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
);
236 EXPORT_SYMBOL_GPL(locks_release_private
);
238 /* Free a lock which is not in use. */
239 void locks_free_lock(struct file_lock
*fl
)
241 BUG_ON(waitqueue_active(&fl
->fl_wait
));
242 BUG_ON(!list_empty(&fl
->fl_block
));
243 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
245 locks_release_private(fl
);
246 kmem_cache_free(filelock_cache
, fl
);
248 EXPORT_SYMBOL(locks_free_lock
);
250 void locks_init_lock(struct file_lock
*fl
)
252 memset(fl
, 0, sizeof(struct file_lock
));
253 locks_init_lock_heads(fl
);
256 EXPORT_SYMBOL(locks_init_lock
);
258 static void locks_copy_private(struct file_lock
*new, struct file_lock
*fl
)
261 if (fl
->fl_ops
->fl_copy_lock
)
262 fl
->fl_ops
->fl_copy_lock(new, fl
);
263 new->fl_ops
= fl
->fl_ops
;
266 new->fl_lmops
= fl
->fl_lmops
;
270 * Initialize a new lock from an existing file_lock structure.
272 void __locks_copy_lock(struct file_lock
*new, const struct file_lock
*fl
)
274 new->fl_owner
= fl
->fl_owner
;
275 new->fl_pid
= fl
->fl_pid
;
277 new->fl_flags
= fl
->fl_flags
;
278 new->fl_type
= fl
->fl_type
;
279 new->fl_start
= fl
->fl_start
;
280 new->fl_end
= fl
->fl_end
;
282 new->fl_lmops
= NULL
;
284 EXPORT_SYMBOL(__locks_copy_lock
);
286 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
288 /* "new" must be a freshly-initialized lock */
289 WARN_ON_ONCE(new->fl_ops
);
291 __locks_copy_lock(new, fl
);
292 new->fl_file
= fl
->fl_file
;
293 new->fl_ops
= fl
->fl_ops
;
294 new->fl_lmops
= fl
->fl_lmops
;
296 locks_copy_private(new, fl
);
299 EXPORT_SYMBOL(locks_copy_lock
);
301 static inline int flock_translate_cmd(int cmd
) {
303 return cmd
& (LOCK_MAND
| LOCK_RW
);
315 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
316 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
319 struct file_lock
*fl
;
320 int type
= flock_translate_cmd(cmd
);
324 fl
= locks_alloc_lock();
330 fl
->fl_pid
= current
->tgid
;
331 fl
->fl_flags
= FL_FLOCK
;
333 fl
->fl_end
= OFFSET_MAX
;
339 static int assign_type(struct file_lock
*fl
, long type
)
353 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
356 switch (l
->l_whence
) {
361 fl
->fl_start
= filp
->f_pos
;
364 fl
->fl_start
= i_size_read(file_inode(filp
));
369 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
371 fl
->fl_start
+= l
->l_start
;
372 if (fl
->fl_start
< 0)
375 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
376 POSIX-2001 defines it. */
378 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
380 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
382 } else if (l
->l_len
< 0) {
383 if (fl
->fl_start
+ l
->l_len
< 0)
385 fl
->fl_end
= fl
->fl_start
- 1;
386 fl
->fl_start
+= l
->l_len
;
388 fl
->fl_end
= OFFSET_MAX
;
390 fl
->fl_owner
= current
->files
;
391 fl
->fl_pid
= current
->tgid
;
393 fl
->fl_flags
= FL_POSIX
;
397 return assign_type(fl
, l
->l_type
);
400 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
403 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
406 struct flock64 ll
= {
408 .l_whence
= l
->l_whence
,
409 .l_start
= l
->l_start
,
413 return flock64_to_posix_lock(filp
, fl
, &ll
);
416 /* default lease lock manager operations */
417 static void lease_break_callback(struct file_lock
*fl
)
419 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
422 static const struct lock_manager_operations lease_manager_ops
= {
423 .lm_break
= lease_break_callback
,
424 .lm_change
= lease_modify
,
428 * Initialize a lease, use the default lock manager operations
430 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
432 if (assign_type(fl
, type
) != 0)
435 fl
->fl_owner
= current
->files
;
436 fl
->fl_pid
= current
->tgid
;
439 fl
->fl_flags
= FL_LEASE
;
441 fl
->fl_end
= OFFSET_MAX
;
443 fl
->fl_lmops
= &lease_manager_ops
;
447 /* Allocate a file_lock initialised to this type of lease */
448 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
450 struct file_lock
*fl
= locks_alloc_lock();
454 return ERR_PTR(error
);
456 error
= lease_init(filp
, type
, fl
);
459 return ERR_PTR(error
);
464 /* Check if two locks overlap each other.
466 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
468 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
469 (fl2
->fl_end
>= fl1
->fl_start
));
473 * Check whether two locks have the same owner.
475 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
477 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
478 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
479 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
480 return fl1
->fl_owner
== fl2
->fl_owner
;
483 /* Must be called with the i_lock held! */
484 static void locks_insert_global_locks(struct file_lock
*fl
)
486 lg_local_lock(&file_lock_lglock
);
487 fl
->fl_link_cpu
= smp_processor_id();
488 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
489 lg_local_unlock(&file_lock_lglock
);
492 /* Must be called with the i_lock held! */
493 static void locks_delete_global_locks(struct file_lock
*fl
)
496 * Avoid taking lock if already unhashed. This is safe since this check
497 * is done while holding the i_lock, and new insertions into the list
498 * also require that it be held.
500 if (hlist_unhashed(&fl
->fl_link
))
502 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
503 hlist_del_init(&fl
->fl_link
);
504 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
508 posix_owner_key(struct file_lock
*fl
)
510 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
511 return fl
->fl_lmops
->lm_owner_key(fl
);
512 return (unsigned long)fl
->fl_owner
;
515 static void locks_insert_global_blocked(struct file_lock
*waiter
)
517 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
520 static void locks_delete_global_blocked(struct file_lock
*waiter
)
522 hash_del(&waiter
->fl_link
);
525 /* Remove waiter from blocker's block list.
