing an existing lock.
1138 locks_copy_lock(new_fl, request);
1141 locks_insert_lock_ctx(request, &fl->fl_list);
1142 locks_delete_lock_ctx(fl, &dispose);
1149 * The above code only modifies existing locks in case of merging or
1150 * replacing. If new lock(s) need to be inserted all modifications are
1151 * done below this, so it's safe yet to bail out.
1153 error = -ENOLCK; /* "no luck" */
1154 if (right && left == right && !new_fl2)
1159 if (request->fl_type == F_UNLCK) {
1160 if (request->fl_flags & FL_EXISTS)
1169 locks_copy_lock(new_fl, request);
1170 locks_insert_lock_ctx(new_fl, &fl->fl_list);
1175 if (left == right) {
1176 /* The new lock breaks the old one in two pieces,
1177 * so we have to use the second new lock.
1181 locks_copy_lock(left, right);
1182 locks_insert_lock_ctx(left, &fl->fl_list);
1184 right->fl_start = request->fl_end + 1;
1185 locks_wake_up_blocks(right);
1188 left->fl_end = request->fl_start - 1;
1189 locks_wake_up_blocks(left);
1192 spin_unlock(&ctx->flc_lock);
1193 percpu_up_read_preempt_enable(&file_rwsem);
1195 * Free any unused locks.
1198 locks_free_lock(new_fl);
1200 locks_free_lock(new_fl2);
1201 locks_dispose_list(&dispose);
1202 trace_posix_lock_inode(inode, request, error);
1208 * posix_lock_file - Apply a POSIX-style lock to a file
1209 * @filp: The file to apply the lock to
1210 * @fl: The lock to be applied
1211 * @conflock: Place to return a copy of the conflicting lock, if found.
1213 * Add a POSIX style lock to a file.
1214 * We merge adjacent & overlapping locks whenever possible.
1215 * POSIX locks are sorted by owner task, then by starting address
1217 * Note that if called with an FL_EXISTS argument, the caller may determine
1218 * whether or not a lock was successfully freed by testing the return
1219 * value for -ENOENT.
1221 int posix_lock_file(struct file *filp, struct file_lock *fl,
1222 struct file_lock *conflock)
1224 return posix_lock_inode(locks_inode(filp), fl, conflock);
1226 EXPORT_SYMBOL(posix_lock_file);
1229 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1230 * @inode: inode of file to which lock request should be applied
1231 * @fl: The lock to be applied
1233 * Apply a POSIX style lock request to an inode.
1235 static int posix_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1240 error = posix_lock_inode(inode, fl, NULL);
1241 if (error != FILE_LOCK_DEFERRED)
1243 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1247 locks_delete_block(fl);
1253 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1255 * locks_mandatory_locked - Check for an active lock
1256 * @file: the file to check
1258 * Searches the inode's list of locks to find any POSIX locks which conflict.
1259 * This function is called from locks_verify_locked() only.
1261 int locks_mandatory_locked(struct file *file)
1264 struct inode *inode = locks_inode(file);
1265 struct file_lock_context *ctx;
1266 struct file_lock *fl;
1268 ctx = smp_load_acquire(&inode->i_flctx);
1269 if (!ctx || list_empty_careful(&ctx->flc_posix))
1273 * Search the lock list for this inode for any POSIX locks.
1275 spin_lock(&ctx->flc_lock);
1277 list_for_each_entry(fl, &ctx->flc_posix, fl_list) {
1278 if (fl->fl_owner != current->files &&
1279 fl->fl_owner != file) {
1284 spin_unlock(&ctx->flc_lock);
1289 * locks_mandatory_area - Check for a conflicting lock
1290 * @inode: the file to check
1291 * @filp: how the file was opened (if it was)
1292 * @start: first byte in the file to check
1293 * @end: lastbyte in the file to check
1294 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1296 * Searches the inode's list of locks to find any POSIX locks which conflict.
1298 int locks_mandatory_area(struct inode *inode, struct file *filp, loff_t start,
1299 loff_t end, unsigned char type)
1301 struct file_lock fl;
1305 locks_init_lock(&fl);
1306 fl.fl_pid = current->tgid;
1308 fl.fl_flags = FL_POSIX | FL_ACCESS;
1309 if (filp && !(filp->f_flags & O_NONBLOCK))
1312 fl.fl_start = start;
1318 fl.fl_flags &= ~FL_SLEEP;
1319 error = posix_lock_inode(inode, &fl, NULL);
1325 fl.fl_flags |= FL_SLEEP;
1326 fl.fl_owner = current->files;
1327 error = posix_lock_inode(inode, &fl, NULL);
1328 if (error != FILE_LOCK_DEFERRED)
1330 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1333 * If we've been sleeping someone might have
1334 * changed the permissions behind our back.
