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|FL_LAYOUT))
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;
161 * The global file_lock_list is only used for displaying /proc/locks, so we
162 * keep a list on each CPU, with each list protected by its own spinlock via
163 * the file_lock_lglock. Note that alterations to the list also require that
164 * the relevant flc_lock is held.
166 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
167 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
170 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
171 * It is protected by blocked_lock_lock.
173 * We hash locks by lockowner in order to optimize searching for the lock a
174 * particular lockowner is waiting on.
176 * FIXME: make this value scale via some heuristic? We generally will want more
177 * buckets when we have more lockowners holding locks, but that's a little
178 * difficult to determine without knowing what the workload will look like.
180 #define BLOCKED_HASH_BITS 7
181 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
184 * This lock protects the blocked_hash. Generally, if you're accessing it, you
185 * want to be holding this lock.
187 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
188 * pointer for file_lock structures that are acting as lock requests (in
189 * contrast to those that are acting as records of acquired locks).
191 * Note that when we acquire this lock in order to change the above fields,
192 * we often hold the flc_lock as well. In certain cases, when reading the fields
193 * protected by this lock, we can skip acquiring it iff we already hold the
196 * In particular, adding an entry to the fl_block list requires that you hold
197 * both the flc_lock and the blocked_lock_lock (acquired in that order).
198 * Deleting an entry from the list however only requires the file_lock_lock.
200 static DEFINE_SPINLOCK(blocked_lock_lock
);
202 static struct kmem_cache
*flctx_cache __read_mostly
;
203 static struct kmem_cache
*filelock_cache __read_mostly
;
205 static struct file_lock_context
*
206 locks_get_lock_context(struct inode
*inode
, int type
)
208 struct file_lock_context
*new;
210 if (likely(inode
->i_flctx
) || type
== F_UNLCK
)
213 new = kmem_cache_alloc(flctx_cache
, GFP_KERNEL
);
217 spin_lock_init(&new->flc_lock
);
218 INIT_LIST_HEAD(&new->flc_flock
);
219 INIT_LIST_HEAD(&new->flc_posix
);
220 INIT_LIST_HEAD(&new->flc_lease
);
223 * Assign the pointer if it's not already assigned. If it is, then
224 * free the context we just allocated.
226 spin_lock(&inode
->i_lock
);
227 if (likely(!inode
->i_flctx
)) {
228 inode
->i_flctx
= new;
231 spin_unlock(&inode
->i_lock
);
234 kmem_cache_free(flctx_cache
, new);
236 return inode
->i_flctx
;
240 locks_free_lock_context(struct file_lock_context
*ctx
)
243 WARN_ON_ONCE(!list_empty(&ctx
->flc_flock
));
244 WARN_ON_ONCE(!list_empty(&ctx
->flc_posix
));
245 WARN_ON_ONCE(!list_empty(&ctx
->flc_lease
));
246 kmem_cache_free(flctx_cache
, ctx
);
250 static void locks_init_lock_heads(struct file_lock
*fl
)
252 INIT_HLIST_NODE(&fl
->fl_link
);
253 INIT_LIST_HEAD(&fl
->fl_list
);
254 INIT_LIST_HEAD(&fl
->fl_block
);
255 init_waitqueue_head(&fl
->fl_wait
);
258 /* Allocate an empty lock structure. */
259 struct file_lock
*locks_alloc_lock(void)
261 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
264 locks_init_lock_heads(fl
);
268 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
270 void locks_release_private(struct file_lock
*fl
)
273 if (fl
->fl_ops
->fl_release_private
)
274 fl
->fl_ops
->fl_release_private(fl
);
279 if (fl
->fl_lmops
->lm_put_owner
) {
280 fl
->fl_lmops
->lm_put_owner(fl
->fl_owner
);
286 EXPORT_SYMBOL_GPL(locks_release_private
);
288 /* Free a lock which is not in use. */
289 void locks_free_lock(struct file_lock
*fl
)
291 BUG_ON(waitqueue_active(&fl
->fl_wait
));
292 BUG_ON(!list_empty(&fl
->fl_list
));
293 BUG_ON(!list_empty(&fl
->fl_block
));
294 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
296 locks_release_private(fl
);
297 kmem_cache_free(filelock_cache
, fl
);
299 EXPORT_SYMBOL(locks_free_lock
);
302 locks_dispose_list(struct list_head
*dispose
)
304 struct file_lock
*fl
;
306 while (!list_empty(dispose
)) {
307 fl
= list_first_entry(dispose
, struct file_lock
, fl_list
);
308 list_del_init(&fl
->fl_list
);
313 void locks_init_lock(struct file_lock
*fl
)
315 memset(fl
, 0, sizeof(struct file_lock
));
316 locks_init_lock_heads(fl
);
319 EXPORT_SYMBOL(locks_init_lock
);
322 * Initialize a new lock from an existing file_lock structure.
324 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
326 new->fl_owner
= fl
->fl_owner
;
327 new->fl_pid
= fl
->fl_pid
;
329 new->fl_flags
= fl
->fl_flags
;
330 new->fl_type
= fl
->fl_type
;
331 new->fl_start
= fl
->fl_start
;
332 new->fl_end
= fl
->fl_end
;
333 new->fl_lmops
= fl
->fl_lmops
;
337 if (fl
->fl_lmops
->lm_get_owner
)
338 fl
->fl_lmops
->lm_get_owner(fl
->fl_owner
);
341 EXPORT_SYMBOL(locks_copy_conflock
);
343 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
345 /* "new" must be a freshly-initialized lock */
346 WARN_ON_ONCE(new->fl_ops
);
348 locks_copy_conflock(new, fl
);
350 new->fl_file
= fl
->fl_file
;
351 new->fl_ops
= fl
->fl_ops
;
354 if (fl
->fl_ops
->fl_copy_lock
)
355 fl
->fl_ops
->fl_copy_lock(new, fl
);
359 EXPORT_SYMBOL(locks_copy_lock
);
361 static inline int flock_translate_cmd(int cmd
) {
363 return cmd
& (LOCK_MAND
| LOCK_RW
);
375 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
376 static struct file_lock
*
377 flock_make_lock(struct file
*filp
, unsigned int cmd
)
379 struct file_lock
*fl
;
380 int type
= flock_translate_cmd(cmd
);
383 return ERR_PTR(type
);
385 fl
= locks_alloc_lock();
387 return ERR_PTR(-ENOMEM
);
391 fl
->fl_pid
= current
->tgid
;
392 fl
->fl_flags
= FL_FLOCK
;
394 fl
->fl_end
= OFFSET_MAX
;
399 static int assign_type(struct file_lock
*fl
, long type
)
413 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
416 switch (l
->l_whence
) {
421 fl
->fl_start
= filp
->f_pos
;
424 fl
->fl_start
= i_size_read(file_inode(filp
));
429 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
431 fl
->fl_start
+= l
->l_start
;
432 if (fl
->fl_start
< 0)
435 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
436 POSIX-2001 defines it. */
438 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
440 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
442 } else if (l
->l_len
< 0) {
443 if (fl
->fl_start
+ l
->l_len
< 0)
445 fl
->fl_end
= fl
->fl_start
- 1;
446 fl
->fl_start
+= l
->l_len
;
448 fl
->fl_end
= OFFSET_MAX
;
450 fl
->fl_owner
= current
->files
;
451 fl
->fl_pid
= current
->tgid
;
453 fl
->fl_flags
= FL_POSIX
;
457 return assign_type(fl
, l
->l_type
);
460 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
463 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
466 struct flock64 ll
= {
468 .l_whence
= l
->l_whence
,
469 .l_start
= l
->l_start
,
473 return flock64_to_posix_lock(filp
, fl
, &ll
);
476 /* default lease lock manager operations */
478 lease_break_callback(struct file_lock
*fl
)
480 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
485 lease_setup(struct file_lock
*fl
, void **priv
)
487 struct file
*filp
= fl
->fl_file
;
488 struct fasync_struct
*fa
= *priv
;
491 * fasync_insert_entry() returns the old entry if any. If there was no
492 * old entry, then it used "priv" and inserted it into the fasync list.
