4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
7 * Deadlock detection added.
8 * FIXME: one thing isn't handled yet:
9 * - mandatory locks (requires lots of changes elsewhere)
10 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
15 * Converted file_lock_table to a linked list from an array, which eliminates
16 * the limits on how many active file locks are open.
17 * Chad Page (pageone@netcom.com), November 27, 1994
19 * Removed dependency on file descriptors. dup()'ed file descriptors now
20 * get the same locks as the original file descriptors, and a close() on
21 * any file descriptor removes ALL the locks on the file for the current
22 * process. Since locks still depend on the process id, locks are inherited
23 * after an exec() but not after a fork(). This agrees with POSIX, and both
24 * BSD and SVR4 practice.
25 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
27 * Scrapped free list which is redundant now that we allocate locks
28 * dynamically with kmalloc()/kfree().
29 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
36 * FL_FLOCK locks are created with calls to flock(), through the flock()
37 * system call, which is new. Old C libraries implement flock() via fcntl()
38 * and will continue to use the old, broken implementation.
40 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
41 * with a file pointer (filp). As a result they can be shared by a parent
42 * process and its children after a fork(). They are removed when the last
43 * file descriptor referring to the file pointer is closed (unless explicitly
46 * FL_FLOCK locks never deadlock, an existing lock is always removed before
47 * upgrading from shared to exclusive (or vice versa). When this happens
48 * any processes blocked by the current lock are woken up and allowed to
49 * run before the new lock is applied.
50 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
52 * Removed some race conditions in flock_lock_file(), marked other possible
53 * races. Just grep for FIXME to see them.
54 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
56 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
57 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
58 * once we've checked for blocking and deadlocking.
59 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
66 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
67 * check if a file has mandatory locks, used by mmap(), open() and creat() to
68 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
75 * Fixed deadlock condition for pathological code that mixes calls to
76 * flock() and fcntl().
77 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
79 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
80 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
81 * guarantee sensible behaviour in the case where file system modules might
82 * be compiled with different options than the kernel itself.
83 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
85 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
86 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
87 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
89 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
90 * locks. Changed process synchronisation to avoid dereferencing locks that
91 * have already been freed.
92 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
97 * Made mandatory locking a mount option. Default is not to allow mandatory
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
104 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
105 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
107 * Use slab allocator instead of kmalloc/kfree.
108 * Use generic list implementation from <linux/list.h>.
109 * Sped up posix_locks_deadlock by only considering blocked locks.
110 * Matthew Wilcox <willy@debian.org>, March, 2000.
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
117 #include <linux/capability.h>
118 #include <linux/file.h>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/module.h>
123 #include <linux/security.h>
124 #include <linux/slab.h>
125 #include <linux/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/rcupdate.h>
128 #include <linux/pid_namespace.h>
129 #include <linux/hashtable.h>
130 #include <linux/percpu.h>
131 #include <linux/lglock.h>
133 #define CREATE_TRACE_POINTS
134 #include <trace/events/filelock.h>
136 #include <asm/uaccess.h>
138 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
139 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
140 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG))
141 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
143 static bool lease_breaking(struct file_lock
*fl
)
145 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
148 static int target_leasetype(struct file_lock
*fl
)
150 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
152 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
157 int leases_enable
= 1;
158 int lease_break_time
= 45;
160 #define for_each_lock(inode, lockp) \
161 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
164 * The global file_lock_list is only used for displaying /proc/locks, so we
165 * keep a list on each CPU, with each list protected by its own spinlock via
166 * the file_lock_lglock. Note that alterations to the list also require that
167 * the relevant i_lock is held.
169 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
170 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
173 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
174 * It is protected by blocked_lock_lock.
176 * We hash locks by lockowner in order to optimize searching for the lock a
177 * particular lockowner is waiting on.
179 * FIXME: make this value scale via some heuristic? We generally will want more
180 * buckets when we have more lockowners holding locks, but that's a little
181 * difficult to determine without knowing what the workload will look like.
183 #define BLOCKED_HASH_BITS 7
184 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
187 * This lock protects the blocked_hash. Generally, if you're accessing it, you
188 * want to be holding this lock.
190 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
191 * pointer for file_lock structures that are acting as lock requests (in
192 * contrast to those that are acting as records of acquired locks).
194 * Note that when we acquire this lock in order to change the above fields,
195 * we often hold the i_lock as well. In certain cases, when reading the fields
196 * protected by this lock, we can skip acquiring it iff we already hold the
199 * In particular, adding an entry to the fl_block list requires that you hold
200 * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting
201 * an entry from the list however only requires the file_lock_lock.
203 static DEFINE_SPINLOCK(blocked_lock_lock
);
205 static struct kmem_cache
*filelock_cache __read_mostly
;
207 static void locks_init_lock_heads(struct file_lock
*fl
)
209 INIT_HLIST_NODE(&fl
->fl_link
);
210 INIT_LIST_HEAD(&fl
->fl_block
);
211 init_waitqueue_head(&fl
->fl_wait
);
214 /* Allocate an empty lock structure. */
215 struct file_lock
*locks_alloc_lock(void)
217 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
220 locks_init_lock_heads(fl
);
224 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
226 void locks_release_private(struct file_lock
*fl
)
229 if (fl
->fl_ops
->fl_release_private
)
230 fl
->fl_ops
->fl_release_private(fl
);
235 if (fl
->fl_lmops
->lm_put_owner
)
236 fl
->fl_lmops
->lm_put_owner(fl
);
240 EXPORT_SYMBOL_GPL(locks_release_private
);
242 /* Free a lock which is not in use. */
243 void locks_free_lock(struct file_lock
*fl
)
245 BUG_ON(waitqueue_active(&fl
->fl_wait
));
246 BUG_ON(!list_empty(&fl
->fl_block
));
247 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
249 locks_release_private(fl
);
250 kmem_cache_free(filelock_cache
, fl
);
252 EXPORT_SYMBOL(locks_free_lock
);
255 locks_dispose_list(struct list_head
*dispose
)
257 struct file_lock
*fl
;
259 while (!list_empty(dispose
)) {
260 fl
= list_first_entry(dispose
, struct file_lock
, fl_block
);
261 list_del_init(&fl
->fl_block
);
266 void locks_init_lock(struct file_lock
*fl
)
268 memset(fl
, 0, sizeof(struct file_lock
));
269 locks_init_lock_heads(fl
);
272 EXPORT_SYMBOL(locks_init_lock
);
274 static void locks_copy_private(struct file_lock
*new, struct file_lock
*fl
)
277 if (fl
->fl_ops
->fl_copy_lock
)
278 fl
->fl_ops
->fl_copy_lock(new, fl
);
279 new->fl_ops
= fl
->fl_ops
;
283 if (fl
->fl_lmops
->lm_get_owner
)
284 fl
->fl_lmops
->lm_get_owner(new, fl
);
285 new->fl_lmops
= fl
->fl_lmops
;
290 * Initialize a new lock from an existing file_lock structure.