1336 if (__mandatory_lock(inode))
1340 locks_delete_block(&fl);
1347 EXPORT_SYMBOL(locks_mandatory_area);
1348 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1350 static void lease_clear_pending(struct file_lock *fl, int arg)
1354 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1357 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1361 /* We already had a lease on this file; just change its type */
1362 int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose)
1364 int error = assign_type(fl, arg);
1368 lease_clear_pending(fl, arg);
1369 locks_wake_up_blocks(fl);
1370 if (arg == F_UNLCK) {
1371 struct file *filp = fl->fl_file;
1374 filp->f_owner.signum = 0;
1375 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1376 if (fl->fl_fasync != NULL) {
1377 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1378 fl->fl_fasync = NULL;
1380 locks_delete_lock_ctx(fl, dispose);
1384 EXPORT_SYMBOL(lease_modify);
1386 static bool past_time(unsigned long then)
1389 /* 0 is a special value meaning "this never expires": */
1391 return time_after(jiffies, then);
1394 static void time_out_leases(struct inode *inode, struct list_head *dispose)
1396 struct file_lock_context *ctx = inode->i_flctx;
1397 struct file_lock *fl, *tmp;
1399 lockdep_assert_held(&ctx->flc_lock);
1401 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1402 trace_time_out_leases(inode, fl);
1403 if (past_time(fl->fl_downgrade_time))
1404 lease_modify(fl, F_RDLCK, dispose);
1405 if (past_time(fl->fl_break_time))
1406 lease_modify(fl, F_UNLCK, dispose);
1410 static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker)
1412 if ((breaker->fl_flags & FL_LAYOUT) != (lease->fl_flags & FL_LAYOUT))
1414 if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE))
1416 return locks_conflict(breaker, lease);
1420 any_leases_conflict(struct inode *inode, struct file_lock *breaker)
1422 struct file_lock_context *ctx = inode->i_flctx;
1423 struct file_lock *fl;
1425 lockdep_assert_held(&ctx->flc_lock);
1427 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1428 if (leases_conflict(fl, breaker))
1435 * __break_lease - revoke all outstanding leases on file
1436 * @inode: the inode of the file to return
1437 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1439 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1442 * break_lease (inlined for speed) has checked there already is at least
1443 * some kind of lock (maybe a lease) on this file. Leases are broken on
1444 * a call to open() or truncate(). This function can sleep unless you
1445 * specified %O_NONBLOCK to your open().
1447 int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1450 struct file_lock_context *ctx;
1451 struct file_lock *new_fl, *fl, *tmp;
1452 unsigned long break_time;
1453 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1456 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1458 return PTR_ERR(new_fl);
1459 new_fl->fl_flags = type;
1461 /* typically we will check that ctx is non-NULL before calling */
1462 ctx = smp_load_acquire(&inode->i_flctx);
1468 percpu_down_read_preempt_disable(&file_rwsem);
1469 spin_lock(&ctx->flc_lock);
1471 time_out_leases(inode, &dispose);
1473 if (!any_leases_conflict(inode, new_fl))
1477 if (lease_break_time > 0) {
1478 break_time = jiffies + lease_break_time * HZ;
1479 if (break_time == 0)
1480 break_time++; /* so that 0 means no break time */
1483 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) {
1484 if (!leases_conflict(fl, new_fl))
1487 if (fl->fl_flags & FL_UNLOCK_PENDING)
1489 fl->fl_flags |= FL_UNLOCK_PENDING;
1490 fl->fl_break_time = break_time;
1492 if (lease_breaking(fl))
1494 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1495 fl->fl_downgrade_time = break_time;
1497 if (fl->fl_lmops->lm_break(fl))
1498 locks_delete_lock_ctx(fl, &dispose);
1501 if (list_empty(&ctx->flc_lease))
1504 if (mode & O_NONBLOCK) {
1505 trace_break_lease_noblock(inode, new_fl);
1506 error = -EWOULDBLOCK;
1511 fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list);
1512 break_time = fl->fl_break_time;
1513 if (break_time != 0)
1514 break_time -= jiffies;
1515 if (break_time == 0)
1517 locks_insert_block(fl, new_fl);
1518 trace_break_lease_block(inode, new_fl);
1519 spin_unlock(&ctx->flc_lock);
1520 percpu_up_read_preempt_enable(&file_rwsem);
1522 locks_dispose_list(&dispose);
1523 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1524 !new_fl->fl_next, break_time);
1526 percpu_down_read_preempt_disable(&file_rwsem);
1527 spin_lock(&ctx->flc_lock);
1528 trace_break_lease_unblock(inode, new_fl);
1529 locks_delete_block(new_fl);
1532 * Wait for the next conflicting lease that has not been
1536 time_out_leases(inode, &dispose);
1537 if (any_leases_conflict(inode, new_fl))
1542 spin_unlock(&ctx->flc_lock);
1543 percpu_up_read_preempt_enable(&file_rwsem);
1544 locks_dispose_list(&dispose);
1545 locks_free_lock(new_fl);
1549 EXPORT_SYMBOL(__break_lease);
1552 * lease_get_mtime - update modified time of an inode with exclusive lease
1554 * @time: pointer to a timespec which contains the last modified time
1556 * This is to force NFS clients to flush their caches for files with
1557 * exclusive leases. The justification is that if someone has an
1558 * exclusive lease, then they could be modifying it.
