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>
130 #include <asm/uaccess.h>
132 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
133 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
134 #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE)
136 static bool lease_breaking(struct file_lock
*fl
)
138 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
141 static int target_leasetype(struct file_lock
*fl
)
143 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
145 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
150 int leases_enable
= 1;
151 int lease_break_time
= 45;
153 #define for_each_lock(inode, lockp) \
154 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
156 static LIST_HEAD(file_lock_list
);
157 static LIST_HEAD(blocked_list
);
158 static DEFINE_SPINLOCK(file_lock_lock
);
161 * Protects the two list heads above, plus the inode->i_flock list
163 void lock_flocks(void)
165 spin_lock(&file_lock_lock
);
167 EXPORT_SYMBOL_GPL(lock_flocks
);
169 void unlock_flocks(void)
171 spin_unlock(&file_lock_lock
);
173 EXPORT_SYMBOL_GPL(unlock_flocks
);
175 static struct kmem_cache
*filelock_cache __read_mostly
;
177 static void locks_init_lock_heads(struct file_lock
*fl
)
179 INIT_LIST_HEAD(&fl
->fl_link
);
180 INIT_LIST_HEAD(&fl
->fl_block
);
181 init_waitqueue_head(&fl
->fl_wait
);
184 /* Allocate an empty lock structure. */
185 struct file_lock
*locks_alloc_lock(void)
187 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
190 locks_init_lock_heads(fl
);
194 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
196 void locks_release_private(struct file_lock
*fl
)
199 if (fl
->fl_ops
->fl_release_private
)
200 fl
->fl_ops
->fl_release_private(fl
);
204 if (fl
->fl_lmops
->lm_release_private
)
205 fl
->fl_lmops
->lm_release_private(fl
);
210 EXPORT_SYMBOL_GPL(locks_release_private
);
212 /* Free a lock which is not in use. */
213 void locks_free_lock(struct file_lock
*fl
)
215 BUG_ON(waitqueue_active(&fl
->fl_wait
));
216 BUG_ON(!list_empty(&fl
->fl_block
));
217 BUG_ON(!list_empty(&fl
->fl_link
));
219 locks_release_private(fl
);
220 kmem_cache_free(filelock_cache
, fl
);
222 EXPORT_SYMBOL(locks_free_lock
);
224 void locks_init_lock(struct file_lock
*fl
)
226 memset(fl
, 0, sizeof(struct file_lock
));
227 locks_init_lock_heads(fl
);
230 EXPORT_SYMBOL(locks_init_lock
);
232 static void locks_copy_private(struct file_lock
*new, struct file_lock
*fl
)
235 if (fl
->fl_ops
->fl_copy_lock
)
236 fl
->fl_ops
->fl_copy_lock(new, fl
);
237 new->fl_ops
= fl
->fl_ops
;
240 new->fl_lmops
= fl
->fl_lmops
;
244 * Initialize a new lock from an existing file_lock structure.
246 void __locks_copy_lock(struct file_lock
*new, const struct file_lock
*fl
)
248 new->fl_owner
= fl
->fl_owner
;
249 new->fl_pid
= fl
->fl_pid
;
251 new->fl_flags
= fl
->fl_flags
;
252 new->fl_type
= fl
->fl_type
;
253 new->fl_start
= fl
->fl_start
;
254 new->fl_end
= fl
->fl_end
;
256 new->fl_lmops
= NULL
;
258 EXPORT_SYMBOL(__locks_copy_lock
);
260 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
262 locks_release_private(new);
264 __locks_copy_lock(new, fl
);
265 new->fl_file
= fl
->fl_file
;
266 new->fl_ops
= fl
->fl_ops
;
267 new->fl_lmops
= fl
->fl_lmops
;
269 locks_copy_private(new, fl
);
272 EXPORT_SYMBOL(locks_copy_lock
);
274 static inline int flock_translate_cmd(int cmd
) {
276 return cmd
& (LOCK_MAND
| LOCK_RW
);
288 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
289 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
292 struct file_lock
*fl
;
293 int type
= flock_translate_cmd(cmd
);
297 fl
= locks_alloc_lock();
302 fl
->fl_pid
= current
->tgid
;
303 fl
->fl_flags
= FL_FLOCK
;
305 fl
->fl_end
= OFFSET_MAX
;
311 static int assign_type(struct file_lock
*fl
, long type
)
325 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
328 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
333 switch (l
->l_whence
) {
341 start
= i_size_read(filp
->f_path
.dentry
->d_inode
);
347 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
348 POSIX-2001 defines it. */
352 fl
->fl_end
= OFFSET_MAX
;
354 end
= start
+ l
->l_len
- 1;
356 } else if (l
->l_len
< 0) {
363 fl
->fl_start
= start
; /* we record the absolute position */
364 if (fl
->fl_end
< fl
->fl_start
)
367 fl
->fl_owner
= current
->files
;
368 fl
->fl_pid
= current
->tgid
;
370 fl
->fl_flags
= FL_POSIX
;
374 return assign_type(fl
, l
->l_type
);
377 #if BITS_PER_LONG == 32
378 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
383 switch (l
->l_whence
) {
391 start
= i_size_read(filp
->f_path
.dentry
->d_inode
);
400 fl
->fl_end
= OFFSET_MAX
;
402 fl
->fl_end
= start
+ l
->l_len
- 1;
403 } else if (l
->l_len
< 0) {
404 fl
->fl_end
= start
- 1;
409 fl
->fl_start
= start
; /* we record the absolute position */
410 if (fl
->fl_end
< fl
->fl_start
)
413 fl
->fl_owner
= current
->files
;
414 fl
->fl_pid
= current
->tgid
;
416 fl
->fl_flags
= FL_POSIX
;
420 return assign_type(fl
, l
->l_type
);
424 /* default lease lock manager operations */
425 static void lease_break_callback(struct file_lock
*fl
)
427 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
430 static const struct lock_manager_operations lease_manager_ops
= {
431 .lm_break
= lease_break_callback
,
432 .lm_change
= lease_modify
,
436 * Initialize a lease, use the default lock manager operations
438 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
440 if (assign_type(fl
, type
) != 0)
443 fl
->fl_owner
= current
->files
;
444 fl
->fl_pid
= current
->tgid
;
447 fl
->fl_flags
= FL_LEASE
;
449 fl
->fl_end
= OFFSET_MAX
;
451 fl
->fl_lmops
= &lease_manager_ops
;
455 /* Allocate a file_lock initialised to this type of lease */
456 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
458 struct file_lock
*fl
= locks_alloc_lock();
462 return ERR_PTR(error
);
464 error
= lease_init(filp
, type
, fl
);
467 return ERR_PTR(error
);
472 /* Check if two locks overlap each other.
