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 #include <asm/uaccess.h>
135 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
136 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
137 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG))
139 static bool lease_breaking(struct file_lock
*fl
)
141 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
144 static int target_leasetype(struct file_lock
*fl
)
146 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
148 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
153 int leases_enable
= 1;
154 int lease_break_time
= 45;
156 #define for_each_lock(inode, lockp) \
157 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
160 * The global file_lock_list is only used for displaying /proc/locks, so we
161 * keep a list on each CPU, with each list protected by its own spinlock via
162 * the file_lock_lglock. Note that alterations to the list also require that
163 * the relevant i_lock is held.
165 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
166 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
169 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
170 * It is protected by blocked_lock_lock.
172 * We hash locks by lockowner in order to optimize searching for the lock a
173 * particular lockowner is waiting on.
175 * FIXME: make this value scale via some heuristic? We generally will want more
176 * buckets when we have more lockowners holding locks, but that's a little
177 * difficult to determine without knowing what the workload will look like.
179 #define BLOCKED_HASH_BITS 7
180 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
183 * This lock protects the blocked_hash. Generally, if you're accessing it, you
184 * want to be holding this lock.
186 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
187 * pointer for file_lock structures that are acting as lock requests (in
188 * contrast to those that are acting as records of acquired locks).
190 * Note that when we acquire this lock in order to change the above fields,
191 * we often hold the i_lock as well. In certain cases, when reading the fields
192 * protected by this lock, we can skip acquiring it iff we already hold the
195 * In particular, adding an entry to the fl_block list requires that you hold
196 * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting
197 * an entry from the list however only requires the file_lock_lock.
199 static DEFINE_SPINLOCK(blocked_lock_lock
);
201 static struct kmem_cache
*filelock_cache __read_mostly
;
203 static void locks_init_lock_heads(struct file_lock
*fl
)
205 INIT_HLIST_NODE(&fl
->fl_link
);
206 INIT_LIST_HEAD(&fl
->fl_block
);
207 init_waitqueue_head(&fl
->fl_wait
);
210 /* Allocate an empty lock structure. */
211 struct file_lock
*locks_alloc_lock(void)
213 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
216 locks_init_lock_heads(fl
);
220 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
222 void locks_release_private(struct file_lock
*fl
)
225 if (fl
->fl_ops
->fl_release_private
)
226 fl
->fl_ops
->fl_release_private(fl
);
232 EXPORT_SYMBOL_GPL(locks_release_private
);
234 /* Free a lock which is not in use. */
235 void locks_free_lock(struct file_lock
*fl
)
237 BUG_ON(waitqueue_active(&fl
->fl_wait
));
238 BUG_ON(!list_empty(&fl
->fl_block
));
239 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
241 locks_release_private(fl
);
242 kmem_cache_free(filelock_cache
, fl
);
244 EXPORT_SYMBOL(locks_free_lock
);
246 void locks_init_lock(struct file_lock
*fl
)
248 memset(fl
, 0, sizeof(struct file_lock
));
249 locks_init_lock_heads(fl
);
252 EXPORT_SYMBOL(locks_init_lock
);
254 static void locks_copy_private(struct file_lock
*new, struct file_lock
*fl
)
257 if (fl
->fl_ops
->fl_copy_lock
)
258 fl
->fl_ops
->fl_copy_lock(new, fl
);
259 new->fl_ops
= fl
->fl_ops
;
262 new->fl_lmops
= fl
->fl_lmops
;
266 * Initialize a new lock from an existing file_lock structure.
268 void __locks_copy_lock(struct file_lock
*new, const struct file_lock
*fl
)
270 new->fl_owner
= fl
->fl_owner
;
271 new->fl_pid
= fl
->fl_pid
;
273 new->fl_flags
= fl
->fl_flags
;
274 new->fl_type
= fl
->fl_type
;
275 new->fl_start
= fl
->fl_start
;
276 new->fl_end
= fl
->fl_end
;
278 new->fl_lmops
= NULL
;
280 EXPORT_SYMBOL(__locks_copy_lock
);
282 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
284 locks_release_private(new);
286 __locks_copy_lock(new, fl
);
287 new->fl_file
= fl
->fl_file
;
288 new->fl_ops
= fl
->fl_ops
;
289 new->fl_lmops
= fl
->fl_lmops
;
291 locks_copy_private(new, fl
);
294 EXPORT_SYMBOL(locks_copy_lock
);
296 static inline int flock_translate_cmd(int cmd
) {
298 return cmd
& (LOCK_MAND
| LOCK_RW
);
310 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
311 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
314 struct file_lock
*fl
;
315 int type
= flock_translate_cmd(cmd
);
319 fl
= locks_alloc_lock();
324 fl
->fl_pid
= current
->tgid
;
325 fl
->fl_flags
= FL_FLOCK
;
327 fl
->fl_end
= OFFSET_MAX
;
333 static int assign_type(struct file_lock
*fl
, long type
)
347 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
350 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
355 switch (l
->l_whence
) {
363 start
= i_size_read(file_inode(filp
));
369 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
370 POSIX-2001 defines it. */
374 fl
->fl_end
= OFFSET_MAX
;
376 end
= start
+ l
->l_len
- 1;
378 } else if (l
->l_len
< 0) {
385 fl
->fl_start
= start
; /* we record the absolute position */
386 if (fl
->fl_end
< fl
->fl_start
)
389 fl
->fl_owner
= current
->files
;
390 fl
->fl_pid
= current
->tgid
;
392 fl
->fl_flags
= FL_POSIX
;
396 return assign_type(fl
, l
->l_type
);
399 #if BITS_PER_LONG == 32
400 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
405 switch (l
->l_whence
) {
413 start
= i_size_read(file_inode(filp
));
422 fl
->fl_end
= OFFSET_MAX
;
424 fl
->fl_end
= start
+ l
->l_len
- 1;
425 } else if (l
->l_len
< 0) {
426 fl
->fl_end
= start
- 1;
431 fl
->fl_start
= start
; /* we record the absolute position */
432 if (fl
->fl_end
< fl
->fl_start
)
435 fl
->fl_owner
= current
->files
;
436 fl
->fl_pid
= current
->tgid
;
438 fl
->fl_flags
= FL_POSIX
;
442 return assign_type(fl
, l
->l_type
);
446 /* default lease lock manager operations */
447 static void lease_break_callback(struct file_lock
*fl
)
449 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
452 static const struct lock_manager_operations lease_manager_ops
= {
453 .lm_break
= lease_break_callback
,
454 .lm_change
= lease_modify
,
458 * Initialize a lease, use the default lock manager operations
460 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
462 if (assign_type(fl
, type
) != 0)
465 fl
->fl_owner
= current
->files
;
466 fl
->fl_pid
= current
->tgid
;
469 fl
->fl_flags
= FL_LEASE
;
471 fl
->fl_end
= OFFSET_MAX
;
473 fl
->fl_lmops
= &lease_manager_ops
;
477 /* Allocate a file_lock initialised to this type of lease */
478 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
480 struct file_lock
*fl
= locks_alloc_lock();
484 return ERR_PTR(error
);
486 error
= lease_init(filp
, type
, fl
);
489 return ERR_PTR(error
);
494 /* Check if two locks overlap each other.