493 * Clear the pointer to indicate that it shouldn't be freed.
495 if (!fasync_insert_entry(fa
->fa_fd
, filp
, &fl
->fl_fasync
, fa
))
498 __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
501 static const struct lock_manager_operations lease_manager_ops
= {
502 .lm_break
= lease_break_callback
,
503 .lm_change
= lease_modify
,
504 .lm_setup
= lease_setup
,
508 * Initialize a lease, use the default lock manager operations
510 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
512 if (assign_type(fl
, type
) != 0)
516 fl
->fl_pid
= current
->tgid
;
519 fl
->fl_flags
= FL_LEASE
;
521 fl
->fl_end
= OFFSET_MAX
;
523 fl
->fl_lmops
= &lease_manager_ops
;
527 /* Allocate a file_lock initialised to this type of lease */
528 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
530 struct file_lock
*fl
= locks_alloc_lock();
534 return ERR_PTR(error
);
536 error
= lease_init(filp
, type
, fl
);
539 return ERR_PTR(error
);
544 /* Check if two locks overlap each other.
546 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
548 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
549 (fl2
->fl_end
>= fl1
->fl_start
));
553 * Check whether two locks have the same owner.
555 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
557 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
558 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
559 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
560 return fl1
->fl_owner
== fl2
->fl_owner
;
563 /* Must be called with the flc_lock held! */
564 static void locks_insert_global_locks(struct file_lock
*fl
)
566 lg_local_lock(&file_lock_lglock
);
567 fl
->fl_link_cpu
= smp_processor_id();
568 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
569 lg_local_unlock(&file_lock_lglock
);
572 /* Must be called with the flc_lock held! */
573 static void locks_delete_global_locks(struct file_lock
*fl
)
576 * Avoid taking lock if already unhashed. This is safe since this check
577 * is done while holding the flc_lock, and new insertions into the list
578 * also require that it be held.
580 if (hlist_unhashed(&fl
->fl_link
))
582 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
583 hlist_del_init(&fl
->fl_link
);
584 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
588 posix_owner_key(struct file_lock
*fl
)
590 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
591 return fl
->fl_lmops
->lm_owner_key(fl
);
592 return (unsigned long)fl
->fl_owner
;
595 static void locks_insert_global_blocked(struct file_lock
*waiter
)
597 lockdep_assert_held(&blocked_lock_lock
);
599 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
602 static void locks_delete_global_blocked(struct file_lock
*waiter
)
604 lockdep_assert_held(&blocked_lock_lock
);
606 hash_del(&waiter
->fl_link
);
609 /* Remove waiter from blocker's block list.
610 * When blocker ends up pointing to itself then the list is empty.
612 * Must be called with blocked_lock_lock held.
614 static void __locks_delete_block(struct file_lock
*waiter
)
616 locks_delete_global_blocked(waiter
);
617 list_del_init(&waiter
->fl_block
);
618 waiter
->fl_next
= NULL
;
621 static void locks_delete_block(struct file_lock
*waiter
)
623 spin_lock(&blocked_lock_lock
);
624 __locks_delete_block(waiter
);
625 spin_unlock(&blocked_lock_lock
);
628 /* Insert waiter into blocker's block list.
629 * We use a circular list so that processes can be easily woken up in
630 * the order they blocked. The documentation doesn't require this but
631 * it seems like the reasonable thing to do.
633 * Must be called with both the flc_lock and blocked_lock_lock held. The
634 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
635 * that the flc_lock is also held on insertions we can avoid taking the
636 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
638 static void __locks_insert_block(struct file_lock
*blocker
,
639 struct file_lock
*waiter
)
641 BUG_ON(!list_empty(&waiter
->fl_block
));
642 waiter
->fl_next
= blocker
;
643 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
644 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
645 locks_insert_global_blocked(waiter
);
648 /* Must be called with flc_lock held. */
649 static void locks_insert_block(struct file_lock
*blocker
,
650 struct file_lock
*waiter
)
652 spin_lock(&blocked_lock_lock
);
653 __locks_insert_block(blocker
, waiter
);
654 spin_unlock(&blocked_lock_lock
);
658 * Wake up processes blocked waiting for blocker.
660 * Must be called with the inode->flc_lock held!
662 static void locks_wake_up_blocks(struct file_lock
*blocker
)
665 * Avoid taking global lock if list is empty. This is safe since new
666 * blocked requests are only added to the list under the flc_lock, and
667 * the flc_lock is always held here. Note that removal from the fl_block
668 * list does not require the flc_lock, so we must recheck list_empty()
669 * after acquiring the blocked_lock_lock.
671 if (list_empty(&blocker
->fl_block
))
674 spin_lock(&blocked_lock_lock
);
675 while (!list_empty(&blocker
->fl_block
)) {
676 struct file_lock
*waiter
;
678 waiter
= list_first_entry(&blocker
->fl_block
,
679 struct file_lock
, fl_block
);
680 __locks_delete_block(waiter
);
681 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
682 waiter
->fl_lmops
->lm_notify(waiter
);
684 wake_up(&waiter
->fl_wait
);
686 spin_unlock(&blocked_lock_lock
);
690 locks_insert_lock_ctx(struct file_lock
*fl
, struct list_head
*before
)
692 fl
->fl_nspid
= get_pid(task_tgid(current
));
693 list_add_tail(&fl
->fl_list
, before
);
694 locks_insert_global_locks(fl
);
698 locks_unlink_lock_ctx(struct file_lock
*fl
)
700 locks_delete_global_locks(fl
);
701 list_del_init(&fl
->fl_list
);
703 put_pid(fl
->fl_nspid
);
706 locks_wake_up_blocks(fl
);
710 locks_delete_lock_ctx(struct file_lock
*fl
, struct list_head
*dispose
)
712 locks_unlink_lock_ctx(fl
);
714 list_add(&fl
->fl_list
, dispose
);
719 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
720 * checks for shared/exclusive status of overlapping locks.
722 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
724 if (sys_fl
->fl_type
== F_WRLCK
)
726 if (caller_fl
->fl_type
== F_WRLCK
)
731 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
732 * checking before calling the locks_conflict().
734 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
736 /* POSIX locks owned by the same process do not conflict with
739 if (posix_same_owner(caller_fl
, sys_fl
))
742 /* Check whether they overlap */
743 if (!locks_overlap(caller_fl
, sys_fl
))
746 return (locks_conflict(caller_fl
, sys_fl
));
749 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
750 * checking before calling the locks_conflict().