1560 void lease_get_mtime(struct inode *inode, struct timespec64 *time)
1562 bool has_lease = false;
1563 struct file_lock_context *ctx;
1564 struct file_lock *fl;
1566 ctx = smp_load_acquire(&inode->i_flctx);
1567 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1568 spin_lock(&ctx->flc_lock);
1569 fl = list_first_entry_or_null(&ctx->flc_lease,
1570 struct file_lock, fl_list);
1571 if (fl && (fl->fl_type == F_WRLCK))
1573 spin_unlock(&ctx->flc_lock);
1577 *time = current_time(inode);
1580 EXPORT_SYMBOL(lease_get_mtime);
1583 * fcntl_getlease - Enquire what lease is currently active
1586 * The value returned by this function will be one of
1587 * (if no lease break is pending):
1589 * %F_RDLCK to indicate a shared lease is held.
1591 * %F_WRLCK to indicate an exclusive lease is held.
1593 * %F_UNLCK to indicate no lease is held.
1595 * (if a lease break is pending):
1597 * %F_RDLCK to indicate an exclusive lease needs to be
1598 * changed to a shared lease (or removed).
1600 * %F_UNLCK to indicate the lease needs to be removed.
1602 * XXX: sfr & willy disagree over whether F_INPROGRESS
1603 * should be returned to userspace.
1605 int fcntl_getlease(struct file *filp)
1607 struct file_lock *fl;
1608 struct inode *inode = locks_inode(filp);
1609 struct file_lock_context *ctx;
1613 ctx = smp_load_acquire(&inode->i_flctx);
1614 if (ctx && !list_empty_careful(&ctx->flc_lease)) {
1615 percpu_down_read_preempt_disable(&file_rwsem);
1616 spin_lock(&ctx->flc_lock);
1617 time_out_leases(inode, &dispose);
1618 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1619 if (fl->fl_file != filp)
1621 type = target_leasetype(fl);
1624 spin_unlock(&ctx->flc_lock);
1625 percpu_up_read_preempt_enable(&file_rwsem);
1627 locks_dispose_list(&dispose);
1633 * check_conflicting_open - see if the given dentry points to a file that has
1634 * an existing open that would conflict with the
1636 * @dentry: dentry to check
1637 * @arg: type of lease that we're trying to acquire
1638 * @flags: current lock flags
1640 * Check to see if there's an existing open fd on this file that would
1641 * conflict with the lease we're trying to set.
1644 check_conflicting_open(const struct dentry *dentry, const long arg, int flags)
1647 struct inode *inode = dentry->d_inode;
1649 if (flags & FL_LAYOUT)
1652 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1655 if ((arg == F_WRLCK) && ((d_count(dentry) > 1) ||
1656 (atomic_read(&inode->i_count) > 1)))
1663 generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv)
1665 struct file_lock *fl, *my_fl = NULL, *lease;
1666 struct dentry *dentry = filp->f_path.dentry;
1667 struct inode *inode = dentry->d_inode;
1668 struct file_lock_context *ctx;
1669 bool is_deleg = (*flp)->fl_flags & FL_DELEG;
1674 trace_generic_add_lease(inode, lease);
1676 /* Note that arg is never F_UNLCK here */
1677 ctx = locks_get_lock_context(inode, arg);
1682 * In the delegation case we need mutual exclusion with
1683 * a number of operations that take the i_mutex. We trylock
1684 * because delegations are an optional optimization, and if
1685 * there's some chance of a conflict--we'd rather not
1686 * bother, maybe that's a sign this just isn't a good file to
1687 * hand out a delegation on.
1689 if (is_deleg && !inode_trylock(inode))
1692 if (is_deleg && arg == F_WRLCK) {
1693 /* Write delegations are not currently supported: */
1694 inode_unlock(inode);
1699 percpu_down_read_preempt_disable(&file_rwsem);
1700 spin_lock(&ctx->flc_lock);
1701 time_out_leases(inode, &dispose);
1702 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1707 * At this point, we know that if there is an exclusive
1708 * lease on this file, then we hold it on this filp
1709 * (otherwise our open of this file would have blocked).
1710 * And if we are trying to acquire an exclusive lease,
1711 * then the file is not open by anyone (including us)
1712 * except for this filp.
1715 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1716 if (fl->fl_file == filp &&
1717 fl->fl_owner == lease->fl_owner) {
1723 * No exclusive leases if someone else has a lease on
1729 * Modifying our existing lease is OK, but no getting a
1730 * new lease if someone else is opening for write:
1732 if (fl->fl_flags & FL_UNLOCK_PENDING)
1736 if (my_fl != NULL) {
1738 error = lease->fl_lmops->lm_change(lease, arg, &dispose);
1748 locks_insert_lock_ctx(lease, &ctx->flc_lease);
1750 * The check in break_lease() is lockless. It's possible for another
1751 * open to race in after we did the earlier check for a conflicting
1752 * open but before the lease was inserted. Check again for a
1753 * conflicting open and cancel the lease if there is one.
1755 * We also add a barrier here to ensure that the insertion of the lock
1756 * precedes these checks.