526 * When blocker ends up pointing to itself then the list is empty.
528 * Must be called with blocked_lock_lock held.
530 static void __locks_delete_block(struct file_lock
*waiter
)
532 locks_delete_global_blocked(waiter
);
533 list_del_init(&waiter
->fl_block
);
534 waiter
->fl_next
= NULL
;
537 static void locks_delete_block(struct file_lock
*waiter
)
539 spin_lock(&blocked_lock_lock
);
540 __locks_delete_block(waiter
);
541 spin_unlock(&blocked_lock_lock
);
544 /* Insert waiter into blocker's block list.
545 * We use a circular list so that processes can be easily woken up in
546 * the order they blocked. The documentation doesn't require this but
547 * it seems like the reasonable thing to do.
549 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
550 * list itself is protected by the blocked_lock_lock, but by ensuring that the
551 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
552 * in some cases when we see that the fl_block list is empty.
554 static void __locks_insert_block(struct file_lock
*blocker
,
555 struct file_lock
*waiter
)
557 BUG_ON(!list_empty(&waiter
->fl_block
));
558 waiter
->fl_next
= blocker
;
559 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
560 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
561 locks_insert_global_blocked(waiter
);
564 /* Must be called with i_lock held. */
565 static void locks_insert_block(struct file_lock
*blocker
,
566 struct file_lock
*waiter
)
568 spin_lock(&blocked_lock_lock
);
569 __locks_insert_block(blocker
, waiter
);
570 spin_unlock(&blocked_lock_lock
);
574 * Wake up processes blocked waiting for blocker.
576 * Must be called with the inode->i_lock held!
578 static void locks_wake_up_blocks(struct file_lock
*blocker
)
581 * Avoid taking global lock if list is empty. This is safe since new
582 * blocked requests are only added to the list under the i_lock, and
583 * the i_lock is always held here. Note that removal from the fl_block
584 * list does not require the i_lock, so we must recheck list_empty()
585 * after acquiring the blocked_lock_lock.
587 if (list_empty(&blocker
->fl_block
))
590 spin_lock(&blocked_lock_lock
);
591 while (!list_empty(&blocker
->fl_block
)) {
592 struct file_lock
*waiter
;
594 waiter
= list_first_entry(&blocker
->fl_block
,
595 struct file_lock
, fl_block
);
596 __locks_delete_block(waiter
);
597 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
598 waiter
->fl_lmops
->lm_notify(waiter
);
600 wake_up(&waiter
->fl_wait
);
602 spin_unlock(&blocked_lock_lock
);
605 /* Insert file lock fl into an inode's lock list at the position indicated
606 * by pos. At the same time add the lock to the global file lock list.
608 * Must be called with the i_lock held!
610 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
612 fl
->fl_nspid
= get_pid(task_tgid(current
));
614 /* insert into file's list */
618 locks_insert_global_locks(fl
);
622 * locks_delete_lock - Delete a lock and then free it.
623 * @thisfl_p: pointer that points to the fl_next field of the previous
624 * inode->i_flock list entry
626 * Unlink a lock from all lists and free the namespace reference, but don't
627 * free it yet. Wake up processes that are blocked waiting for this lock and
628 * notify the FS that the lock has been cleared.
630 * Must be called with the i_lock held!
632 static void locks_unlink_lock(struct file_lock
**thisfl_p
)
634 struct file_lock
*fl
= *thisfl_p
;
636 locks_delete_global_locks(fl
);
638 *thisfl_p
= fl
->fl_next
;
642 put_pid(fl
->fl_nspid
);
646 locks_wake_up_blocks(fl
);
650 * Unlink a lock from all lists and free it.
652 * Must be called with i_lock held!
654 static void locks_delete_lock(struct file_lock
**thisfl_p
)
656 struct file_lock
*fl
= *thisfl_p
;
658 locks_unlink_lock(thisfl_p
);
662 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
663 * checks for shared/exclusive status of overlapping locks.
665 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
667 if (sys_fl
->fl_type
== F_WRLCK
)
669 if (caller_fl
->fl_type
== F_WRLCK
)
674 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
675 * checking before calling the locks_conflict().
677 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
679 /* POSIX locks owned by the same process do not conflict with
682 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
685 /* Check whether they overlap */
686 if (!locks_overlap(caller_fl
, sys_fl
))
689 return (locks_conflict(caller_fl
, sys_fl
));
692 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
693 * checking before calling the locks_conflict().
695 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
697 /* FLOCK locks referring to the same filp do not conflict with
700 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
702 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
705 return (locks_conflict(caller_fl
, sys_fl
));
709 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
711 struct file_lock
*cfl
;
712 struct inode
*inode
= file_inode(filp
);
714 spin_lock(&inode
->i_lock
);
715 for (cfl
= file_inode(filp
)->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
718 if (posix_locks_conflict(fl
, cfl
))
722 __locks_copy_lock(fl
, cfl
);
724 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
726 fl
->fl_type
= F_UNLCK
;
727 spin_unlock(&inode
->i_lock
);
730 EXPORT_SYMBOL(posix_test_lock
);
733 * Deadlock detection:
735 * We attempt to detect deadlocks that are due purely to posix file
738 * We assume that a task can be waiting for at most one lock at a time.
739 * So for any acquired lock, the process holding that lock may be
740 * waiting on at most one other lock. That lock in turns may be held by
741 * someone waiting for at most one other lock. Given a requested lock
742 * caller_fl which is about to wait for a conflicting lock block_fl, we
743 * follow this chain of waiters to ensure we are not about to create a
746 * Since we do this before we ever put a process to sleep on a lock, we
747 * are ensured that there is never a cycle; that is what guarantees that
748 * the while() loop in posix_locks_deadlock() eventually completes.
750 * Note: the above assumption may not be true when handling lock
751 * requests from a broken NFS client. It may also fail in the presence
752 * of tasks (such as posix threads) sharing the same open file table.
