4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
7 * Deadlock detection added.
8 * FIXME: one thing isn't handled yet:
9 * - mandatory locks (requires lots of changes elsewhere)
10 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
15 * Converted file_lock_table to a linked list from an array, which eliminates
16 * the limits on how many active file locks are open.
17 * Chad Page (pageone@netcom.com), November 27, 1994
19 * Removed dependency on file descriptors. dup()'ed file descriptors now
20 * get the same locks as the original file descriptors, and a close() on
21 * any file descriptor removes ALL the locks on the file for the current
22 * process. Since locks still depend on the process id, locks are inherited
23 * after an exec() but not after a fork(). This agrees with POSIX, and both
24 * BSD and SVR4 practice.
25 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
27 * Scrapped free list which is redundant now that we allocate locks
28 * dynamically with kmalloc()/kfree().
29 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
36 * FL_FLOCK locks are created with calls to flock(), through the flock()
37 * system call, which is new. Old C libraries implement flock() via fcntl()
38 * and will continue to use the old, broken implementation.
40 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
41 * with a file pointer (filp). As a result they can be shared by a parent
42 * process and its children after a fork(). They are removed when the last
43 * file descriptor referring to the file pointer is closed (unless explicitly
46 * FL_FLOCK locks never deadlock, an existing lock is always removed before
47 * upgrading from shared to exclusive (or vice versa). When this happens
48 * any processes blocked by the current lock are woken up and allowed to
49 * run before the new lock is applied.
50 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
52 * Removed some race conditions in flock_lock_file(), marked other possible
53 * races. Just grep for FIXME to see them.
54 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
56 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
57 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
58 * once we've checked for blocking and deadlocking.
59 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
66 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
67 * check if a file has mandatory locks, used by mmap(), open() and creat() to
68 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
75 * Fixed deadlock condition for pathological code that mixes calls to
76 * flock() and fcntl().
77 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
79 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
80 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
81 * guarantee sensible behaviour in the case where file system modules might
82 * be compiled with different options than the kernel itself.
83 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
85 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
86 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
87 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
89 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
90 * locks. Changed process synchronisation to avoid dereferencing locks that
91 * have already been freed.
92 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
97 * Made mandatory locking a mount option. Default is not to allow mandatory
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
104 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
105 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
107 * Use slab allocator instead of kmalloc/kfree.
108 * Use generic list implementation from <linux/list.h>.
109 * Sped up posix_locks_deadlock by only considering blocked locks.
110 * Matthew Wilcox <willy@debian.org>, March, 2000.
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
117 #include <linux/capability.h>
118 #include <linux/file.h>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/module.h>
123 #include <linux/security.h>
124 #include <linux/slab.h>
125 #include <linux/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/rcupdate.h>
128 #include <linux/pid_namespace.h>
129 #include <linux/hashtable.h>
130 #include <linux/percpu.h>
131 #include <linux/lglock.h>
133 #define CREATE_TRACE_POINTS
134 #include <trace/events/filelock.h>
136 #include <asm/uaccess.h>
138 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
139 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
140 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG))
141 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
143 static bool lease_breaking(struct file_lock
*fl
)
145 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
148 static int target_leasetype(struct file_lock
*fl
)
150 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
152 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
157 int leases_enable
= 1;
158 int lease_break_time
= 45;
160 #define for_each_lock(inode, lockp) \
161 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
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 via
166 * the file_lock_lglock. Note that alterations to the list also require that
167 * the relevant i_lock is held.
169 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
170 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
173 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
174 * It is protected by blocked_lock_lock.
176 * We hash locks by lockowner in order to optimize searching for the lock a
177 * particular lockowner is waiting on.
179 * FIXME: make this value scale via some heuristic? We generally will want more
180 * buckets when we have more lockowners holding locks, but that's a little
181 * difficult to determine without knowing what the workload will look like.
183 #define BLOCKED_HASH_BITS 7
184 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
187 * This lock protects the blocked_hash. Generally, if you're accessing it, you
188 * want to be holding this lock.
190 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
191 * pointer for file_lock structures that are acting as lock requests (in
192 * contrast to those that are acting as records of acquired locks).
194 * Note that when we acquire this lock in order to change the above fields,
195 * we often hold the i_lock as well. In certain cases, when reading the fields
196 * protected by this lock, we can skip acquiring it iff we already hold the
199 * In particular, adding an entry to the fl_block list requires that you hold
200 * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting
201 * an entry from the list however only requires the file_lock_lock.
203 static DEFINE_SPINLOCK(blocked_lock_lock
);
205 static struct kmem_cache
*filelock_cache __read_mostly
;
207 static void locks_init_lock_heads(struct file_lock
*fl
)
209 INIT_HLIST_NODE(&fl
->fl_link
);
210 INIT_LIST_HEAD(&fl
->fl_block
);
211 init_waitqueue_head(&fl
->fl_wait
);
214 /* Allocate an empty lock structure. */
215 struct file_lock
*locks_alloc_lock(void)
217 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
220 locks_init_lock_heads(fl
);
224 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
226 void locks_release_private(struct file_lock
*fl
)
229 if (fl
->fl_ops
->fl_release_private
)
230 fl
->fl_ops
->fl_release_private(fl
);
235 if (fl
->fl_lmops
->lm_put_owner
)
236 fl
->fl_lmops
->lm_put_owner(fl
);
240 EXPORT_SYMBOL_GPL(locks_release_private
);
242 /* Free a lock which is not in use. */
243 void locks_free_lock(struct file_lock
*fl
)
245 BUG_ON(waitqueue_active(&fl
->fl_wait
));
246 BUG_ON(!list_empty(&fl
->fl_block
));
247 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
249 locks_release_private(fl
);
250 kmem_cache_free(filelock_cache
, fl
);
252 EXPORT_SYMBOL(locks_free_lock
);
255 locks_dispose_list(struct list_head
*dispose
)
257 struct file_lock
*fl
;
259 while (!list_empty(dispose
)) {
260 fl
= list_first_entry(dispose
, struct file_lock
, fl_block
);
261 list_del_init(&fl
->fl_block
);
266 void locks_init_lock(struct file_lock
*fl
)
268 memset(fl
, 0, sizeof(struct file_lock
));
269 locks_init_lock_heads(fl
);
272 EXPORT_SYMBOL(locks_init_lock
);
275 * Initialize a new lock from an existing file_lock structure.