1759 error = check_conflicting_open(dentry, arg, lease->fl_flags);
1761 locks_unlink_lock_ctx(lease);
1766 if (lease->fl_lmops->lm_setup)
1767 lease->fl_lmops->lm_setup(lease, priv);
1769 spin_unlock(&ctx->flc_lock);
1770 percpu_up_read_preempt_enable(&file_rwsem);
1771 locks_dispose_list(&dispose);
1773 inode_unlock(inode);
1774 if (!error && !my_fl)
1779 static int generic_delete_lease(struct file *filp, void *owner)
1781 int error = -EAGAIN;
1782 struct file_lock *fl, *victim = NULL;
1783 struct inode *inode = locks_inode(filp);
1784 struct file_lock_context *ctx;
1787 ctx = smp_load_acquire(&inode->i_flctx);
1789 trace_generic_delete_lease(inode, NULL);
1793 percpu_down_read_preempt_disable(&file_rwsem);
1794 spin_lock(&ctx->flc_lock);
1795 list_for_each_entry(fl, &ctx->flc_lease, fl_list) {
1796 if (fl->fl_file == filp &&
1797 fl->fl_owner == owner) {
1802 trace_generic_delete_lease(inode, victim);
1804 error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose);
1805 spin_unlock(&ctx->flc_lock);
1806 percpu_up_read_preempt_enable(&file_rwsem);
1807 locks_dispose_list(&dispose);
1812 * generic_setlease - sets a lease on an open file
1813 * @filp: file pointer
1814 * @arg: type of lease to obtain
1815 * @flp: input - file_lock to use, output - file_lock inserted
1816 * @priv: private data for lm_setup (may be NULL if lm_setup
1817 * doesn't require it)
1819 * The (input) flp->fl_lmops->lm_break function is required
1822 int generic_setlease(struct file *filp, long arg, struct file_lock **flp,
1825 struct inode *inode = locks_inode(filp);
1828 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1830 if (!S_ISREG(inode->i_mode))
1832 error = security_file_lock(filp, arg);
1838 return generic_delete_lease(filp, *priv);
1841 if (!(*flp)->fl_lmops->lm_break) {
1846 return generic_add_lease(filp, arg, flp, priv);
1851 EXPORT_SYMBOL(generic_setlease);
1854 * vfs_setlease - sets a lease on an open file
1855 * @filp: file pointer
1856 * @arg: type of lease to obtain
1857 * @lease: file_lock to use when adding a lease
1858 * @priv: private info for lm_setup when adding a lease (may be
1859 * NULL if lm_setup doesn't require it)
1861 * Call this to establish a lease on the file. The "lease" argument is not
1862 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1863 * an existing lease, the ``(*lease)->fl_lmops->lm_break`` operation must be
1864 * set; if not, this function will return -ENOLCK (and generate a scary-looking
1867 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1868 * may be NULL if the lm_setup operation doesn't require it.
1871 vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv)
1873 if (filp->f_op->setlease)
1874 return filp->f_op->setlease(filp, arg, lease, priv);
1876 return generic_setlease(filp, arg, lease, priv);
1878 EXPORT_SYMBOL_GPL(vfs_setlease);
1880 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1882 struct file_lock *fl;
1883 struct fasync_struct *new;
1886 fl = lease_alloc(filp, arg);
1890 new = fasync_alloc();
1892 locks_free_lock(fl);
1897 error = vfs_setlease(filp, arg, &fl, (void **)&new);
1899 locks_free_lock(fl);
1906 * fcntl_setlease - sets a lease on an open file
1907 * @fd: open file descriptor
1908 * @filp: file pointer
1909 * @arg: type of lease to obtain
1911 * Call this fcntl to establish a lease on the file.
1912 * Note that you also need to call %F_SETSIG to
1913 * receive a signal when the lease is broken.
1915 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1918 return vfs_setlease(filp, F_UNLCK, NULL, (void **)&filp);
1919 return do_fcntl_add_lease(fd, filp, arg);
1923 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1924 * @inode: inode of the file to apply to
1925 * @fl: The lock to be applied
1927 * Apply a FLOCK style lock request to an inode.
1929 static int flock_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1934 error = flock_lock_inode(inode, fl);
1935 if (error != FILE_LOCK_DEFERRED)
1937 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1941 locks_delete_block(fl);
1948 * locks_lock_inode_wait - Apply a lock to an inode
1949 * @inode: inode of the file to apply to
1950 * @fl: The lock to be applied
1952 * Apply a POSIX or FLOCK style lock request to an inode.
1954 int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1957 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
1959 res = posix_lock_inode_wait(inode, fl);
1962 res = flock_lock_inode_wait(inode, fl);
1969 EXPORT_SYMBOL(locks_lock_inode_wait);
1972 * sys_flock: - flock() system call.
1973 * @fd: the file descriptor to lock.
1974 * @cmd: the type of lock to apply.
1976 * Apply a %FL_FLOCK style lock to an open file descriptor.
1977 * The @cmd can be one of:
1979 * - %LOCK_SH -- a shared lock.
1980 * - %LOCK_EX -- an exclusive lock.