474 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
476 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
477 (fl2
->fl_end
>= fl1
->fl_start
));
481 * Check whether two locks have the same owner.
483 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
485 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
486 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
487 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
488 return fl1
->fl_owner
== fl2
->fl_owner
;
491 /* Remove waiter from blocker's block list.
492 * When blocker ends up pointing to itself then the list is empty.
494 static void __locks_delete_block(struct file_lock
*waiter
)
496 list_del_init(&waiter
->fl_block
);
497 list_del_init(&waiter
->fl_link
);
498 waiter
->fl_next
= NULL
;
503 void locks_delete_block(struct file_lock
*waiter
)
506 __locks_delete_block(waiter
);
509 EXPORT_SYMBOL(locks_delete_block
);
511 /* Insert waiter into blocker's block list.
512 * We use a circular list so that processes can be easily woken up in
513 * the order they blocked. The documentation doesn't require this but
514 * it seems like the reasonable thing to do.
516 static void locks_insert_block(struct file_lock
*blocker
,
517 struct file_lock
*waiter
)
519 BUG_ON(!list_empty(&waiter
->fl_block
));
520 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
521 waiter
->fl_next
= blocker
;
522 if (IS_POSIX(blocker
))
523 list_add(&waiter
->fl_link
, &blocked_list
);
526 /* Wake up processes blocked waiting for blocker.
527 * If told to wait then schedule the processes until the block list
528 * is empty, otherwise empty the block list ourselves.
530 static void locks_wake_up_blocks(struct file_lock
*blocker
)
532 while (!list_empty(&blocker
->fl_block
)) {
533 struct file_lock
*waiter
;
535 waiter
= list_first_entry(&blocker
->fl_block
,
536 struct file_lock
, fl_block
);
537 __locks_delete_block(waiter
);
538 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
539 waiter
->fl_lmops
->lm_notify(waiter
);
541 wake_up(&waiter
->fl_wait
);
545 /* Insert file lock fl into an inode's lock list at the position indicated
546 * by pos. At the same time add the lock to the global file lock list.
548 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
550 list_add(&fl
->fl_link
, &file_lock_list
);
552 fl
->fl_nspid
= get_pid(task_tgid(current
));
554 /* insert into file's list */
560 * Delete a lock and then free it.
561 * Wake up processes that are blocked waiting for this lock,
562 * notify the FS that the lock has been cleared and
563 * finally free the lock.
565 static void locks_delete_lock(struct file_lock
**thisfl_p
)
567 struct file_lock
*fl
= *thisfl_p
;
569 *thisfl_p
= fl
->fl_next
;
571 list_del_init(&fl
->fl_link
);
574 put_pid(fl
->fl_nspid
);
578 locks_wake_up_blocks(fl
);
582 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
583 * checks for shared/exclusive status of overlapping locks.
585 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
587 if (sys_fl
->fl_type
== F_WRLCK
)
589 if (caller_fl
->fl_type
== F_WRLCK
)
594 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
595 * checking before calling the locks_conflict().
597 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
599 /* POSIX locks owned by the same process do not conflict with
602 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
605 /* Check whether they overlap */
606 if (!locks_overlap(caller_fl
, sys_fl
))
609 return (locks_conflict(caller_fl
, sys_fl
));
612 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
613 * checking before calling the locks_conflict().
615 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
617 /* FLOCK locks referring to the same filp do not conflict with
620 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
622 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
625 return (locks_conflict(caller_fl
, sys_fl
));
629 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
631 struct file_lock
*cfl
;
634 for (cfl
= filp
->f_path
.dentry
->d_inode
->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
637 if (posix_locks_conflict(fl
, cfl
))
641 __locks_copy_lock(fl
, cfl
);
643 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
645 fl
->fl_type
= F_UNLCK
;
649 EXPORT_SYMBOL(posix_test_lock
);
652 * Deadlock detection:
654 * We attempt to detect deadlocks that are due purely to posix file
657 * We assume that a task can be waiting for at most one lock at a time.
658 * So for any acquired lock, the process holding that lock may be
659 * waiting on at most one other lock. That lock in turns may be held by
660 * someone waiting for at most one other lock. Given a requested lock
661 * caller_fl which is about to wait for a conflicting lock block_fl, we
662 * follow this chain of waiters to ensure we are not about to create a
665 * Since we do this before we ever put a process to sleep on a lock, we
666 * are ensured that there is never a cycle; that is what guarantees that
667 * the while() loop in posix_locks_deadlock() eventually completes.
669 * Note: the above assumption may not be true when handling lock
670 * requests from a broken NFS client. It may also fail in the presence
671 * of tasks (such as posix threads) sharing the same open file table.
673 * To handle those cases, we just bail out after a few iterations.
676 #define MAX_DEADLK_ITERATIONS 10
678 /* Find a lock that the owner of the given block_fl is blocking on. */
679 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
681 struct file_lock
*fl
;
683 list_for_each_entry(fl
, &blocked_list
, fl_link
) {
684 if (posix_same_owner(fl
, block_fl
))
690 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
691 struct file_lock
*block_fl
)
695 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
696 if (i
++ > MAX_DEADLK_ITERATIONS
)
698 if (posix_same_owner(caller_fl
, block_fl
))
704 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
705 * after any leases, but before any posix locks.