496 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
498 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
499 (fl2
->fl_end
>= fl1
->fl_start
));
503 * Check whether two locks have the same owner.
505 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
507 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
508 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
509 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
510 return fl1
->fl_owner
== fl2
->fl_owner
;
513 /* Must be called with the i_lock held! */
515 locks_insert_global_locks(struct file_lock
*fl
)
517 lg_local_lock(&file_lock_lglock
);
518 fl
->fl_link_cpu
= smp_processor_id();
519 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
520 lg_local_unlock(&file_lock_lglock
);
523 /* Must be called with the i_lock held! */
525 locks_delete_global_locks(struct file_lock
*fl
)
528 * Avoid taking lock if already unhashed. This is safe since this check
529 * is done while holding the i_lock, and new insertions into the list
530 * also require that it be held.
532 if (hlist_unhashed(&fl
->fl_link
))
534 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
535 hlist_del_init(&fl
->fl_link
);
536 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
540 posix_owner_key(struct file_lock
*fl
)
542 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
543 return fl
->fl_lmops
->lm_owner_key(fl
);
544 return (unsigned long)fl
->fl_owner
;
548 locks_insert_global_blocked(struct file_lock
*waiter
)
550 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
554 locks_delete_global_blocked(struct file_lock
*waiter
)
556 hash_del(&waiter
->fl_link
);
559 /* Remove waiter from blocker's block list.
560 * When blocker ends up pointing to itself then the list is empty.
562 * Must be called with blocked_lock_lock held.
564 static void __locks_delete_block(struct file_lock
*waiter
)
566 locks_delete_global_blocked(waiter
);
567 list_del_init(&waiter
->fl_block
);
568 waiter
->fl_next
= NULL
;
571 static void locks_delete_block(struct file_lock
*waiter
)
573 spin_lock(&blocked_lock_lock
);
574 __locks_delete_block(waiter
);
575 spin_unlock(&blocked_lock_lock
);
578 /* Insert waiter into blocker's block list.
579 * We use a circular list so that processes can be easily woken up in
580 * the order they blocked. The documentation doesn't require this but
581 * it seems like the reasonable thing to do.
583 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
584 * list itself is protected by the blocked_lock_lock, but by ensuring that the
585 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
586 * in some cases when we see that the fl_block list is empty.
588 static void __locks_insert_block(struct file_lock
*blocker
,
589 struct file_lock
*waiter
)
591 BUG_ON(!list_empty(&waiter
->fl_block
));
592 waiter
->fl_next
= blocker
;
593 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
594 if (IS_POSIX(blocker
))
595 locks_insert_global_blocked(waiter
);
598 /* Must be called with i_lock held. */
599 static void locks_insert_block(struct file_lock
*blocker
,
600 struct file_lock
*waiter
)
602 spin_lock(&blocked_lock_lock
);
603 __locks_insert_block(blocker
, waiter
);
604 spin_unlock(&blocked_lock_lock
);
608 * Wake up processes blocked waiting for blocker.
610 * Must be called with the inode->i_lock held!
612 static void locks_wake_up_blocks(struct file_lock
*blocker
)
615 * Avoid taking global lock if list is empty. This is safe since new
616 * blocked requests are only added to the list under the i_lock, and
617 * the i_lock is always held here. Note that removal from the fl_block
618 * list does not require the i_lock, so we must recheck list_empty()
619 * after acquiring the blocked_lock_lock.
621 if (list_empty(&blocker
->fl_block
))
624 spin_lock(&blocked_lock_lock
);
625 while (!list_empty(&blocker
->fl_block
)) {
626 struct file_lock
*waiter
;
628 waiter
= list_first_entry(&blocker
->fl_block
,
629 struct file_lock
, fl_block
);
630 __locks_delete_block(waiter
);
631 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
632 waiter
->fl_lmops
->lm_notify(waiter
);
634 wake_up(&waiter
->fl_wait
);
636 spin_unlock(&blocked_lock_lock
);
639 /* Insert file lock fl into an inode's lock list at the position indicated
640 * by pos. At the same time add the lock to the global file lock list.
642 * Must be called with the i_lock held!
644 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
646 fl
->fl_nspid
= get_pid(task_tgid(current
));
648 /* insert into file's list */
652 locks_insert_global_locks(fl
);
656 * locks_delete_lock - Delete a lock and then free it.
657 * @thisfl_p: pointer that points to the fl_next field of the previous
658 * inode->i_flock list entry
660 * Unlink a lock from all lists and free the namespace reference, but don't
661 * free it yet. Wake up processes that are blocked waiting for this lock and
662 * notify the FS that the lock has been cleared.
664 * Must be called with the i_lock held!
666 static void locks_unlink_lock(struct file_lock
**thisfl_p
)
668 struct file_lock
*fl
= *thisfl_p
;
670 locks_delete_global_locks(fl
);
672 *thisfl_p
= fl
->fl_next
;
676 put_pid(fl
->fl_nspid
);
680 locks_wake_up_blocks(fl
);
684 * Unlink a lock from all lists and free it.
686 * Must be called with i_lock held!
688 static void locks_delete_lock(struct file_lock
**thisfl_p
)
690 struct file_lock
*fl
= *thisfl_p
;
692 locks_unlink_lock(thisfl_p
);
696 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
697 * checks for shared/exclusive status of overlapping locks.
699 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
701 if (sys_fl
->fl_type
== F_WRLCK
)
703 if (caller_fl
->fl_type
== F_WRLCK
)
708 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
709 * checking before calling the locks_conflict().
711 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
713 /* POSIX locks owned by the same process do not conflict with
716 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
719 /* Check whether they overlap */
720 if (!locks_overlap(caller_fl
, sys_fl
))
723 return (locks_conflict(caller_fl
, sys_fl
));
726 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
727 * checking before calling the locks_conflict().
729 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
731 /* FLOCK locks referring to the same filp do not conflict with
734 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
736 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
739 return (locks_conflict(caller_fl
, sys_fl
));
743 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
745 struct file_lock
*cfl
;
746 struct inode
*inode
= file_inode(filp
);
748 spin_lock(&inode
->i_lock
);
749 for (cfl
= file_inode(filp
)->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
752 if (posix_locks_conflict(fl
, cfl
))
756 __locks_copy_lock(fl
, cfl
);
758 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
760 fl
->fl_type
= F_UNLCK
;
761 spin_unlock(&inode
->i_lock
);
764 EXPORT_SYMBOL(posix_test_lock
);
767 * Deadlock detection:
769 * We attempt to detect deadlocks that are due purely to posix file
772 * We assume that a task can be waiting for at most one lock at a time.