277 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
279 new->fl_owner
= fl
->fl_owner
;
280 new->fl_pid
= fl
->fl_pid
;
282 new->fl_flags
= fl
->fl_flags
;
283 new->fl_type
= fl
->fl_type
;
284 new->fl_start
= fl
->fl_start
;
285 new->fl_end
= fl
->fl_end
;
286 new->fl_lmops
= fl
->fl_lmops
;
290 if (fl
->fl_lmops
->lm_get_owner
)
291 fl
->fl_lmops
->lm_get_owner(new, fl
);
294 EXPORT_SYMBOL(locks_copy_conflock
);
296 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
298 /* "new" must be a freshly-initialized lock */
299 WARN_ON_ONCE(new->fl_ops
);
301 locks_copy_conflock(new, fl
);
303 new->fl_file
= fl
->fl_file
;
304 new->fl_ops
= fl
->fl_ops
;
307 if (fl
->fl_ops
->fl_copy_lock
)
308 fl
->fl_ops
->fl_copy_lock(new, fl
);
312 EXPORT_SYMBOL(locks_copy_lock
);
314 static inline int flock_translate_cmd(int cmd
) {
316 return cmd
& (LOCK_MAND
| LOCK_RW
);
328 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
329 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
332 struct file_lock
*fl
;
333 int type
= flock_translate_cmd(cmd
);
337 fl
= locks_alloc_lock();
343 fl
->fl_pid
= current
->tgid
;
344 fl
->fl_flags
= FL_FLOCK
;
346 fl
->fl_end
= OFFSET_MAX
;
352 static int assign_type(struct file_lock
*fl
, long type
)
366 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
369 switch (l
->l_whence
) {
374 fl
->fl_start
= filp
->f_pos
;
377 fl
->fl_start
= i_size_read(file_inode(filp
));
382 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
384 fl
->fl_start
+= l
->l_start
;
385 if (fl
->fl_start
< 0)
388 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
389 POSIX-2001 defines it. */
391 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
393 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
395 } else if (l
->l_len
< 0) {
396 if (fl
->fl_start
+ l
->l_len
< 0)
398 fl
->fl_end
= fl
->fl_start
- 1;
399 fl
->fl_start
+= l
->l_len
;
401 fl
->fl_end
= OFFSET_MAX
;
403 fl
->fl_owner
= current
->files
;
404 fl
->fl_pid
= current
->tgid
;
406 fl
->fl_flags
= FL_POSIX
;
410 return assign_type(fl
, l
->l_type
);
413 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
416 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
419 struct flock64 ll
= {
421 .l_whence
= l
->l_whence
,
422 .l_start
= l
->l_start
,
426 return flock64_to_posix_lock(filp
, fl
, &ll
);
429 /* default lease lock manager operations */
430 static void lease_break_callback(struct file_lock
*fl
)
432 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
435 static const struct lock_manager_operations lease_manager_ops
= {
436 .lm_break
= lease_break_callback
,
437 .lm_change
= lease_modify
,
441 * Initialize a lease, use the default lock manager operations
443 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
445 if (assign_type(fl
, type
) != 0)
448 fl
->fl_owner
= current
->files
;
449 fl
->fl_pid
= current
->tgid
;
452 fl
->fl_flags
= FL_LEASE
;
454 fl
->fl_end
= OFFSET_MAX
;
456 fl
->fl_lmops
= &lease_manager_ops
;
460 /* Allocate a file_lock initialised to this type of lease */
461 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
463 struct file_lock
*fl
= locks_alloc_lock();
467 return ERR_PTR(error
);
469 error
= lease_init(filp
, type
, fl
);
472 return ERR_PTR(error
);
477 /* Check if two locks overlap each other.
479 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
481 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
482 (fl2
->fl_end
>= fl1
->fl_start
));
486 * Check whether two locks have the same owner.
488 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
490 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
491 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
492 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
493 return fl1
->fl_owner
== fl2
->fl_owner
;
496 /* Must be called with the i_lock held! */
497 static void locks_insert_global_locks(struct file_lock
*fl
)
499 lg_local_lock(&file_lock_lglock
);
500 fl
->fl_link_cpu
= smp_processor_id();
501 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
502 lg_local_unlock(&file_lock_lglock
);
505 /* Must be called with the i_lock held! */
506 static void locks_delete_global_locks(struct file_lock
*fl
)
509 * Avoid taking lock if already unhashed. This is safe since this check
510 * is done while holding the i_lock, and new insertions into the list
511 * also require that it be held.
513 if (hlist_unhashed(&fl
->fl_link
))
515 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
516 hlist_del_init(&fl
->fl_link
);
517 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
521 posix_owner_key(struct file_lock
*fl
)
523 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
524 return fl
->fl_lmops
->lm_owner_key(fl
);
525 return (unsigned long)fl
->fl_owner
;
528 static void locks_insert_global_blocked(struct file_lock
*waiter
)
530 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
533 static void locks_delete_global_blocked(struct file_lock
*waiter
)
535 hash_del(&waiter
->fl_link
);
538 /* Remove waiter from blocker's block list.
539 * When blocker ends up pointing to itself then the list is empty.
541 * Must be called with blocked_lock_lock held.
543 static void __locks_delete_block(struct file_lock
*waiter
)
545 locks_delete_global_blocked(waiter
);
546 list_del_init(&waiter
->fl_block
);
547 waiter
->fl_next
= NULL
;
550 static void locks_delete_block(struct file_lock
*waiter
)
552 spin_lock(&blocked_lock_lock
);
553 __locks_delete_block(waiter
);
554 spin_unlock(&blocked_lock_lock
);
557 /* Insert waiter into blocker's block list.
558 * We use a circular list so that processes can be easily woken up in
559 * the order they blocked. The documentation doesn't require this but
560 * it seems like the reasonable thing to do.
562 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
563 * list itself is protected by the blocked_lock_lock, but by ensuring that the
564 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
565 * in some cases when we see that the fl_block list is empty.
567 static void __locks_insert_block(struct file_lock
*blocker
,
568 struct file_lock
*waiter
)
570 BUG_ON(!list_empty(&waiter
->fl_block
));
571 waiter
->fl_next
= blocker
;
572 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
573 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
574 locks_insert_global_blocked(waiter
);
577 /* Must be called with i_lock held. */
578 static void locks_insert_block(struct file_lock
*blocker
,
579 struct file_lock
*waiter
)
581 spin_lock(&blocked_lock_lock
);
582 __locks_insert_block(blocker
, waiter
);
583 spin_unlock(&blocked_lock_lock
);
587 * Wake up processes blocked waiting for blocker.
589 * Must be called with the inode->i_lock held!
591 static void locks_wake_up_blocks(struct file_lock
*blocker
)
594 * Avoid taking global lock if list is empty. This is safe since new
595 * blocked requests are only added to the list under the i_lock, and
596 * the i_lock is always held here. Note that removal from the fl_block
597 * list does not require the i_lock, so we must recheck list_empty()
598 * after acquiring the blocked_lock_lock.
600 if (list_empty(&blocker
->fl_block
))
603 spin_lock(&blocked_lock_lock
);
604 while (!list_empty(&blocker
->fl_block
)) {
605 struct file_lock
*waiter
;
607 waiter
= list_first_entry(&blocker
->fl_block
,
608 struct file_lock
, fl_block
);
609 __locks_delete_block(waiter
);
610 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
611 waiter
->fl_lmops
->lm_notify(waiter
);
613 wake_up(&waiter
->fl_wait
);
615 spin_unlock(&blocked_lock_lock
);
618 /* Insert file lock fl into an inode's lock list at the position indicated
619 * by pos. At the same time add the lock to the global file lock list.
621 * Must be called with the i_lock held!
623 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
625 fl
->fl_nspid
= get_pid(task_tgid(current
));
627 /* insert into file's list */
631 locks_insert_global_locks(fl
);
635 * locks_delete_lock - Delete a lock and then free it.