753 * To handle those cases, we just bail out after a few iterations.
755 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
756 * Because the owner is not even nominally tied to a thread of
757 * execution, the deadlock detection below can't reasonably work well. Just
760 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
761 * locks that just checks for the case where two tasks are attempting to
762 * upgrade from read to write locks on the same inode.
765 #define MAX_DEADLK_ITERATIONS 10
767 /* Find a lock that the owner of the given block_fl is blocking on. */
768 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
770 struct file_lock
*fl
;
772 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
773 if (posix_same_owner(fl
, block_fl
))
779 /* Must be called with the blocked_lock_lock held! */
780 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
781 struct file_lock
*block_fl
)
786 * This deadlock detector can't reasonably detect deadlocks with
787 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
789 if (IS_OFDLCK(caller_fl
))
792 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
793 if (i
++ > MAX_DEADLK_ITERATIONS
)
795 if (posix_same_owner(caller_fl
, block_fl
))
801 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
802 * after any leases, but before any posix locks.
804 * Note that if called with an FL_EXISTS argument, the caller may determine
805 * whether or not a lock was successfully freed by testing the return
808 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
810 struct file_lock
*new_fl
= NULL
;
811 struct file_lock
**before
;
812 struct inode
* inode
= file_inode(filp
);
816 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
817 new_fl
= locks_alloc_lock();
822 spin_lock(&inode
->i_lock
);
823 if (request
->fl_flags
& FL_ACCESS
)
826 for_each_lock(inode
, before
) {
827 struct file_lock
*fl
= *before
;
832 if (filp
!= fl
->fl_file
)
834 if (request
->fl_type
== fl
->fl_type
)
837 locks_delete_lock(before
);
841 if (request
->fl_type
== F_UNLCK
) {
842 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
848 * If a higher-priority process was blocked on the old file lock,
849 * give it the opportunity to lock the file.
852 spin_unlock(&inode
->i_lock
);
854 spin_lock(&inode
->i_lock
);
858 for_each_lock(inode
, before
) {
859 struct file_lock
*fl
= *before
;
864 if (!flock_locks_conflict(request
, fl
))
867 if (!(request
->fl_flags
& FL_SLEEP
))
869 error
= FILE_LOCK_DEFERRED
;
870 locks_insert_block(fl
, request
);
873 if (request
->fl_flags
& FL_ACCESS
)
875 locks_copy_lock(new_fl
, request
);
876 locks_insert_lock(before
, new_fl
);
881 spin_unlock(&inode
->i_lock
);
883 locks_free_lock(new_fl
);
887 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
889 struct file_lock
*fl
;
890 struct file_lock
*new_fl
= NULL
;
891 struct file_lock
*new_fl2
= NULL
;
892 struct file_lock
*left
= NULL
;
893 struct file_lock
*right
= NULL
;
894 struct file_lock
**before
;
899 * We may need two file_lock structures for this operation,
900 * so we get them in advance to avoid races.
902 * In some cases we can be sure, that no new locks will be needed
904 if (!(request
->fl_flags
& FL_ACCESS
) &&
905 (request
->fl_type
!= F_UNLCK
||
906 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
907 new_fl
= locks_alloc_lock();
908 new_fl2
= locks_alloc_lock();
911 spin_lock(&inode
->i_lock
);
913 * New lock request. Walk all POSIX locks and look for conflicts. If
914 * there are any, either return error or put the request on the
915 * blocker's list of waiters and the global blocked_hash.
917 if (request
->fl_type
!= F_UNLCK
) {
918 for_each_lock(inode
, before
) {
922 if (!posix_locks_conflict(request
, fl
))
925 __locks_copy_lock(conflock
, fl
);
927 if (!(request
->fl_flags
& FL_SLEEP
))
930 * Deadlock detection and insertion into the blocked
931 * locks list must be done while holding the same lock!
934 spin_lock(&blocked_lock_lock
);
935 if (likely(!posix_locks_deadlock(request
, fl
))) {
936 error
= FILE_LOCK_DEFERRED
;
937 __locks_insert_block(fl
, request
);
939 spin_unlock(&blocked_lock_lock
);
944 /* If we're just looking for a conflict, we're done. */
946 if (request
->fl_flags
& FL_ACCESS
)
950 * Find the first old lock with the same owner as the new lock.
953 before
= &inode
->i_flock
;
955 /* First skip locks owned by other processes. */
956 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
957 !posix_same_owner(request
, fl
))) {
958 before
= &fl
->fl_next
;
961 /* Process locks with this owner. */
962 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
963 /* Detect adjacent or overlapping regions (if same lock type)
965 if (request
->fl_type
== fl
->fl_type
) {
966 /* In all comparisons of start vs end, use
967 * "start - 1" rather than "end + 1". If end
968 * is OFFSET_MAX, end + 1 will become negative.
970 if (fl
->fl_end
< request
->fl_start
- 1)
972 /* If the next lock in the list has entirely bigger
973 * addresses than the new one, insert the lock here.
975 if (fl
->fl_start
- 1 > request
->fl_end
)
978 /* If we come here, the new and old lock are of the
979 * same type and adjacent or overlapping. Make one
980 * lock yielding from the lower start address of both
981 * locks to the higher end address.
983 if (fl
->fl_start
> request
->fl_start
)
984 fl
->fl_start
= request
->fl_start
;
986 request
->fl_start
= fl
->fl_start
;
987 if (fl
->fl_end
< request
->fl_end
)
988 fl
->fl_end
= request
->fl_end
;
990 request
->fl_end
= fl
->fl_end
;
992 locks_delete_lock(before
);
999 /* Processing for different lock types is a bit
1002 if (fl
->fl_end
< request
->fl_start
)
1004 if (fl
->fl_start
> request
->fl_end
)
1006 if (request
->fl_type
== F_UNLCK
)
1008 if (fl
->fl_start
< request
->fl_start
)
1010 /* If the next lock in the list has a higher end
1011 * address than the new one, insert the new one here.