773 * So for any acquired lock, the process holding that lock may be
774 * waiting on at most one other lock. That lock in turns may be held by
775 * someone waiting for at most one other lock. Given a requested lock
776 * caller_fl which is about to wait for a conflicting lock block_fl, we
777 * follow this chain of waiters to ensure we are not about to create a
780 * Since we do this before we ever put a process to sleep on a lock, we
781 * are ensured that there is never a cycle; that is what guarantees that
782 * the while() loop in posix_locks_deadlock() eventually completes.
784 * Note: the above assumption may not be true when handling lock
785 * requests from a broken NFS client. It may also fail in the presence
786 * of tasks (such as posix threads) sharing the same open file table.
788 * To handle those cases, we just bail out after a few iterations.
791 #define MAX_DEADLK_ITERATIONS 10
793 /* Find a lock that the owner of the given block_fl is blocking on. */
794 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
796 struct file_lock
*fl
;
798 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
799 if (posix_same_owner(fl
, block_fl
))
805 /* Must be called with the blocked_lock_lock held! */
806 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
807 struct file_lock
*block_fl
)
811 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
812 if (i
++ > MAX_DEADLK_ITERATIONS
)
814 if (posix_same_owner(caller_fl
, block_fl
))
820 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
821 * after any leases, but before any posix locks.
823 * Note that if called with an FL_EXISTS argument, the caller may determine
824 * whether or not a lock was successfully freed by testing the return
827 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
829 struct file_lock
*new_fl
= NULL
;
830 struct file_lock
**before
;
831 struct inode
* inode
= file_inode(filp
);
835 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
836 new_fl
= locks_alloc_lock();
841 spin_lock(&inode
->i_lock
);
842 if (request
->fl_flags
& FL_ACCESS
)
845 for_each_lock(inode
, before
) {
846 struct file_lock
*fl
= *before
;
851 if (filp
!= fl
->fl_file
)
853 if (request
->fl_type
== fl
->fl_type
)
856 locks_delete_lock(before
);
860 if (request
->fl_type
== F_UNLCK
) {
861 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
867 * If a higher-priority process was blocked on the old file lock,
868 * give it the opportunity to lock the file.
871 spin_unlock(&inode
->i_lock
);
873 spin_lock(&inode
->i_lock
);
877 for_each_lock(inode
, before
) {
878 struct file_lock
*fl
= *before
;
883 if (!flock_locks_conflict(request
, fl
))
886 if (!(request
->fl_flags
& FL_SLEEP
))
888 error
= FILE_LOCK_DEFERRED
;
889 locks_insert_block(fl
, request
);
892 if (request
->fl_flags
& FL_ACCESS
)
894 locks_copy_lock(new_fl
, request
);
895 locks_insert_lock(before
, new_fl
);
900 spin_unlock(&inode
->i_lock
);
902 locks_free_lock(new_fl
);
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
;
918 * We may need two file_lock structures for this operation,
919 * so we get them in advance to avoid races.
921 * In some cases we can be sure, that no new locks will be needed
923 if (!(request
->fl_flags
& FL_ACCESS
) &&
924 (request
->fl_type
!= F_UNLCK
||
925 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
926 new_fl
= locks_alloc_lock();
927 new_fl2
= locks_alloc_lock();
930 spin_lock(&inode
->i_lock
);
932 * New lock request. Walk all POSIX locks and look for conflicts. If
933 * there are any, either return error or put the request on the
934 * blocker's list of waiters and the global blocked_hash.
936 if (request
->fl_type
!= F_UNLCK
) {
937 for_each_lock(inode
, before
) {
941 if (!posix_locks_conflict(request
, fl
))
944 __locks_copy_lock(conflock
, fl
);
946 if (!(request
->fl_flags
& FL_SLEEP
))
949 * Deadlock detection and insertion into the blocked
950 * locks list must be done while holding the same lock!
953 spin_lock(&blocked_lock_lock
);
954 if (likely(!posix_locks_deadlock(request
, fl
))) {
955 error
= FILE_LOCK_DEFERRED
;
956 __locks_insert_block(fl
, request
);
958 spin_unlock(&blocked_lock_lock
);
963 /* If we're just looking for a conflict, we're done. */
965 if (request
->fl_flags
& FL_ACCESS
)
969 * Find the first old lock with the same owner as the new lock.
972 before
= &inode
->i_flock
;
974 /* First skip locks owned by other processes. */
975 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
976 !posix_same_owner(request
, fl
))) {
977 before
= &fl
->fl_next
;
980 /* Process locks with this owner. */
981 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
982 /* Detect adjacent or overlapping regions (if same lock type)
984 if (request
->fl_type
== fl
->fl_type
) {
985 /* In all comparisons of start vs end, use
986 * "start - 1" rather than "end + 1". If end
987 * is OFFSET_MAX, end + 1 will become negative.
989 if (fl
->fl_end
< request
->fl_start
- 1)
991 /* If the next lock in the list has entirely bigger
992 * addresses than the new one, insert the lock here.
994 if (fl
->fl_start
- 1 > request
->fl_end
)
997 /* If we come here, the new and old lock are of the
998 * same type and adjacent or overlapping. Make one
999 * lock yielding from the lower start address of both
1000 * locks to the higher end address.
1002 if (fl
->fl_start
> request
->fl_start
)
1003 fl
->fl_start
= request
->fl_start
;
1005 request
->fl_start
= fl
->fl_start
;
1006 if (fl
->fl_end
< request
->fl_end
)
1007 fl
->fl_end
= request
->fl_end
;
1009 request
->fl_end
= fl
->fl_end
;
1011 locks_delete_lock(before
);
1018 /* Processing for different lock types is a bit
1021 if (fl
->fl_end
< request
->fl_start
)
1023 if (fl
->fl_start
> request
->fl_end
)
1025 if (request
->fl_type
== F_UNLCK
)
1027 if (fl
->fl_start
< request
->fl_start
)
1029 /* If the next lock in the list has a higher end
1030 * address than the new one, insert the new one here.
1032 if (fl
->fl_end
> request
->fl_end
) {
1036 if (fl
->fl_start
>= request
->fl_start
) {
1037 /* The new lock completely replaces an old
1038 * one (This may happen several times).
1041 locks_delete_lock(before
);
1044 /* Replace the old lock with the new one.