636 * @thisfl_p: pointer that points to the fl_next field of the previous
637 * inode->i_flock list entry
639 * Unlink a lock from all lists and free the namespace reference, but don't
640 * free it yet. Wake up processes that are blocked waiting for this lock and
641 * notify the FS that the lock has been cleared.
643 * Must be called with the i_lock held!
645 static void locks_unlink_lock(struct file_lock
**thisfl_p
)
647 struct file_lock
*fl
= *thisfl_p
;
649 locks_delete_global_locks(fl
);
651 *thisfl_p
= fl
->fl_next
;
655 put_pid(fl
->fl_nspid
);
659 locks_wake_up_blocks(fl
);
663 * Unlink a lock from all lists and free it.
665 * Must be called with i_lock held!
667 static void locks_delete_lock(struct file_lock
**thisfl_p
,
668 struct list_head
*dispose
)
670 struct file_lock
*fl
= *thisfl_p
;
672 locks_unlink_lock(thisfl_p
);
674 list_add(&fl
->fl_block
, dispose
);
679 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
680 * checks for shared/exclusive status of overlapping locks.
682 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
684 if (sys_fl
->fl_type
== F_WRLCK
)
686 if (caller_fl
->fl_type
== F_WRLCK
)
691 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
692 * checking before calling the locks_conflict().
694 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
696 /* POSIX locks owned by the same process do not conflict with
699 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
702 /* Check whether they overlap */
703 if (!locks_overlap(caller_fl
, sys_fl
))
706 return (locks_conflict(caller_fl
, sys_fl
));
709 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
710 * checking before calling the locks_conflict().
712 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
714 /* FLOCK locks referring to the same filp do not conflict with
717 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
719 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
722 return (locks_conflict(caller_fl
, sys_fl
));
726 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
728 struct file_lock
*cfl
;
729 struct inode
*inode
= file_inode(filp
);
731 spin_lock(&inode
->i_lock
);
732 for (cfl
= file_inode(filp
)->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
735 if (posix_locks_conflict(fl
, cfl
))
739 locks_copy_conflock(fl
, cfl
);
741 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
743 fl
->fl_type
= F_UNLCK
;
744 spin_unlock(&inode
->i_lock
);
747 EXPORT_SYMBOL(posix_test_lock
);
750 * Deadlock detection:
752 * We attempt to detect deadlocks that are due purely to posix file
755 * We assume that a task can be waiting for at most one lock at a time.
756 * So for any acquired lock, the process holding that lock may be
757 * waiting on at most one other lock. That lock in turns may be held by
758 * someone waiting for at most one other lock. Given a requested lock
759 * caller_fl which is about to wait for a conflicting lock block_fl, we
760 * follow this chain of waiters to ensure we are not about to create a
763 * Since we do this before we ever put a process to sleep on a lock, we
764 * are ensured that there is never a cycle; that is what guarantees that
765 * the while() loop in posix_locks_deadlock() eventually completes.
767 * Note: the above assumption may not be true when handling lock
768 * requests from a broken NFS client. It may also fail in the presence
769 * of tasks (such as posix threads) sharing the same open file table.
770 * To handle those cases, we just bail out after a few iterations.
772 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
773 * Because the owner is not even nominally tied to a thread of
774 * execution, the deadlock detection below can't reasonably work well. Just
777 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
778 * locks that just checks for the case where two tasks are attempting to
779 * upgrade from read to write locks on the same inode.
782 #define MAX_DEADLK_ITERATIONS 10
784 /* Find a lock that the owner of the given block_fl is blocking on. */
785 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
787 struct file_lock
*fl
;
789 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
790 if (posix_same_owner(fl
, block_fl
))
796 /* Must be called with the blocked_lock_lock held! */
797 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
798 struct file_lock
*block_fl
)
803 * This deadlock detector can't reasonably detect deadlocks with
804 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
806 if (IS_OFDLCK(caller_fl
))
809 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
810 if (i
++ > MAX_DEADLK_ITERATIONS
)
812 if (posix_same_owner(caller_fl
, block_fl
))
818 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
819 * after any leases, but before any posix locks.
821 * Note that if called with an FL_EXISTS argument, the caller may determine
822 * whether or not a lock was successfully freed by testing the return
825 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
827 struct file_lock
*new_fl
= NULL
;
828 struct file_lock
**before
;
829 struct inode
* inode
= file_inode(filp
);
834 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
835 new_fl
= locks_alloc_lock();
840 spin_lock(&inode
->i_lock
);
841 if (request
->fl_flags
& FL_ACCESS
)
844 for_each_lock(inode
, before
) {
845 struct file_lock
*fl
= *before
;
850 if (filp
!= fl
->fl_file
)
852 if (request
->fl_type
== fl
->fl_type
)
855 locks_delete_lock(before
, &dispose
);
859 if (request
->fl_type
== F_UNLCK
) {
860 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
866 * If a higher-priority process was blocked on the old file lock,
867 * give it the opportunity to lock the file.
870 spin_unlock(&inode
->i_lock
);
872 spin_lock(&inode
->i_lock
);
876 for_each_lock(inode
, before
) {
877 struct file_lock
*fl
= *before
;
882 if (!flock_locks_conflict(request
, fl
))
885 if (!(request
->fl_flags
& FL_SLEEP
))
887 error
= FILE_LOCK_DEFERRED
;
888 locks_insert_block(fl
, request
);
891 if (request
->fl_flags
& FL_ACCESS
)
893 locks_copy_lock(new_fl
, request
);
894 locks_insert_lock(before
, new_fl
);
899 spin_unlock(&inode
->i_lock
);
901 locks_free_lock(new_fl
);
902 locks_dispose_list(&dispose
);
906 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
908 struct file_lock
*fl
;
909 struct file_lock
*new_fl
= NULL
;
910 struct file_lock
*new_fl2
= NULL
;
911 struct file_lock
*left
= NULL
;
912 struct file_lock
*right
= NULL
;
913 struct file_lock
**before
;
919 * We may need two file_lock structures for this operation,
920 * so we get them in advance to avoid races.
922 * In some cases we can be sure, that no new locks will be needed
924 if (!(request
->fl_flags
& FL_ACCESS
) &&
925 (request
->fl_type
!= F_UNLCK
||
926 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
927 new_fl
= locks_alloc_lock();
928 new_fl2
= locks_alloc_lock();
931 spin_lock(&inode
->i_lock
);
933 * New lock request. Walk all POSIX locks and look for conflicts. If
934 * there are any, either return error or put the request on the
935 * blocker's list of waiters and the global blocked_hash.
937 if (request
->fl_type
!= F_UNLCK
) {
938 for_each_lock(inode
, before
) {
942 if (!posix_locks_conflict(request
, fl
))
945 locks_copy_conflock(conflock
, fl
);
947 if (!(request
->fl_flags
& FL_SLEEP
))
950 * Deadlock detection and insertion into the blocked
951 * locks list must be done while holding the same lock!
954 spin_lock(&blocked_lock_lock
);
955 if (likely(!posix_locks_deadlock(request
, fl
))) {
956 error
= FILE_LOCK_DEFERRED
;
957 __locks_insert_block(fl
, request
);
959 spin_unlock(&blocked_lock_lock
);
964 /* If we're just looking for a conflict, we're done. */
966 if (request
->fl_flags
& FL_ACCESS
)
970 * Find the first old lock with the same owner as the new lock.