1013 if (fl
->fl_end
> request
->fl_end
) {
1017 if (fl
->fl_start
>= request
->fl_start
) {
1018 /* The new lock completely replaces an old
1019 * one (This may happen several times).
1022 locks_delete_lock(before
);
1025 /* Replace the old lock with the new one.
1026 * Wake up anybody waiting for the old one,
1027 * as the change in lock type might satisfy
1030 locks_wake_up_blocks(fl
);
1031 fl
->fl_start
= request
->fl_start
;
1032 fl
->fl_end
= request
->fl_end
;
1033 fl
->fl_type
= request
->fl_type
;
1034 locks_release_private(fl
);
1035 locks_copy_private(fl
, request
);
1040 /* Go on to next lock.
1043 before
= &fl
->fl_next
;
1047 * The above code only modifies existing locks in case of merging or
1048 * replacing. If new lock(s) need to be inserted all modifications are
1049 * done below this, so it's safe yet to bail out.
1051 error
= -ENOLCK
; /* "no luck" */
1052 if (right
&& left
== right
&& !new_fl2
)
1057 if (request
->fl_type
== F_UNLCK
) {
1058 if (request
->fl_flags
& FL_EXISTS
)
1067 locks_copy_lock(new_fl
, request
);
1068 locks_insert_lock(before
, new_fl
);
1072 if (left
== right
) {
1073 /* The new lock breaks the old one in two pieces,
1074 * so we have to use the second new lock.
1078 locks_copy_lock(left
, right
);
1079 locks_insert_lock(before
, left
);
1081 right
->fl_start
= request
->fl_end
+ 1;
1082 locks_wake_up_blocks(right
);
1085 left
->fl_end
= request
->fl_start
- 1;
1086 locks_wake_up_blocks(left
);
1089 spin_unlock(&inode
->i_lock
);
1091 * Free any unused locks.
1094 locks_free_lock(new_fl
);
1096 locks_free_lock(new_fl2
);
1101 * posix_lock_file - Apply a POSIX-style lock to a file
1102 * @filp: The file to apply the lock to
1103 * @fl: The lock to be applied
1104 * @conflock: Place to return a copy of the conflicting lock, if found.
1106 * Add a POSIX style lock to a file.
1107 * We merge adjacent & overlapping locks whenever possible.
1108 * POSIX locks are sorted by owner task, then by starting address
1110 * Note that if called with an FL_EXISTS argument, the caller may determine
1111 * whether or not a lock was successfully freed by testing the return
1112 * value for -ENOENT.
1114 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1115 struct file_lock
*conflock
)
1117 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1119 EXPORT_SYMBOL(posix_lock_file
);
1122 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1123 * @filp: The file to apply the lock to
1124 * @fl: The lock to be applied
1126 * Add a POSIX style lock to a file.
1127 * We merge adjacent & overlapping locks whenever possible.
1128 * POSIX locks are sorted by owner task, then by starting address
1130 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1135 error
= posix_lock_file(filp
, fl
, NULL
);
1136 if (error
!= FILE_LOCK_DEFERRED
)
1138 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1142 locks_delete_block(fl
);
1147 EXPORT_SYMBOL(posix_lock_file_wait
);
1150 * locks_mandatory_locked - Check for an active lock
1151 * @file: the file to check
1153 * Searches the inode's list of locks to find any POSIX locks which conflict.
1154 * This function is called from locks_verify_locked() only.
1156 int locks_mandatory_locked(struct file
*file
)
1158 struct inode
*inode
= file_inode(file
);
1159 struct file_lock
*fl
;
1162 * Search the lock list for this inode for any POSIX locks.
1164 spin_lock(&inode
->i_lock
);
1165 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1168 if (fl
->fl_owner
!= current
->files
&&
1169 fl
->fl_owner
!= file
)
1172 spin_unlock(&inode
->i_lock
);
1173 return fl
? -EAGAIN
: 0;
1177 * locks_mandatory_area - Check for a conflicting lock
1178 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1180 * @inode: the file to check
1181 * @filp: how the file was opened (if it was)
1182 * @offset: start of area to check
1183 * @count: length of area to check
1185 * Searches the inode's list of locks to find any POSIX locks which conflict.
1186 * This function is called from rw_verify_area() and
1187 * locks_verify_truncate().
1189 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1190 struct file
*filp
, loff_t offset
,
1193 struct file_lock fl
;
1197 locks_init_lock(&fl
);
1198 fl
.fl_pid
= current
->tgid
;
1200 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1201 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1203 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1204 fl
.fl_start
= offset
;
1205 fl
.fl_end
= offset
+ count
- 1;
1210 fl
.fl_flags
&= ~FL_SLEEP
;
1211 error
= __posix_lock_file(inode
, &fl
, NULL
);
1217 fl
.fl_flags
|= FL_SLEEP
;
1218 fl
.fl_owner
= current
->files
;
1219 error
= __posix_lock_file(inode
, &fl
, NULL
);
1220 if (error
!= FILE_LOCK_DEFERRED
)
1222 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1225 * If we've been sleeping someone might have
1226 * changed the permissions behind our back.