707 * Note that if called with an FL_EXISTS argument, the caller may determine
708 * whether or not a lock was successfully freed by testing the return
711 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
713 struct file_lock
*new_fl
= NULL
;
714 struct file_lock
**before
;
715 struct inode
* inode
= filp
->f_path
.dentry
->d_inode
;
719 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
720 new_fl
= locks_alloc_lock();
726 if (request
->fl_flags
& FL_ACCESS
)
729 for_each_lock(inode
, before
) {
730 struct file_lock
*fl
= *before
;
735 if (filp
!= fl
->fl_file
)
737 if (request
->fl_type
== fl
->fl_type
)
740 locks_delete_lock(before
);
744 if (request
->fl_type
== F_UNLCK
) {
745 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
751 * If a higher-priority process was blocked on the old file lock,
752 * give it the opportunity to lock the file.
761 for_each_lock(inode
, before
) {
762 struct file_lock
*fl
= *before
;
767 if (!flock_locks_conflict(request
, fl
))
770 if (!(request
->fl_flags
& FL_SLEEP
))
772 error
= FILE_LOCK_DEFERRED
;
773 locks_insert_block(fl
, request
);
776 if (request
->fl_flags
& FL_ACCESS
)
778 locks_copy_lock(new_fl
, request
);
779 locks_insert_lock(before
, new_fl
);
786 locks_free_lock(new_fl
);
790 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
792 struct file_lock
*fl
;
793 struct file_lock
*new_fl
= NULL
;
794 struct file_lock
*new_fl2
= NULL
;
795 struct file_lock
*left
= NULL
;
796 struct file_lock
*right
= NULL
;
797 struct file_lock
**before
;
798 int error
, added
= 0;
801 * We may need two file_lock structures for this operation,
802 * so we get them in advance to avoid races.
804 * In some cases we can be sure, that no new locks will be needed
806 if (!(request
->fl_flags
& FL_ACCESS
) &&
807 (request
->fl_type
!= F_UNLCK
||
808 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
809 new_fl
= locks_alloc_lock();
810 new_fl2
= locks_alloc_lock();
814 if (request
->fl_type
!= F_UNLCK
) {
815 for_each_lock(inode
, before
) {
819 if (!posix_locks_conflict(request
, fl
))
822 __locks_copy_lock(conflock
, fl
);
824 if (!(request
->fl_flags
& FL_SLEEP
))
827 if (posix_locks_deadlock(request
, fl
))
829 error
= FILE_LOCK_DEFERRED
;
830 locks_insert_block(fl
, request
);
835 /* If we're just looking for a conflict, we're done. */
837 if (request
->fl_flags
& FL_ACCESS
)
841 * Find the first old lock with the same owner as the new lock.
844 before
= &inode
->i_flock
;
846 /* First skip locks owned by other processes. */
847 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
848 !posix_same_owner(request
, fl
))) {
849 before
= &fl
->fl_next
;
852 /* Process locks with this owner. */
853 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
854 /* Detect adjacent or overlapping regions (if same lock type)
856 if (request
->fl_type
== fl
->fl_type
) {
857 /* In all comparisons of start vs end, use
858 * "start - 1" rather than "end + 1". If end
859 * is OFFSET_MAX, end + 1 will become negative.
861 if (fl
->fl_end
< request
->fl_start
- 1)
863 /* If the next lock in the list has entirely bigger
864 * addresses than the new one, insert the lock here.
866 if (fl
->fl_start
- 1 > request
->fl_end
)
869 /* If we come here, the new and old lock are of the
870 * same type and adjacent or overlapping. Make one
871 * lock yielding from the lower start address of both
872 * locks to the higher end address.
874 if (fl
->fl_start
> request
->fl_start
)
875 fl
->fl_start
= request
->fl_start
;
877 request
->fl_start
= fl
->fl_start
;
878 if (fl
->fl_end
< request
->fl_end
)
879 fl
->fl_end
= request
->fl_end
;
881 request
->fl_end
= fl
->fl_end
;
883 locks_delete_lock(before
);
890 /* Processing for different lock types is a bit
893 if (fl
->fl_end
< request
->fl_start
)
895 if (fl
->fl_start
> request
->fl_end
)
897 if (request
->fl_type
== F_UNLCK
)
899 if (fl
->fl_start
< request
->fl_start
)
901 /* If the next lock in the list has a higher end
902 * address than the new one, insert the new one here.
904 if (fl
->fl_end
> request
->fl_end
) {
908 if (fl
->fl_start
>= request
->fl_start
) {
909 /* The new lock completely replaces an old
910 * one (This may happen several times).
913 locks_delete_lock(before
);
916 /* Replace the old lock with the new one.
917 * Wake up anybody waiting for the old one,
918 * as the change in lock type might satisfy
921 locks_wake_up_blocks(fl
);
922 fl
->fl_start
= request
->fl_start
;
923 fl
->fl_end
= request
->fl_end
;
924 fl
->fl_type
= request
->fl_type
;
925 locks_release_private(fl
);
926 locks_copy_private(fl
, request
);
931 /* Go on to next lock.
934 before
= &fl
->fl_next
;
938 * The above code only modifies existing locks in case of
939 * merging or replacing. If new lock(s) need to be inserted
940 * all modifications are done bellow this, so it's safe yet to
943 error
= -ENOLCK
; /* "no luck" */
944 if (right
&& left
== right
&& !new_fl2
)
949 if (request
->fl_type
== F_UNLCK
) {
950 if (request
->fl_flags
& FL_EXISTS
)
959 locks_copy_lock(new_fl
, request
);
960 locks_insert_lock(before
, new_fl
);
965 /* The new lock breaks the old one in two pieces,
966 * so we have to use the second new lock.
970 locks_copy_lock(left
, right
);
971 locks_insert_lock(before
, left
);
973 right
->fl_start
= request
->fl_end
+ 1;
974 locks_wake_up_blocks(right
);
977 left
->fl_end
= request
->fl_start
- 1;
978 locks_wake_up_blocks(left
);
983 * Free any unused locks.