1045 * Wake up anybody waiting for the old one,
1046 * as the change in lock type might satisfy
1049 locks_wake_up_blocks(fl
);
1050 fl
->fl_start
= request
->fl_start
;
1051 fl
->fl_end
= request
->fl_end
;
1052 fl
->fl_type
= request
->fl_type
;
1053 locks_release_private(fl
);
1054 locks_copy_private(fl
, request
);
1059 /* Go on to next lock.
1062 before
= &fl
->fl_next
;
1066 * The above code only modifies existing locks in case of merging or
1067 * replacing. If new lock(s) need to be inserted all modifications are
1068 * done below this, so it's safe yet to bail out.
1070 error
= -ENOLCK
; /* "no luck" */
1071 if (right
&& left
== right
&& !new_fl2
)
1076 if (request
->fl_type
== F_UNLCK
) {
1077 if (request
->fl_flags
& FL_EXISTS
)
1086 locks_copy_lock(new_fl
, request
);
1087 locks_insert_lock(before
, new_fl
);
1091 if (left
== right
) {
1092 /* The new lock breaks the old one in two pieces,
1093 * so we have to use the second new lock.
1097 locks_copy_lock(left
, right
);
1098 locks_insert_lock(before
, left
);
1100 right
->fl_start
= request
->fl_end
+ 1;
1101 locks_wake_up_blocks(right
);
1104 left
->fl_end
= request
->fl_start
- 1;
1105 locks_wake_up_blocks(left
);
1108 spin_unlock(&inode
->i_lock
);
1110 * Free any unused locks.
1113 locks_free_lock(new_fl
);
1115 locks_free_lock(new_fl2
);
1120 * posix_lock_file - Apply a POSIX-style lock to a file
1121 * @filp: The file to apply the lock to
1122 * @fl: The lock to be applied
1123 * @conflock: Place to return a copy of the conflicting lock, if found.
1125 * Add a POSIX style lock to a file.
1126 * We merge adjacent & overlapping locks whenever possible.
1127 * POSIX locks are sorted by owner task, then by starting address
1129 * Note that if called with an FL_EXISTS argument, the caller may determine
1130 * whether or not a lock was successfully freed by testing the return
1131 * value for -ENOENT.
1133 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1134 struct file_lock
*conflock
)
1136 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1138 EXPORT_SYMBOL(posix_lock_file
);
1141 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1142 * @filp: The file to apply the lock to
1143 * @fl: The lock to be applied
1145 * Add a POSIX style lock to a file.
1146 * We merge adjacent & overlapping locks whenever possible.
1147 * POSIX locks are sorted by owner task, then by starting address
1149 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1154 error
= posix_lock_file(filp
, fl
, NULL
);
1155 if (error
!= FILE_LOCK_DEFERRED
)
1157 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1161 locks_delete_block(fl
);
1166 EXPORT_SYMBOL(posix_lock_file_wait
);
1169 * locks_mandatory_locked - Check for an active lock
1170 * @inode: the file to check
1172 * Searches the inode's list of locks to find any POSIX locks which conflict.
1173 * This function is called from locks_verify_locked() only.
1175 int locks_mandatory_locked(struct inode
*inode
)
1177 fl_owner_t owner
= current
->files
;
1178 struct file_lock
*fl
;
1181 * Search the lock list for this inode for any POSIX locks.
1183 spin_lock(&inode
->i_lock
);
1184 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1187 if (fl
->fl_owner
!= owner
)
1190 spin_unlock(&inode
->i_lock
);
1191 return fl
? -EAGAIN
: 0;
1195 * locks_mandatory_area - Check for a conflicting lock
1196 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1198 * @inode: the file to check
1199 * @filp: how the file was opened (if it was)
1200 * @offset: start of area to check
1201 * @count: length of area to check
1203 * Searches the inode's list of locks to find any POSIX locks which conflict.
1204 * This function is called from rw_verify_area() and
1205 * locks_verify_truncate().
1207 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1208 struct file
*filp
, loff_t offset
,
1211 struct file_lock fl
;
1214 locks_init_lock(&fl
);
1215 fl
.fl_owner
= current
->files
;
1216 fl
.fl_pid
= current
->tgid
;
1218 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1219 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1220 fl
.fl_flags
|= FL_SLEEP
;
1221 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1222 fl
.fl_start
= offset
;
1223 fl
.fl_end
= offset
+ count
- 1;
1226 error
= __posix_lock_file(inode
, &fl
, NULL
);
1227 if (error
!= FILE_LOCK_DEFERRED
)
1229 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1232 * If we've been sleeping someone might have
1233 * changed the permissions behind our back.
1235 if (__mandatory_lock(inode
))
1239 locks_delete_block(&fl
);
1246 EXPORT_SYMBOL(locks_mandatory_area
);
1248 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1252 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1255 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1259 /* We already had a lease on this file; just change its type */
1260 int lease_modify(struct file_lock
**before
, int arg
)
1262 struct file_lock
*fl
= *before
;
1263 int error
= assign_type(fl
, arg
);
1267 lease_clear_pending(fl
, arg
);
1268 locks_wake_up_blocks(fl
);
1269 if (arg
== F_UNLCK
) {
1270 struct file
*filp
= fl
->fl_file
;
1273 filp
->f_owner
.signum
= 0;
1274 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1275 if (fl
->fl_fasync
!= NULL
) {
1276 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1277 fl
->fl_fasync
= NULL
;
1279 locks_delete_lock(before
);
1284 EXPORT_SYMBOL(lease_modify
);
1286 static bool past_time(unsigned long then
)
1289 /* 0 is a special value meaning "this never expires": */
1291 return time_after(jiffies
, then
);
1294 static void time_out_leases(struct inode
*inode
)
1296 struct file_lock
**before
;
1297 struct file_lock
*fl
;
1299 before
= &inode
->i_flock
;
1300 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1301 if (past_time(fl
->fl_downgrade_time
))
1302 lease_modify(before
, F_RDLCK
);
1303 if (past_time(fl
->fl_break_time
))
1304 lease_modify(before
, F_UNLCK
);
1305 if (fl
== *before
) /* lease_modify may have freed fl */
1306 before
= &fl
->fl_next
;
1310 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1312 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1314 return locks_conflict(breaker
, lease
);
1318 * __break_lease - revoke all outstanding leases on file
1319 * @inode: the inode of the file to return
1320 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1322 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1325 * break_lease (inlined for speed) has checked there already is at least
1326 * some kind of lock (maybe a lease) on this file. Leases are broken on
1327 * a call to open() or truncate(). This function can sleep unless you
1328 * specified %O_NONBLOCK to your open().