1228 if (__mandatory_lock(inode
))
1232 locks_delete_block(&fl
);
1239 EXPORT_SYMBOL(locks_mandatory_area
);
1241 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1245 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1248 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1252 /* We already had a lease on this file; just change its type */
1253 int lease_modify(struct file_lock
**before
, int arg
)
1255 struct file_lock
*fl
= *before
;
1256 int error
= assign_type(fl
, arg
);
1260 lease_clear_pending(fl
, arg
);
1261 locks_wake_up_blocks(fl
);
1262 if (arg
== F_UNLCK
) {
1263 struct file
*filp
= fl
->fl_file
;
1266 filp
->f_owner
.signum
= 0;
1267 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1268 if (fl
->fl_fasync
!= NULL
) {
1269 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1270 fl
->fl_fasync
= NULL
;
1272 locks_delete_lock(before
);
1277 EXPORT_SYMBOL(lease_modify
);
1279 static bool past_time(unsigned long then
)
1282 /* 0 is a special value meaning "this never expires": */
1284 return time_after(jiffies
, then
);
1287 static void time_out_leases(struct inode
*inode
)
1289 struct file_lock
**before
;
1290 struct file_lock
*fl
;
1292 before
= &inode
->i_flock
;
1293 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1294 trace_time_out_leases(inode
, fl
);
1295 if (past_time(fl
->fl_downgrade_time
))
1296 lease_modify(before
, F_RDLCK
);
1297 if (past_time(fl
->fl_break_time
))
1298 lease_modify(before
, F_UNLCK
);
1299 if (fl
== *before
) /* lease_modify may have freed fl */
1300 before
= &fl
->fl_next
;
1304 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1306 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1308 return locks_conflict(breaker
, lease
);
1312 * __break_lease - revoke all outstanding leases on file
1313 * @inode: the inode of the file to return
1314 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1316 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1319 * break_lease (inlined for speed) has checked there already is at least
1320 * some kind of lock (maybe a lease) on this file. Leases are broken on
1321 * a call to open() or truncate(). This function can sleep unless you
1322 * specified %O_NONBLOCK to your open().
1324 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1327 struct file_lock
*new_fl
, *flock
;
1328 struct file_lock
*fl
;
1329 unsigned long break_time
;
1330 int i_have_this_lease
= 0;
1331 bool lease_conflict
= false;
1332 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1334 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1336 return PTR_ERR(new_fl
);
1337 new_fl
->fl_flags
= type
;
1339 spin_lock(&inode
->i_lock
);
1341 time_out_leases(inode
);
1343 flock
= inode
->i_flock
;
1344 if ((flock
== NULL
) || !IS_LEASE(flock
))
1347 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1348 if (leases_conflict(fl
, new_fl
)) {
1349 lease_conflict
= true;
1350 if (fl
->fl_owner
== current
->files
)
1351 i_have_this_lease
= 1;
1354 if (!lease_conflict
)
1358 if (lease_break_time
> 0) {
1359 break_time
= jiffies
+ lease_break_time
* HZ
;
1360 if (break_time
== 0)
1361 break_time
++; /* so that 0 means no break time */
1364 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1365 if (!leases_conflict(fl
, new_fl
))
1368 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1370 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1371 fl
->fl_break_time
= break_time
;
1373 if (lease_breaking(flock
))
1375 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1376 fl
->fl_downgrade_time
= break_time
;
1378 fl
->fl_lmops
->lm_break(fl
);
1381 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1382 trace_break_lease_noblock(inode
, new_fl
);
1383 error
= -EWOULDBLOCK
;
1388 break_time
= flock
->fl_break_time
;
1389 if (break_time
!= 0)
1390 break_time
-= jiffies
;
1391 if (break_time
== 0)
1393 locks_insert_block(flock
, new_fl
);
1394 trace_break_lease_block(inode
, new_fl
);
1395 spin_unlock(&inode
->i_lock
);
1396 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1397 !new_fl
->fl_next
, break_time
);
1398 spin_lock(&inode
->i_lock
);
1399 trace_break_lease_unblock(inode
, new_fl
);
1400 locks_delete_block(new_fl
);
1403 time_out_leases(inode
);
1405 * Wait for the next conflicting lease that has not been
1408 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1409 flock
= flock
->fl_next
) {
1410 if (leases_conflict(new_fl
, flock
))
1417 spin_unlock(&inode
->i_lock
);
1418 locks_free_lock(new_fl
);
1422 EXPORT_SYMBOL(__break_lease
);
1425 * lease_get_mtime - get the last modified time of an inode
1427 * @time: pointer to a timespec which will contain the last modified time
1429 * This is to force NFS clients to flush their caches for files with
1430 * exclusive leases. The justification is that if someone has an
1431 * exclusive lease, then they could be modifying it.
1433 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1435 struct file_lock
*flock
= inode
->i_flock
;
1436 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
== F_WRLCK
))
1437 *time
= current_fs_time(inode
->i_sb
);
1439 *time
= inode
->i_mtime
;
1442 EXPORT_SYMBOL(lease_get_mtime
);
1445 * fcntl_getlease - Enquire what lease is currently active
1448 * The value returned by this function will be one of
1449 * (if no lease break is pending):
1451 * %F_RDLCK to indicate a shared lease is held.
1453 * %F_WRLCK to indicate an exclusive lease is held.
1455 * %F_UNLCK to indicate no lease is held.
1457 * (if a lease break is pending):
1459 * %F_RDLCK to indicate an exclusive lease needs to be
1460 * changed to a shared lease (or removed).
1462 * %F_UNLCK to indicate the lease needs to be removed.
1464 * XXX: sfr & willy disagree over whether F_INPROGRESS
1465 * should be returned to userspace.
1467 int fcntl_getlease(struct file
*filp
)
1469 struct file_lock
*fl
;
1470 struct inode
*inode
= file_inode(filp
);
1473 spin_lock(&inode
->i_lock
);
1474 time_out_leases(file_inode(filp
));
1475 for (fl
= file_inode(filp
)->i_flock
; fl
&& IS_LEASE(fl
);
1477 if (fl
->fl_file
== filp
) {
1478 type
= target_leasetype(fl
);
1482 spin_unlock(&inode
->i_lock
);
1487 * check_conflicting_open - see if the given dentry points to a file that has
1488 * an existing open that would conflict with the
1490 * @dentry: dentry to check
1491 * @arg: type of lease that we're trying to acquire
1493 * Check to see if there's an existing open fd on this file that would
1494 * conflict with the lease we're trying to set.