292 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
294 new->fl_owner
= fl
->fl_owner
;
295 new->fl_pid
= fl
->fl_pid
;
297 new->fl_flags
= fl
->fl_flags
;
298 new->fl_type
= fl
->fl_type
;
299 new->fl_start
= fl
->fl_start
;
300 new->fl_end
= fl
->fl_end
;
302 new->fl_lmops
= NULL
;
304 EXPORT_SYMBOL(locks_copy_conflock
);
306 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
308 /* "new" must be a freshly-initialized lock */
309 WARN_ON_ONCE(new->fl_ops
);
311 locks_copy_conflock(new, fl
);
312 new->fl_file
= fl
->fl_file
;
313 new->fl_ops
= fl
->fl_ops
;
314 new->fl_lmops
= fl
->fl_lmops
;
316 locks_copy_private(new, fl
);
319 EXPORT_SYMBOL(locks_copy_lock
);
321 static inline int flock_translate_cmd(int cmd
) {
323 return cmd
& (LOCK_MAND
| LOCK_RW
);
335 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
336 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
339 struct file_lock
*fl
;
340 int type
= flock_translate_cmd(cmd
);
344 fl
= locks_alloc_lock();
350 fl
->fl_pid
= current
->tgid
;
351 fl
->fl_flags
= FL_FLOCK
;
353 fl
->fl_end
= OFFSET_MAX
;
359 static int assign_type(struct file_lock
*fl
, long type
)
373 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
376 switch (l
->l_whence
) {
381 fl
->fl_start
= filp
->f_pos
;
384 fl
->fl_start
= i_size_read(file_inode(filp
));
389 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
391 fl
->fl_start
+= l
->l_start
;
392 if (fl
->fl_start
< 0)
395 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
396 POSIX-2001 defines it. */
398 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
400 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
402 } else if (l
->l_len
< 0) {
403 if (fl
->fl_start
+ l
->l_len
< 0)
405 fl
->fl_end
= fl
->fl_start
- 1;
406 fl
->fl_start
+= l
->l_len
;
408 fl
->fl_end
= OFFSET_MAX
;
410 fl
->fl_owner
= current
->files
;
411 fl
->fl_pid
= current
->tgid
;
413 fl
->fl_flags
= FL_POSIX
;
417 return assign_type(fl
, l
->l_type
);
420 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
423 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
426 struct flock64 ll
= {
428 .l_whence
= l
->l_whence
,
429 .l_start
= l
->l_start
,
433 return flock64_to_posix_lock(filp
, fl
, &ll
);
436 /* default lease lock manager operations */
437 static void lease_break_callback(struct file_lock
*fl
)
439 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
442 static const struct lock_manager_operations lease_manager_ops
= {
443 .lm_break
= lease_break_callback
,
444 .lm_change
= lease_modify
,
448 * Initialize a lease, use the default lock manager operations
450 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
452 if (assign_type(fl
, type
) != 0)
455 fl
->fl_owner
= current
->files
;
456 fl
->fl_pid
= current
->tgid
;
459 fl
->fl_flags
= FL_LEASE
;
461 fl
->fl_end
= OFFSET_MAX
;
463 fl
->fl_lmops
= &lease_manager_ops
;
467 /* Allocate a file_lock initialised to this type of lease */
468 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
470 struct file_lock
*fl
= locks_alloc_lock();
474 return ERR_PTR(error
);
476 error
= lease_init(filp
, type
, fl
);
479 return ERR_PTR(error
);
484 /* Check if two locks overlap each other.
486 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
488 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
489 (fl2
->fl_end
>= fl1
->fl_start
));
493 * Check whether two locks have the same owner.
495 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
497 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
498 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
499 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
500 return fl1
->fl_owner
== fl2
->fl_owner
;
503 /* Must be called with the i_lock held! */
504 static void locks_insert_global_locks(struct file_lock
*fl
)
506 lg_local_lock(&file_lock_lglock
);
507 fl
->fl_link_cpu
= smp_processor_id();
508 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
509 lg_local_unlock(&file_lock_lglock
);
512 /* Must be called with the i_lock held! */
513 static void locks_delete_global_locks(struct file_lock
*fl
)
516 * Avoid taking lock if already unhashed. This is safe since this check
517 * is done while holding the i_lock, and new insertions into the list
518 * also require that it be held.
520 if (hlist_unhashed(&fl
->fl_link
))
522 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
523 hlist_del_init(&fl
->fl_link
);
524 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
528 posix_owner_key(struct file_lock
*fl
)
530 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
531 return fl
->fl_lmops
->lm_owner_key(fl
);
532 return (unsigned long)fl
->fl_owner
;
535 static void locks_insert_global_blocked(struct file_lock
*waiter
)
537 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
540 static void locks_delete_global_blocked(struct file_lock
*waiter
)
542 hash_del(&waiter
->fl_link
);
545 /* Remove waiter from blocker's block list.
546 * When blocker ends up pointing to itself then the list is empty.
548 * Must be called with blocked_lock_lock held.
550 static void __locks_delete_block(struct file_lock
*waiter
)
552 locks_delete_global_blocked(waiter
);
553 list_del_init(&waiter
->fl_block
);
554 waiter
->fl_next
= NULL
;
557 static void locks_delete_block(struct file_lock
*waiter
)
559 spin_lock(&blocked_lock_lock
);
560 __locks_delete_block(waiter
);
561 spin_unlock(&blocked_lock_lock
);
564 /* Insert waiter into blocker's block list.
565 * We use a circular list so that processes can be easily woken up in
566 * the order they blocked. The documentation doesn't require this but
567 * it seems like the reasonable thing to do.
569 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
570 * list itself is protected by the blocked_lock_lock, but by ensuring that the
571 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
572 * in some cases when we see that the fl_block list is empty.
574 static void __locks_insert_block(struct file_lock
*blocker
,
575 struct file_lock
*waiter
)
577 BUG_ON(!list_empty(&waiter
->fl_block
));
578 waiter
->fl_next
= blocker
;
579 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
580 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
581 locks_insert_global_blocked(waiter
);
584 /* Must be called with i_lock held. */
585 static void locks_insert_block(struct file_lock
*blocker
,
586 struct file_lock
*waiter
)
588 spin_lock(&blocked_lock_lock
);
589 __locks_insert_block(blocker
, waiter
);
590 spin_unlock(&blocked_lock_lock
);
594 * Wake up processes blocked waiting for blocker.
596 * Must be called with the inode->i_lock held!
598 static void locks_wake_up_blocks(struct file_lock
*blocker
)
601 * Avoid taking global lock if list is empty. This is safe since new
602 * blocked requests are only added to the list under the i_lock, and
603 * the i_lock is always held here. Note that removal from the fl_block
604 * list does not require the i_lock, so we must recheck list_empty()
605 * after acquiring the blocked_lock_lock.
607 if (list_empty(&blocker
->fl_block
))
610 spin_lock(&blocked_lock_lock
);
611 while (!list_empty(&blocker
->fl_block
)) {
612 struct file_lock
*waiter
;
614 waiter
= list_first_entry(&blocker
->fl_block
,
615 struct file_lock
, fl_block
);
616 __locks_delete_block(waiter
);
617 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
618 waiter
->fl_lmops
->lm_notify(waiter
);
620 wake_up(&waiter
->fl_wait
);
622 spin_unlock(&blocked_lock_lock
);
625 /* Insert file lock fl into an inode's lock list at the position indicated
626 * by pos. At the same time add the lock to the global file lock list.
628 * Must be called with the i_lock held!
630 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
632 fl
->fl_nspid
= get_pid(task_tgid(current
));
634 /* insert into file's list */
638 locks_insert_global_locks(fl
);
642 * locks_delete_lock - Delete a lock and then free it.
643 * @thisfl_p: pointer that points to the fl_next field of the previous
644 * inode->i_flock list entry
646 * Unlink a lock from all lists and free the namespace reference, but don't
647 * free it yet. Wake up processes that are blocked waiting for this lock and
648 * notify the FS that the lock has been cleared.
650 * Must be called with the i_lock held!