1330 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1333 struct file_lock
*new_fl
, *flock
;
1334 struct file_lock
*fl
;
1335 unsigned long break_time
;
1336 int i_have_this_lease
= 0;
1337 bool lease_conflict
= false;
1338 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1340 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1342 return PTR_ERR(new_fl
);
1343 new_fl
->fl_flags
= type
;
1345 spin_lock(&inode
->i_lock
);
1347 time_out_leases(inode
);
1349 flock
= inode
->i_flock
;
1350 if ((flock
== NULL
) || !IS_LEASE(flock
))
1353 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1354 if (leases_conflict(fl
, new_fl
)) {
1355 lease_conflict
= true;
1356 if (fl
->fl_owner
== current
->files
)
1357 i_have_this_lease
= 1;
1360 if (!lease_conflict
)
1364 if (lease_break_time
> 0) {
1365 break_time
= jiffies
+ lease_break_time
* HZ
;
1366 if (break_time
== 0)
1367 break_time
++; /* so that 0 means no break time */
1370 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1371 if (!leases_conflict(fl
, new_fl
))
1374 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1376 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1377 fl
->fl_break_time
= break_time
;
1379 if (lease_breaking(flock
))
1381 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1382 fl
->fl_downgrade_time
= break_time
;
1384 fl
->fl_lmops
->lm_break(fl
);
1387 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1388 error
= -EWOULDBLOCK
;
1393 break_time
= flock
->fl_break_time
;
1394 if (break_time
!= 0) {
1395 break_time
-= jiffies
;
1396 if (break_time
== 0)
1399 locks_insert_block(flock
, new_fl
);
1400 spin_unlock(&inode
->i_lock
);
1401 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1402 !new_fl
->fl_next
, break_time
);
1403 spin_lock(&inode
->i_lock
);
1404 locks_delete_block(new_fl
);
1407 time_out_leases(inode
);
1409 * Wait for the next conflicting lease that has not been
1412 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1413 flock
= flock
->fl_next
) {
1414 if (leases_conflict(new_fl
, flock
))
1421 spin_unlock(&inode
->i_lock
);
1422 locks_free_lock(new_fl
);
1426 EXPORT_SYMBOL(__break_lease
);
1429 * lease_get_mtime - get the last modified time of an inode
1431 * @time: pointer to a timespec which will contain the last modified time
1433 * This is to force NFS clients to flush their caches for files with
1434 * exclusive leases. The justification is that if someone has an
1435 * exclusive lease, then they could be modifying it.
1437 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1439 struct file_lock
*flock
= inode
->i_flock
;
1440 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
== F_WRLCK
))
1441 *time
= current_fs_time(inode
->i_sb
);
1443 *time
= inode
->i_mtime
;
1446 EXPORT_SYMBOL(lease_get_mtime
);
1449 * fcntl_getlease - Enquire what lease is currently active
1452 * The value returned by this function will be one of
1453 * (if no lease break is pending):
1455 * %F_RDLCK to indicate a shared lease is held.
1457 * %F_WRLCK to indicate an exclusive lease is held.
1459 * %F_UNLCK to indicate no lease is held.
1461 * (if a lease break is pending):
1463 * %F_RDLCK to indicate an exclusive lease needs to be
1464 * changed to a shared lease (or removed).
1466 * %F_UNLCK to indicate the lease needs to be removed.
1468 * XXX: sfr & willy disagree over whether F_INPROGRESS
1469 * should be returned to userspace.
1471 int fcntl_getlease(struct file
*filp
)
1473 struct file_lock
*fl
;
1474 struct inode
*inode
= file_inode(filp
);
1477 spin_lock(&inode
->i_lock
);
1478 time_out_leases(file_inode(filp
));
1479 for (fl
= file_inode(filp
)->i_flock
; fl
&& IS_LEASE(fl
);
1481 if (fl
->fl_file
== filp
) {
1482 type
= target_leasetype(fl
);
1486 spin_unlock(&inode
->i_lock
);
1491 * check_conflicting_open - see if the given dentry points to a file that has
1492 * an existing open that would conflict with the
1494 * @dentry: dentry to check
1495 * @arg: type of lease that we're trying to acquire
1497 * Check to see if there's an existing open fd on this file that would
1498 * conflict with the lease we're trying to set.
1501 check_conflicting_open(const struct dentry
*dentry
, const long arg
)
1504 struct inode
*inode
= dentry
->d_inode
;
1506 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1509 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1510 (atomic_read(&inode
->i_count
) > 1)))
1516 static int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1518 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1519 struct dentry
*dentry
= filp
->f_path
.dentry
;
1520 struct inode
*inode
= dentry
->d_inode
;
1521 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1526 * In the delegation case we need mutual exclusion with
1527 * a number of operations that take the i_mutex. We trylock
1528 * because delegations are an optional optimization, and if
1529 * there's some chance of a conflict--we'd rather not
1530 * bother, maybe that's a sign this just isn't a good file to
1531 * hand out a delegation on.
1533 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1536 if (is_deleg
&& arg
== F_WRLCK
) {
1537 /* Write delegations are not currently supported: */
1538 mutex_unlock(&inode
->i_mutex
);
1543 error
= check_conflicting_open(dentry
, arg
);
1548 * At this point, we know that if there is an exclusive
1549 * lease on this file, then we hold it on this filp
1550 * (otherwise our open of this file would have blocked).
1551 * And if we are trying to acquire an exclusive lease,
1552 * then the file is not open by anyone (including us)
1553 * except for this filp.
1556 for (before
= &inode
->i_flock
;
1557 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1558 before
= &fl
->fl_next
) {
1559 if (fl
->fl_file
== filp
) {
1564 * No exclusive leases if someone else has a lease on
1570 * Modifying our existing lease is OK, but no getting a
1571 * new lease if someone else is opening for write:
1573 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1577 if (my_before
!= NULL
) {
1578 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1588 locks_insert_lock(before
, lease
);
1590 * The check in break_lease() is lockless. It's possible for another
1591 * open to race in after we did the earlier check for a conflicting
1592 * open but before the lease was inserted. Check again for a
1593 * conflicting open and cancel the lease if there is one.
1595 * We also add a barrier here to ensure that the insertion of the lock
1596 * precedes these checks.
1599 error
= check_conflicting_open(dentry
, arg
);
1601 locks_unlink_lock(flp
);
1604 mutex_unlock(&inode
->i_mutex
);
1608 static int generic_delete_lease(struct file
*filp
, struct file_lock
**flp
)
1610 struct file_lock
*fl
, **before
;
1611 struct dentry
*dentry
= filp
->f_path
.dentry
;
1612 struct inode
*inode
= dentry
->d_inode
;
1614 for (before
= &inode
->i_flock
;
1615 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1616 before
= &fl
->fl_next
) {
1617 if (fl
->fl_file
!= filp
)
1619 return (*flp
)->fl_lmops
->lm_change(before
, F_UNLCK
);
1625 * generic_setlease - sets a lease on an open file
1626 * @filp: file pointer
1627 * @arg: type of lease to obtain
1628 * @flp: input - file_lock to use, output - file_lock inserted
1630 * The (input) flp->fl_lmops->lm_break function is required
1633 * Called with inode->i_lock held.