1497 check_conflicting_open(const struct dentry
*dentry
, const long arg
)
1500 struct inode
*inode
= dentry
->d_inode
;
1502 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1505 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1506 (atomic_read(&inode
->i_count
) > 1)))
1512 static int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1514 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1515 struct dentry
*dentry
= filp
->f_path
.dentry
;
1516 struct inode
*inode
= dentry
->d_inode
;
1517 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1521 trace_generic_add_lease(inode
, lease
);
1524 * In the delegation case we need mutual exclusion with
1525 * a number of operations that take the i_mutex. We trylock
1526 * because delegations are an optional optimization, and if
1527 * there's some chance of a conflict--we'd rather not
1528 * bother, maybe that's a sign this just isn't a good file to
1529 * hand out a delegation on.
1531 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1534 if (is_deleg
&& arg
== F_WRLCK
) {
1535 /* Write delegations are not currently supported: */
1536 mutex_unlock(&inode
->i_mutex
);
1541 error
= check_conflicting_open(dentry
, arg
);
1546 * At this point, we know that if there is an exclusive
1547 * lease on this file, then we hold it on this filp
1548 * (otherwise our open of this file would have blocked).
1549 * And if we are trying to acquire an exclusive lease,
1550 * then the file is not open by anyone (including us)
1551 * except for this filp.
1554 for (before
= &inode
->i_flock
;
1555 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1556 before
= &fl
->fl_next
) {
1557 if (fl
->fl_file
== filp
) {
1562 * No exclusive leases if someone else has a lease on
1568 * Modifying our existing lease is OK, but no getting a
1569 * new lease if someone else is opening for write:
1571 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1575 if (my_before
!= NULL
) {
1576 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1586 locks_insert_lock(before
, lease
);
1588 * The check in break_lease() is lockless. It's possible for another
1589 * open to race in after we did the earlier check for a conflicting
1590 * open but before the lease was inserted. Check again for a
1591 * conflicting open and cancel the lease if there is one.
1593 * We also add a barrier here to ensure that the insertion of the lock
1594 * precedes these checks.
1597 error
= check_conflicting_open(dentry
, arg
);
1599 locks_unlink_lock(flp
);
1602 mutex_unlock(&inode
->i_mutex
);
1606 static int generic_delete_lease(struct file
*filp
, struct file_lock
**flp
)
1608 struct file_lock
*fl
, **before
;
1609 struct dentry
*dentry
= filp
->f_path
.dentry
;
1610 struct inode
*inode
= dentry
->d_inode
;
1612 trace_generic_delete_lease(inode
, *flp
);
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
= IS_OFDLCK(fl
) ? -1 : 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
= IS_OFDLCK(fl
) ? -1 : 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
, unsigned int cmd
, 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 if (cmd
== F_OFD_GETLK
) {
1947 if (flock
.l_pid
!= 0)
1951 file_lock
.fl_flags
|= FL_OFDLCK
;
1952 file_lock
.fl_owner
= filp
;
1955 error
= vfs_test_lock(filp
, &file_lock
);
1959 flock
.l_type
= file_lock
.fl_type
;
1960 if (file_lock
.fl_type
!= F_UNLCK
) {
1961 error
= posix_lock_to_flock(&flock
, &file_lock
);
1966 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1973 * vfs_lock_file - file byte range lock
1974 * @filp: The file to apply the lock to
1975 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1976 * @fl: The lock to be applied
1977 * @conf: Place to return a copy of the conflicting lock, if found.
1979 * A caller that doesn't care about the conflicting lock may pass NULL
1980 * as the final argument.
1982 * If the filesystem defines a private ->lock() method, then @conf will
1983 * be left unchanged; so a caller that cares should initialize it to
1984 * some acceptable default.
1986 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1987 * locks, the ->lock() interface may return asynchronously, before the lock has
1988 * been granted or denied by the underlying filesystem, if (and only if)
1989 * lm_grant is set. Callers expecting ->lock() to return asynchronously
1990 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1991 * the request is for a blocking lock. When ->lock() does return asynchronously,
1992 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
1993 * request completes.
1994 * If the request is for non-blocking lock the file system should return
1995 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1996 * with the result. If the request timed out the callback routine will return a
1997 * nonzero return code and the file system should release the lock. The file
1998 * system is also responsible to keep a corresponding posix lock when it
1999 * grants a lock so the VFS can find out which locks are locally held and do
2000 * the correct lock cleanup when required.
2001 * The underlying filesystem must not drop the kernel lock or call
2002 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2005 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2007 if (filp
->f_op
->lock
)
2008 return filp
->f_op
->lock(filp
, cmd
, fl
);
2010 return posix_lock_file(filp
, fl
, conf
);
2012 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2014 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2015 struct file_lock
*fl
)
2019 error
= security_file_lock(filp
, fl
->fl_type
);
2024 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2025 if (error
!= FILE_LOCK_DEFERRED
)
2027 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2031 locks_delete_block(fl
);
2038 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2040 check_fmode_for_setlk(struct file_lock
*fl
)
2042 switch (fl
->fl_type
) {
2044 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2048 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2054 /* Apply the lock described by l to an open file descriptor.
2055 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2057 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2058 struct flock __user
*l
)
2060 struct file_lock
*file_lock
= locks_alloc_lock();
2062 struct inode
*inode
;
2066 if (file_lock
== NULL
)
2070 * This might block, so we do it before checking the inode.
2073 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2076 inode
= file_inode(filp
);
2078 /* Don't allow mandatory locks on files that may be memory mapped
2081 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2087 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2091 error
= check_fmode_for_setlk(file_lock
);
2096 * If the cmd is requesting file-private locks, then set the
2097 * FL_OFDLCK flag and override the owner.
2102 if (flock
.l_pid
!= 0)
2106 file_lock
->fl_flags
|= FL_OFDLCK
;
2107 file_lock
->fl_owner
= filp
;
2111 if (flock
.l_pid
!= 0)
2115 file_lock
->fl_flags
|= FL_OFDLCK
;
2116 file_lock
->fl_owner
= filp
;
2119 file_lock
->fl_flags
|= FL_SLEEP
;
2122 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2125 * Attempt to detect a close/fcntl race and recover by
2126 * releasing the lock that was just acquired.