986 locks_free_lock(new_fl
);
988 locks_free_lock(new_fl2
);
993 * posix_lock_file - Apply a POSIX-style lock to a file
994 * @filp: The file to apply the lock to
995 * @fl: The lock to be applied
996 * @conflock: Place to return a copy of the conflicting lock, if found.
998 * Add a POSIX style lock to a file.
999 * We merge adjacent & overlapping locks whenever possible.
1000 * POSIX locks are sorted by owner task, then by starting address
1002 * Note that if called with an FL_EXISTS argument, the caller may determine
1003 * whether or not a lock was successfully freed by testing the return
1004 * value for -ENOENT.
1006 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1007 struct file_lock
*conflock
)
1009 return __posix_lock_file(filp
->f_path
.dentry
->d_inode
, fl
, conflock
);
1011 EXPORT_SYMBOL(posix_lock_file
);
1014 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1015 * @filp: The file to apply the lock to
1016 * @fl: The lock to be applied
1018 * Add a POSIX style lock to a file.
1019 * We merge adjacent & overlapping locks whenever possible.
1020 * POSIX locks are sorted by owner task, then by starting address
1022 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1027 error
= posix_lock_file(filp
, fl
, NULL
);
1028 if (error
!= FILE_LOCK_DEFERRED
)
1030 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1034 locks_delete_block(fl
);
1039 EXPORT_SYMBOL(posix_lock_file_wait
);
1042 * locks_mandatory_locked - Check for an active lock
1043 * @inode: the file to check
1045 * Searches the inode's list of locks to find any POSIX locks which conflict.
1046 * This function is called from locks_verify_locked() only.
1048 int locks_mandatory_locked(struct inode
*inode
)
1050 fl_owner_t owner
= current
->files
;
1051 struct file_lock
*fl
;
1054 * Search the lock list for this inode for any POSIX locks.
1057 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1060 if (fl
->fl_owner
!= owner
)
1064 return fl
? -EAGAIN
: 0;
1068 * locks_mandatory_area - Check for a conflicting lock
1069 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1071 * @inode: the file to check
1072 * @filp: how the file was opened (if it was)
1073 * @offset: start of area to check
1074 * @count: length of area to check
1076 * Searches the inode's list of locks to find any POSIX locks which conflict.
1077 * This function is called from rw_verify_area() and
1078 * locks_verify_truncate().
1080 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1081 struct file
*filp
, loff_t offset
,
1084 struct file_lock fl
;
1087 locks_init_lock(&fl
);
1088 fl
.fl_owner
= current
->files
;
1089 fl
.fl_pid
= current
->tgid
;
1091 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1092 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1093 fl
.fl_flags
|= FL_SLEEP
;
1094 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1095 fl
.fl_start
= offset
;
1096 fl
.fl_end
= offset
+ count
- 1;
1099 error
= __posix_lock_file(inode
, &fl
, NULL
);
1100 if (error
!= FILE_LOCK_DEFERRED
)
1102 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1105 * If we've been sleeping someone might have
1106 * changed the permissions behind our back.
1108 if (__mandatory_lock(inode
))
1112 locks_delete_block(&fl
);
1119 EXPORT_SYMBOL(locks_mandatory_area
);
1121 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1125 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1128 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1132 /* We already had a lease on this file; just change its type */
1133 int lease_modify(struct file_lock
**before
, int arg
)
1135 struct file_lock
*fl
= *before
;
1136 int error
= assign_type(fl
, arg
);
1140 lease_clear_pending(fl
, arg
);
1141 locks_wake_up_blocks(fl
);
1142 if (arg
== F_UNLCK
) {
1143 struct file
*filp
= fl
->fl_file
;
1146 filp
->f_owner
.signum
= 0;
1147 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1148 if (fl
->fl_fasync
!= NULL
) {
1149 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1150 fl
->fl_fasync
= NULL
;
1152 locks_delete_lock(before
);
1157 EXPORT_SYMBOL(lease_modify
);
1159 static bool past_time(unsigned long then
)
1162 /* 0 is a special value meaning "this never expires": */
1164 return time_after(jiffies
, then
);
1167 static void time_out_leases(struct inode
*inode
)
1169 struct file_lock
**before
;
1170 struct file_lock
*fl
;
1172 before
= &inode
->i_flock
;
1173 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1174 if (past_time(fl
->fl_downgrade_time
))
1175 lease_modify(before
, F_RDLCK
);
1176 if (past_time(fl
->fl_break_time
))
1177 lease_modify(before
, F_UNLCK
);
1178 if (fl
== *before
) /* lease_modify may have freed fl */
1179 before
= &fl
->fl_next
;
1184 * __break_lease - revoke all outstanding leases on file
1185 * @inode: the inode of the file to return
1186 * @mode: the open mode (read or write)
1188 * break_lease (inlined for speed) has checked there already is at least
1189 * some kind of lock (maybe a lease) on this file. Leases are broken on
1190 * a call to open() or truncate(). This function can sleep unless you
1191 * specified %O_NONBLOCK to your open().
1193 int __break_lease(struct inode
*inode
, unsigned int mode
)
1196 struct file_lock
*new_fl
, *flock
;
1197 struct file_lock
*fl
;
1198 unsigned long break_time
;
1199 int i_have_this_lease
= 0;
1200 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1202 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1204 return PTR_ERR(new_fl
);
1208 time_out_leases(inode
);
1210 flock
= inode
->i_flock
;
1211 if ((flock
== NULL
) || !IS_LEASE(flock
))
1214 if (!locks_conflict(flock
, new_fl
))
1217 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
)
1218 if (fl
->fl_owner
== current
->files
)
1219 i_have_this_lease
= 1;
1222 if (lease_break_time
> 0) {
1223 break_time
= jiffies
+ lease_break_time
* HZ
;
1224 if (break_time
== 0)
1225 break_time
++; /* so that 0 means no break time */
1228 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1230 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1232 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1233 fl
->fl_break_time
= break_time
;
1235 if (lease_breaking(flock
))
1237 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1238 fl
->fl_downgrade_time
= break_time
;
1240 fl
->fl_lmops
->lm_break(fl
);
1243 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1244 error
= -EWOULDBLOCK
;
1249 break_time
= flock
->fl_break_time
;
1250 if (break_time
!= 0) {
1251 break_time
-= jiffies
;
1252 if (break_time
== 0)
1255 locks_insert_block(flock
, new_fl
);
1257 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1258 !new_fl
->fl_next
, break_time
);
1260 __locks_delete_block(new_fl
);
1263 time_out_leases(inode
);
1265 * Wait for the next conflicting lease that has not been
1268 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1269 flock
= flock
->fl_next
) {
1270 if (locks_conflict(new_fl
, flock
))
1278 locks_free_lock(new_fl
);
1282 EXPORT_SYMBOL(__break_lease
);
1285 * lease_get_mtime - get the last modified time of an inode
1287 * @time: pointer to a timespec which will contain the last modified time
1289 * This is to force NFS clients to flush their caches for files with
1290 * exclusive leases. The justification is that if someone has an
1291 * exclusive lease, then they could be modifying it.