652 static void locks_unlink_lock(struct file_lock
**thisfl_p
)
654 struct file_lock
*fl
= *thisfl_p
;
656 locks_delete_global_locks(fl
);
658 *thisfl_p
= fl
->fl_next
;
662 put_pid(fl
->fl_nspid
);
666 locks_wake_up_blocks(fl
);
670 * Unlink a lock from all lists and free it.
672 * Must be called with i_lock held!
674 static void locks_delete_lock(struct file_lock
**thisfl_p
,
675 struct list_head
*dispose
)
677 struct file_lock
*fl
= *thisfl_p
;
679 locks_unlink_lock(thisfl_p
);
681 list_add(&fl
->fl_block
, dispose
);
686 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
687 * checks for shared/exclusive status of overlapping locks.
689 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
691 if (sys_fl
->fl_type
== F_WRLCK
)
693 if (caller_fl
->fl_type
== F_WRLCK
)
698 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
699 * checking before calling the locks_conflict().
701 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
703 /* POSIX locks owned by the same process do not conflict with
706 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
709 /* Check whether they overlap */
710 if (!locks_overlap(caller_fl
, sys_fl
))
713 return (locks_conflict(caller_fl
, sys_fl
));
716 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
717 * checking before calling the locks_conflict().
719 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
721 /* FLOCK locks referring to the same filp do not conflict with
724 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
726 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
729 return (locks_conflict(caller_fl
, sys_fl
));
733 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
735 struct file_lock
*cfl
;
736 struct inode
*inode
= file_inode(filp
);
738 spin_lock(&inode
->i_lock
);
739 for (cfl
= file_inode(filp
)->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
742 if (posix_locks_conflict(fl
, cfl
))
746 locks_copy_conflock(fl
, cfl
);
748 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
750 fl
->fl_type
= F_UNLCK
;
751 spin_unlock(&inode
->i_lock
);
754 EXPORT_SYMBOL(posix_test_lock
);
757 * Deadlock detection:
759 * We attempt to detect deadlocks that are due purely to posix file
762 * We assume that a task can be waiting for at most one lock at a time.
763 * So for any acquired lock, the process holding that lock may be
764 * waiting on at most one other lock. That lock in turns may be held by
765 * someone waiting for at most one other lock. Given a requested lock
766 * caller_fl which is about to wait for a conflicting lock block_fl, we
767 * follow this chain of waiters to ensure we are not about to create a
770 * Since we do this before we ever put a process to sleep on a lock, we
771 * are ensured that there is never a cycle; that is what guarantees that
772 * the while() loop in posix_locks_deadlock() eventually completes.
774 * Note: the above assumption may not be true when handling lock
775 * requests from a broken NFS client. It may also fail in the presence
776 * of tasks (such as posix threads) sharing the same open file table.
777 * To handle those cases, we just bail out after a few iterations.
779 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
780 * Because the owner is not even nominally tied to a thread of
781 * execution, the deadlock detection below can't reasonably work well. Just
784 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
785 * locks that just checks for the case where two tasks are attempting to
786 * upgrade from read to write locks on the same inode.
789 #define MAX_DEADLK_ITERATIONS 10
791 /* Find a lock that the owner of the given block_fl is blocking on. */
792 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
794 struct file_lock
*fl
;
796 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
797 if (posix_same_owner(fl
, block_fl
))
803 /* Must be called with the blocked_lock_lock held! */
804 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
805 struct file_lock
*block_fl
)
810 * This deadlock detector can't reasonably detect deadlocks with
811 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
813 if (IS_OFDLCK(caller_fl
))
816 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
817 if (i
++ > MAX_DEADLK_ITERATIONS
)
819 if (posix_same_owner(caller_fl
, block_fl
))
825 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
826 * after any leases, but before any posix locks.
828 * Note that if called with an FL_EXISTS argument, the caller may determine
829 * whether or not a lock was successfully freed by testing the return
832 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
834 struct file_lock
*new_fl
= NULL
;
835 struct file_lock
**before
;
836 struct inode
* inode
= file_inode(filp
);
841 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
842 new_fl
= locks_alloc_lock();
847 spin_lock(&inode
->i_lock
);
848 if (request
->fl_flags
& FL_ACCESS
)
851 for_each_lock(inode
, before
) {
852 struct file_lock
*fl
= *before
;
857 if (filp
!= fl
->fl_file
)
859 if (request
->fl_type
== fl
->fl_type
)
862 locks_delete_lock(before
, &dispose
);
866 if (request
->fl_type
== F_UNLCK
) {
867 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
873 * If a higher-priority process was blocked on the old file lock,
874 * give it the opportunity to lock the file.
877 spin_unlock(&inode
->i_lock
);
879 spin_lock(&inode
->i_lock
);
883 for_each_lock(inode
, before
) {
884 struct file_lock
*fl
= *before
;
889 if (!flock_locks_conflict(request
, fl
))
892 if (!(request
->fl_flags
& FL_SLEEP
))
894 error
= FILE_LOCK_DEFERRED
;
895 locks_insert_block(fl
, request
);
898 if (request
->fl_flags
& FL_ACCESS
)
900 locks_copy_lock(new_fl
, request
);
901 locks_insert_lock(before
, new_fl
);
906 spin_unlock(&inode
->i_lock
);
908 locks_free_lock(new_fl
);
909 locks_dispose_list(&dispose
);
913 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
915 struct file_lock
*fl
;
916 struct file_lock
*new_fl
= NULL
;
917 struct file_lock
*new_fl2
= NULL
;
918 struct file_lock
*left
= NULL
;
919 struct file_lock
*right
= NULL
;
920 struct file_lock
**before
;
926 * We may need two file_lock structures for this operation,
927 * so we get them in advance to avoid races.
929 * In some cases we can be sure, that no new locks will be needed
931 if (!(request
->fl_flags
& FL_ACCESS
) &&
932 (request
->fl_type
!= F_UNLCK
||
933 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
934 new_fl
= locks_alloc_lock();
935 new_fl2
= locks_alloc_lock();
938 spin_lock(&inode
->i_lock
);
940 * New lock request. Walk all POSIX locks and look for conflicts. If
941 * there are any, either return error or put the request on the
942 * blocker's list of waiters and the global blocked_hash.
944 if (request
->fl_type
!= F_UNLCK
) {
945 for_each_lock(inode
, before
) {
949 if (!posix_locks_conflict(request
, fl
))
952 locks_copy_conflock(conflock
, fl
);
954 if (!(request
->fl_flags
& FL_SLEEP
))
957 * Deadlock detection and insertion into the blocked
958 * locks list must be done while holding the same lock!
961 spin_lock(&blocked_lock_lock
);
962 if (likely(!posix_locks_deadlock(request
, fl
))) {
963 error
= FILE_LOCK_DEFERRED
;
964 __locks_insert_block(fl
, request
);
966 spin_unlock(&blocked_lock_lock
);
971 /* If we're just looking for a conflict, we're done. */
973 if (request
->fl_flags
& FL_ACCESS
)
977 * Find the first old lock with the same owner as the new lock.
980 before
= &inode
->i_flock
;
982 /* First skip locks owned by other processes. */
983 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
984 !posix_same_owner(request
, fl
))) {
985 before
= &fl
->fl_next
;
988 /* Process locks with this owner. */
989 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
990 /* Detect adjacent or overlapping regions (if same lock type)
992 if (request
->fl_type
== fl
->fl_type
) {
993 /* In all comparisons of start vs end, use
994 * "start - 1" rather than "end + 1". If end
995 * is OFFSET_MAX, end + 1 will become negative.