1981 * - %LOCK_UN -- remove an existing lock.
1982 * - %LOCK_MAND -- a 'mandatory' flock.
1983 * This exists to emulate Windows Share Modes.
1985 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1986 * processes read and write access respectively.
1988 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1990 struct fd f = fdget(fd);
1991 struct file_lock *lock;
1992 int can_sleep, unlock;
1999 can_sleep = !(cmd & LOCK_NB);
2001 unlock = (cmd == LOCK_UN);
2003 if (!unlock && !(cmd & LOCK_MAND) &&
2004 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
2007 lock = flock_make_lock(f.file, cmd);
2009 error = PTR_ERR(lock);
2014 lock->fl_flags |= FL_SLEEP;
2016 error = security_file_lock(f.file, lock->fl_type);
2020 if (f.file->f_op->flock)
2021 error = f.file->f_op->flock(f.file,
2022 (can_sleep) ? F_SETLKW : F_SETLK,
2025 error = locks_lock_file_wait(f.file, lock);
2028 locks_free_lock(lock);
2037 * vfs_test_lock - test file byte range lock
2038 * @filp: The file to test lock for
2039 * @fl: The lock to test; also used to hold result
2041 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2042 * setting conf->fl_type to something other than F_UNLCK.
2044 int vfs_test_lock(struct file *filp, struct file_lock *fl)
2046 if (filp->f_op->lock)
2047 return filp->f_op->lock(filp, F_GETLK, fl);
2048 posix_test_lock(filp, fl);
2051 EXPORT_SYMBOL_GPL(vfs_test_lock);
2054 * locks_translate_pid - translate a file_lock's fl_pid number into a namespace
2055 * @fl: The file_lock who's fl_pid should be translated
2056 * @ns: The namespace into which the pid should be translated
2058 * Used to tranlate a fl_pid into a namespace virtual pid number
2060 static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns)
2067 if (IS_REMOTELCK(fl))
2071 pid = find_pid_ns(fl->fl_pid, &init_pid_ns);
2072 vnr = pid_nr_ns(pid, ns);
2077 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
2079 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2080 #if BITS_PER_LONG == 32
2082 * Make sure we can represent the posix lock via
2083 * legacy 32bit flock.
2085 if (fl->fl_start > OFFT_OFFSET_MAX)
2087 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
2090 flock->l_start = fl->fl_start;
2091 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2092 fl->fl_end - fl->fl_start + 1;
2093 flock->l_whence = 0;
2094 flock->l_type = fl->fl_type;
2098 #if BITS_PER_LONG == 32
2099 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
2101 flock->l_pid = locks_translate_pid(fl, task_active_pid_ns(current));
2102 flock->l_start = fl->fl_start;
2103 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
2104 fl->fl_end - fl->fl_start + 1;
2105 flock->l_whence = 0;
2106 flock->l_type = fl->fl_type;
2110 /* Report the first existing lock that would conflict with l.
2111 * This implements the F_GETLK command of fcntl().
2113 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock *flock)
2115 struct file_lock *fl;
2118 fl = locks_alloc_lock();
2122 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2125 error = flock_to_posix_lock(filp, fl, flock);
2129 if (cmd == F_OFD_GETLK) {
2131 if (flock->l_pid != 0)
2135 fl->fl_flags |= FL_OFDLCK;
2136 fl->fl_owner = filp;
2139 error = vfs_test_lock(filp, fl);
2143 flock->l_type = fl->fl_type;
2144 if (fl->fl_type != F_UNLCK) {
2145 error = posix_lock_to_flock(flock, fl);
2150 locks_free_lock(fl);
2155 * vfs_lock_file - file byte range lock
2156 * @filp: The file to apply the lock to
2157 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2158 * @fl: The lock to be applied
2159 * @conf: Place to return a copy of the conflicting lock, if found.
2161 * A caller that doesn't care about the conflicting lock may pass NULL
2162 * as the final argument.
2164 * If the filesystem defines a private ->lock() method, then @conf will
2165 * be left unchanged; so a caller that cares should initialize it to
2166 * some acceptable default.
2168 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2169 * locks, the ->lock() interface may return asynchronously, before the lock has
2170 * been granted or denied by the underlying filesystem, if (and only if)
2171 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2172 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2173 * the request is for a blocking lock. When ->lock() does return asynchronously,
2174 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2175 * request completes.
2176 * If the request is for non-blocking lock the file system should return
2177 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2178 * with the result. If the request timed out the callback routine will return a
2179 * nonzero return code and the file system should release the lock. The file
2180 * system is also responsible to keep a corresponding posix lock when it
2181 * grants a lock so the VFS can find out which locks are locally held and do
2182 * the correct lock cleanup when required.