973 before
= &inode
->i_flock
;
975 /* First skip locks owned by other processes. */
976 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
977 !posix_same_owner(request
, fl
))) {
978 before
= &fl
->fl_next
;
981 /* Process locks with this owner. */
982 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
983 /* Detect adjacent or overlapping regions (if same lock type)
985 if (request
->fl_type
== fl
->fl_type
) {
986 /* In all comparisons of start vs end, use
987 * "start - 1" rather than "end + 1". If end
988 * is OFFSET_MAX, end + 1 will become negative.
990 if (fl
->fl_end
< request
->fl_start
- 1)
992 /* If the next lock in the list has entirely bigger
993 * addresses than the new one, insert the lock here.
995 if (fl
->fl_start
- 1 > request
->fl_end
)
998 /* If we come here, the new and old lock are of the
999 * same type and adjacent or overlapping. Make one
1000 * lock yielding from the lower start address of both
1001 * locks to the higher end address.
1003 if (fl
->fl_start
> request
->fl_start
)
1004 fl
->fl_start
= request
->fl_start
;
1006 request
->fl_start
= fl
->fl_start
;
1007 if (fl
->fl_end
< request
->fl_end
)
1008 fl
->fl_end
= request
->fl_end
;
1010 request
->fl_end
= fl
->fl_end
;
1012 locks_delete_lock(before
, &dispose
);
1019 /* Processing for different lock types is a bit
1022 if (fl
->fl_end
< request
->fl_start
)
1024 if (fl
->fl_start
> request
->fl_end
)
1026 if (request
->fl_type
== F_UNLCK
)
1028 if (fl
->fl_start
< request
->fl_start
)
1030 /* If the next lock in the list has a higher end
1031 * address than the new one, insert the new one here.
1033 if (fl
->fl_end
> request
->fl_end
) {
1037 if (fl
->fl_start
>= request
->fl_start
) {
1038 /* The new lock completely replaces an old
1039 * one (This may happen several times).
1042 locks_delete_lock(before
, &dispose
);
1046 * Replace the old lock with new_fl, and
1047 * remove the old one. It's safe to do the
1048 * insert here since we know that we won't be
1049 * using new_fl later, and that the lock is
1050 * just replacing an existing lock.
1055 locks_copy_lock(new_fl
, request
);
1058 locks_delete_lock(before
, &dispose
);
1059 locks_insert_lock(before
, request
);
1063 /* Go on to next lock.
1066 before
= &fl
->fl_next
;
1070 * The above code only modifies existing locks in case of merging or
1071 * replacing. If new lock(s) need to be inserted all modifications are
1072 * done below this, so it's safe yet to bail out.
1074 error
= -ENOLCK
; /* "no luck" */
1075 if (right
&& left
== right
&& !new_fl2
)
1080 if (request
->fl_type
== F_UNLCK
) {
1081 if (request
->fl_flags
& FL_EXISTS
)
1090 locks_copy_lock(new_fl
, request
);
1091 locks_insert_lock(before
, new_fl
);
1095 if (left
== right
) {
1096 /* The new lock breaks the old one in two pieces,
1097 * so we have to use the second new lock.
1101 locks_copy_lock(left
, right
);
1102 locks_insert_lock(before
, left
);
1104 right
->fl_start
= request
->fl_end
+ 1;
1105 locks_wake_up_blocks(right
);
1108 left
->fl_end
= request
->fl_start
- 1;
1109 locks_wake_up_blocks(left
);
1112 spin_unlock(&inode
->i_lock
);
1114 * Free any unused locks.
1117 locks_free_lock(new_fl
);
1119 locks_free_lock(new_fl2
);
1120 locks_dispose_list(&dispose
);
1125 * posix_lock_file - Apply a POSIX-style lock to a file
1126 * @filp: The file to apply the lock to
1127 * @fl: The lock to be applied
1128 * @conflock: Place to return a copy of the conflicting lock, if found.
1130 * Add a POSIX style lock to a file.
1131 * We merge adjacent & overlapping locks whenever possible.
1132 * POSIX locks are sorted by owner task, then by starting address
1134 * Note that if called with an FL_EXISTS argument, the caller may determine
1135 * whether or not a lock was successfully freed by testing the return
1136 * value for -ENOENT.
1138 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1139 struct file_lock
*conflock
)
1141 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1143 EXPORT_SYMBOL(posix_lock_file
);
1146 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1147 * @filp: The file to apply the lock to
1148 * @fl: The lock to be applied
1150 * Add a POSIX style lock to a file.
1151 * We merge adjacent & overlapping locks whenever possible.
1152 * POSIX locks are sorted by owner task, then by starting address
1154 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1159 error
= posix_lock_file(filp
, fl
, NULL
);
1160 if (error
!= FILE_LOCK_DEFERRED
)
1162 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1166 locks_delete_block(fl
);
1171 EXPORT_SYMBOL(posix_lock_file_wait
);
1174 * locks_mandatory_locked - Check for an active lock
1175 * @file: the file to check
1177 * Searches the inode's list of locks to find any POSIX locks which conflict.
1178 * This function is called from locks_verify_locked() only.
1180 int locks_mandatory_locked(struct file
*file
)
1182 struct inode
*inode
= file_inode(file
);
1183 struct file_lock
*fl
;
1186 * Search the lock list for this inode for any POSIX locks.
1188 spin_lock(&inode
->i_lock
);
1189 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1192 if (fl
->fl_owner
!= current
->files
&&
1193 fl
->fl_owner
!= file
)
1196 spin_unlock(&inode
->i_lock
);
1197 return fl
? -EAGAIN
: 0;
1201 * locks_mandatory_area - Check for a conflicting lock
1202 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1204 * @inode: the file to check
1205 * @filp: how the file was opened (if it was)
1206 * @offset: start of area to check
1207 * @count: length of area to check
1209 * Searches the inode's list of locks to find any POSIX locks which conflict.
1210 * This function is called from rw_verify_area() and
1211 * locks_verify_truncate().
1213 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1214 struct file
*filp
, loff_t offset
,
1217 struct file_lock fl
;
1221 locks_init_lock(&fl
);
1222 fl
.fl_pid
= current
->tgid
;
1224 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1225 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1227 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1228 fl
.fl_start
= offset
;
1229 fl
.fl_end
= offset
+ count
- 1;
1234 fl
.fl_flags
&= ~FL_SLEEP
;
1235 error
= __posix_lock_file(inode
, &fl
, NULL
);
1241 fl
.fl_flags
|= FL_SLEEP
;
1242 fl
.fl_owner
= current
->files
;
1243 error
= __posix_lock_file(inode
, &fl
, NULL
);
1244 if (error
!= FILE_LOCK_DEFERRED
)
1246 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1249 * If we've been sleeping someone might have
1250 * changed the permissions behind our back.