997 if (fl
->fl_end
< request
->fl_start
- 1)
999 /* If the next lock in the list has entirely bigger
1000 * addresses than the new one, insert the lock here.
1002 if (fl
->fl_start
- 1 > request
->fl_end
)
1005 /* If we come here, the new and old lock are of the
1006 * same type and adjacent or overlapping. Make one
1007 * lock yielding from the lower start address of both
1008 * locks to the higher end address.
1010 if (fl
->fl_start
> request
->fl_start
)
1011 fl
->fl_start
= request
->fl_start
;
1013 request
->fl_start
= fl
->fl_start
;
1014 if (fl
->fl_end
< request
->fl_end
)
1015 fl
->fl_end
= request
->fl_end
;
1017 request
->fl_end
= fl
->fl_end
;
1019 locks_delete_lock(before
, &dispose
);
1026 /* Processing for different lock types is a bit
1029 if (fl
->fl_end
< request
->fl_start
)
1031 if (fl
->fl_start
> request
->fl_end
)
1033 if (request
->fl_type
== F_UNLCK
)
1035 if (fl
->fl_start
< request
->fl_start
)
1037 /* If the next lock in the list has a higher end
1038 * address than the new one, insert the new one here.
1040 if (fl
->fl_end
> request
->fl_end
) {
1044 if (fl
->fl_start
>= request
->fl_start
) {
1045 /* The new lock completely replaces an old
1046 * one (This may happen several times).
1049 locks_delete_lock(before
, &dispose
);
1053 * Replace the old lock with new_fl, and
1054 * remove the old one. It's safe to do the
1055 * insert here since we know that we won't be
1056 * using new_fl later, and that the lock is
1057 * just replacing an existing lock.
1062 locks_copy_lock(new_fl
, request
);
1065 locks_delete_lock(before
, &dispose
);
1066 locks_insert_lock(before
, request
);
1070 /* Go on to next lock.
1073 before
= &fl
->fl_next
;
1077 * The above code only modifies existing locks in case of merging or
1078 * replacing. If new lock(s) need to be inserted all modifications are
1079 * done below this, so it's safe yet to bail out.
1081 error
= -ENOLCK
; /* "no luck" */
1082 if (right
&& left
== right
&& !new_fl2
)
1087 if (request
->fl_type
== F_UNLCK
) {
1088 if (request
->fl_flags
& FL_EXISTS
)
1097 locks_copy_lock(new_fl
, request
);
1098 locks_insert_lock(before
, new_fl
);
1102 if (left
== right
) {
1103 /* The new lock breaks the old one in two pieces,
1104 * so we have to use the second new lock.
1108 locks_copy_lock(left
, right
);
1109 locks_insert_lock(before
, left
);
1111 right
->fl_start
= request
->fl_end
+ 1;
1112 locks_wake_up_blocks(right
);
1115 left
->fl_end
= request
->fl_start
- 1;
1116 locks_wake_up_blocks(left
);
1119 spin_unlock(&inode
->i_lock
);
1121 * Free any unused locks.
1124 locks_free_lock(new_fl
);
1126 locks_free_lock(new_fl2
);
1127 locks_dispose_list(&dispose
);
1132 * posix_lock_file - Apply a POSIX-style lock to a file
1133 * @filp: The file to apply the lock to
1134 * @fl: The lock to be applied
1135 * @conflock: Place to return a copy of the conflicting lock, if found.
1137 * Add a POSIX style lock to a file.
1138 * We merge adjacent & overlapping locks whenever possible.
1139 * POSIX locks are sorted by owner task, then by starting address
1141 * Note that if called with an FL_EXISTS argument, the caller may determine
1142 * whether or not a lock was successfully freed by testing the return
1143 * value for -ENOENT.
1145 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1146 struct file_lock
*conflock
)
1148 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1150 EXPORT_SYMBOL(posix_lock_file
);
1153 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1154 * @filp: The file to apply the lock to
1155 * @fl: The lock to be applied
1157 * Add a POSIX style lock to a file.
1158 * We merge adjacent & overlapping locks whenever possible.
1159 * POSIX locks are sorted by owner task, then by starting address
1161 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1166 error
= posix_lock_file(filp
, fl
, NULL
);
1167 if (error
!= FILE_LOCK_DEFERRED
)
1169 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1173 locks_delete_block(fl
);
1178 EXPORT_SYMBOL(posix_lock_file_wait
);
1181 * locks_mandatory_locked - Check for an active lock
1182 * @file: the file to check
1184 * Searches the inode's list of locks to find any POSIX locks which conflict.
1185 * This function is called from locks_verify_locked() only.
1187 int locks_mandatory_locked(struct file
*file
)
1189 struct inode
*inode
= file_inode(file
);
1190 struct file_lock
*fl
;
1193 * Search the lock list for this inode for any POSIX locks.
1195 spin_lock(&inode
->i_lock
);
1196 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1199 if (fl
->fl_owner
!= current
->files
&&
1200 fl
->fl_owner
!= file
)
1203 spin_unlock(&inode
->i_lock
);
1204 return fl
? -EAGAIN
: 0;
1208 * locks_mandatory_area - Check for a conflicting lock
1209 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1211 * @inode: the file to check
1212 * @filp: how the file was opened (if it was)
1213 * @offset: start of area to check
1214 * @count: length of area to check
1216 * Searches the inode's list of locks to find any POSIX locks which conflict.
1217 * This function is called from rw_verify_area() and
1218 * locks_verify_truncate().
1220 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1221 struct file
*filp
, loff_t offset
,
1224 struct file_lock fl
;
1228 locks_init_lock(&fl
);
1229 fl
.fl_pid
= current
->tgid
;
1231 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1232 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1234 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1235 fl
.fl_start
= offset
;
1236 fl
.fl_end
= offset
+ count
- 1;
1241 fl
.fl_flags
&= ~FL_SLEEP
;
1242 error
= __posix_lock_file(inode
, &fl
, NULL
);
1248 fl
.fl_flags
|= FL_SLEEP
;
1249 fl
.fl_owner
= current
->files
;
1250 error
= __posix_lock_file(inode
, &fl
, NULL
);
1251 if (error
!= FILE_LOCK_DEFERRED
)
1253 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1256 * If we've been sleeping someone might have
1257 * changed the permissions behind our back.
1259 if (__mandatory_lock(inode
))
1263 locks_delete_block(&fl
);
1270 EXPORT_SYMBOL(locks_mandatory_area
);
1272 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1276 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1279 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1283 /* We already had a lease on this file; just change its type */
1284 int lease_modify(struct file_lock
**before
, int arg
)
1286 struct file_lock
*fl
= *before
;
1287 int error
= assign_type(fl
, arg
);
1291 lease_clear_pending(fl
, arg
);
1292 locks_wake_up_blocks(fl
);
1293 if (arg
== F_UNLCK
) {
1294 struct file
*filp
= fl
->fl_file
;
1297 filp
->f_owner
.signum
= 0;
1298 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1299 if (fl
->fl_fasync
!= NULL
) {
1300 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1301 fl
->fl_fasync
= NULL
;
1303 locks_delete_lock(before
, NULL
);
1308 EXPORT_SYMBOL(lease_modify
);
1310 static bool past_time(unsigned long then
)
1313 /* 0 is a special value meaning "this never expires": */
1315 return time_after(jiffies
, then
);
1318 static void time_out_leases(struct inode
*inode
)
1320 struct file_lock
**before
;
1321 struct file_lock
*fl
;
1323 before
= &inode
->i_flock
;
1324 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1325 trace_time_out_leases(inode
, fl
);
1326 if (past_time(fl
->fl_downgrade_time
))
1327 lease_modify(before
, F_RDLCK
);
1328 if (past_time(fl
->fl_break_time
))
1329 lease_modify(before
, F_UNLCK
);
1330 if (fl
== *before
) /* lease_modify may have freed fl */
1331 before
= &fl
->fl_next
;
1335 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1337 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1339 return locks_conflict(breaker
, lease
);
1343 * __break_lease - revoke all outstanding leases on file
1344 * @inode: the inode of the file to return
1345 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1347 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1350 * break_lease (inlined for speed) has checked there already is at least
1351 * some kind of lock (maybe a lease) on this file. Leases are broken on
1352 * a call to open() or truncate(). This function can sleep unless you
1353 * specified %O_NONBLOCK to your open().