1293 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1295 struct file_lock
*flock
= inode
->i_flock
;
1296 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
& F_WRLCK
))
1297 *time
= current_fs_time(inode
->i_sb
);
1299 *time
= inode
->i_mtime
;
1302 EXPORT_SYMBOL(lease_get_mtime
);
1305 * fcntl_getlease - Enquire what lease is currently active
1308 * The value returned by this function will be one of
1309 * (if no lease break is pending):
1311 * %F_RDLCK to indicate a shared lease is held.
1313 * %F_WRLCK to indicate an exclusive lease is held.
1315 * %F_UNLCK to indicate no lease is held.
1317 * (if a lease break is pending):
1319 * %F_RDLCK to indicate an exclusive lease needs to be
1320 * changed to a shared lease (or removed).
1322 * %F_UNLCK to indicate the lease needs to be removed.
1324 * XXX: sfr & willy disagree over whether F_INPROGRESS
1325 * should be returned to userspace.
1327 int fcntl_getlease(struct file
*filp
)
1329 struct file_lock
*fl
;
1333 time_out_leases(filp
->f_path
.dentry
->d_inode
);
1334 for (fl
= filp
->f_path
.dentry
->d_inode
->i_flock
; fl
&& IS_LEASE(fl
);
1336 if (fl
->fl_file
== filp
) {
1337 type
= target_leasetype(fl
);
1345 int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1347 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1348 struct dentry
*dentry
= filp
->f_path
.dentry
;
1349 struct inode
*inode
= dentry
->d_inode
;
1355 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1357 if ((arg
== F_WRLCK
)
1358 && ((dentry
->d_count
> 1)
1359 || (atomic_read(&inode
->i_count
) > 1)))
1363 * At this point, we know that if there is an exclusive
1364 * lease on this file, then we hold it on this filp
1365 * (otherwise our open of this file would have blocked).
1366 * And if we are trying to acquire an exclusive lease,
1367 * then the file is not open by anyone (including us)
1368 * except for this filp.
1371 for (before
= &inode
->i_flock
;
1372 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1373 before
= &fl
->fl_next
) {
1374 if (fl
->fl_file
== filp
) {
1379 * No exclusive leases if someone else has a lease on
1385 * Modifying our existing lease is OK, but no getting a
1386 * new lease if someone else is opening for write:
1388 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1392 if (my_before
!= NULL
) {
1393 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1403 locks_insert_lock(before
, lease
);
1410 int generic_delete_lease(struct file
*filp
, struct file_lock
**flp
)
1412 struct file_lock
*fl
, **before
;
1413 struct dentry
*dentry
= filp
->f_path
.dentry
;
1414 struct inode
*inode
= dentry
->d_inode
;
1416 for (before
= &inode
->i_flock
;
1417 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1418 before
= &fl
->fl_next
) {
1419 if (fl
->fl_file
!= filp
)
1421 return (*flp
)->fl_lmops
->lm_change(before
, F_UNLCK
);
1427 * generic_setlease - sets a lease on an open file
1428 * @filp: file pointer
1429 * @arg: type of lease to obtain
1430 * @flp: input - file_lock to use, output - file_lock inserted
1432 * The (input) flp->fl_lmops->lm_break function is required
1435 * Called with file_lock_lock held.
1437 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1439 struct dentry
*dentry
= filp
->f_path
.dentry
;
1440 struct inode
*inode
= dentry
->d_inode
;
1443 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1445 if (!S_ISREG(inode
->i_mode
))
1447 error
= security_file_lock(filp
, arg
);
1451 time_out_leases(inode
);
1453 BUG_ON(!(*flp
)->fl_lmops
->lm_break
);
1457 return generic_delete_lease(filp
, flp
);
1460 return generic_add_lease(filp
, arg
, flp
);
1465 EXPORT_SYMBOL(generic_setlease
);
1467 static int __vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1469 if (filp
->f_op
&& filp
->f_op
->setlease
)
1470 return filp
->f_op
->setlease(filp
, arg
, lease
);
1472 return generic_setlease(filp
, arg
, lease
);
1476 * vfs_setlease - sets a lease on an open file
1477 * @filp: file pointer
1478 * @arg: type of lease to obtain
1479 * @lease: file_lock to use
1481 * Call this to establish a lease on the file.
1482 * The (*lease)->fl_lmops->lm_break operation must be set; if not,
1483 * break_lease will oops!
1485 * This will call the filesystem's setlease file method, if
1486 * defined. Note that there is no getlease method; instead, the
1487 * filesystem setlease method should call back to setlease() to
1488 * add a lease to the inode's lease list, where fcntl_getlease() can
1489 * find it. Since fcntl_getlease() only reports whether the current
1490 * task holds a lease, a cluster filesystem need only do this for
1491 * leases held by processes on this node.
1493 * There is also no break_lease method; filesystems that
1494 * handle their own leases should break leases themselves from the
1495 * filesystem's open, create, and (on truncate) setattr methods.
1497 * Warning: the only current setlease methods exist only to disable
1498 * leases in certain cases. More vfs changes may be required to
1499 * allow a full filesystem lease implementation.