752 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
754 /* FLOCK locks referring to the same filp do not conflict with
757 if (caller_fl
->fl_file
== sys_fl
->fl_file
)
759 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
762 return (locks_conflict(caller_fl
, sys_fl
));
766 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
768 struct file_lock
*cfl
;
769 struct file_lock_context
*ctx
;
770 struct inode
*inode
= file_inode(filp
);
772 ctx
= inode
->i_flctx
;
773 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
)) {
774 fl
->fl_type
= F_UNLCK
;
778 spin_lock(&ctx
->flc_lock
);
779 list_for_each_entry(cfl
, &ctx
->flc_posix
, fl_list
) {
780 if (posix_locks_conflict(fl
, cfl
)) {
781 locks_copy_conflock(fl
, cfl
);
783 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
787 fl
->fl_type
= F_UNLCK
;
789 spin_unlock(&ctx
->flc_lock
);
792 EXPORT_SYMBOL(posix_test_lock
);
795 * Deadlock detection:
797 * We attempt to detect deadlocks that are due purely to posix file
800 * We assume that a task can be waiting for at most one lock at a time.
801 * So for any acquired lock, the process holding that lock may be
802 * waiting on at most one other lock. That lock in turns may be held by
803 * someone waiting for at most one other lock. Given a requested lock
804 * caller_fl which is about to wait for a conflicting lock block_fl, we
805 * follow this chain of waiters to ensure we are not about to create a
808 * Since we do this before we ever put a process to sleep on a lock, we
809 * are ensured that there is never a cycle; that is what guarantees that
810 * the while() loop in posix_locks_deadlock() eventually completes.
812 * Note: the above assumption may not be true when handling lock
813 * requests from a broken NFS client. It may also fail in the presence
814 * of tasks (such as posix threads) sharing the same open file table.
815 * To handle those cases, we just bail out after a few iterations.
817 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
818 * Because the owner is not even nominally tied to a thread of
819 * execution, the deadlock detection below can't reasonably work well. Just
822 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
823 * locks that just checks for the case where two tasks are attempting to
824 * upgrade from read to write locks on the same inode.
827 #define MAX_DEADLK_ITERATIONS 10
829 /* Find a lock that the owner of the given block_fl is blocking on. */
830 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
832 struct file_lock
*fl
;
834 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
835 if (posix_same_owner(fl
, block_fl
))
841 /* Must be called with the blocked_lock_lock held! */
842 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
843 struct file_lock
*block_fl
)
847 lockdep_assert_held(&blocked_lock_lock
);
850 * This deadlock detector can't reasonably detect deadlocks with
851 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
853 if (IS_OFDLCK(caller_fl
))
856 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
857 if (i
++ > MAX_DEADLK_ITERATIONS
)
859 if (posix_same_owner(caller_fl
, block_fl
))
865 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
866 * after any leases, but before any posix locks.
868 * Note that if called with an FL_EXISTS argument, the caller may determine
869 * whether or not a lock was successfully freed by testing the return
872 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
874 struct file_lock
*new_fl
= NULL
;
875 struct file_lock
*fl
;
876 struct file_lock_context
*ctx
;
877 struct inode
*inode
= file_inode(filp
);
882 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
884 if (request
->fl_type
!= F_UNLCK
)
886 return (request
->fl_flags
& FL_EXISTS
) ? -ENOENT
: 0;
889 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
890 new_fl
= locks_alloc_lock();
895 spin_lock(&ctx
->flc_lock
);
896 if (request
->fl_flags
& FL_ACCESS
)
899 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
900 if (filp
!= fl
->fl_file
)
902 if (request
->fl_type
== fl
->fl_type
)
905 locks_delete_lock_ctx(fl
, &dispose
);
909 if (request
->fl_type
== F_UNLCK
) {
910 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
916 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
917 if (!flock_locks_conflict(request
, fl
))
920 if (!(request
->fl_flags
& FL_SLEEP
))
922 error
= FILE_LOCK_DEFERRED
;
923 locks_insert_block(fl
, request
);
926 if (request
->fl_flags
& FL_ACCESS
)
928 locks_copy_lock(new_fl
, request
);
929 locks_insert_lock_ctx(new_fl
, &ctx
->flc_flock
);
934 spin_unlock(&ctx
->flc_lock
);
936 locks_free_lock(new_fl
);
937 locks_dispose_list(&dispose
);
941 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
943 struct file_lock
*fl
, *tmp
;
944 struct file_lock
*new_fl
= NULL
;
945 struct file_lock
*new_fl2
= NULL
;
946 struct file_lock
*left
= NULL
;
947 struct file_lock
*right
= NULL
;
948 struct file_lock_context
*ctx
;
953 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
955 return (request
->fl_type
== F_UNLCK
) ? 0 : -ENOMEM
;
958 * We may need two file_lock structures for this operation,
959 * so we get them in advance to avoid races.
961 * In some cases we can be sure, that no new locks will be needed
963 if (!(request
->fl_flags
& FL_ACCESS
) &&
964 (request
->fl_type
!= F_UNLCK
||
965 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
966 new_fl
= locks_alloc_lock();
967 new_fl2
= locks_alloc_lock();
970 spin_lock(&ctx
->flc_lock
);
972 * New lock request. Walk all POSIX locks and look for conflicts. If
973 * there are any, either return error or put the request on the
974 * blocker's list of waiters and the global blocked_hash.
976 if (request
->fl_type
!= F_UNLCK
) {
977 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
978 if (!posix_locks_conflict(request
, fl
))
981 locks_copy_conflock(conflock
, fl
);
983 if (!(request
->fl_flags
& FL_SLEEP
))
986 * Deadlock detection and insertion into the blocked
987 * locks list must be done while holding the same lock!
990 spin_lock(&blocked_lock_lock
);
991 if (likely(!posix_locks_deadlock(request
, fl
))) {
992 error
= FILE_LOCK_DEFERRED
;
993 __locks_insert_block(fl
, request
);
995 spin_unlock(&blocked_lock_lock
);
1000 /* If we're just looking for a conflict, we're done. */
1002 if (request
->fl_flags
& FL_ACCESS
)
1005 /* Find the first old lock with the same owner as the new lock */
1006 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1007 if (posix_same_owner(request
, fl
))
1011 /* Process locks with this owner. */
1012 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1013 if (!posix_same_owner(request
, fl
))
1016 /* Detect adjacent or overlapping regions (if same lock type) */
1017 if (request
->fl_type
== fl
->fl_type
) {
1018 /* In all comparisons of start vs end, use
1019 * "start - 1" rather than "end + 1". If end
1020 * is OFFSET_MAX, end + 1 will become negative.
1022 if (fl
->fl_end
< request
->fl_start
- 1)
1024 /* If the next lock in the list has entirely bigger
1025 * addresses than the new one, insert the lock here.
1027 if (fl
->fl_start
- 1 > request
->fl_end
)
1030 /* If we come here, the new and old lock are of the
1031 * same type and adjacent or overlapping. Make one
1032 * lock yielding from the lower start address of both
1033 * locks to the higher end address.