2129 * we need that spin_lock here - it prevents reordering between
2130 * update of inode->i_flock and check for it done in close().
2131 * rcu_read_lock() wouldn't do.
2133 spin_lock(¤t
->files
->file_lock
);
2135 spin_unlock(¤t
->files
->file_lock
);
2136 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2137 flock
.l_type
= F_UNLCK
;
2142 locks_free_lock(file_lock
);
2146 #if BITS_PER_LONG == 32
2147 /* Report the first existing lock that would conflict with l.
2148 * This implements the F_GETLK command of fcntl().
2150 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2152 struct file_lock file_lock
;
2153 struct flock64 flock
;
2157 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2160 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2163 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2167 if (cmd
== F_OFD_GETLK
) {
2169 if (flock
.l_pid
!= 0)
2173 file_lock
.fl_flags
|= FL_OFDLCK
;
2174 file_lock
.fl_owner
= filp
;
2177 error
= vfs_test_lock(filp
, &file_lock
);
2181 flock
.l_type
= file_lock
.fl_type
;
2182 if (file_lock
.fl_type
!= F_UNLCK
)
2183 posix_lock_to_flock64(&flock
, &file_lock
);
2186 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2193 /* Apply the lock described by l to an open file descriptor.
2194 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2196 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2197 struct flock64 __user
*l
)
2199 struct file_lock
*file_lock
= locks_alloc_lock();
2200 struct flock64 flock
;
2201 struct inode
*inode
;
2205 if (file_lock
== NULL
)
2209 * This might block, so we do it before checking the inode.
2212 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2215 inode
= file_inode(filp
);
2217 /* Don't allow mandatory locks on files that may be memory mapped
2220 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2226 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2230 error
= check_fmode_for_setlk(file_lock
);
2235 * If the cmd is requesting file-private locks, then set the
2236 * FL_OFDLCK flag and override the owner.
2241 if (flock
.l_pid
!= 0)
2245 file_lock
->fl_flags
|= FL_OFDLCK
;
2246 file_lock
->fl_owner
= filp
;
2250 if (flock
.l_pid
!= 0)
2254 file_lock
->fl_flags
|= FL_OFDLCK
;
2255 file_lock
->fl_owner
= filp
;
2258 file_lock
->fl_flags
|= FL_SLEEP
;
2261 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2264 * Attempt to detect a close/fcntl race and recover by
2265 * releasing the lock that was just acquired.
2267 spin_lock(¤t
->files
->file_lock
);
2269 spin_unlock(¤t
->files
->file_lock
);
2270 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2271 flock
.l_type
= F_UNLCK
;
2276 locks_free_lock(file_lock
);
2279 #endif /* BITS_PER_LONG == 32 */
2282 * This function is called when the file is being removed
2283 * from the task's fd array. POSIX locks belonging to this task
2284 * are deleted at this time.
2286 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2288 struct file_lock lock
;
2291 * If there are no locks held on this file, we don't need to call
2292 * posix_lock_file(). Another process could be setting a lock on this
2293 * file at the same time, but we wouldn't remove that lock anyway.
2295 if (!file_inode(filp
)->i_flock
)
2298 lock
.fl_type
= F_UNLCK
;
2299 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2301 lock
.fl_end
= OFFSET_MAX
;
2302 lock
.fl_owner
= owner
;
2303 lock
.fl_pid
= current
->tgid
;
2304 lock
.fl_file
= filp
;
2306 lock
.fl_lmops
= NULL
;
2308 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2310 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2311 lock
.fl_ops
->fl_release_private(&lock
);
2314 EXPORT_SYMBOL(locks_remove_posix
);
2317 * This function is called on the last close of an open file.
2319 void locks_remove_file(struct file
*filp
)
2321 struct inode
* inode
= file_inode(filp
);
2322 struct file_lock
*fl
;
2323 struct file_lock
**before
;
2325 if (!inode
->i_flock
)
2328 locks_remove_posix(filp
, filp
);
2330 if (filp
->f_op
->flock
) {
2331 struct file_lock fl
= {
2333 .fl_pid
= current
->tgid
,
2335 .fl_flags
= FL_FLOCK
,
2337 .fl_end
= OFFSET_MAX
,
2339 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2340 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2341 fl
.fl_ops
->fl_release_private(&fl
);
2344 spin_lock(&inode
->i_lock
);
2345 before
= &inode
->i_flock
;
2347 while ((fl
= *before
) != NULL
) {
2348 if (fl
->fl_file
== filp
) {
2350 lease_modify(before
, F_UNLCK
);
2355 * There's a leftover lock on the list of a type that
2356 * we didn't expect to see. Most likely a classic
2357 * POSIX lock that ended up not getting released
2358 * properly, or that raced onto the list somehow. Log
2359 * some info about it and then just remove it from
2363 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2364 MAJOR(inode
->i_sb
->s_dev
),
2365 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
,
2366 fl
->fl_type
, fl
->fl_flags
,
2367 fl
->fl_start
, fl
->fl_end
);
2369 locks_delete_lock(before
);
2372 before
= &fl
->fl_next
;
2374 spin_unlock(&inode
->i_lock
);
2378 * posix_unblock_lock - stop waiting for a file lock
2379 * @waiter: the lock which was waiting
2381 * lockd needs to block waiting for locks.