1635 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1637 struct dentry
*dentry
= filp
->f_path
.dentry
;
1638 struct inode
*inode
= dentry
->d_inode
;
1641 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1643 if (!S_ISREG(inode
->i_mode
))
1645 error
= security_file_lock(filp
, arg
);
1649 time_out_leases(inode
);
1651 BUG_ON(!(*flp
)->fl_lmops
->lm_break
);
1655 return generic_delete_lease(filp
, flp
);
1658 return generic_add_lease(filp
, arg
, flp
);
1663 EXPORT_SYMBOL(generic_setlease
);
1665 static int __vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1667 if (filp
->f_op
->setlease
)
1668 return filp
->f_op
->setlease(filp
, arg
, lease
);
1670 return generic_setlease(filp
, arg
, lease
);
1674 * vfs_setlease - sets a lease on an open file
1675 * @filp: file pointer
1676 * @arg: type of lease to obtain
1677 * @lease: file_lock to use
1679 * Call this to establish a lease on the file.
1680 * The (*lease)->fl_lmops->lm_break operation must be set; if not,
1681 * break_lease will oops!
1683 * This will call the filesystem's setlease file method, if
1684 * defined. Note that there is no getlease method; instead, the
1685 * filesystem setlease method should call back to setlease() to
1686 * add a lease to the inode's lease list, where fcntl_getlease() can
1687 * find it. Since fcntl_getlease() only reports whether the current
1688 * task holds a lease, a cluster filesystem need only do this for
1689 * leases held by processes on this node.
1691 * There is also no break_lease method; filesystems that
1692 * handle their own leases should break leases themselves from the
1693 * filesystem's open, create, and (on truncate) setattr methods.
1695 * Warning: the only current setlease methods exist only to disable
1696 * leases in certain cases. More vfs changes may be required to
1697 * allow a full filesystem lease implementation.
1700 int vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1702 struct inode
*inode
= file_inode(filp
);
1705 spin_lock(&inode
->i_lock
);
1706 error
= __vfs_setlease(filp
, arg
, lease
);
1707 spin_unlock(&inode
->i_lock
);
1711 EXPORT_SYMBOL_GPL(vfs_setlease
);
1713 static int do_fcntl_delete_lease(struct file
*filp
)
1715 struct file_lock fl
, *flp
= &fl
;
1717 lease_init(filp
, F_UNLCK
, flp
);
1719 return vfs_setlease(filp
, F_UNLCK
, &flp
);
1722 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1724 struct file_lock
*fl
, *ret
;
1725 struct inode
*inode
= file_inode(filp
);
1726 struct fasync_struct
*new;
1729 fl
= lease_alloc(filp
, arg
);
1733 new = fasync_alloc();
1735 locks_free_lock(fl
);
1739 spin_lock(&inode
->i_lock
);
1740 error
= __vfs_setlease(filp
, arg
, &ret
);
1742 spin_unlock(&inode
->i_lock
);
1743 locks_free_lock(fl
);
1744 goto out_free_fasync
;
1747 locks_free_lock(fl
);
1750 * fasync_insert_entry() returns the old entry if any.
1751 * If there was no old entry, then it used 'new' and
1752 * inserted it into the fasync list. Clear new so that
1753 * we don't release it here.
1755 if (!fasync_insert_entry(fd
, filp
, &ret
->fl_fasync
, new))
1758 error
= __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
1759 spin_unlock(&inode
->i_lock
);
1768 * fcntl_setlease - sets a lease on an open file
1769 * @fd: open file descriptor
1770 * @filp: file pointer
1771 * @arg: type of lease to obtain
1773 * Call this fcntl to establish a lease on the file.
1774 * Note that you also need to call %F_SETSIG to
1775 * receive a signal when the lease is broken.
1777 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1780 return do_fcntl_delete_lease(filp
);
1781 return do_fcntl_add_lease(fd
, filp
, arg
);
1785 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1786 * @filp: The file to apply the lock to
1787 * @fl: The lock to be applied
1789 * Add a FLOCK style lock to a file.
1791 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1796 error
= flock_lock_file(filp
, fl
);
1797 if (error
!= FILE_LOCK_DEFERRED
)
1799 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1803 locks_delete_block(fl
);
1809 EXPORT_SYMBOL(flock_lock_file_wait
);
1812 * sys_flock: - flock() system call.
1813 * @fd: the file descriptor to lock.
1814 * @cmd: the type of lock to apply.
1816 * Apply a %FL_FLOCK style lock to an open file descriptor.
1817 * The @cmd can be one of
1819 * %LOCK_SH -- a shared lock.
1821 * %LOCK_EX -- an exclusive lock.
1823 * %LOCK_UN -- remove an existing lock.
1825 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1827 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1828 * processes read and write access respectively.
1830 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1832 struct fd f
= fdget(fd
);
1833 struct file_lock
*lock
;
1834 int can_sleep
, unlock
;
1841 can_sleep
= !(cmd
& LOCK_NB
);
1843 unlock
= (cmd
== LOCK_UN
);
1845 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1846 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1849 error
= flock_make_lock(f
.file
, &lock
, cmd
);
1853 lock
->fl_flags
|= FL_SLEEP
;
1855 error
= security_file_lock(f
.file
, lock
->fl_type
);
1859 if (f
.file
->f_op
->flock
)
1860 error
= f
.file
->f_op
->flock(f
.file
,
1861 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1864 error
= flock_lock_file_wait(f
.file
, lock
);
1867 locks_free_lock(lock
);
1876 * vfs_test_lock - test file byte range lock
1877 * @filp: The file to test lock for
1878 * @fl: The lock to test; also used to hold result
1880 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1881 * setting conf->fl_type to something other than F_UNLCK.
1883 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1885 if (filp
->f_op
->lock
)
1886 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1887 posix_test_lock(filp
, fl
);
1890 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1892 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1894 flock
->l_pid
= fl
->fl_pid
;
1895 #if BITS_PER_LONG == 32
1897 * Make sure we can represent the posix lock via
1898 * legacy 32bit flock.
1900 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1902 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1905 flock
->l_start
= fl
->fl_start
;
1906 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1907 fl
->fl_end
- fl
->fl_start
+ 1;
1908 flock
->l_whence
= 0;
1909 flock
->l_type
= fl
->fl_type
;
1913 #if BITS_PER_LONG == 32
1914 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1916 flock
->l_pid
= fl
->fl_pid
;
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 /* Report the first existing lock that would conflict with l.
1926 * This implements the F_GETLK command of fcntl().