1035 if (fl
->fl_start
> request
->fl_start
)
1036 fl
->fl_start
= request
->fl_start
;
1038 request
->fl_start
= fl
->fl_start
;
1039 if (fl
->fl_end
< request
->fl_end
)
1040 fl
->fl_end
= request
->fl_end
;
1042 request
->fl_end
= fl
->fl_end
;
1044 locks_delete_lock_ctx(fl
, &dispose
);
1050 /* Processing for different lock types is a bit
1053 if (fl
->fl_end
< request
->fl_start
)
1055 if (fl
->fl_start
> request
->fl_end
)
1057 if (request
->fl_type
== F_UNLCK
)
1059 if (fl
->fl_start
< request
->fl_start
)
1061 /* If the next lock in the list has a higher end
1062 * address than the new one, insert the new one here.
1064 if (fl
->fl_end
> request
->fl_end
) {
1068 if (fl
->fl_start
>= request
->fl_start
) {
1069 /* The new lock completely replaces an old
1070 * one (This may happen several times).
1073 locks_delete_lock_ctx(fl
, &dispose
);
1077 * Replace the old lock with new_fl, and
1078 * remove the old one. It's safe to do the
1079 * insert here since we know that we won't be
1080 * using new_fl later, and that the lock is
1081 * just replacing an existing lock.
1086 locks_copy_lock(new_fl
, request
);
1089 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1090 locks_delete_lock_ctx(fl
, &dispose
);
1097 * The above code only modifies existing locks in case of merging or
1098 * replacing. If new lock(s) need to be inserted all modifications are
1099 * done below this, so it's safe yet to bail out.
1101 error
= -ENOLCK
; /* "no luck" */
1102 if (right
&& left
== right
&& !new_fl2
)
1107 if (request
->fl_type
== F_UNLCK
) {
1108 if (request
->fl_flags
& FL_EXISTS
)
1117 locks_copy_lock(new_fl
, request
);
1118 locks_insert_lock_ctx(new_fl
, &fl
->fl_list
);
1123 if (left
== right
) {
1124 /* The new lock breaks the old one in two pieces,
1125 * so we have to use the second new lock.
1129 locks_copy_lock(left
, right
);
1130 locks_insert_lock_ctx(left
, &fl
->fl_list
);
1132 right
->fl_start
= request
->fl_end
+ 1;
1133 locks_wake_up_blocks(right
);
1136 left
->fl_end
= request
->fl_start
- 1;
1137 locks_wake_up_blocks(left
);
1140 spin_unlock(&ctx
->flc_lock
);
1142 * Free any unused locks.
1145 locks_free_lock(new_fl
);
1147 locks_free_lock(new_fl2
);
1148 locks_dispose_list(&dispose
);
1153 * posix_lock_file - Apply a POSIX-style lock to a file
1154 * @filp: The file to apply the lock to
1155 * @fl: The lock to be applied
1156 * @conflock: Place to return a copy of the conflicting lock, if found.
1158 * Add a POSIX style lock to a file.
1159 * We merge adjacent & overlapping locks whenever possible.
1160 * POSIX locks are sorted by owner task, then by starting address
1162 * Note that if called with an FL_EXISTS argument, the caller may determine
1163 * whether or not a lock was successfully freed by testing the return
1164 * value for -ENOENT.
1166 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1167 struct file_lock
*conflock
)
1169 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1171 EXPORT_SYMBOL(posix_lock_file
);
1174 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1175 * @filp: The file to apply the lock to
1176 * @fl: The lock to be applied
1178 * Add a POSIX style lock to a file.
1179 * We merge adjacent & overlapping locks whenever possible.
1180 * POSIX locks are sorted by owner task, then by starting address
1182 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1187 error
= posix_lock_file(filp
, fl
, NULL
);
1188 if (error
!= FILE_LOCK_DEFERRED
)
1190 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1194 locks_delete_block(fl
);
1199 EXPORT_SYMBOL(posix_lock_file_wait
);
1202 * locks_mandatory_locked - Check for an active lock
1203 * @file: the file to check
1205 * Searches the inode's list of locks to find any POSIX locks which conflict.
1206 * This function is called from locks_verify_locked() only.
1208 int locks_mandatory_locked(struct file
*file
)
1211 struct inode
*inode
= file_inode(file
);
1212 struct file_lock_context
*ctx
;
1213 struct file_lock
*fl
;
1215 ctx
= inode
->i_flctx
;
1216 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
))
1220 * Search the lock list for this inode for any POSIX locks.
1222 spin_lock(&ctx
->flc_lock
);
1224 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1225 if (fl
->fl_owner
!= current
->files
&&
1226 fl
->fl_owner
!= file
) {
1231 spin_unlock(&ctx
->flc_lock
);
1236 * locks_mandatory_area - Check for a conflicting lock
1237 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1239 * @inode: the file to check
1240 * @filp: how the file was opened (if it was)
1241 * @offset: start of area to check
1242 * @count: length of area to check
1244 * Searches the inode's list of locks to find any POSIX locks which conflict.
1245 * This function is called from rw_verify_area() and
1246 * locks_verify_truncate().
1248 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1249 struct file
*filp
, loff_t offset
,
1252 struct file_lock fl
;
1256 locks_init_lock(&fl
);
1257 fl
.fl_pid
= current
->tgid
;
1259 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1260 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1262 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1263 fl
.fl_start
= offset
;
1264 fl
.fl_end
= offset
+ count
- 1;
1269 fl
.fl_flags
&= ~FL_SLEEP
;
1270 error
= __posix_lock_file(inode
, &fl
, NULL
);
1276 fl
.fl_flags
|= FL_SLEEP
;
1277 fl
.fl_owner
= current
->files
;
1278 error
= __posix_lock_file(inode
, &fl
, NULL
);
1279 if (error
!= FILE_LOCK_DEFERRED
)
1281 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1284 * If we've been sleeping someone might have
1285 * changed the permissions behind our back.
1287 if (__mandatory_lock(inode
))
1291 locks_delete_block(&fl
);
1298 EXPORT_SYMBOL(locks_mandatory_area
);
1300 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1304 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1307 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1311 /* We already had a lease on this file; just change its type */
1312 int lease_modify(struct file_lock
*fl
, int arg
, struct list_head
*dispose
)
1314 int error
= assign_type(fl
, arg
);
1318 lease_clear_pending(fl
, arg
);
1319 locks_wake_up_blocks(fl
);
1320 if (arg
== F_UNLCK
) {
1321 struct file
*filp
= fl
->fl_file
;
1324 filp
->f_owner
.signum
= 0;
1325 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1326 if (fl
->fl_fasync
!= NULL
) {
1327 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1328 fl
->fl_fasync
= NULL
;
1330 locks_delete_lock_ctx(fl
, dispose
);
1334 EXPORT_SYMBOL(lease_modify
);
1336 static bool past_time(unsigned long then
)
1339 /* 0 is a special value meaning "this never expires": */
1341 return time_after(jiffies
, then
);
1344 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1346 struct file_lock_context
*ctx
= inode
->i_flctx
;
1347 struct file_lock
*fl
, *tmp
;
1349 lockdep_assert_held(&ctx
->flc_lock
);
1351 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1352 trace_time_out_leases(inode
, fl
);
1353 if (past_time(fl
->fl_downgrade_time
))
1354 lease_modify(fl
, F_RDLCK
, dispose
);
1355 if (past_time(fl
->fl_break_time
))
1356 lease_modify(fl
, F_UNLCK
, dispose
);
1360 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1362 if ((breaker
->fl_flags
& FL_LAYOUT
) != (lease
->fl_flags
& FL_LAYOUT
))
1364 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1366 return locks_conflict(breaker
, lease
);
1370 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1372 struct file_lock_context
*ctx
= inode
->i_flctx
;
1373 struct file_lock
*fl
;
1375 lockdep_assert_held(&ctx
->flc_lock
);
1377 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1378 if (leases_conflict(fl
, breaker
))
1385 * __break_lease - revoke all outstanding leases on file
1386 * @inode: the inode of the file to return
1387 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1389 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1392 * break_lease (inlined for speed) has checked there already is at least
1393 * some kind of lock (maybe a lease) on this file. Leases are broken on
1394 * a call to open() or truncate(). This function can sleep unless you
1395 * specified %O_NONBLOCK to your open().