1502 int vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1507 error
= __vfs_setlease(filp
, arg
, lease
);
1512 EXPORT_SYMBOL_GPL(vfs_setlease
);
1514 static int do_fcntl_delete_lease(struct file
*filp
)
1516 struct file_lock fl
, *flp
= &fl
;
1518 lease_init(filp
, F_UNLCK
, flp
);
1520 return vfs_setlease(filp
, F_UNLCK
, &flp
);
1523 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1525 struct file_lock
*fl
, *ret
;
1526 struct fasync_struct
*new;
1529 fl
= lease_alloc(filp
, arg
);
1533 new = fasync_alloc();
1535 locks_free_lock(fl
);
1540 error
= __vfs_setlease(filp
, arg
, &ret
);
1543 locks_free_lock(fl
);
1544 goto out_free_fasync
;
1547 locks_free_lock(fl
);
1550 * fasync_insert_entry() returns the old entry if any.
1551 * If there was no old entry, then it used 'new' and
1552 * inserted it into the fasync list. Clear new so that
1553 * we don't release it here.
1555 if (!fasync_insert_entry(fd
, filp
, &ret
->fl_fasync
, new))
1558 error
= __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
1568 * fcntl_setlease - sets a lease on an open file
1569 * @fd: open file descriptor
1570 * @filp: file pointer
1571 * @arg: type of lease to obtain
1573 * Call this fcntl to establish a lease on the file.
1574 * Note that you also need to call %F_SETSIG to
1575 * receive a signal when the lease is broken.
1577 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1580 return do_fcntl_delete_lease(filp
);
1581 return do_fcntl_add_lease(fd
, filp
, arg
);
1585 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1586 * @filp: The file to apply the lock to
1587 * @fl: The lock to be applied
1589 * Add a FLOCK style lock to a file.
1591 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1596 error
= flock_lock_file(filp
, fl
);
1597 if (error
!= FILE_LOCK_DEFERRED
)
1599 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1603 locks_delete_block(fl
);
1609 EXPORT_SYMBOL(flock_lock_file_wait
);
1612 * sys_flock: - flock() system call.
1613 * @fd: the file descriptor to lock.
1614 * @cmd: the type of lock to apply.
1616 * Apply a %FL_FLOCK style lock to an open file descriptor.
1617 * The @cmd can be one of
1619 * %LOCK_SH -- a shared lock.
1621 * %LOCK_EX -- an exclusive lock.
1623 * %LOCK_UN -- remove an existing lock.
1625 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1627 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1628 * processes read and write access respectively.
1630 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1634 struct file_lock
*lock
;
1635 int can_sleep
, unlock
;
1639 filp
= fget_light(fd
, &fput_needed
);
1643 can_sleep
= !(cmd
& LOCK_NB
);
1645 unlock
= (cmd
== LOCK_UN
);
1647 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1648 !(filp
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1651 error
= flock_make_lock(filp
, &lock
, cmd
);
1655 lock
->fl_flags
|= FL_SLEEP
;
1657 error
= security_file_lock(filp
, lock
->fl_type
);
1661 if (filp
->f_op
&& filp
->f_op
->flock
)
1662 error
= filp
->f_op
->flock(filp
,
1663 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1666 error
= flock_lock_file_wait(filp
, lock
);
1669 locks_free_lock(lock
);
1672 fput_light(filp
, fput_needed
);
1678 * vfs_test_lock - test file byte range lock
1679 * @filp: The file to test lock for
1680 * @fl: The lock to test; also used to hold result
1682 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1683 * setting conf->fl_type to something other than F_UNLCK.
1685 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1687 if (filp
->f_op
&& filp
->f_op
->lock
)
1688 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1689 posix_test_lock(filp
, fl
);
1692 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1694 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1696 flock
->l_pid
= fl
->fl_pid
;
1697 #if BITS_PER_LONG == 32
1699 * Make sure we can represent the posix lock via
1700 * legacy 32bit flock.
1702 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1704 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1707 flock
->l_start
= fl
->fl_start
;
1708 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1709 fl
->fl_end
- fl
->fl_start
+ 1;
1710 flock
->l_whence
= 0;
1711 flock
->l_type
= fl
->fl_type
;
1715 #if BITS_PER_LONG == 32
1716 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1718 flock
->l_pid
= fl
->fl_pid
;
1719 flock
->l_start
= fl
->fl_start
;
1720 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1721 fl
->fl_end
- fl
->fl_start
+ 1;
1722 flock
->l_whence
= 0;
1723 flock
->l_type
= fl
->fl_type
;
1727 /* Report the first existing lock that would conflict with l.
1728 * This implements the F_GETLK command of fcntl().
1730 int fcntl_getlk(struct file
*filp
, struct flock __user
*l
)
1732 struct file_lock file_lock
;
1737 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1740 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1743 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1747 error
= vfs_test_lock(filp
, &file_lock
);
1751 flock
.l_type
= file_lock
.fl_type
;
1752 if (file_lock
.fl_type
!= F_UNLCK
) {
1753 error
= posix_lock_to_flock(&flock
, &file_lock
);
1758 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1765 * vfs_lock_file - file byte range lock
1766 * @filp: The file to apply the lock to
1767 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1768 * @fl: The lock to be applied
1769 * @conf: Place to return a copy of the conflicting lock, if found.
1771 * A caller that doesn't care about the conflicting lock may pass NULL
1772 * as the final argument.
1774 * If the filesystem defines a private ->lock() method, then @conf will
1775 * be left unchanged; so a caller that cares should initialize it to
1776 * some acceptable default.
1778 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1779 * locks, the ->lock() interface may return asynchronously, before the lock has
1780 * been granted or denied by the underlying filesystem, if (and only if)
1781 * lm_grant is set. Callers expecting ->lock() to return asynchronously
1782 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1783 * the request is for a blocking lock. When ->lock() does return asynchronously,
1784 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
1785 * request completes.
1786 * If the request is for non-blocking lock the file system should return
1787 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1788 * with the result. If the request timed out the callback routine will return a
1789 * nonzero return code and the file system should release the lock. The file
1790 * system is also responsible to keep a corresponding posix lock when it
1791 * grants a lock so the VFS can find out which locks are locally held and do
1792 * the correct lock cleanup when required.