1355 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1358 struct file_lock
*new_fl
, *flock
;
1359 struct file_lock
*fl
;
1360 unsigned long break_time
;
1361 int i_have_this_lease
= 0;
1362 bool lease_conflict
= false;
1363 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1365 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1367 return PTR_ERR(new_fl
);
1368 new_fl
->fl_flags
= type
;
1370 spin_lock(&inode
->i_lock
);
1372 time_out_leases(inode
);
1374 flock
= inode
->i_flock
;
1375 if ((flock
== NULL
) || !IS_LEASE(flock
))
1378 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1379 if (leases_conflict(fl
, new_fl
)) {
1380 lease_conflict
= true;
1381 if (fl
->fl_owner
== current
->files
)
1382 i_have_this_lease
= 1;
1385 if (!lease_conflict
)
1389 if (lease_break_time
> 0) {
1390 break_time
= jiffies
+ lease_break_time
* HZ
;
1391 if (break_time
== 0)
1392 break_time
++; /* so that 0 means no break time */
1395 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1396 if (!leases_conflict(fl
, new_fl
))
1399 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1401 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1402 fl
->fl_break_time
= break_time
;
1404 if (lease_breaking(flock
))
1406 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1407 fl
->fl_downgrade_time
= break_time
;
1409 fl
->fl_lmops
->lm_break(fl
);
1412 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1413 trace_break_lease_noblock(inode
, new_fl
);
1414 error
= -EWOULDBLOCK
;
1419 break_time
= flock
->fl_break_time
;
1420 if (break_time
!= 0)
1421 break_time
-= jiffies
;
1422 if (break_time
== 0)
1424 locks_insert_block(flock
, new_fl
);
1425 trace_break_lease_block(inode
, new_fl
);
1426 spin_unlock(&inode
->i_lock
);
1427 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1428 !new_fl
->fl_next
, break_time
);
1429 spin_lock(&inode
->i_lock
);
1430 trace_break_lease_unblock(inode
, new_fl
);
1431 locks_delete_block(new_fl
);
1434 time_out_leases(inode
);
1436 * Wait for the next conflicting lease that has not been
1439 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1440 flock
= flock
->fl_next
) {
1441 if (leases_conflict(new_fl
, flock
))
1448 spin_unlock(&inode
->i_lock
);
1449 locks_free_lock(new_fl
);
1453 EXPORT_SYMBOL(__break_lease
);
1456 * lease_get_mtime - get the last modified time of an inode
1458 * @time: pointer to a timespec which will contain the last modified time
1460 * This is to force NFS clients to flush their caches for files with
1461 * exclusive leases. The justification is that if someone has an
1462 * exclusive lease, then they could be modifying it.
1464 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1466 struct file_lock
*flock
= inode
->i_flock
;
1467 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
== F_WRLCK
))
1468 *time
= current_fs_time(inode
->i_sb
);
1470 *time
= inode
->i_mtime
;
1473 EXPORT_SYMBOL(lease_get_mtime
);
1476 * fcntl_getlease - Enquire what lease is currently active
1479 * The value returned by this function will be one of
1480 * (if no lease break is pending):
1482 * %F_RDLCK to indicate a shared lease is held.
1484 * %F_WRLCK to indicate an exclusive lease is held.
1486 * %F_UNLCK to indicate no lease is held.
1488 * (if a lease break is pending):
1490 * %F_RDLCK to indicate an exclusive lease needs to be
1491 * changed to a shared lease (or removed).
1493 * %F_UNLCK to indicate the lease needs to be removed.
1495 * XXX: sfr & willy disagree over whether F_INPROGRESS
1496 * should be returned to userspace.
1498 int fcntl_getlease(struct file
*filp
)
1500 struct file_lock
*fl
;
1501 struct inode
*inode
= file_inode(filp
);
1504 spin_lock(&inode
->i_lock
);
1505 time_out_leases(file_inode(filp
));
1506 for (fl
= file_inode(filp
)->i_flock
; fl
&& IS_LEASE(fl
);
1508 if (fl
->fl_file
== filp
) {
1509 type
= target_leasetype(fl
);
1513 spin_unlock(&inode
->i_lock
);
1518 * check_conflicting_open - see if the given dentry points to a file that has
1519 * an existing open that would conflict with the
1521 * @dentry: dentry to check
1522 * @arg: type of lease that we're trying to acquire
1524 * Check to see if there's an existing open fd on this file that would
1525 * conflict with the lease we're trying to set.
1528 check_conflicting_open(const struct dentry
*dentry
, const long arg
)
1531 struct inode
*inode
= dentry
->d_inode
;
1533 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1536 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1537 (atomic_read(&inode
->i_count
) > 1)))
1543 static int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1545 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1546 struct dentry
*dentry
= filp
->f_path
.dentry
;
1547 struct inode
*inode
= dentry
->d_inode
;
1548 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1552 trace_generic_add_lease(inode
, lease
);
1555 * In the delegation case we need mutual exclusion with
1556 * a number of operations that take the i_mutex. We trylock
1557 * because delegations are an optional optimization, and if
1558 * there's some chance of a conflict--we'd rather not
1559 * bother, maybe that's a sign this just isn't a good file to
1560 * hand out a delegation on.
1562 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1565 if (is_deleg
&& arg
== F_WRLCK
) {
1566 /* Write delegations are not currently supported: */
1567 mutex_unlock(&inode
->i_mutex
);
1572 error
= check_conflicting_open(dentry
, arg
);
1577 * At this point, we know that if there is an exclusive
1578 * lease on this file, then we hold it on this filp
1579 * (otherwise our open of this file would have blocked).
1580 * And if we are trying to acquire an exclusive lease,
1581 * then the file is not open by anyone (including us)
1582 * except for this filp.
1585 for (before
= &inode
->i_flock
;
1586 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1587 before
= &fl
->fl_next
) {
1588 if (fl
->fl_file
== filp
) {
1593 * No exclusive leases if someone else has a lease on
1599 * Modifying our existing lease is OK, but no getting a
1600 * new lease if someone else is opening for write:
1602 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1606 if (my_before
!= NULL
) {
1607 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1617 locks_insert_lock(before
, lease
);
1619 * The check in break_lease() is lockless. It's possible for another
1620 * open to race in after we did the earlier check for a conflicting
1621 * open but before the lease was inserted. Check again for a
1622 * conflicting open and cancel the lease if there is one.
1624 * We also add a barrier here to ensure that the insertion of the lock
1625 * precedes these checks.