1397 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1400 struct file_lock_context
*ctx
= inode
->i_flctx
;
1401 struct file_lock
*new_fl
, *fl
, *tmp
;
1402 unsigned long break_time
;
1403 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1406 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1408 return PTR_ERR(new_fl
);
1409 new_fl
->fl_flags
= type
;
1411 /* typically we will check that ctx is non-NULL before calling */
1417 spin_lock(&ctx
->flc_lock
);
1419 time_out_leases(inode
, &dispose
);
1421 if (!any_leases_conflict(inode
, new_fl
))
1425 if (lease_break_time
> 0) {
1426 break_time
= jiffies
+ lease_break_time
* HZ
;
1427 if (break_time
== 0)
1428 break_time
++; /* so that 0 means no break time */
1431 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1432 if (!leases_conflict(fl
, new_fl
))
1435 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1437 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1438 fl
->fl_break_time
= break_time
;
1440 if (lease_breaking(fl
))
1442 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1443 fl
->fl_downgrade_time
= break_time
;
1445 if (fl
->fl_lmops
->lm_break(fl
))
1446 locks_delete_lock_ctx(fl
, &dispose
);
1449 if (list_empty(&ctx
->flc_lease
))
1452 if (mode
& O_NONBLOCK
) {
1453 trace_break_lease_noblock(inode
, new_fl
);
1454 error
= -EWOULDBLOCK
;
1459 fl
= list_first_entry(&ctx
->flc_lease
, struct file_lock
, fl_list
);
1460 break_time
= fl
->fl_break_time
;
1461 if (break_time
!= 0)
1462 break_time
-= jiffies
;
1463 if (break_time
== 0)
1465 locks_insert_block(fl
, new_fl
);
1466 trace_break_lease_block(inode
, new_fl
);
1467 spin_unlock(&ctx
->flc_lock
);
1468 locks_dispose_list(&dispose
);
1469 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1470 !new_fl
->fl_next
, break_time
);
1471 spin_lock(&ctx
->flc_lock
);
1472 trace_break_lease_unblock(inode
, new_fl
);
1473 locks_delete_block(new_fl
);
1476 * Wait for the next conflicting lease that has not been
1480 time_out_leases(inode
, &dispose
);
1481 if (any_leases_conflict(inode
, new_fl
))
1486 spin_unlock(&ctx
->flc_lock
);
1487 locks_dispose_list(&dispose
);
1488 locks_free_lock(new_fl
);
1492 EXPORT_SYMBOL(__break_lease
);
1495 * lease_get_mtime - get the last modified time of an inode
1497 * @time: pointer to a timespec which will contain the last modified time
1499 * This is to force NFS clients to flush their caches for files with
1500 * exclusive leases. The justification is that if someone has an
1501 * exclusive lease, then they could be modifying it.
1503 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1505 bool has_lease
= false;
1506 struct file_lock_context
*ctx
= inode
->i_flctx
;
1507 struct file_lock
*fl
;
1509 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1510 spin_lock(&ctx
->flc_lock
);
1511 if (!list_empty(&ctx
->flc_lease
)) {
1512 fl
= list_first_entry(&ctx
->flc_lease
,
1513 struct file_lock
, fl_list
);
1514 if (fl
->fl_type
== F_WRLCK
)
1517 spin_unlock(&ctx
->flc_lock
);
1521 *time
= current_fs_time(inode
->i_sb
);
1523 *time
= inode
->i_mtime
;
1526 EXPORT_SYMBOL(lease_get_mtime
);
1529 * fcntl_getlease - Enquire what lease is currently active
1532 * The value returned by this function will be one of
1533 * (if no lease break is pending):
1535 * %F_RDLCK to indicate a shared lease is held.
1537 * %F_WRLCK to indicate an exclusive lease is held.
1539 * %F_UNLCK to indicate no lease is held.
1541 * (if a lease break is pending):
1543 * %F_RDLCK to indicate an exclusive lease needs to be
1544 * changed to a shared lease (or removed).
1546 * %F_UNLCK to indicate the lease needs to be removed.
1548 * XXX: sfr & willy disagree over whether F_INPROGRESS
1549 * should be returned to userspace.
1551 int fcntl_getlease(struct file
*filp
)
1553 struct file_lock
*fl
;
1554 struct inode
*inode
= file_inode(filp
);
1555 struct file_lock_context
*ctx
= inode
->i_flctx
;
1559 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1560 spin_lock(&ctx
->flc_lock
);
1561 time_out_leases(file_inode(filp
), &dispose
);
1562 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1563 if (fl
->fl_file
!= filp
)
1565 type
= target_leasetype(fl
);
1568 spin_unlock(&ctx
->flc_lock
);
1569 locks_dispose_list(&dispose
);
1575 * check_conflicting_open - see if the given dentry points to a file that has
1576 * an existing open that would conflict with the
1578 * @dentry: dentry to check
1579 * @arg: type of lease that we're trying to acquire
1581 * Check to see if there's an existing open fd on this file that would
1582 * conflict with the lease we're trying to set.
1585 check_conflicting_open(const struct dentry
*dentry
, const long arg
, int flags
)
1588 struct inode
*inode
= dentry
->d_inode
;
1590 if (flags
& FL_LAYOUT
)
1593 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1596 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1597 (atomic_read(&inode
->i_count
) > 1)))
1604 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1606 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1607 struct dentry
*dentry
= filp
->f_path
.dentry
;
1608 struct inode
*inode
= dentry
->d_inode
;
1609 struct file_lock_context
*ctx
;
1610 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1615 trace_generic_add_lease(inode
, lease
);
1617 /* Note that arg is never F_UNLCK here */
1618 ctx
= locks_get_lock_context(inode
, arg
);
1623 * In the delegation case we need mutual exclusion with
1624 * a number of operations that take the i_mutex. We trylock
1625 * because delegations are an optional optimization, and if
1626 * there's some chance of a conflict--we'd rather not
1627 * bother, maybe that's a sign this just isn't a good file to
1628 * hand out a delegation on.