1252 if (__mandatory_lock(inode
))
1256 locks_delete_block(&fl
);
1263 EXPORT_SYMBOL(locks_mandatory_area
);
1265 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1269 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1272 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1276 /* We already had a lease on this file; just change its type */
1277 int lease_modify(struct file_lock
**before
, int arg
)
1279 struct file_lock
*fl
= *before
;
1280 int error
= assign_type(fl
, arg
);
1284 lease_clear_pending(fl
, arg
);
1285 locks_wake_up_blocks(fl
);
1286 if (arg
== F_UNLCK
) {
1287 struct file
*filp
= fl
->fl_file
;
1290 filp
->f_owner
.signum
= 0;
1291 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1292 if (fl
->fl_fasync
!= NULL
) {
1293 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1294 fl
->fl_fasync
= NULL
;
1296 locks_delete_lock(before
, NULL
);
1301 EXPORT_SYMBOL(lease_modify
);
1303 static bool past_time(unsigned long then
)
1306 /* 0 is a special value meaning "this never expires": */
1308 return time_after(jiffies
, then
);
1311 static void time_out_leases(struct inode
*inode
)
1313 struct file_lock
**before
;
1314 struct file_lock
*fl
;
1316 before
= &inode
->i_flock
;
1317 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1318 trace_time_out_leases(inode
, fl
);
1319 if (past_time(fl
->fl_downgrade_time
))
1320 lease_modify(before
, F_RDLCK
);
1321 if (past_time(fl
->fl_break_time
))
1322 lease_modify(before
, F_UNLCK
);
1323 if (fl
== *before
) /* lease_modify may have freed fl */
1324 before
= &fl
->fl_next
;
1328 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1330 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1332 return locks_conflict(breaker
, lease
);
1336 * __break_lease - revoke all outstanding leases on file
1337 * @inode: the inode of the file to return
1338 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1340 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1343 * break_lease (inlined for speed) has checked there already is at least
1344 * some kind of lock (maybe a lease) on this file. Leases are broken on
1345 * a call to open() or truncate(). This function can sleep unless you
1346 * specified %O_NONBLOCK to your open().
1348 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1351 struct file_lock
*new_fl
, *flock
;
1352 struct file_lock
*fl
;
1353 unsigned long break_time
;
1354 int i_have_this_lease
= 0;
1355 bool lease_conflict
= false;
1356 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1358 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1360 return PTR_ERR(new_fl
);
1361 new_fl
->fl_flags
= type
;
1363 spin_lock(&inode
->i_lock
);
1365 time_out_leases(inode
);
1367 flock
= inode
->i_flock
;
1368 if ((flock
== NULL
) || !IS_LEASE(flock
))
1371 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1372 if (leases_conflict(fl
, new_fl
)) {
1373 lease_conflict
= true;
1374 if (fl
->fl_owner
== current
->files
)
1375 i_have_this_lease
= 1;
1378 if (!lease_conflict
)
1382 if (lease_break_time
> 0) {
1383 break_time
= jiffies
+ lease_break_time
* HZ
;
1384 if (break_time
== 0)
1385 break_time
++; /* so that 0 means no break time */
1388 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1389 if (!leases_conflict(fl
, new_fl
))
1392 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1394 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1395 fl
->fl_break_time
= break_time
;
1397 if (lease_breaking(flock
))
1399 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1400 fl
->fl_downgrade_time
= break_time
;
1402 fl
->fl_lmops
->lm_break(fl
);
1405 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1406 trace_break_lease_noblock(inode
, new_fl
);
1407 error
= -EWOULDBLOCK
;
1412 break_time
= flock
->fl_break_time
;
1413 if (break_time
!= 0)
1414 break_time
-= jiffies
;
1415 if (break_time
== 0)
1417 locks_insert_block(flock
, new_fl
);
1418 trace_break_lease_block(inode
, new_fl
);
1419 spin_unlock(&inode
->i_lock
);
1420 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1421 !new_fl
->fl_next
, break_time
);
1422 spin_lock(&inode
->i_lock
);
1423 trace_break_lease_unblock(inode
, new_fl
);
1424 locks_delete_block(new_fl
);
1427 time_out_leases(inode
);
1429 * Wait for the next conflicting lease that has not been
1432 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1433 flock
= flock
->fl_next
) {
1434 if (leases_conflict(new_fl
, flock
))
1441 spin_unlock(&inode
->i_lock
);
1442 locks_free_lock(new_fl
);
1446 EXPORT_SYMBOL(__break_lease
);
1449 * lease_get_mtime - get the last modified time of an inode
1451 * @time: pointer to a timespec which will contain the last modified time
1453 * This is to force NFS clients to flush their caches for files with
1454 * exclusive leases. The justification is that if someone has an
1455 * exclusive lease, then they could be modifying it.
1457 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1459 bool has_lease
= false;
1460 struct file_lock
*flock
;
1462 if (inode
->i_flock
) {
1463 spin_lock(&inode
->i_lock
);
1464 flock
= inode
->i_flock
;
1465 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
== F_WRLCK
))
1467 spin_unlock(&inode
->i_lock
);
1471 *time
= current_fs_time(inode
->i_sb
);
1473 *time
= inode
->i_mtime
;
1476 EXPORT_SYMBOL(lease_get_mtime
);
1479 * fcntl_getlease - Enquire what lease is currently active
1482 * The value returned by this function will be one of
1483 * (if no lease break is pending):
1485 * %F_RDLCK to indicate a shared lease is held.
1487 * %F_WRLCK to indicate an exclusive lease is held.
1489 * %F_UNLCK to indicate no lease is held.
1491 * (if a lease break is pending):
1493 * %F_RDLCK to indicate an exclusive lease needs to be
1494 * changed to a shared lease (or removed).
1496 * %F_UNLCK to indicate the lease needs to be removed.
1498 * XXX: sfr & willy disagree over whether F_INPROGRESS
1499 * should be returned to userspace.
1501 int fcntl_getlease(struct file
*filp
)
1503 struct file_lock
*fl
;
1504 struct inode
*inode
= file_inode(filp
);
1507 spin_lock(&inode
->i_lock
);
1508 time_out_leases(file_inode(filp
));
1509 for (fl
= file_inode(filp
)->i_flock
; fl
&& IS_LEASE(fl
);
1511 if (fl
->fl_file
== filp
) {
1512 type
= target_leasetype(fl
);
1516 spin_unlock(&inode
->i_lock
);
1521 * check_conflicting_open - see if the given dentry points to a file that has
1522 * an existing open that would conflict with the
1524 * @dentry: dentry to check
1525 * @arg: type of lease that we're trying to acquire
1527 * Check to see if there's an existing open fd on this file that would
1528 * conflict with the lease we're trying to set.
1531 check_conflicting_open(const struct dentry
*dentry
, const long arg
)
1534 struct inode
*inode
= dentry
->d_inode
;
1536 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1539 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1540 (atomic_read(&inode
->i_count
) > 1)))
1546 static int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1548 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1549 struct dentry
*dentry
= filp
->f_path
.dentry
;
1550 struct inode
*inode
= dentry
->d_inode
;
1551 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1555 trace_generic_add_lease(inode
, lease
);
1558 * In the delegation case we need mutual exclusion with
1559 * a number of operations that take the i_mutex. We trylock
1560 * because delegations are an optional optimization, and if
1561 * there's some chance of a conflict--we'd rather not
1562 * bother, maybe that's a sign this just isn't a good file to
1563 * hand out a delegation on.