1628 error
= check_conflicting_open(dentry
, arg
);
1630 locks_unlink_lock(before
);
1633 mutex_unlock(&inode
->i_mutex
);
1637 static int generic_delete_lease(struct file
*filp
, struct file_lock
**flp
)
1639 struct file_lock
*fl
, **before
;
1640 struct dentry
*dentry
= filp
->f_path
.dentry
;
1641 struct inode
*inode
= dentry
->d_inode
;
1643 trace_generic_delete_lease(inode
, *flp
);
1645 for (before
= &inode
->i_flock
;
1646 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1647 before
= &fl
->fl_next
) {
1648 if (fl
->fl_file
!= filp
)
1650 return (*flp
)->fl_lmops
->lm_change(before
, F_UNLCK
);
1656 * generic_setlease - sets a lease on an open file
1657 * @filp: file pointer
1658 * @arg: type of lease to obtain
1659 * @flp: input - file_lock to use, output - file_lock inserted
1661 * The (input) flp->fl_lmops->lm_break function is required
1664 * Called with inode->i_lock held.
1666 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1668 struct dentry
*dentry
= filp
->f_path
.dentry
;
1669 struct inode
*inode
= dentry
->d_inode
;
1672 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1674 if (!S_ISREG(inode
->i_mode
))
1676 error
= security_file_lock(filp
, arg
);
1680 time_out_leases(inode
);
1682 BUG_ON(!(*flp
)->fl_lmops
->lm_break
);
1686 return generic_delete_lease(filp
, flp
);
1689 return generic_add_lease(filp
, arg
, flp
);
1694 EXPORT_SYMBOL(generic_setlease
);
1696 static int __vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1698 if (filp
->f_op
->setlease
)
1699 return filp
->f_op
->setlease(filp
, arg
, lease
);
1701 return generic_setlease(filp
, arg
, lease
);
1705 * vfs_setlease - sets a lease on an open file
1706 * @filp: file pointer
1707 * @arg: type of lease to obtain
1708 * @lease: file_lock to use
1710 * Call this to establish a lease on the file.
1711 * The (*lease)->fl_lmops->lm_break operation must be set; if not,
1712 * break_lease will oops!
1714 * This will call the filesystem's setlease file method, if
1715 * defined. Note that there is no getlease method; instead, the
1716 * filesystem setlease method should call back to setlease() to
1717 * add a lease to the inode's lease list, where fcntl_getlease() can
1718 * find it. Since fcntl_getlease() only reports whether the current
1719 * task holds a lease, a cluster filesystem need only do this for
1720 * leases held by processes on this node.
1722 * There is also no break_lease method; filesystems that
1723 * handle their own leases should break leases themselves from the
1724 * filesystem's open, create, and (on truncate) setattr methods.
1726 * Warning: the only current setlease methods exist only to disable
1727 * leases in certain cases. More vfs changes may be required to
1728 * allow a full filesystem lease implementation.
1731 int vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1733 struct inode
*inode
= file_inode(filp
);
1736 spin_lock(&inode
->i_lock
);
1737 error
= __vfs_setlease(filp
, arg
, lease
);
1738 spin_unlock(&inode
->i_lock
);
1742 EXPORT_SYMBOL_GPL(vfs_setlease
);
1744 static int do_fcntl_delete_lease(struct file
*filp
)
1746 struct file_lock fl
, *flp
= &fl
;
1748 lease_init(filp
, F_UNLCK
, flp
);
1750 return vfs_setlease(filp
, F_UNLCK
, &flp
);
1753 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1755 struct file_lock
*fl
, *ret
;
1756 struct inode
*inode
= file_inode(filp
);
1757 struct fasync_struct
*new;
1760 fl
= lease_alloc(filp
, arg
);
1764 new = fasync_alloc();
1766 locks_free_lock(fl
);
1770 spin_lock(&inode
->i_lock
);
1771 error
= __vfs_setlease(filp
, arg
, &ret
);
1778 * fasync_insert_entry() returns the old entry if any.
1779 * If there was no old entry, then it used 'new' and
1780 * inserted it into the fasync list. Clear new so that
1781 * we don't release it here.
1783 if (!fasync_insert_entry(fd
, filp
, &ret
->fl_fasync
, new))
1786 error
= __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
1788 spin_unlock(&inode
->i_lock
);
1790 locks_free_lock(fl
);
1797 * fcntl_setlease - sets a lease on an open file
1798 * @fd: open file descriptor
1799 * @filp: file pointer
1800 * @arg: type of lease to obtain
1802 * Call this fcntl to establish a lease on the file.
1803 * Note that you also need to call %F_SETSIG to
1804 * receive a signal when the lease is broken.
1806 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1809 return do_fcntl_delete_lease(filp
);
1810 return do_fcntl_add_lease(fd
, filp
, arg
);
1814 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1815 * @filp: The file to apply the lock to
1816 * @fl: The lock to be applied
1818 * Add a FLOCK style lock to a file.
1820 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1825 error
= flock_lock_file(filp
, fl
);
1826 if (error
!= FILE_LOCK_DEFERRED
)
1828 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1832 locks_delete_block(fl
);
1838 EXPORT_SYMBOL(flock_lock_file_wait
);
1841 * sys_flock: - flock() system call.
1842 * @fd: the file descriptor to lock.
1843 * @cmd: the type of lock to apply.
1845 * Apply a %FL_FLOCK style lock to an open file descriptor.
1846 * The @cmd can be one of
1848 * %LOCK_SH -- a shared lock.
1850 * %LOCK_EX -- an exclusive lock.
1852 * %LOCK_UN -- remove an existing lock.
1854 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1856 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1857 * processes read and write access respectively.
1859 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1861 struct fd f
= fdget(fd
);
1862 struct file_lock
*lock
;
1863 int can_sleep
, unlock
;
1870 can_sleep
= !(cmd
& LOCK_NB
);
1872 unlock
= (cmd
== LOCK_UN
);
1874 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1875 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1878 error
= flock_make_lock(f
.file
, &lock
, cmd
);
1882 lock
->fl_flags
|= FL_SLEEP
;
1884 error
= security_file_lock(f
.file
, lock
->fl_type
);
1888 if (f
.file
->f_op
->flock
)
1889 error
= f
.file
->f_op
->flock(f
.file
,
1890 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1893 error
= flock_lock_file_wait(f
.file
, lock
);
1896 locks_free_lock(lock
);
1905 * vfs_test_lock - test file byte range lock
1906 * @filp: The file to test lock for
1907 * @fl: The lock to test; also used to hold result
1909 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1910 * setting conf->fl_type to something other than F_UNLCK.
1912 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1914 if (filp
->f_op
->lock
)
1915 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1916 posix_test_lock(filp
, fl
);
1919 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1921 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1923 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1924 #if BITS_PER_LONG == 32
1926 * Make sure we can represent the posix lock via
1927 * legacy 32bit flock.
1929 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1931 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1934 flock
->l_start
= fl
->fl_start
;
1935 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1936 fl
->fl_end
- fl
->fl_start
+ 1;
1937 flock
->l_whence
= 0;
1938 flock
->l_type
= fl
->fl_type
;
1942 #if BITS_PER_LONG == 32
1943 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1945 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1946 flock
->l_start
= fl
->fl_start
;
1947 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1948 fl
->fl_end
- fl
->fl_start
+ 1;
1949 flock
->l_whence
= 0;
1950 flock
->l_type
= fl
->fl_type
;
1954 /* Report the first existing lock that would conflict with l.
1955 * This implements the F_GETLK command of fcntl().