1630 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1633 if (is_deleg
&& arg
== F_WRLCK
) {
1634 /* Write delegations are not currently supported: */
1635 mutex_unlock(&inode
->i_mutex
);
1640 spin_lock(&ctx
->flc_lock
);
1641 time_out_leases(inode
, &dispose
);
1642 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1647 * At this point, we know that if there is an exclusive
1648 * lease on this file, then we hold it on this filp
1649 * (otherwise our open of this file would have blocked).
1650 * And if we are trying to acquire an exclusive lease,
1651 * then the file is not open by anyone (including us)
1652 * except for this filp.
1655 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1656 if (fl
->fl_file
== filp
&&
1657 fl
->fl_owner
== lease
->fl_owner
) {
1663 * No exclusive leases if someone else has a lease on
1669 * Modifying our existing lease is OK, but no getting a
1670 * new lease if someone else is opening for write:
1672 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1676 if (my_fl
!= NULL
) {
1678 error
= lease
->fl_lmops
->lm_change(lease
, arg
, &dispose
);
1688 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1690 * The check in break_lease() is lockless. It's possible for another
1691 * open to race in after we did the earlier check for a conflicting
1692 * open but before the lease was inserted. Check again for a
1693 * conflicting open and cancel the lease if there is one.
1695 * We also add a barrier here to ensure that the insertion of the lock
1696 * precedes these checks.
1699 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1701 locks_unlink_lock_ctx(lease
);
1706 if (lease
->fl_lmops
->lm_setup
)
1707 lease
->fl_lmops
->lm_setup(lease
, priv
);
1709 spin_unlock(&ctx
->flc_lock
);
1710 locks_dispose_list(&dispose
);
1712 mutex_unlock(&inode
->i_mutex
);
1713 if (!error
&& !my_fl
)
1718 static int generic_delete_lease(struct file
*filp
, void *owner
)
1720 int error
= -EAGAIN
;
1721 struct file_lock
*fl
, *victim
= NULL
;
1722 struct dentry
*dentry
= filp
->f_path
.dentry
;
1723 struct inode
*inode
= dentry
->d_inode
;
1724 struct file_lock_context
*ctx
= inode
->i_flctx
;
1728 trace_generic_delete_lease(inode
, NULL
);
1732 spin_lock(&ctx
->flc_lock
);
1733 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1734 if (fl
->fl_file
== filp
&&
1735 fl
->fl_owner
== owner
) {
1740 trace_generic_delete_lease(inode
, victim
);
1742 error
= fl
->fl_lmops
->lm_change(victim
, F_UNLCK
, &dispose
);
1743 spin_unlock(&ctx
->flc_lock
);
1744 locks_dispose_list(&dispose
);
1749 * generic_setlease - sets a lease on an open file
1750 * @filp: file pointer
1751 * @arg: type of lease to obtain
1752 * @flp: input - file_lock to use, output - file_lock inserted
1753 * @priv: private data for lm_setup (may be NULL if lm_setup
1754 * doesn't require it)
1756 * The (input) flp->fl_lmops->lm_break function is required
1759 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1762 struct dentry
*dentry
= filp
->f_path
.dentry
;
1763 struct inode
*inode
= dentry
->d_inode
;
1766 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1768 if (!S_ISREG(inode
->i_mode
))
1770 error
= security_file_lock(filp
, arg
);
1776 return generic_delete_lease(filp
, *priv
);
1779 if (!(*flp
)->fl_lmops
->lm_break
) {
1784 return generic_add_lease(filp
, arg
, flp
, priv
);
1789 EXPORT_SYMBOL(generic_setlease
);
1792 * vfs_setlease - sets a lease on an open file
1793 * @filp: file pointer
1794 * @arg: type of lease to obtain
1795 * @lease: file_lock to use when adding a lease
1796 * @priv: private info for lm_setup when adding a lease (may be
1797 * NULL if lm_setup doesn't require it)
1799 * Call this to establish a lease on the file. The "lease" argument is not
1800 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1801 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1802 * if not, this function will return -ENOLCK (and generate a scary-looking
1805 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1806 * may be NULL if the lm_setup operation doesn't require it.
1809 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1811 if (filp
->f_op
->setlease
)
1812 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1814 return generic_setlease(filp
, arg
, lease
, priv
);
1816 EXPORT_SYMBOL_GPL(vfs_setlease
);
1818 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1820 struct file_lock
*fl
;
1821 struct fasync_struct
*new;
1824 fl
= lease_alloc(filp
, arg
);
1828 new = fasync_alloc();
1830 locks_free_lock(fl
);
1835 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1837 locks_free_lock(fl
);
1844 * fcntl_setlease - sets a lease on an open file
1845 * @fd: open file descriptor
1846 * @filp: file pointer
1847 * @arg: type of lease to obtain
1849 * Call this fcntl to establish a lease on the file.
1850 * Note that you also need to call %F_SETSIG to
1851 * receive a signal when the lease is broken.
1853 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1856 return vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&filp
);
1857 return do_fcntl_add_lease(fd
, filp
, arg
);
1861 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1862 * @filp: The file to apply the lock to
1863 * @fl: The lock to be applied
1865 * Add a FLOCK style lock to a file.
1867 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1872 error
= flock_lock_file(filp
, fl
);
1873 if (error
!= FILE_LOCK_DEFERRED
)
1875 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1879 locks_delete_block(fl
);
1885 EXPORT_SYMBOL(flock_lock_file_wait
);
1888 * sys_flock: - flock() system call.
1889 * @fd: the file descriptor to lock.
1890 * @cmd: the type of lock to apply.
1892 * Apply a %FL_FLOCK style lock to an open file descriptor.
1893 * The @cmd can be one of
1895 * %LOCK_SH -- a shared lock.
1897 * %LOCK_EX -- an exclusive lock.
1899 * %LOCK_UN -- remove an existing lock.
1901 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1903 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1904 * processes read and write access respectively.
1906 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1908 struct fd f
= fdget(fd
);
1909 struct file_lock
*lock
;
1910 int can_sleep
, unlock
;
1917 can_sleep
= !(cmd
& LOCK_NB
);
1919 unlock
= (cmd
== LOCK_UN
);
1921 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1922 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1925 lock
= flock_make_lock(f
.file
, cmd
);
1927 error
= PTR_ERR(lock
);
1932 lock
->fl_flags
|= FL_SLEEP
;
1934 error
= security_file_lock(f
.file
, lock
->fl_type
);
1938 if (f
.file
->f_op
->flock
)
1939 error
= f
.file
->f_op
->flock(f
.file
,
1940 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1943 error
= flock_lock_file_wait(f
.file
, lock
);
1946 locks_free_lock(lock
);
1955 * vfs_test_lock - test file byte range lock
1956 * @filp: The file to test lock for
1957 * @fl: The lock to test; also used to hold result
1959 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1960 * setting conf->fl_type to something other than F_UNLCK.
1962 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1964 if (filp
->f_op
->lock
)
1965 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1966 posix_test_lock(filp
, fl
);
1969 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1971 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1973 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1974 #if BITS_PER_LONG == 32
1976 * Make sure we can represent the posix lock via
1977 * legacy 32bit flock.