1793 * The underlying filesystem must not drop the kernel lock or call
1794 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
1797 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
1799 if (filp
->f_op
&& filp
->f_op
->lock
)
1800 return filp
->f_op
->lock(filp
, cmd
, fl
);
1802 return posix_lock_file(filp
, fl
, conf
);
1804 EXPORT_SYMBOL_GPL(vfs_lock_file
);
1806 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
1807 struct file_lock
*fl
)
1811 error
= security_file_lock(filp
, fl
->fl_type
);
1816 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
1817 if (error
!= FILE_LOCK_DEFERRED
)
1819 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1823 locks_delete_block(fl
);
1830 /* Apply the lock described by l to an open file descriptor.
1831 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1833 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
1834 struct flock __user
*l
)
1836 struct file_lock
*file_lock
= locks_alloc_lock();
1838 struct inode
*inode
;
1842 if (file_lock
== NULL
)
1846 * This might block, so we do it before checking the inode.
1849 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1852 inode
= filp
->f_path
.dentry
->d_inode
;
1854 /* Don't allow mandatory locks on files that may be memory mapped
1857 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
1863 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
1866 if (cmd
== F_SETLKW
) {
1867 file_lock
->fl_flags
|= FL_SLEEP
;
1871 switch (flock
.l_type
) {
1873 if (!(filp
->f_mode
& FMODE_READ
))
1877 if (!(filp
->f_mode
& FMODE_WRITE
))
1887 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
1890 * Attempt to detect a close/fcntl race and recover by
1891 * releasing the lock that was just acquired.
1894 * we need that spin_lock here - it prevents reordering between
1895 * update of inode->i_flock and check for it done in close().
1896 * rcu_read_lock() wouldn't do.
1898 spin_lock(¤t
->files
->file_lock
);
1900 spin_unlock(¤t
->files
->file_lock
);
1901 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
1902 flock
.l_type
= F_UNLCK
;
1907 locks_free_lock(file_lock
);
1911 #if BITS_PER_LONG == 32
1912 /* Report the first existing lock that would conflict with l.
1913 * This implements the F_GETLK command of fcntl().
1915 int fcntl_getlk64(struct file
*filp
, struct flock64 __user
*l
)
1917 struct file_lock file_lock
;
1918 struct flock64 flock
;
1922 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1925 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1928 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
1932 error
= vfs_test_lock(filp
, &file_lock
);
1936 flock
.l_type
= file_lock
.fl_type
;
1937 if (file_lock
.fl_type
!= F_UNLCK
)
1938 posix_lock_to_flock64(&flock
, &file_lock
);
1941 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1948 /* Apply the lock described by l to an open file descriptor.
1949 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1951 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
1952 struct flock64 __user
*l
)
1954 struct file_lock
*file_lock
= locks_alloc_lock();
1955 struct flock64 flock
;
1956 struct inode
*inode
;
1960 if (file_lock
== NULL
)
1964 * This might block, so we do it before checking the inode.
1967 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1970 inode
= filp
->f_path
.dentry
->d_inode
;
1972 /* Don't allow mandatory locks on files that may be memory mapped
1975 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
1981 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
1984 if (cmd
== F_SETLKW64
) {
1985 file_lock
->fl_flags
|= FL_SLEEP
;
1989 switch (flock
.l_type
) {
1991 if (!(filp
->f_mode
& FMODE_READ
))
1995 if (!(filp
->f_mode
& FMODE_WRITE
))
2005 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2008 * Attempt to detect a close/fcntl race and recover by
2009 * releasing the lock that was just acquired.
2011 spin_lock(¤t
->files
->file_lock
);
2013 spin_unlock(¤t
->files
->file_lock
);
2014 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2015 flock
.l_type
= F_UNLCK
;
2020 locks_free_lock(file_lock
);
2023 #endif /* BITS_PER_LONG == 32 */
2026 * This function is called when the file is being removed
2027 * from the task's fd array. POSIX locks belonging to this task
2028 * are deleted at this time.
2030 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2032 struct file_lock lock
;
2035 * If there are no locks held on this file, we don't need to call
2036 * posix_lock_file(). Another process could be setting a lock on this
2037 * file at the same time, but we wouldn't remove that lock anyway.
2039 if (!filp
->f_path
.dentry
->d_inode
->i_flock
)
2042 lock
.fl_type
= F_UNLCK
;
2043 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2045 lock
.fl_end
= OFFSET_MAX
;
2046 lock
.fl_owner
= owner
;
2047 lock
.fl_pid
= current
->tgid
;
2048 lock
.fl_file
= filp
;
2050 lock
.fl_lmops
= NULL
;
2052 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2054 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2055 lock
.fl_ops
->fl_release_private(&lock
);
2058 EXPORT_SYMBOL(locks_remove_posix
);
2061 * This function is called on the last close of an open file.
2063 void locks_remove_flock(struct file
*filp
)
2065 struct inode
* inode
= filp
->f_path
.dentry
->d_inode
;
2066 struct file_lock
*fl
;
2067 struct file_lock
**before
;
2069 if (!inode
->i_flock
)
2072 if (filp
->f_op
&& filp
->f_op
->flock
) {
2073 struct file_lock fl
= {
2074 .fl_pid
= current
->tgid
,
2076 .fl_flags
= FL_FLOCK
,
2078 .fl_end
= OFFSET_MAX
,
2080 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2081 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2082 fl
.fl_ops
->fl_release_private(&fl
);
2086 before
= &inode
->i_flock
;
2088 while ((fl
= *before
) != NULL
) {
2089 if (fl
->fl_file
== filp
) {
2091 locks_delete_lock(before
);
2095 lease_modify(before
, F_UNLCK
);
2101 before
= &fl
->fl_next
;
2107 * posix_unblock_lock - stop waiting for a file lock
2108 * @filp: how the file was opened
2109 * @waiter: the lock which was waiting
2111 * lockd needs to block waiting for locks.