2384 posix_unblock_lock(struct file_lock
*waiter
)
2388 spin_lock(&blocked_lock_lock
);
2389 if (waiter
->fl_next
)
2390 __locks_delete_block(waiter
);
2393 spin_unlock(&blocked_lock_lock
);
2396 EXPORT_SYMBOL(posix_unblock_lock
);
2399 * vfs_cancel_lock - file byte range unblock lock
2400 * @filp: The file to apply the unblock to
2401 * @fl: The lock to be unblocked
2403 * Used by lock managers to cancel blocked requests
2405 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2407 if (filp
->f_op
->lock
)
2408 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2412 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2414 #ifdef CONFIG_PROC_FS
2415 #include <linux/proc_fs.h>
2416 #include <linux/seq_file.h>
2418 struct locks_iterator
{
2423 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2424 loff_t id
, char *pfx
)
2426 struct inode
*inode
= NULL
;
2427 unsigned int fl_pid
;
2430 fl_pid
= pid_vnr(fl
->fl_nspid
);
2432 fl_pid
= fl
->fl_pid
;
2434 if (fl
->fl_file
!= NULL
)
2435 inode
= file_inode(fl
->fl_file
);
2437 seq_printf(f
, "%lld:%s ", id
, pfx
);
2439 if (fl
->fl_flags
& FL_ACCESS
)
2440 seq_puts(f
, "ACCESS");
2441 else if (IS_OFDLCK(fl
))
2442 seq_puts(f
, "OFDLCK");
2444 seq_puts(f
, "POSIX ");
2446 seq_printf(f
, " %s ",
2447 (inode
== NULL
) ? "*NOINODE*" :
2448 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2449 } else if (IS_FLOCK(fl
)) {
2450 if (fl
->fl_type
& LOCK_MAND
) {
2451 seq_puts(f
, "FLOCK MSNFS ");
2453 seq_puts(f
, "FLOCK ADVISORY ");
2455 } else if (IS_LEASE(fl
)) {
2456 if (fl
->fl_flags
& FL_DELEG
)
2457 seq_puts(f
, "DELEG ");
2459 seq_puts(f
, "LEASE ");
2461 if (lease_breaking(fl
))
2462 seq_puts(f
, "BREAKING ");
2463 else if (fl
->fl_file
)
2464 seq_puts(f
, "ACTIVE ");
2466 seq_puts(f
, "BREAKER ");
2468 seq_puts(f
, "UNKNOWN UNKNOWN ");
2470 if (fl
->fl_type
& LOCK_MAND
) {
2471 seq_printf(f
, "%s ",
2472 (fl
->fl_type
& LOCK_READ
)
2473 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2474 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2476 seq_printf(f
, "%s ",
2477 (lease_breaking(fl
))
2478 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2479 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2482 #ifdef WE_CAN_BREAK_LSLK_NOW
2483 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2484 inode
->i_sb
->s_id
, inode
->i_ino
);
2486 /* userspace relies on this representation of dev_t ;-( */
2487 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2488 MAJOR(inode
->i_sb
->s_dev
),
2489 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2492 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2495 if (fl
->fl_end
== OFFSET_MAX
)
2496 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2498 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2500 seq_puts(f
, "0 EOF\n");
2504 static int locks_show(struct seq_file
*f
, void *v
)
2506 struct locks_iterator
*iter
= f
->private;
2507 struct file_lock
*fl
, *bfl
;
2509 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2511 lock_get_status(f
, fl
, iter
->li_pos
, "");
2513 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2514 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2519 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2520 __acquires(&blocked_lock_lock
)
2522 struct locks_iterator
*iter
= f
->private;
2524 iter
->li_pos
= *pos
+ 1;
2525 lg_global_lock(&file_lock_lglock
);
2526 spin_lock(&blocked_lock_lock
);
2527 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2530 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2532 struct locks_iterator
*iter
= f
->private;
2535 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2538 static void locks_stop(struct seq_file
*f
, void *v
)
2539 __releases(&blocked_lock_lock
)
2541 spin_unlock(&blocked_lock_lock
);
2542 lg_global_unlock(&file_lock_lglock
);
2545 static const struct seq_operations locks_seq_operations
= {
2546 .start
= locks_start
,
2552 static int locks_open(struct inode
*inode
, struct file
*filp
)
2554 return seq_open_private(filp
, &locks_seq_operations
,
2555 sizeof(struct locks_iterator
));
2558 static const struct file_operations proc_locks_operations
= {
2561 .llseek
= seq_lseek
,
2562 .release
= seq_release_private
,
2565 static int __init
proc_locks_init(void)
2567 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2570 module_init(proc_locks_init
);
2574 * lock_may_read - checks that the region is free of locks
2575 * @inode: the inode that is being read
2576 * @start: the first byte to read
2577 * @len: the number of bytes to read
2579 * Emulates Windows locking requirements. Whole-file
2580 * mandatory locks (share modes) can prohibit a read and
2581 * byte-range POSIX locks can prohibit a read if they overlap.
2583 * N.B. this function is only ever called
2584 * from knfsd and ownership of locks is never checked.
2586 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
2588 struct file_lock
*fl
;
2591 spin_lock(&inode
->i_lock
);
2592 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2594 if (fl
->fl_type
== F_RDLCK
)
2596 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2598 } else if (IS_FLOCK(fl
)) {
2599 if (!(fl
->fl_type
& LOCK_MAND
))
2601 if (fl
->fl_type
& LOCK_READ
)
2608 spin_unlock(&inode
->i_lock
);
2612 EXPORT_SYMBOL(lock_may_read
);
2615 * lock_may_write - checks that the region is free of locks
2616 * @inode: the inode that is being written
2617 * @start: the first byte to write
2618 * @len: the number of bytes to write
2620 * Emulates Windows locking requirements. Whole-file
2621 * mandatory locks (share modes) can prohibit a write and
2622 * byte-range POSIX locks can prohibit a write if they overlap.
2624 * N.B. this function is only ever called
2625 * from knfsd and ownership of locks is never checked.
2627 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2629 struct file_lock
*fl
;
2632 spin_lock(&inode
->i_lock
);
2633 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2635 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2637 } else if (IS_FLOCK(fl
)) {
2638 if (!(fl
->fl_type
& LOCK_MAND
))
2640 if (fl
->fl_type
& LOCK_WRITE
)
2647 spin_unlock(&inode
->i_lock
);
2651 EXPORT_SYMBOL(lock_may_write
);
2653 static int __init
filelock_init(void)
2657 filelock_cache
= kmem_cache_create("file_lock_cache",
2658 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2660 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2662 for_each_possible_cpu(i
)
2663 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2668 core_initcall(filelock_init
);