1565 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1568 if (is_deleg
&& arg
== F_WRLCK
) {
1569 /* Write delegations are not currently supported: */
1570 mutex_unlock(&inode
->i_mutex
);
1575 error
= check_conflicting_open(dentry
, arg
);
1580 * At this point, we know that if there is an exclusive
1581 * lease on this file, then we hold it on this filp
1582 * (otherwise our open of this file would have blocked).
1583 * And if we are trying to acquire an exclusive lease,
1584 * then the file is not open by anyone (including us)
1585 * except for this filp.
1588 for (before
= &inode
->i_flock
;
1589 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1590 before
= &fl
->fl_next
) {
1591 if (fl
->fl_file
== filp
) {
1596 * No exclusive leases if someone else has a lease on
1602 * Modifying our existing lease is OK, but no getting a
1603 * new lease if someone else is opening for write:
1605 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1609 if (my_before
!= NULL
) {
1610 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1620 locks_insert_lock(before
, lease
);
1622 * The check in break_lease() is lockless. It's possible for another
1623 * open to race in after we did the earlier check for a conflicting
1624 * open but before the lease was inserted. Check again for a
1625 * conflicting open and cancel the lease if there is one.
1627 * We also add a barrier here to ensure that the insertion of the lock
1628 * precedes these checks.
1631 error
= check_conflicting_open(dentry
, arg
);
1633 locks_unlink_lock(before
);
1636 mutex_unlock(&inode
->i_mutex
);
1640 static int generic_delete_lease(struct file
*filp
)
1642 int error
= -EAGAIN
;
1643 struct file_lock
*fl
, **before
;
1644 struct dentry
*dentry
= filp
->f_path
.dentry
;
1645 struct inode
*inode
= dentry
->d_inode
;
1647 for (before
= &inode
->i_flock
;
1648 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1649 before
= &fl
->fl_next
) {
1650 if (fl
->fl_file
== filp
)
1653 trace_generic_delete_lease(inode
, fl
);
1655 error
= fl
->fl_lmops
->lm_change(before
, F_UNLCK
);
1660 * generic_setlease - sets a lease on an open file
1661 * @filp: file pointer
1662 * @arg: type of lease to obtain
1663 * @flp: input - file_lock to use, output - file_lock inserted
1665 * The (input) flp->fl_lmops->lm_break function is required
1668 * Called with inode->i_lock held.
1670 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1672 struct dentry
*dentry
= filp
->f_path
.dentry
;
1673 struct inode
*inode
= dentry
->d_inode
;
1676 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1678 if (!S_ISREG(inode
->i_mode
))
1680 error
= security_file_lock(filp
, arg
);
1684 time_out_leases(inode
);
1688 return generic_delete_lease(filp
);
1691 if (!(*flp
)->fl_lmops
->lm_break
) {
1695 return generic_add_lease(filp
, arg
, flp
);
1700 EXPORT_SYMBOL(generic_setlease
);
1702 static int __vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1704 if (filp
->f_op
->setlease
)
1705 return filp
->f_op
->setlease(filp
, arg
, lease
);
1707 return generic_setlease(filp
, arg
, lease
);
1711 * vfs_setlease - sets a lease on an open file
1712 * @filp: file pointer
1713 * @arg: type of lease to obtain
1714 * @lease: file_lock to use when adding a lease
1716 * Call this to establish a lease on the file. The "lease" argument is not
1717 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1718 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1719 * if not, this function will return -ENOLCK (and generate a scary-looking
1723 int vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1725 struct inode
*inode
= file_inode(filp
);
1728 spin_lock(&inode
->i_lock
);
1729 error
= __vfs_setlease(filp
, arg
, lease
);
1730 spin_unlock(&inode
->i_lock
);
1734 EXPORT_SYMBOL_GPL(vfs_setlease
);
1736 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1738 struct file_lock
*fl
, *ret
;
1739 struct inode
*inode
= file_inode(filp
);
1740 struct fasync_struct
*new;
1743 fl
= lease_alloc(filp
, arg
);
1747 new = fasync_alloc();
1749 locks_free_lock(fl
);
1753 spin_lock(&inode
->i_lock
);
1754 error
= __vfs_setlease(filp
, arg
, &ret
);
1761 * fasync_insert_entry() returns the old entry if any.
1762 * If there was no old entry, then it used 'new' and
1763 * inserted it into the fasync list. Clear new so that
1764 * we don't release it here.
1766 if (!fasync_insert_entry(fd
, filp
, &ret
->fl_fasync
, new))
1769 __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
1771 spin_unlock(&inode
->i_lock
);
1773 locks_free_lock(fl
);
1780 * fcntl_setlease - sets a lease on an open file
1781 * @fd: open file descriptor
1782 * @filp: file pointer
1783 * @arg: type of lease to obtain
1785 * Call this fcntl to establish a lease on the file.
1786 * Note that you also need to call %F_SETSIG to
1787 * receive a signal when the lease is broken.
1789 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1792 return vfs_setlease(filp
, F_UNLCK
, NULL
);
1793 return do_fcntl_add_lease(fd
, filp
, arg
);
1797 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1798 * @filp: The file to apply the lock to
1799 * @fl: The lock to be applied
1801 * Add a FLOCK style lock to a file.
1803 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1808 error
= flock_lock_file(filp
, fl
);
1809 if (error
!= FILE_LOCK_DEFERRED
)
1811 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1815 locks_delete_block(fl
);
1821 EXPORT_SYMBOL(flock_lock_file_wait
);
1824 * sys_flock: - flock() system call.
1825 * @fd: the file descriptor to lock.
1826 * @cmd: the type of lock to apply.
1828 * Apply a %FL_FLOCK style lock to an open file descriptor.
1829 * The @cmd can be one of
1831 * %LOCK_SH -- a shared lock.
1833 * %LOCK_EX -- an exclusive lock.
1835 * %LOCK_UN -- remove an existing lock.
1837 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1839 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1840 * processes read and write access respectively.
1842 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1844 struct fd f
= fdget(fd
);
1845 struct file_lock
*lock
;
1846 int can_sleep
, unlock
;
1853 can_sleep
= !(cmd
& LOCK_NB
);
1855 unlock
= (cmd
== LOCK_UN
);
1857 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1858 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1861 error
= flock_make_lock(f
.file
, &lock
, cmd
);
1865 lock
->fl_flags
|= FL_SLEEP
;
1867 error
= security_file_lock(f
.file
, lock
->fl_type
);
1871 if (f
.file
->f_op
->flock
)
1872 error
= f
.file
->f_op
->flock(f
.file
,
1873 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1876 error
= flock_lock_file_wait(f
.file
, lock
);
1879 locks_free_lock(lock
);
1888 * vfs_test_lock - test file byte range lock
1889 * @filp: The file to test lock for
1890 * @fl: The lock to test; also used to hold result
1892 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1893 * setting conf->fl_type to something other than F_UNLCK.
1895 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1897 if (filp
->f_op
->lock
)
1898 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1899 posix_test_lock(filp
, fl
);
1902 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1904 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1906 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1907 #if BITS_PER_LONG == 32
1909 * Make sure we can represent the posix lock via
1910 * legacy 32bit flock.