1957 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
1959 struct file_lock file_lock
;
1964 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1967 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1970 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1974 if (cmd
== F_OFD_GETLK
) {
1976 if (flock
.l_pid
!= 0)
1980 file_lock
.fl_flags
|= FL_OFDLCK
;
1981 file_lock
.fl_owner
= filp
;
1984 error
= vfs_test_lock(filp
, &file_lock
);
1988 flock
.l_type
= file_lock
.fl_type
;
1989 if (file_lock
.fl_type
!= F_UNLCK
) {
1990 error
= posix_lock_to_flock(&flock
, &file_lock
);
1995 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2002 * vfs_lock_file - file byte range lock
2003 * @filp: The file to apply the lock to
2004 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2005 * @fl: The lock to be applied
2006 * @conf: Place to return a copy of the conflicting lock, if found.
2008 * A caller that doesn't care about the conflicting lock may pass NULL
2009 * as the final argument.
2011 * If the filesystem defines a private ->lock() method, then @conf will
2012 * be left unchanged; so a caller that cares should initialize it to
2013 * some acceptable default.
2015 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2016 * locks, the ->lock() interface may return asynchronously, before the lock has
2017 * been granted or denied by the underlying filesystem, if (and only if)
2018 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2019 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2020 * the request is for a blocking lock. When ->lock() does return asynchronously,
2021 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2022 * request completes.
2023 * If the request is for non-blocking lock the file system should return
2024 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2025 * with the result. If the request timed out the callback routine will return a
2026 * nonzero return code and the file system should release the lock. The file
2027 * system is also responsible to keep a corresponding posix lock when it
2028 * grants a lock so the VFS can find out which locks are locally held and do
2029 * the correct lock cleanup when required.
2030 * The underlying filesystem must not drop the kernel lock or call
2031 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2034 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2036 if (filp
->f_op
->lock
)
2037 return filp
->f_op
->lock(filp
, cmd
, fl
);
2039 return posix_lock_file(filp
, fl
, conf
);
2041 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2043 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2044 struct file_lock
*fl
)
2048 error
= security_file_lock(filp
, fl
->fl_type
);
2053 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2054 if (error
!= FILE_LOCK_DEFERRED
)
2056 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2060 locks_delete_block(fl
);
2067 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2069 check_fmode_for_setlk(struct file_lock
*fl
)
2071 switch (fl
->fl_type
) {
2073 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2077 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2083 /* Apply the lock described by l to an open file descriptor.
2084 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2086 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2087 struct flock __user
*l
)
2089 struct file_lock
*file_lock
= locks_alloc_lock();
2091 struct inode
*inode
;
2095 if (file_lock
== NULL
)
2099 * This might block, so we do it before checking the inode.
2102 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2105 inode
= file_inode(filp
);
2107 /* Don't allow mandatory locks on files that may be memory mapped
2110 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2116 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2120 error
= check_fmode_for_setlk(file_lock
);
2125 * If the cmd is requesting file-private locks, then set the
2126 * FL_OFDLCK flag and override the owner.
2131 if (flock
.l_pid
!= 0)
2135 file_lock
->fl_flags
|= FL_OFDLCK
;
2136 file_lock
->fl_owner
= filp
;
2140 if (flock
.l_pid
!= 0)
2144 file_lock
->fl_flags
|= FL_OFDLCK
;
2145 file_lock
->fl_owner
= filp
;
2148 file_lock
->fl_flags
|= FL_SLEEP
;
2151 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2154 * Attempt to detect a close/fcntl race and recover by
2155 * releasing the lock that was just acquired.
2158 * we need that spin_lock here - it prevents reordering between
2159 * update of inode->i_flock and check for it done in close().
2160 * rcu_read_lock() wouldn't do.
2162 spin_lock(¤t
->files
->file_lock
);
2164 spin_unlock(¤t
->files
->file_lock
);
2165 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2166 flock
.l_type
= F_UNLCK
;
2171 locks_free_lock(file_lock
);
2175 #if BITS_PER_LONG == 32
2176 /* Report the first existing lock that would conflict with l.
2177 * This implements the F_GETLK command of fcntl().
2179 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2181 struct file_lock file_lock
;
2182 struct flock64 flock
;
2186 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2189 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2192 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2196 if (cmd
== F_OFD_GETLK
) {
2198 if (flock
.l_pid
!= 0)
2202 file_lock
.fl_flags
|= FL_OFDLCK
;
2203 file_lock
.fl_owner
= filp
;
2206 error
= vfs_test_lock(filp
, &file_lock
);
2210 flock
.l_type
= file_lock
.fl_type
;
2211 if (file_lock
.fl_type
!= F_UNLCK
)
2212 posix_lock_to_flock64(&flock
, &file_lock
);
2215 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2222 /* Apply the lock described by l to an open file descriptor.
2223 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2225 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2226 struct flock64 __user
*l
)
2228 struct file_lock
*file_lock
= locks_alloc_lock();
2229 struct flock64 flock
;
2230 struct inode
*inode
;
2234 if (file_lock
== NULL
)
2238 * This might block, so we do it before checking the inode.
2241 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2244 inode
= file_inode(filp
);
2246 /* Don't allow mandatory locks on files that may be memory mapped
2249 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2255 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2259 error
= check_fmode_for_setlk(file_lock
);
2264 * If the cmd is requesting file-private locks, then set the
2265 * FL_OFDLCK flag and override the owner.
2270 if (flock
.l_pid
!= 0)
2274 file_lock
->fl_flags
|= FL_OFDLCK
;
2275 file_lock
->fl_owner
= filp
;
2279 if (flock
.l_pid
!= 0)
2283 file_lock
->fl_flags
|= FL_OFDLCK
;
2284 file_lock
->fl_owner
= filp
;
2287 file_lock
->fl_flags
|= FL_SLEEP
;
2290 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2293 * Attempt to detect a close/fcntl race and recover by
2294 * releasing the lock that was just acquired.
2296 spin_lock(¤t
->files
->file_lock
);
2298 spin_unlock(¤t
->files
->file_lock
);
2299 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2300 flock
.l_type
= F_UNLCK
;
2305 locks_free_lock(file_lock
);
2308 #endif /* BITS_PER_LONG == 32 */
2311 * This function is called when the file is being removed
2312 * from the task's fd array. POSIX locks belonging to this task
2313 * are deleted at this time.
2315 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2317 struct file_lock lock
;
2320 * If there are no locks held on this file, we don't need to call
2321 * posix_lock_file(). Another process could be setting a lock on this
2322 * file at the same time, but we wouldn't remove that lock anyway.
2324 if (!file_inode(filp
)->i_flock
)
2327 lock
.fl_type
= F_UNLCK
;
2328 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2330 lock
.fl_end
= OFFSET_MAX
;
2331 lock
.fl_owner
= owner
;
2332 lock
.fl_pid
= current
->tgid
;
2333 lock
.fl_file
= filp
;
2335 lock
.fl_lmops
= NULL
;
2337 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2339 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2340 lock
.fl_ops
->fl_release_private(&lock
);
2343 EXPORT_SYMBOL(locks_remove_posix
);
2346 * This function is called on the last close of an open file.