1979 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1981 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1984 flock
->l_start
= fl
->fl_start
;
1985 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1986 fl
->fl_end
- fl
->fl_start
+ 1;
1987 flock
->l_whence
= 0;
1988 flock
->l_type
= fl
->fl_type
;
1992 #if BITS_PER_LONG == 32
1993 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1995 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1996 flock
->l_start
= fl
->fl_start
;
1997 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1998 fl
->fl_end
- fl
->fl_start
+ 1;
1999 flock
->l_whence
= 0;
2000 flock
->l_type
= fl
->fl_type
;
2004 /* Report the first existing lock that would conflict with l.
2005 * This implements the F_GETLK command of fcntl().
2007 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2009 struct file_lock file_lock
;
2014 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2017 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2020 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2024 if (cmd
== F_OFD_GETLK
) {
2026 if (flock
.l_pid
!= 0)
2030 file_lock
.fl_flags
|= FL_OFDLCK
;
2031 file_lock
.fl_owner
= filp
;
2034 error
= vfs_test_lock(filp
, &file_lock
);
2038 flock
.l_type
= file_lock
.fl_type
;
2039 if (file_lock
.fl_type
!= F_UNLCK
) {
2040 error
= posix_lock_to_flock(&flock
, &file_lock
);
2045 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2048 locks_release_private(&file_lock
);
2054 * vfs_lock_file - file byte range lock
2055 * @filp: The file to apply the lock to
2056 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2057 * @fl: The lock to be applied
2058 * @conf: Place to return a copy of the conflicting lock, if found.
2060 * A caller that doesn't care about the conflicting lock may pass NULL
2061 * as the final argument.
2063 * If the filesystem defines a private ->lock() method, then @conf will
2064 * be left unchanged; so a caller that cares should initialize it to
2065 * some acceptable default.
2067 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2068 * locks, the ->lock() interface may return asynchronously, before the lock has
2069 * been granted or denied by the underlying filesystem, if (and only if)
2070 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2071 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2072 * the request is for a blocking lock. When ->lock() does return asynchronously,
2073 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2074 * request completes.
2075 * If the request is for non-blocking lock the file system should return
2076 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2077 * with the result. If the request timed out the callback routine will return a
2078 * nonzero return code and the file system should release the lock. The file
2079 * system is also responsible to keep a corresponding posix lock when it
2080 * grants a lock so the VFS can find out which locks are locally held and do
2081 * the correct lock cleanup when required.
2082 * The underlying filesystem must not drop the kernel lock or call
2083 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2086 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2088 if (filp
->f_op
->lock
)
2089 return filp
->f_op
->lock(filp
, cmd
, fl
);
2091 return posix_lock_file(filp
, fl
, conf
);
2093 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2095 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2096 struct file_lock
*fl
)
2100 error
= security_file_lock(filp
, fl
->fl_type
);
2105 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2106 if (error
!= FILE_LOCK_DEFERRED
)
2108 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2112 locks_delete_block(fl
);
2119 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2121 check_fmode_for_setlk(struct file_lock
*fl
)
2123 switch (fl
->fl_type
) {
2125 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2129 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2135 /* Apply the lock described by l to an open file descriptor.
2136 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2138 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2139 struct flock __user
*l
)
2141 struct file_lock
*file_lock
= locks_alloc_lock();
2143 struct inode
*inode
;
2147 if (file_lock
== NULL
)
2151 * This might block, so we do it before checking the inode.
2154 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2157 inode
= file_inode(filp
);
2159 /* Don't allow mandatory locks on files that may be memory mapped
2162 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2168 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2172 error
= check_fmode_for_setlk(file_lock
);
2177 * If the cmd is requesting file-private locks, then set the
2178 * FL_OFDLCK flag and override the owner.
2183 if (flock
.l_pid
!= 0)
2187 file_lock
->fl_flags
|= FL_OFDLCK
;
2188 file_lock
->fl_owner
= filp
;
2192 if (flock
.l_pid
!= 0)
2196 file_lock
->fl_flags
|= FL_OFDLCK
;
2197 file_lock
->fl_owner
= filp
;
2200 file_lock
->fl_flags
|= FL_SLEEP
;
2203 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2206 * Attempt to detect a close/fcntl race and recover by
2207 * releasing the lock that was just acquired.
2210 * we need that spin_lock here - it prevents reordering between
2211 * update of i_flctx->flc_posix and check for it done in close().
2212 * rcu_read_lock() wouldn't do.
2214 spin_lock(¤t
->files
->file_lock
);
2216 spin_unlock(¤t
->files
->file_lock
);
2217 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2218 flock
.l_type
= F_UNLCK
;
2223 locks_free_lock(file_lock
);
2227 #if BITS_PER_LONG == 32
2228 /* Report the first existing lock that would conflict with l.
2229 * This implements the F_GETLK command of fcntl().
2231 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2233 struct file_lock file_lock
;
2234 struct flock64 flock
;
2238 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2241 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2244 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2248 if (cmd
== F_OFD_GETLK
) {
2250 if (flock
.l_pid
!= 0)
2254 file_lock
.fl_flags
|= FL_OFDLCK
;
2255 file_lock
.fl_owner
= filp
;
2258 error
= vfs_test_lock(filp
, &file_lock
);
2262 flock
.l_type
= file_lock
.fl_type
;
2263 if (file_lock
.fl_type
!= F_UNLCK
)
2264 posix_lock_to_flock64(&flock
, &file_lock
);
2267 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2270 locks_release_private(&file_lock
);
2275 /* Apply the lock described by l to an open file descriptor.
2276 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2278 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2279 struct flock64 __user
*l
)
2281 struct file_lock
*file_lock
= locks_alloc_lock();
2282 struct flock64 flock
;
2283 struct inode
*inode
;
2287 if (file_lock
== NULL
)
2291 * This might block, so we do it before checking the inode.
2294 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2297 inode
= file_inode(filp
);
2299 /* Don't allow mandatory locks on files that may be memory mapped
2302 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2308 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2312 error
= check_fmode_for_setlk(file_lock
);
2317 * If the cmd is requesting file-private locks, then set the
2318 * FL_OFDLCK flag and override the owner.
2323 if (flock
.l_pid
!= 0)
2327 file_lock
->fl_flags
|= FL_OFDLCK
;
2328 file_lock
->fl_owner
= filp
;
2332 if (flock
.l_pid
!= 0)
2336 file_lock
->fl_flags
|= FL_OFDLCK
;
2337 file_lock
->fl_owner
= filp
;
2340 file_lock
->fl_flags
|= FL_SLEEP
;
2343 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2346 * Attempt to detect a close/fcntl race and recover by
2347 * releasing the lock that was just acquired.
2349 spin_lock(¤t
->files
->file_lock
);
2351 spin_unlock(¤t
->files
->file_lock
);
2352 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2353 flock
.l_type
= F_UNLCK
;
2358 locks_free_lock(file_lock
);
2361 #endif /* BITS_PER_LONG == 32 */
2364 * This function is called when the file is being removed
2365 * from the task's fd array. POSIX locks belonging to this task
2366 * are deleted at this time.
2368 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2370 struct file_lock lock
;
2371 struct file_lock_context
*ctx
= file_inode(filp
)->i_flctx
;
2374 * If there are no locks held on this file, we don't need to call
2375 * posix_lock_file(). Another process could be setting a lock on this
2376 * file at the same time, but we wouldn't remove that lock anyway.