1912 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1914 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1917 flock
->l_start
= fl
->fl_start
;
1918 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1919 fl
->fl_end
- fl
->fl_start
+ 1;
1920 flock
->l_whence
= 0;
1921 flock
->l_type
= fl
->fl_type
;
1925 #if BITS_PER_LONG == 32
1926 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1928 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1929 flock
->l_start
= fl
->fl_start
;
1930 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1931 fl
->fl_end
- fl
->fl_start
+ 1;
1932 flock
->l_whence
= 0;
1933 flock
->l_type
= fl
->fl_type
;
1937 /* Report the first existing lock that would conflict with l.
1938 * This implements the F_GETLK command of fcntl().
1940 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
1942 struct file_lock file_lock
;
1947 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1950 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1953 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1957 if (cmd
== F_OFD_GETLK
) {
1959 if (flock
.l_pid
!= 0)
1963 file_lock
.fl_flags
|= FL_OFDLCK
;
1964 file_lock
.fl_owner
= filp
;
1967 error
= vfs_test_lock(filp
, &file_lock
);
1971 flock
.l_type
= file_lock
.fl_type
;
1972 if (file_lock
.fl_type
!= F_UNLCK
) {
1973 error
= posix_lock_to_flock(&flock
, &file_lock
);
1978 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1981 locks_release_private(&file_lock
);
1987 * vfs_lock_file - file byte range lock
1988 * @filp: The file to apply the lock to
1989 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1990 * @fl: The lock to be applied
1991 * @conf: Place to return a copy of the conflicting lock, if found.
1993 * A caller that doesn't care about the conflicting lock may pass NULL
1994 * as the final argument.
1996 * If the filesystem defines a private ->lock() method, then @conf will
1997 * be left unchanged; so a caller that cares should initialize it to
1998 * some acceptable default.
2000 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2001 * locks, the ->lock() interface may return asynchronously, before the lock has
2002 * been granted or denied by the underlying filesystem, if (and only if)
2003 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2004 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2005 * the request is for a blocking lock. When ->lock() does return asynchronously,
2006 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2007 * request completes.
2008 * If the request is for non-blocking lock the file system should return
2009 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2010 * with the result. If the request timed out the callback routine will return a
2011 * nonzero return code and the file system should release the lock. The file
2012 * system is also responsible to keep a corresponding posix lock when it
2013 * grants a lock so the VFS can find out which locks are locally held and do
2014 * the correct lock cleanup when required.
2015 * The underlying filesystem must not drop the kernel lock or call
2016 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2019 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2021 if (filp
->f_op
->lock
)
2022 return filp
->f_op
->lock(filp
, cmd
, fl
);
2024 return posix_lock_file(filp
, fl
, conf
);
2026 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2028 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2029 struct file_lock
*fl
)
2033 error
= security_file_lock(filp
, fl
->fl_type
);
2038 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2039 if (error
!= FILE_LOCK_DEFERRED
)
2041 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2045 locks_delete_block(fl
);
2052 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2054 check_fmode_for_setlk(struct file_lock
*fl
)
2056 switch (fl
->fl_type
) {
2058 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2062 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2068 /* Apply the lock described by l to an open file descriptor.
2069 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2071 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2072 struct flock __user
*l
)
2074 struct file_lock
*file_lock
= locks_alloc_lock();
2076 struct inode
*inode
;
2080 if (file_lock
== NULL
)
2084 * This might block, so we do it before checking the inode.
2087 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2090 inode
= file_inode(filp
);
2092 /* Don't allow mandatory locks on files that may be memory mapped
2095 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2101 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2105 error
= check_fmode_for_setlk(file_lock
);
2110 * If the cmd is requesting file-private locks, then set the
2111 * FL_OFDLCK flag and override the owner.
2116 if (flock
.l_pid
!= 0)
2120 file_lock
->fl_flags
|= FL_OFDLCK
;
2121 file_lock
->fl_owner
= filp
;
2125 if (flock
.l_pid
!= 0)
2129 file_lock
->fl_flags
|= FL_OFDLCK
;
2130 file_lock
->fl_owner
= filp
;
2133 file_lock
->fl_flags
|= FL_SLEEP
;
2136 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2139 * Attempt to detect a close/fcntl race and recover by
2140 * releasing the lock that was just acquired.
2143 * we need that spin_lock here - it prevents reordering between
2144 * update of inode->i_flock and check for it done in close().
2145 * rcu_read_lock() wouldn't do.
2147 spin_lock(¤t
->files
->file_lock
);
2149 spin_unlock(¤t
->files
->file_lock
);
2150 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2151 flock
.l_type
= F_UNLCK
;
2156 locks_free_lock(file_lock
);
2160 #if BITS_PER_LONG == 32
2161 /* Report the first existing lock that would conflict with l.
2162 * This implements the F_GETLK command of fcntl().
2164 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2166 struct file_lock file_lock
;
2167 struct flock64 flock
;
2171 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2174 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2177 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2181 if (cmd
== F_OFD_GETLK
) {
2183 if (flock
.l_pid
!= 0)
2187 file_lock
.fl_flags
|= FL_OFDLCK
;
2188 file_lock
.fl_owner
= filp
;
2191 error
= vfs_test_lock(filp
, &file_lock
);
2195 flock
.l_type
= file_lock
.fl_type
;
2196 if (file_lock
.fl_type
!= F_UNLCK
)
2197 posix_lock_to_flock64(&flock
, &file_lock
);
2200 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2203 locks_release_private(&file_lock
);
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_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2212 struct flock64 __user
*l
)
2214 struct file_lock
*file_lock
= locks_alloc_lock();
2215 struct flock64 flock
;
2216 struct inode
*inode
;
2220 if (file_lock
== NULL
)
2224 * This might block, so we do it before checking the inode.
2227 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2230 inode
= file_inode(filp
);
2232 /* Don't allow mandatory locks on files that may be memory mapped
2235 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2241 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2245 error
= check_fmode_for_setlk(file_lock
);
2250 * If the cmd is requesting file-private locks, then set the
2251 * FL_OFDLCK flag and override the owner.
2256 if (flock
.l_pid
!= 0)
2260 file_lock
->fl_flags
|= FL_OFDLCK
;
2261 file_lock
->fl_owner
= filp
;
2265 if (flock
.l_pid
!= 0)
2269 file_lock
->fl_flags
|= FL_OFDLCK
;
2270 file_lock
->fl_owner
= filp
;
2273 file_lock
->fl_flags
|= FL_SLEEP
;
2276 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2279 * Attempt to detect a close/fcntl race and recover by
2280 * releasing the lock that was just acquired.
2282 spin_lock(¤t
->files
->file_lock
);
2284 spin_unlock(¤t
->files
->file_lock
);
2285 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2286 flock
.l_type
= F_UNLCK
;
2291 locks_free_lock(file_lock
);
2294 #endif /* BITS_PER_LONG == 32 */
2297 * This function is called when the file is being removed
2298 * from the task's fd array. POSIX locks belonging to this task
2299 * are deleted at this time.
2301 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2303 struct file_lock lock
;
2306 * If there are no locks held on this file, we don't need to call
2307 * posix_lock_file(). Another process could be setting a lock on this
2308 * file at the same time, but we wouldn't remove that lock anyway.