2183 * The underlying filesystem must not drop the kernel lock or call
2184 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2187 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
2189 if (filp->f_op->lock)
2190 return filp->f_op->lock(filp, cmd, fl);
2192 return posix_lock_file(filp, fl, conf);
2194 EXPORT_SYMBOL_GPL(vfs_lock_file);
2196 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
2197 struct file_lock *fl)
2201 error = security_file_lock(filp, fl->fl_type);
2206 error = vfs_lock_file(filp, cmd, fl, NULL);
2207 if (error != FILE_LOCK_DEFERRED)
2209 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
2213 locks_delete_block(fl);
2220 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2222 check_fmode_for_setlk(struct file_lock *fl)
2224 switch (fl->fl_type) {
2226 if (!(fl->fl_file->f_mode & FMODE_READ))
2230 if (!(fl->fl_file->f_mode & FMODE_WRITE))
2236 /* Apply the lock described by l to an open file descriptor.
2237 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2239 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
2240 struct flock *flock)
2242 struct file_lock *file_lock = locks_alloc_lock();
2243 struct inode *inode = locks_inode(filp);
2247 if (file_lock == NULL)
2250 /* Don't allow mandatory locks on files that may be memory mapped
2253 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2258 error = flock_to_posix_lock(filp, file_lock, flock);
2262 error = check_fmode_for_setlk(file_lock);
2267 * If the cmd is requesting file-private locks, then set the
2268 * FL_OFDLCK flag and override the owner.
2273 if (flock->l_pid != 0)
2277 file_lock->fl_flags |= FL_OFDLCK;
2278 file_lock->fl_owner = filp;
2282 if (flock->l_pid != 0)
2286 file_lock->fl_flags |= FL_OFDLCK;
2287 file_lock->fl_owner = filp;
2290 file_lock->fl_flags |= FL_SLEEP;
2293 error = do_lock_file_wait(filp, cmd, file_lock);
2296 * Attempt to detect a close/fcntl race and recover by releasing the
2297 * lock that was just acquired. There is no need to do that when we're
2298 * unlocking though, or for OFD locks.
2300 if (!error && file_lock->fl_type != F_UNLCK &&
2301 !(file_lock->fl_flags & FL_OFDLCK)) {
2303 * We need that spin_lock here - it prevents reordering between
2304 * update of i_flctx->flc_posix and check for it done in
2305 * close(). rcu_read_lock() wouldn't do.
2307 spin_lock(¤t->files->file_lock);
2309 spin_unlock(¤t->files->file_lock);
2311 file_lock->fl_type = F_UNLCK;
2312 error = do_lock_file_wait(filp, cmd, file_lock);
2313 WARN_ON_ONCE(error);
2318 trace_fcntl_setlk(inode, file_lock, error);
2319 locks_free_lock(file_lock);
2323 #if BITS_PER_LONG == 32
2324 /* Report the first existing lock that would conflict with l.
2325 * This implements the F_GETLK command of fcntl().
2327 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 *flock)
2329 struct file_lock *fl;
2332 fl = locks_alloc_lock();
2337 if (flock->l_type != F_RDLCK && flock->l_type != F_WRLCK)
2340 error = flock64_to_posix_lock(filp, fl, flock);
2344 if (cmd == F_OFD_GETLK) {
2346 if (flock->l_pid != 0)
2350 fl->fl_flags |= FL_OFDLCK;
2351 fl->fl_owner = filp;
2354 error = vfs_test_lock(filp, fl);
2358 flock->l_type = fl->fl_type;
2359 if (fl->fl_type != F_UNLCK)
2360 posix_lock_to_flock64(flock, fl);
2363 locks_free_lock(fl);
2367 /* Apply the lock described by l to an open file descriptor.
2368 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2370 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
2371 struct flock64 *flock)
2373 struct file_lock *file_lock = locks_alloc_lock();
2374 struct inode *inode = locks_inode(filp);
2378 if (file_lock == NULL)
2381 /* Don't allow mandatory locks on files that may be memory mapped
2384 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2389 error = flock64_to_posix_lock(filp, file_lock, flock);
2393 error = check_fmode_for_setlk(file_lock);
2398 * If the cmd is requesting file-private locks, then set the
2399 * FL_OFDLCK flag and override the owner.
2404 if (flock->l_pid != 0)
2408 file_lock->fl_flags |= FL_OFDLCK;
2409 file_lock->fl_owner = filp;
2413 if (flock->l_pid != 0)
2417 file_lock->fl_flags |= FL_OFDLCK;
2418 file_lock->fl_owner = filp;
2421 file_lock->fl_flags |= FL_SLEEP;
2424 error = do_lock_file_wait(filp, cmd, file_lock);
2427 * Attempt to detect a close/fcntl race and recover by releasing the
2428 * lock that was just acquired. There is no need to do that when we're
2429 * unlocking though, or for OFD locks.
2431 if (!error && file_lock->fl_type != F_UNLCK &&
2432 !(file_lock->fl_flags & FL_OFDLCK)) {
2434 * We need that spin_lock here - it prevents reordering between
2435 * update of i_flctx->flc_posix and check for it done in
2436 * close(). rcu_read_lock() wouldn't do.
2438 spin_lock(¤t->files->file_lock);
2440 spin_unlock(¤t->files->file_lock);
2442 file_lock->fl_type = F_UNLCK;
2443 error = do_lock_file_wait(filp, cmd, file_lock);
2444 WARN_ON_ONCE(error);
2449 locks_free_lock(file_lock);
2452 #endif /* BITS_PER_LONG == 32 */
2455 * This function is called when the file is being removed
2456 * from the task's fd array. POSIX locks belonging to this task
2457 * are deleted at this time.