1928 int fcntl_getlk(struct file
*filp
, struct flock __user
*l
)
1930 struct file_lock file_lock
;
1935 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1938 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1941 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1945 error
= vfs_test_lock(filp
, &file_lock
);
1949 flock
.l_type
= file_lock
.fl_type
;
1950 if (file_lock
.fl_type
!= F_UNLCK
) {
1951 error
= posix_lock_to_flock(&flock
, &file_lock
);
1956 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1963 * vfs_lock_file - file byte range lock
1964 * @filp: The file to apply the lock to
1965 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1966 * @fl: The lock to be applied
1967 * @conf: Place to return a copy of the conflicting lock, if found.
1969 * A caller that doesn't care about the conflicting lock may pass NULL
1970 * as the final argument.
1972 * If the filesystem defines a private ->lock() method, then @conf will
1973 * be left unchanged; so a caller that cares should initialize it to
1974 * some acceptable default.
1976 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1977 * locks, the ->lock() interface may return asynchronously, before the lock has
1978 * been granted or denied by the underlying filesystem, if (and only if)
1979 * lm_grant is set. Callers expecting ->lock() to return asynchronously
1980 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1981 * the request is for a blocking lock. When ->lock() does return asynchronously,
1982 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
1983 * request completes.
1984 * If the request is for non-blocking lock the file system should return
1985 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1986 * with the result. If the request timed out the callback routine will return a
1987 * nonzero return code and the file system should release the lock. The file
1988 * system is also responsible to keep a corresponding posix lock when it
1989 * grants a lock so the VFS can find out which locks are locally held and do
1990 * the correct lock cleanup when required.
1991 * The underlying filesystem must not drop the kernel lock or call
1992 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
1995 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
1997 if (filp
->f_op
->lock
)
1998 return filp
->f_op
->lock(filp
, cmd
, fl
);
2000 return posix_lock_file(filp
, fl
, conf
);
2002 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2004 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2005 struct file_lock
*fl
)
2009 error
= security_file_lock(filp
, fl
->fl_type
);
2014 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2015 if (error
!= FILE_LOCK_DEFERRED
)
2017 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2021 locks_delete_block(fl
);
2028 /* Apply the lock described by l to an open file descriptor.
2029 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2031 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2032 struct flock __user
*l
)
2034 struct file_lock
*file_lock
= locks_alloc_lock();
2036 struct inode
*inode
;
2040 if (file_lock
== NULL
)
2044 * This might block, so we do it before checking the inode.
2047 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2050 inode
= file_inode(filp
);
2052 /* Don't allow mandatory locks on files that may be memory mapped
2055 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2061 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2064 if (cmd
== F_SETLKW
) {
2065 file_lock
->fl_flags
|= FL_SLEEP
;
2069 switch (flock
.l_type
) {
2071 if (!(filp
->f_mode
& FMODE_READ
))
2075 if (!(filp
->f_mode
& FMODE_WRITE
))
2085 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2088 * Attempt to detect a close/fcntl race and recover by
2089 * releasing the lock that was just acquired.
2092 * we need that spin_lock here - it prevents reordering between
2093 * update of inode->i_flock and check for it done in close().
2094 * rcu_read_lock() wouldn't do.
2096 spin_lock(¤t
->files
->file_lock
);
2098 spin_unlock(¤t
->files
->file_lock
);
2099 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2100 flock
.l_type
= F_UNLCK
;
2105 locks_free_lock(file_lock
);
2109 #if BITS_PER_LONG == 32
2110 /* Report the first existing lock that would conflict with l.
2111 * This implements the F_GETLK command of fcntl().
2113 int fcntl_getlk64(struct file
*filp
, struct flock64 __user
*l
)
2115 struct file_lock file_lock
;
2116 struct flock64 flock
;
2120 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2123 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2126 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2130 error
= vfs_test_lock(filp
, &file_lock
);
2134 flock
.l_type
= file_lock
.fl_type
;
2135 if (file_lock
.fl_type
!= F_UNLCK
)
2136 posix_lock_to_flock64(&flock
, &file_lock
);
2139 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2146 /* Apply the lock described by l to an open file descriptor.
2147 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2149 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2150 struct flock64 __user
*l
)
2152 struct file_lock
*file_lock
= locks_alloc_lock();
2153 struct flock64 flock
;
2154 struct inode
*inode
;
2158 if (file_lock
== NULL
)
2162 * This might block, so we do it before checking the inode.
2165 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2168 inode
= file_inode(filp
);
2170 /* Don't allow mandatory locks on files that may be memory mapped
2173 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2179 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2182 if (cmd
== F_SETLKW64
) {
2183 file_lock
->fl_flags
|= FL_SLEEP
;
2187 switch (flock
.l_type
) {
2189 if (!(filp
->f_mode
& FMODE_READ
))
2193 if (!(filp
->f_mode
& FMODE_WRITE
))
2203 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2206 * Attempt to detect a close/fcntl race and recover by
2207 * releasing the lock that was just acquired.
2209 spin_lock(¤t
->files
->file_lock
);
2211 spin_unlock(¤t
->files
->file_lock
);
2212 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2213 flock
.l_type
= F_UNLCK
;
2218 locks_free_lock(file_lock
);
2221 #endif /* BITS_PER_LONG == 32 */
2224 * This function is called when the file is being removed
2225 * from the task's fd array. POSIX locks belonging to this task
2226 * are deleted at this time.
2228 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2230 struct file_lock lock
;
2233 * If there are no locks held on this file, we don't need to call
2234 * posix_lock_file(). Another process could be setting a lock on this
2235 * file at the same time, but we wouldn't remove that lock anyway.
2237 if (!file_inode(filp
)->i_flock
)
2240 lock
.fl_type
= F_UNLCK
;
2241 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2243 lock
.fl_end
= OFFSET_MAX
;
2244 lock
.fl_owner
= owner
;
2245 lock
.fl_pid
= current
->tgid
;
2246 lock
.fl_file
= filp
;
2248 lock
.fl_lmops
= NULL
;
2250 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2252 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2253 lock
.fl_ops
->fl_release_private(&lock
);
2256 EXPORT_SYMBOL(locks_remove_posix
);
2259 * This function is called on the last close of an open file.
2261 void locks_remove_flock(struct file
*filp
)
2263 struct inode
* inode
= file_inode(filp
);
2264 struct file_lock
*fl
;
2265 struct file_lock
**before
;
2267 if (!inode
->i_flock
)
2270 if (filp
->f_op
->flock
) {
2271 struct file_lock fl
= {
2272 .fl_pid
= current
->tgid
,
2274 .fl_flags
= FL_FLOCK
,
2276 .fl_end
= OFFSET_MAX
,
2278 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2279 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2280 fl
.fl_ops
->fl_release_private(&fl
);
2283 spin_lock(&inode
->i_lock
);
2284 before
= &inode
->i_flock
;
2286 while ((fl
= *before
) != NULL
) {
2287 if (fl
->fl_file
== filp
) {
2289 locks_delete_lock(before
);
2293 lease_modify(before
, F_UNLCK
);
2299 before
= &fl
->fl_next
;
2301 spin_unlock(&inode
->i_lock
);
2305 * posix_unblock_lock - stop waiting for a file lock
2306 * @waiter: the lock which was waiting
2308 * lockd needs to block waiting for locks.