2378 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2381 lock
.fl_type
= F_UNLCK
;
2382 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2384 lock
.fl_end
= OFFSET_MAX
;
2385 lock
.fl_owner
= owner
;
2386 lock
.fl_pid
= current
->tgid
;
2387 lock
.fl_file
= filp
;
2389 lock
.fl_lmops
= NULL
;
2391 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2393 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2394 lock
.fl_ops
->fl_release_private(&lock
);
2397 EXPORT_SYMBOL(locks_remove_posix
);
2399 /* The i_flctx must be valid when calling into here */
2401 locks_remove_flock(struct file
*filp
)
2403 struct file_lock fl
= {
2405 .fl_pid
= current
->tgid
,
2407 .fl_flags
= FL_FLOCK
,
2409 .fl_end
= OFFSET_MAX
,
2411 struct file_lock_context
*flctx
= file_inode(filp
)->i_flctx
;
2413 if (list_empty(&flctx
->flc_flock
))
2416 if (filp
->f_op
->flock
)
2417 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2419 flock_lock_file(filp
, &fl
);
2421 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2422 fl
.fl_ops
->fl_release_private(&fl
);
2425 /* The i_flctx must be valid when calling into here */
2427 locks_remove_lease(struct file
*filp
)
2429 struct inode
*inode
= file_inode(filp
);
2430 struct file_lock_context
*ctx
= inode
->i_flctx
;
2431 struct file_lock
*fl
, *tmp
;
2434 if (list_empty(&ctx
->flc_lease
))
2437 spin_lock(&ctx
->flc_lock
);
2438 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2439 if (filp
== fl
->fl_file
)
2440 lease_modify(fl
, F_UNLCK
, &dispose
);
2441 spin_unlock(&ctx
->flc_lock
);
2442 locks_dispose_list(&dispose
);
2446 * This function is called on the last close of an open file.
2448 void locks_remove_file(struct file
*filp
)
2450 if (!file_inode(filp
)->i_flctx
)
2453 /* remove any OFD locks */
2454 locks_remove_posix(filp
, filp
);
2456 /* remove flock locks */
2457 locks_remove_flock(filp
);
2459 /* remove any leases */
2460 locks_remove_lease(filp
);
2464 * posix_unblock_lock - stop waiting for a file lock
2465 * @waiter: the lock which was waiting
2467 * lockd needs to block waiting for locks.
2470 posix_unblock_lock(struct file_lock
*waiter
)
2474 spin_lock(&blocked_lock_lock
);
2475 if (waiter
->fl_next
)
2476 __locks_delete_block(waiter
);
2479 spin_unlock(&blocked_lock_lock
);
2482 EXPORT_SYMBOL(posix_unblock_lock
);
2485 * vfs_cancel_lock - file byte range unblock lock
2486 * @filp: The file to apply the unblock to
2487 * @fl: The lock to be unblocked
2489 * Used by lock managers to cancel blocked requests
2491 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2493 if (filp
->f_op
->lock
)
2494 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2498 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2500 #ifdef CONFIG_PROC_FS
2501 #include <linux/proc_fs.h>
2502 #include <linux/seq_file.h>
2504 struct locks_iterator
{
2509 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2510 loff_t id
, char *pfx
)
2512 struct inode
*inode
= NULL
;
2513 unsigned int fl_pid
;
2516 fl_pid
= pid_vnr(fl
->fl_nspid
);
2518 fl_pid
= fl
->fl_pid
;
2520 if (fl
->fl_file
!= NULL
)
2521 inode
= file_inode(fl
->fl_file
);
2523 seq_printf(f
, "%lld:%s ", id
, pfx
);
2525 if (fl
->fl_flags
& FL_ACCESS
)
2526 seq_puts(f
, "ACCESS");
2527 else if (IS_OFDLCK(fl
))
2528 seq_puts(f
, "OFDLCK");
2530 seq_puts(f
, "POSIX ");
2532 seq_printf(f
, " %s ",
2533 (inode
== NULL
) ? "*NOINODE*" :
2534 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2535 } else if (IS_FLOCK(fl
)) {
2536 if (fl
->fl_type
& LOCK_MAND
) {
2537 seq_puts(f
, "FLOCK MSNFS ");
2539 seq_puts(f
, "FLOCK ADVISORY ");
2541 } else if (IS_LEASE(fl
)) {
2542 if (fl
->fl_flags
& FL_DELEG
)
2543 seq_puts(f
, "DELEG ");
2545 seq_puts(f
, "LEASE ");
2547 if (lease_breaking(fl
))
2548 seq_puts(f
, "BREAKING ");
2549 else if (fl
->fl_file
)
2550 seq_puts(f
, "ACTIVE ");
2552 seq_puts(f
, "BREAKER ");
2554 seq_puts(f
, "UNKNOWN UNKNOWN ");
2556 if (fl
->fl_type
& LOCK_MAND
) {
2557 seq_printf(f
, "%s ",
2558 (fl
->fl_type
& LOCK_READ
)
2559 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2560 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2562 seq_printf(f
, "%s ",
2563 (lease_breaking(fl
))
2564 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2565 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2568 #ifdef WE_CAN_BREAK_LSLK_NOW
2569 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2570 inode
->i_sb
->s_id
, inode
->i_ino
);
2572 /* userspace relies on this representation of dev_t ;-( */
2573 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2574 MAJOR(inode
->i_sb
->s_dev
),
2575 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2578 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2581 if (fl
->fl_end
== OFFSET_MAX
)
2582 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2584 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2586 seq_puts(f
, "0 EOF\n");
2590 static int locks_show(struct seq_file
*f
, void *v
)
2592 struct locks_iterator
*iter
= f
->private;
2593 struct file_lock
*fl
, *bfl
;
2595 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2597 lock_get_status(f
, fl
, iter
->li_pos
, "");
2599 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2600 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2605 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2606 __acquires(&blocked_lock_lock
)
2608 struct locks_iterator
*iter
= f
->private;
2610 iter
->li_pos
= *pos
+ 1;
2611 lg_global_lock(&file_lock_lglock
);
2612 spin_lock(&blocked_lock_lock
);
2613 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2616 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2618 struct locks_iterator
*iter
= f
->private;
2621 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2624 static void locks_stop(struct seq_file
*f
, void *v
)
2625 __releases(&blocked_lock_lock
)
2627 spin_unlock(&blocked_lock_lock
);
2628 lg_global_unlock(&file_lock_lglock
);
2631 static const struct seq_operations locks_seq_operations
= {
2632 .start
= locks_start
,
2638 static int locks_open(struct inode
*inode
, struct file
*filp
)
2640 return seq_open_private(filp
, &locks_seq_operations
,
2641 sizeof(struct locks_iterator
));
2644 static const struct file_operations proc_locks_operations
= {
2647 .llseek
= seq_lseek
,
2648 .release
= seq_release_private
,
2651 static int __init
proc_locks_init(void)
2653 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2656 module_init(proc_locks_init
);
2659 static int __init
filelock_init(void)
2663 flctx_cache
= kmem_cache_create("file_lock_ctx",
2664 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2666 filelock_cache
= kmem_cache_create("file_lock_cache",
2667 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2669 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2671 for_each_possible_cpu(i
)
2672 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2677 core_initcall(filelock_init
);