2310 if (!file_inode(filp
)->i_flock
)
2313 lock
.fl_type
= F_UNLCK
;
2314 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2316 lock
.fl_end
= OFFSET_MAX
;
2317 lock
.fl_owner
= owner
;
2318 lock
.fl_pid
= current
->tgid
;
2319 lock
.fl_file
= filp
;
2321 lock
.fl_lmops
= NULL
;
2323 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2325 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2326 lock
.fl_ops
->fl_release_private(&lock
);
2329 EXPORT_SYMBOL(locks_remove_posix
);
2332 * This function is called on the last close of an open file.
2334 void locks_remove_file(struct file
*filp
)
2336 struct inode
* inode
= file_inode(filp
);
2337 struct file_lock
*fl
;
2338 struct file_lock
**before
;
2341 if (!inode
->i_flock
)
2344 locks_remove_posix(filp
, filp
);
2346 if (filp
->f_op
->flock
) {
2347 struct file_lock fl
= {
2349 .fl_pid
= current
->tgid
,
2351 .fl_flags
= FL_FLOCK
,
2353 .fl_end
= OFFSET_MAX
,
2355 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2356 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2357 fl
.fl_ops
->fl_release_private(&fl
);
2360 spin_lock(&inode
->i_lock
);
2361 before
= &inode
->i_flock
;
2363 while ((fl
= *before
) != NULL
) {
2364 if (fl
->fl_file
== filp
) {
2366 lease_modify(before
, F_UNLCK
);
2371 * There's a leftover lock on the list of a type that
2372 * we didn't expect to see. Most likely a classic
2373 * POSIX lock that ended up not getting released
2374 * properly, or that raced onto the list somehow. Log
2375 * some info about it and then just remove it from
2379 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2380 MAJOR(inode
->i_sb
->s_dev
),
2381 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
,
2382 fl
->fl_type
, fl
->fl_flags
,
2383 fl
->fl_start
, fl
->fl_end
);
2385 locks_delete_lock(before
, &dispose
);
2388 before
= &fl
->fl_next
;
2390 spin_unlock(&inode
->i_lock
);
2391 locks_dispose_list(&dispose
);
2395 * posix_unblock_lock - stop waiting for a file lock
2396 * @waiter: the lock which was waiting
2398 * lockd needs to block waiting for locks.
2401 posix_unblock_lock(struct file_lock
*waiter
)
2405 spin_lock(&blocked_lock_lock
);
2406 if (waiter
->fl_next
)
2407 __locks_delete_block(waiter
);
2410 spin_unlock(&blocked_lock_lock
);
2413 EXPORT_SYMBOL(posix_unblock_lock
);
2416 * vfs_cancel_lock - file byte range unblock lock
2417 * @filp: The file to apply the unblock to
2418 * @fl: The lock to be unblocked
2420 * Used by lock managers to cancel blocked requests
2422 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2424 if (filp
->f_op
->lock
)
2425 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2429 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2431 #ifdef CONFIG_PROC_FS
2432 #include <linux/proc_fs.h>
2433 #include <linux/seq_file.h>
2435 struct locks_iterator
{
2440 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2441 loff_t id
, char *pfx
)
2443 struct inode
*inode
= NULL
;
2444 unsigned int fl_pid
;
2447 fl_pid
= pid_vnr(fl
->fl_nspid
);
2449 fl_pid
= fl
->fl_pid
;
2451 if (fl
->fl_file
!= NULL
)
2452 inode
= file_inode(fl
->fl_file
);
2454 seq_printf(f
, "%lld:%s ", id
, pfx
);
2456 if (fl
->fl_flags
& FL_ACCESS
)
2457 seq_puts(f
, "ACCESS");
2458 else if (IS_OFDLCK(fl
))
2459 seq_puts(f
, "OFDLCK");
2461 seq_puts(f
, "POSIX ");
2463 seq_printf(f
, " %s ",
2464 (inode
== NULL
) ? "*NOINODE*" :
2465 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2466 } else if (IS_FLOCK(fl
)) {
2467 if (fl
->fl_type
& LOCK_MAND
) {
2468 seq_puts(f
, "FLOCK MSNFS ");
2470 seq_puts(f
, "FLOCK ADVISORY ");
2472 } else if (IS_LEASE(fl
)) {
2473 if (fl
->fl_flags
& FL_DELEG
)
2474 seq_puts(f
, "DELEG ");
2476 seq_puts(f
, "LEASE ");
2478 if (lease_breaking(fl
))
2479 seq_puts(f
, "BREAKING ");
2480 else if (fl
->fl_file
)
2481 seq_puts(f
, "ACTIVE ");
2483 seq_puts(f
, "BREAKER ");
2485 seq_puts(f
, "UNKNOWN UNKNOWN ");
2487 if (fl
->fl_type
& LOCK_MAND
) {
2488 seq_printf(f
, "%s ",
2489 (fl
->fl_type
& LOCK_READ
)
2490 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2491 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2493 seq_printf(f
, "%s ",
2494 (lease_breaking(fl
))
2495 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2496 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2499 #ifdef WE_CAN_BREAK_LSLK_NOW
2500 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2501 inode
->i_sb
->s_id
, inode
->i_ino
);
2503 /* userspace relies on this representation of dev_t ;-( */
2504 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2505 MAJOR(inode
->i_sb
->s_dev
),
2506 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2509 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2512 if (fl
->fl_end
== OFFSET_MAX
)
2513 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2515 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2517 seq_puts(f
, "0 EOF\n");
2521 static int locks_show(struct seq_file
*f
, void *v
)
2523 struct locks_iterator
*iter
= f
->private;
2524 struct file_lock
*fl
, *bfl
;
2526 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2528 lock_get_status(f
, fl
, iter
->li_pos
, "");
2530 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2531 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2536 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2537 __acquires(&blocked_lock_lock
)
2539 struct locks_iterator
*iter
= f
->private;
2541 iter
->li_pos
= *pos
+ 1;
2542 lg_global_lock(&file_lock_lglock
);
2543 spin_lock(&blocked_lock_lock
);
2544 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2547 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2549 struct locks_iterator
*iter
= f
->private;
2552 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2555 static void locks_stop(struct seq_file
*f
, void *v
)
2556 __releases(&blocked_lock_lock
)
2558 spin_unlock(&blocked_lock_lock
);
2559 lg_global_unlock(&file_lock_lglock
);
2562 static const struct seq_operations locks_seq_operations
= {
2563 .start
= locks_start
,
2569 static int locks_open(struct inode
*inode
, struct file
*filp
)
2571 return seq_open_private(filp
, &locks_seq_operations
,
2572 sizeof(struct locks_iterator
));
2575 static const struct file_operations proc_locks_operations
= {
2578 .llseek
= seq_lseek
,
2579 .release
= seq_release_private
,
2582 static int __init
proc_locks_init(void)
2584 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2587 module_init(proc_locks_init
);
2590 static int __init
filelock_init(void)
2594 filelock_cache
= kmem_cache_create("file_lock_cache",
2595 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2597 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2599 for_each_possible_cpu(i
)
2600 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2605 core_initcall(filelock_init
);