2114 posix_unblock_lock(struct file
*filp
, struct file_lock
*waiter
)
2119 if (waiter
->fl_next
)
2120 __locks_delete_block(waiter
);
2127 EXPORT_SYMBOL(posix_unblock_lock
);
2130 * vfs_cancel_lock - file byte range unblock lock
2131 * @filp: The file to apply the unblock to
2132 * @fl: The lock to be unblocked
2134 * Used by lock managers to cancel blocked requests
2136 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2138 if (filp
->f_op
&& filp
->f_op
->lock
)
2139 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2143 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2145 #ifdef CONFIG_PROC_FS
2146 #include <linux/proc_fs.h>
2147 #include <linux/seq_file.h>
2149 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2150 loff_t id
, char *pfx
)
2152 struct inode
*inode
= NULL
;
2153 unsigned int fl_pid
;
2156 fl_pid
= pid_vnr(fl
->fl_nspid
);
2158 fl_pid
= fl
->fl_pid
;
2160 if (fl
->fl_file
!= NULL
)
2161 inode
= fl
->fl_file
->f_path
.dentry
->d_inode
;
2163 seq_printf(f
, "%lld:%s ", id
, pfx
);
2165 seq_printf(f
, "%6s %s ",
2166 (fl
->fl_flags
& FL_ACCESS
) ? "ACCESS" : "POSIX ",
2167 (inode
== NULL
) ? "*NOINODE*" :
2168 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2169 } else if (IS_FLOCK(fl
)) {
2170 if (fl
->fl_type
& LOCK_MAND
) {
2171 seq_printf(f
, "FLOCK MSNFS ");
2173 seq_printf(f
, "FLOCK ADVISORY ");
2175 } else if (IS_LEASE(fl
)) {
2176 seq_printf(f
, "LEASE ");
2177 if (lease_breaking(fl
))
2178 seq_printf(f
, "BREAKING ");
2179 else if (fl
->fl_file
)
2180 seq_printf(f
, "ACTIVE ");
2182 seq_printf(f
, "BREAKER ");
2184 seq_printf(f
, "UNKNOWN UNKNOWN ");
2186 if (fl
->fl_type
& LOCK_MAND
) {
2187 seq_printf(f
, "%s ",
2188 (fl
->fl_type
& LOCK_READ
)
2189 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2190 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2192 seq_printf(f
, "%s ",
2193 (lease_breaking(fl
))
2194 ? (fl
->fl_type
& F_UNLCK
) ? "UNLCK" : "READ "
2195 : (fl
->fl_type
& F_WRLCK
) ? "WRITE" : "READ ");
2198 #ifdef WE_CAN_BREAK_LSLK_NOW
2199 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2200 inode
->i_sb
->s_id
, inode
->i_ino
);
2202 /* userspace relies on this representation of dev_t ;-( */
2203 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2204 MAJOR(inode
->i_sb
->s_dev
),
2205 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2208 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2211 if (fl
->fl_end
== OFFSET_MAX
)
2212 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2214 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2216 seq_printf(f
, "0 EOF\n");
2220 static int locks_show(struct seq_file
*f
, void *v
)
2222 struct file_lock
*fl
, *bfl
;
2224 fl
= list_entry(v
, struct file_lock
, fl_link
);
2226 lock_get_status(f
, fl
, *((loff_t
*)f
->private), "");
2228 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2229 lock_get_status(f
, bfl
, *((loff_t
*)f
->private), " ->");
2234 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2236 loff_t
*p
= f
->private;
2240 return seq_list_start(&file_lock_list
, *pos
);
2243 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2245 loff_t
*p
= f
->private;
2247 return seq_list_next(v
, &file_lock_list
, pos
);
2250 static void locks_stop(struct seq_file
*f
, void *v
)
2255 static const struct seq_operations locks_seq_operations
= {
2256 .start
= locks_start
,
2262 static int locks_open(struct inode
*inode
, struct file
*filp
)
2264 return seq_open_private(filp
, &locks_seq_operations
, sizeof(loff_t
));
2267 static const struct file_operations proc_locks_operations
= {
2270 .llseek
= seq_lseek
,
2271 .release
= seq_release_private
,
2274 static int __init
proc_locks_init(void)
2276 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2279 module_init(proc_locks_init
);
2283 * lock_may_read - checks that the region is free of locks
2284 * @inode: the inode that is being read
2285 * @start: the first byte to read
2286 * @len: the number of bytes to read
2288 * Emulates Windows locking requirements. Whole-file
2289 * mandatory locks (share modes) can prohibit a read and
2290 * byte-range POSIX locks can prohibit a read if they overlap.
2292 * N.B. this function is only ever called
2293 * from knfsd and ownership of locks is never checked.
2295 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
2297 struct file_lock
*fl
;
2300 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2302 if (fl
->fl_type
== F_RDLCK
)
2304 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2306 } else if (IS_FLOCK(fl
)) {
2307 if (!(fl
->fl_type
& LOCK_MAND
))
2309 if (fl
->fl_type
& LOCK_READ
)
2320 EXPORT_SYMBOL(lock_may_read
);
2323 * lock_may_write - checks that the region is free of locks
2324 * @inode: the inode that is being written
2325 * @start: the first byte to write
2326 * @len: the number of bytes to write
2328 * Emulates Windows locking requirements. Whole-file
2329 * mandatory locks (share modes) can prohibit a write and
2330 * byte-range POSIX locks can prohibit a write if they overlap.
2332 * N.B. this function is only ever called
2333 * from knfsd and ownership of locks is never checked.
2335 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2337 struct file_lock
*fl
;
2340 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2342 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2344 } else if (IS_FLOCK(fl
)) {
2345 if (!(fl
->fl_type
& LOCK_MAND
))
2347 if (fl
->fl_type
& LOCK_WRITE
)
2358 EXPORT_SYMBOL(lock_may_write
);
2360 static int __init
filelock_init(void)
2362 filelock_cache
= kmem_cache_create("file_lock_cache",
2363 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2368 core_initcall(filelock_init
);