2311 posix_unblock_lock(struct file_lock
*waiter
)
2315 spin_lock(&blocked_lock_lock
);
2316 if (waiter
->fl_next
)
2317 __locks_delete_block(waiter
);
2320 spin_unlock(&blocked_lock_lock
);
2323 EXPORT_SYMBOL(posix_unblock_lock
);
2326 * vfs_cancel_lock - file byte range unblock lock
2327 * @filp: The file to apply the unblock to
2328 * @fl: The lock to be unblocked
2330 * Used by lock managers to cancel blocked requests
2332 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2334 if (filp
->f_op
->lock
)
2335 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2339 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2341 #ifdef CONFIG_PROC_FS
2342 #include <linux/proc_fs.h>
2343 #include <linux/seq_file.h>
2345 struct locks_iterator
{
2350 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2351 loff_t id
, char *pfx
)
2353 struct inode
*inode
= NULL
;
2354 unsigned int fl_pid
;
2357 fl_pid
= pid_vnr(fl
->fl_nspid
);
2359 fl_pid
= fl
->fl_pid
;
2361 if (fl
->fl_file
!= NULL
)
2362 inode
= file_inode(fl
->fl_file
);
2364 seq_printf(f
, "%lld:%s ", id
, pfx
);
2366 seq_printf(f
, "%6s %s ",
2367 (fl
->fl_flags
& FL_ACCESS
) ? "ACCESS" : "POSIX ",
2368 (inode
== NULL
) ? "*NOINODE*" :
2369 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2370 } else if (IS_FLOCK(fl
)) {
2371 if (fl
->fl_type
& LOCK_MAND
) {
2372 seq_printf(f
, "FLOCK MSNFS ");
2374 seq_printf(f
, "FLOCK ADVISORY ");
2376 } else if (IS_LEASE(fl
)) {
2377 seq_printf(f
, "LEASE ");
2378 if (lease_breaking(fl
))
2379 seq_printf(f
, "BREAKING ");
2380 else if (fl
->fl_file
)
2381 seq_printf(f
, "ACTIVE ");
2383 seq_printf(f
, "BREAKER ");
2385 seq_printf(f
, "UNKNOWN UNKNOWN ");
2387 if (fl
->fl_type
& LOCK_MAND
) {
2388 seq_printf(f
, "%s ",
2389 (fl
->fl_type
& LOCK_READ
)
2390 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2391 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2393 seq_printf(f
, "%s ",
2394 (lease_breaking(fl
))
2395 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2396 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2399 #ifdef WE_CAN_BREAK_LSLK_NOW
2400 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2401 inode
->i_sb
->s_id
, inode
->i_ino
);
2403 /* userspace relies on this representation of dev_t ;-( */
2404 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2405 MAJOR(inode
->i_sb
->s_dev
),
2406 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2409 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2412 if (fl
->fl_end
== OFFSET_MAX
)
2413 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2415 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2417 seq_printf(f
, "0 EOF\n");
2421 static int locks_show(struct seq_file
*f
, void *v
)
2423 struct locks_iterator
*iter
= f
->private;
2424 struct file_lock
*fl
, *bfl
;
2426 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2428 lock_get_status(f
, fl
, iter
->li_pos
, "");
2430 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2431 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2436 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2438 struct locks_iterator
*iter
= f
->private;
2440 iter
->li_pos
= *pos
+ 1;
2441 lg_global_lock(&file_lock_lglock
);
2442 spin_lock(&blocked_lock_lock
);
2443 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2446 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2448 struct locks_iterator
*iter
= f
->private;
2451 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2454 static void locks_stop(struct seq_file
*f
, void *v
)
2456 spin_unlock(&blocked_lock_lock
);
2457 lg_global_unlock(&file_lock_lglock
);
2460 static const struct seq_operations locks_seq_operations
= {
2461 .start
= locks_start
,
2467 static int locks_open(struct inode
*inode
, struct file
*filp
)
2469 return seq_open_private(filp
, &locks_seq_operations
,
2470 sizeof(struct locks_iterator
));
2473 static const struct file_operations proc_locks_operations
= {
2476 .llseek
= seq_lseek
,
2477 .release
= seq_release_private
,
2480 static int __init
proc_locks_init(void)
2482 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2485 module_init(proc_locks_init
);
2489 * lock_may_read - checks that the region is free of locks
2490 * @inode: the inode that is being read
2491 * @start: the first byte to read
2492 * @len: the number of bytes to read
2494 * Emulates Windows locking requirements. Whole-file
2495 * mandatory locks (share modes) can prohibit a read and
2496 * byte-range POSIX locks can prohibit a read if they overlap.
2498 * N.B. this function is only ever called
2499 * from knfsd and ownership of locks is never checked.
2501 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
2503 struct file_lock
*fl
;
2506 spin_lock(&inode
->i_lock
);
2507 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2509 if (fl
->fl_type
== F_RDLCK
)
2511 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2513 } else if (IS_FLOCK(fl
)) {
2514 if (!(fl
->fl_type
& LOCK_MAND
))
2516 if (fl
->fl_type
& LOCK_READ
)
2523 spin_unlock(&inode
->i_lock
);
2527 EXPORT_SYMBOL(lock_may_read
);
2530 * lock_may_write - checks that the region is free of locks
2531 * @inode: the inode that is being written
2532 * @start: the first byte to write
2533 * @len: the number of bytes to write
2535 * Emulates Windows locking requirements. Whole-file
2536 * mandatory locks (share modes) can prohibit a write and
2537 * byte-range POSIX locks can prohibit a write if they overlap.
2539 * N.B. this function is only ever called
2540 * from knfsd and ownership of locks is never checked.
2542 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2544 struct file_lock
*fl
;
2547 spin_lock(&inode
->i_lock
);
2548 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2550 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2552 } else if (IS_FLOCK(fl
)) {
2553 if (!(fl
->fl_type
& LOCK_MAND
))
2555 if (fl
->fl_type
& LOCK_WRITE
)
2562 spin_unlock(&inode
->i_lock
);
2566 EXPORT_SYMBOL(lock_may_write
);
2568 static int __init
filelock_init(void)
2572 filelock_cache
= kmem_cache_create("file_lock_cache",
2573 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2575 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2577 for_each_possible_cpu(i
)
2578 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2583 core_initcall(filelock_init
);