2459 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2462 struct inode *inode = locks_inode(filp);
2463 struct file_lock lock;
2464 struct file_lock_context *ctx;
2467 * If there are no locks held on this file, we don't need to call
2468 * posix_lock_file(). Another process could be setting a lock on this
2469 * file at the same time, but we wouldn't remove that lock anyway.
2471 ctx = smp_load_acquire(&inode->i_flctx);
2472 if (!ctx || list_empty(&ctx->flc_posix))
2475 lock.fl_type = F_UNLCK;
2476 lock.fl_flags = FL_POSIX | FL_CLOSE;
2478 lock.fl_end = OFFSET_MAX;
2479 lock.fl_owner = owner;
2480 lock.fl_pid = current->tgid;
2481 lock.fl_file = filp;
2483 lock.fl_lmops = NULL;
2485 error = vfs_lock_file(filp, F_SETLK, &lock, NULL);
2487 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2488 lock.fl_ops->fl_release_private(&lock);
2489 trace_locks_remove_posix(inode, &lock, error);
2492 EXPORT_SYMBOL(locks_remove_posix);
2494 /* The i_flctx must be valid when calling into here */
2496 locks_remove_flock(struct file *filp, struct file_lock_context *flctx)
2498 struct file_lock fl = {
2500 .fl_pid = current->tgid,
2502 .fl_flags = FL_FLOCK | FL_CLOSE,
2504 .fl_end = OFFSET_MAX,
2506 struct inode *inode = locks_inode(filp);
2508 if (list_empty(&flctx->flc_flock))
2511 if (filp->f_op->flock)
2512 filp->f_op->flock(filp, F_SETLKW, &fl);
2514 flock_lock_inode(inode, &fl);
2516 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2517 fl.fl_ops->fl_release_private(&fl);
2520 /* The i_flctx must be valid when calling into here */
2522 locks_remove_lease(struct file *filp, struct file_lock_context *ctx)
2524 struct file_lock *fl, *tmp;
2527 if (list_empty(&ctx->flc_lease))
2530 percpu_down_read_preempt_disable(&file_rwsem);
2531 spin_lock(&ctx->flc_lock);
2532 list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list)
2533 if (filp == fl->fl_file)
2534 lease_modify(fl, F_UNLCK, &dispose);
2535 spin_unlock(&ctx->flc_lock);
2536 percpu_up_read_preempt_enable(&file_rwsem);
2538 locks_dispose_list(&dispose);
2542 * This function is called on the last close of an open file.
2544 void locks_remove_file(struct file *filp)
2546 struct file_lock_context *ctx;
2548 ctx = smp_load_acquire(&locks_inode(filp)->i_flctx);
2552 /* remove any OFD locks */
2553 locks_remove_posix(filp, filp);
2555 /* remove flock locks */
2556 locks_remove_flock(filp, ctx);
2558 /* remove any leases */
2559 locks_remove_lease(filp, ctx);
2561 spin_lock(&ctx->flc_lock);
2562 locks_check_ctx_file_list(filp, &ctx->flc_posix, "POSIX");
2563 locks_check_ctx_file_list(filp, &ctx->flc_flock, "FLOCK");
2564 locks_check_ctx_file_list(filp, &ctx->flc_lease, "LEASE");
2565 spin_unlock(&ctx->flc_lock);
2569 * posix_unblock_lock - stop waiting for a file lock
2570 * @waiter: the lock which was waiting
2572 * lockd needs to block waiting for locks.