2348 void locks_remove_file(struct file
*filp
)
2350 struct inode
* inode
= file_inode(filp
);
2351 struct file_lock
*fl
;
2352 struct file_lock
**before
;
2355 if (!inode
->i_flock
)
2358 locks_remove_posix(filp
, filp
);
2360 if (filp
->f_op
->flock
) {
2361 struct file_lock fl
= {
2363 .fl_pid
= current
->tgid
,
2365 .fl_flags
= FL_FLOCK
,
2367 .fl_end
= OFFSET_MAX
,
2369 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2370 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2371 fl
.fl_ops
->fl_release_private(&fl
);
2374 spin_lock(&inode
->i_lock
);
2375 before
= &inode
->i_flock
;
2377 while ((fl
= *before
) != NULL
) {
2378 if (fl
->fl_file
== filp
) {
2380 lease_modify(before
, F_UNLCK
);
2385 * There's a leftover lock on the list of a type that
2386 * we didn't expect to see. Most likely a classic
2387 * POSIX lock that ended up not getting released
2388 * properly, or that raced onto the list somehow. Log
2389 * some info about it and then just remove it from
2393 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2394 MAJOR(inode
->i_sb
->s_dev
),
2395 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
,
2396 fl
->fl_type
, fl
->fl_flags
,
2397 fl
->fl_start
, fl
->fl_end
);
2399 locks_delete_lock(before
, &dispose
);
2402 before
= &fl
->fl_next
;
2404 spin_unlock(&inode
->i_lock
);
2405 locks_dispose_list(&dispose
);
2409 * posix_unblock_lock - stop waiting for a file lock
2410 * @waiter: the lock which was waiting
2412 * lockd needs to block waiting for locks.
2415 posix_unblock_lock(struct file_lock
*waiter
)
2419 spin_lock(&blocked_lock_lock
);
2420 if (waiter
->fl_next
)
2421 __locks_delete_block(waiter
);
2424 spin_unlock(&blocked_lock_lock
);
2427 EXPORT_SYMBOL(posix_unblock_lock
);
2430 * vfs_cancel_lock - file byte range unblock lock
2431 * @filp: The file to apply the unblock to
2432 * @fl: The lock to be unblocked
2434 * Used by lock managers to cancel blocked requests
2436 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2438 if (filp
->f_op
->lock
)
2439 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2443 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2445 #ifdef CONFIG_PROC_FS
2446 #include <linux/proc_fs.h>
2447 #include <linux/seq_file.h>
2449 struct locks_iterator
{
2454 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2455 loff_t id
, char *pfx
)
2457 struct inode
*inode
= NULL
;
2458 unsigned int fl_pid
;
2461 fl_pid
= pid_vnr(fl
->fl_nspid
);
2463 fl_pid
= fl
->fl_pid
;
2465 if (fl
->fl_file
!= NULL
)
2466 inode
= file_inode(fl
->fl_file
);
2468 seq_printf(f
, "%lld:%s ", id
, pfx
);
2470 if (fl
->fl_flags
& FL_ACCESS
)
2471 seq_puts(f
, "ACCESS");
2472 else if (IS_OFDLCK(fl
))
2473 seq_puts(f
, "OFDLCK");
2475 seq_puts(f
, "POSIX ");
2477 seq_printf(f
, " %s ",
2478 (inode
== NULL
) ? "*NOINODE*" :
2479 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2480 } else if (IS_FLOCK(fl
)) {
2481 if (fl
->fl_type
& LOCK_MAND
) {
2482 seq_puts(f
, "FLOCK MSNFS ");
2484 seq_puts(f
, "FLOCK ADVISORY ");
2486 } else if (IS_LEASE(fl
)) {
2487 if (fl
->fl_flags
& FL_DELEG
)
2488 seq_puts(f
, "DELEG ");
2490 seq_puts(f
, "LEASE ");
2492 if (lease_breaking(fl
))
2493 seq_puts(f
, "BREAKING ");
2494 else if (fl
->fl_file
)
2495 seq_puts(f
, "ACTIVE ");
2497 seq_puts(f
, "BREAKER ");
2499 seq_puts(f
, "UNKNOWN UNKNOWN ");
2501 if (fl
->fl_type
& LOCK_MAND
) {
2502 seq_printf(f
, "%s ",
2503 (fl
->fl_type
& LOCK_READ
)
2504 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2505 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2507 seq_printf(f
, "%s ",
2508 (lease_breaking(fl
))
2509 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2510 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2513 #ifdef WE_CAN_BREAK_LSLK_NOW
2514 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2515 inode
->i_sb
->s_id
, inode
->i_ino
);
2517 /* userspace relies on this representation of dev_t ;-( */
2518 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2519 MAJOR(inode
->i_sb
->s_dev
),
2520 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2523 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2526 if (fl
->fl_end
== OFFSET_MAX
)
2527 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2529 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2531 seq_puts(f
, "0 EOF\n");
2535 static int locks_show(struct seq_file
*f
, void *v
)
2537 struct locks_iterator
*iter
= f
->private;
2538 struct file_lock
*fl
, *bfl
;
2540 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2542 lock_get_status(f
, fl
, iter
->li_pos
, "");
2544 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2545 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2550 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2551 __acquires(&blocked_lock_lock
)
2553 struct locks_iterator
*iter
= f
->private;
2555 iter
->li_pos
= *pos
+ 1;
2556 lg_global_lock(&file_lock_lglock
);
2557 spin_lock(&blocked_lock_lock
);
2558 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2561 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2563 struct locks_iterator
*iter
= f
->private;
2566 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2569 static void locks_stop(struct seq_file
*f
, void *v
)
2570 __releases(&blocked_lock_lock
)
2572 spin_unlock(&blocked_lock_lock
);
2573 lg_global_unlock(&file_lock_lglock
);
2576 static const struct seq_operations locks_seq_operations
= {
2577 .start
= locks_start
,
2583 static int locks_open(struct inode
*inode
, struct file
*filp
)
2585 return seq_open_private(filp
, &locks_seq_operations
,
2586 sizeof(struct locks_iterator
));
2589 static const struct file_operations proc_locks_operations
= {
2592 .llseek
= seq_lseek
,
2593 .release
= seq_release_private
,
2596 static int __init
proc_locks_init(void)
2598 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2601 module_init(proc_locks_init
);
2605 * lock_may_read - checks that the region is free of locks
2606 * @inode: the inode that is being read
2607 * @start: the first byte to read
2608 * @len: the number of bytes to read
2610 * Emulates Windows locking requirements. Whole-file
2611 * mandatory locks (share modes) can prohibit a read and
2612 * byte-range POSIX locks can prohibit a read if they overlap.
2614 * N.B. this function is only ever called
2615 * from knfsd and ownership of locks is never checked.
2617 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
2619 struct file_lock
*fl
;
2622 spin_lock(&inode
->i_lock
);
2623 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2625 if (fl
->fl_type
== F_RDLCK
)
2627 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2629 } else if (IS_FLOCK(fl
)) {
2630 if (!(fl
->fl_type
& LOCK_MAND
))
2632 if (fl
->fl_type
& LOCK_READ
)
2639 spin_unlock(&inode
->i_lock
);
2643 EXPORT_SYMBOL(lock_may_read
);
2646 * lock_may_write - checks that the region is free of locks
2647 * @inode: the inode that is being written
2648 * @start: the first byte to write
2649 * @len: the number of bytes to write
2651 * Emulates Windows locking requirements. Whole-file
2652 * mandatory locks (share modes) can prohibit a write and
2653 * byte-range POSIX locks can prohibit a write if they overlap.
2655 * N.B. this function is only ever called
2656 * from knfsd and ownership of locks is never checked.
2658 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2660 struct file_lock
*fl
;
2663 spin_lock(&inode
->i_lock
);
2664 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2666 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2668 } else if (IS_FLOCK(fl
)) {
2669 if (!(fl
->fl_type
& LOCK_MAND
))
2671 if (fl
->fl_type
& LOCK_WRITE
)
2678 spin_unlock(&inode
->i_lock
);
2682 EXPORT_SYMBOL(lock_may_write
);
2684 static int __init
filelock_init(void)
2688 filelock_cache
= kmem_cache_create("file_lock_cache",
2689 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2691 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2693 for_each_possible_cpu(i
)
2694 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2699 core_initcall(filelock_init
);