2575 posix_unblock_lock(struct file_lock *waiter)
2579 spin_lock(&blocked_lock_lock);
2580 if (waiter->fl_next)
2581 __locks_delete_block(waiter);
2584 spin_unlock(&blocked_lock_lock);
2587 EXPORT_SYMBOL(posix_unblock_lock);
2590 * vfs_cancel_lock - file byte range unblock lock
2591 * @filp: The file to apply the unblock to
2592 * @fl: The lock to be unblocked
2594 * Used by lock managers to cancel blocked requests
2596 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2598 if (filp->f_op->lock)
2599 return filp->f_op->lock(filp, F_CANCELLK, fl);
2603 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2605 #ifdef CONFIG_PROC_FS
2606 #include <linux/proc_fs.h>
2607 #include <linux/seq_file.h>
2609 struct locks_iterator {
2614 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2615 loff_t id, char *pfx)
2617 struct inode *inode = NULL;
2618 unsigned int fl_pid;
2619 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2621 fl_pid = locks_translate_pid(fl, proc_pidns);
2623 * If there isn't a fl_pid don't display who is waiting on
2624 * the lock if we are called from locks_show, or if we are
2625 * called from __show_fd_info - skip lock entirely
2630 if (fl->fl_file != NULL)
2631 inode = locks_inode(fl->fl_file);
2633 seq_printf(f, "%lld:%s ", id, pfx);
2635 if (fl->fl_flags & FL_ACCESS)
2636 seq_puts(f, "ACCESS");
2637 else if (IS_OFDLCK(fl))
2638 seq_puts(f, "OFDLCK");
2640 seq_puts(f, "POSIX ");
2642 seq_printf(f, " %s ",
2643 (inode == NULL) ? "*NOINODE*" :
2644 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2645 } else if (IS_FLOCK(fl)) {
2646 if (fl->fl_type & LOCK_MAND) {
2647 seq_puts(f, "FLOCK MSNFS ");
2649 seq_puts(f, "FLOCK ADVISORY ");
2651 } else if (IS_LEASE(fl)) {
2652 if (fl->fl_flags & FL_DELEG)
2653 seq_puts(f, "DELEG ");
2655 seq_puts(f, "LEASE ");
2657 if (lease_breaking(fl))
2658 seq_puts(f, "BREAKING ");
2659 else if (fl->fl_file)
2660 seq_puts(f, "ACTIVE ");
2662 seq_puts(f, "BREAKER ");
2664 seq_puts(f, "UNKNOWN UNKNOWN ");
2666 if (fl->fl_type & LOCK_MAND) {
2667 seq_printf(f, "%s ",
2668 (fl->fl_type & LOCK_READ)
2669 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2670 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2672 seq_printf(f, "%s ",
2673 (lease_breaking(fl))
2674 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2675 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2678 /* userspace relies on this representation of dev_t */
2679 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2680 MAJOR(inode->i_sb->s_dev),
2681 MINOR(inode->i_sb->s_dev), inode->i_ino);
2683 seq_printf(f, "%d <none>:0 ", fl_pid);
2686 if (fl->fl_end == OFFSET_MAX)
2687 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2689 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2691 seq_puts(f, "0 EOF\n");
2695 static int locks_show(struct seq_file *f, void *v)
2697 struct locks_iterator *iter = f->private;
2698 struct file_lock *fl, *bfl;
2699 struct pid_namespace *proc_pidns = file_inode(f->file)->i_sb->s_fs_info;
2701 fl = hlist_entry(v, struct file_lock, fl_link);
2703 if (locks_translate_pid(fl, proc_pidns) == 0)
2706 lock_get_status(f, fl, iter->li_pos, "");
2708 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2709 lock_get_status(f, bfl, iter->li_pos, " ->");
2714 static void __show_fd_locks(struct seq_file *f,
2715 struct list_head *head, int *id,
2716 struct file *filp, struct files_struct *files)
2718 struct file_lock *fl;
2720 list_for_each_entry(fl, head, fl_list) {
2722 if (filp != fl->fl_file)
2724 if (fl->fl_owner != files &&
2725 fl->fl_owner != filp)
2729 seq_puts(f, "lock:\t");
2730 lock_get_status(f, fl, *id, "");
2734 void show_fd_locks(struct seq_file *f,
2735 struct file *filp, struct files_struct *files)
2737 struct inode *inode = locks_inode(filp);
2738 struct file_lock_context *ctx;
2741 ctx = smp_load_acquire(&inode->i_flctx);
2745 spin_lock(&ctx->flc_lock);
2746 __show_fd_locks(f, &ctx->flc_flock, &id, filp, files);
2747 __show_fd_locks(f, &ctx->flc_posix, &id, filp, files);
2748 __show_fd_locks(f, &ctx->flc_lease, &id, filp, files);
2749 spin_unlock(&ctx->flc_lock);
2752 static void *locks_start(struct seq_file *f, loff_t *pos)
2753 __acquires(&blocked_lock_lock)
2755 struct locks_iterator *iter = f->private;
2757 iter->li_pos = *pos + 1;
2758 percpu_down_write(&file_rwsem);
2759 spin_lock(&blocked_lock_lock);
2760 return seq_hlist_start_percpu(&file_lock_list.hlist, &iter->li_cpu, *pos);
2763 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2765 struct locks_iterator *iter = f->private;
2768 return seq_hlist_next_percpu(v, &file_lock_list.hlist, &iter->li_cpu, pos);
2771 static void locks_stop(struct seq_file *f, void *v)
2772 __releases(&blocked_lock_lock)
2774 spin_unlock(&blocked_lock_lock);
2775 percpu_up_write(&file_rwsem);
2778 static const struct seq_operations locks_seq_operations = {
2779 .start = locks_start,
2785 static int __init proc_locks_init(void)
2787 proc_create_seq_private("locks", 0, NULL, &locks_seq_operations,
2788 sizeof(struct locks_iterator), NULL);
2791 fs_initcall(proc_locks_init);
2794 static int __init filelock_init(void)
2798 flctx_cache = kmem_cache_create("file_lock_ctx",
2799 sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL);
2801 filelock_cache = kmem_cache_create("file_lock_cache",
2802 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2805 for_each_possible_cpu(i) {
2806 struct file_lock_list_struct *fll = per_cpu_ptr(&file_lock_list, i);
2808 spin_lock_init(&fll->lock);
2809 INIT_HLIST_HEAD(&fll->hlist);
2815 core_initcall(filelock_init);