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;
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
)
208 struct file_lock_context
*new;
210 if (likely(inode
->i_flctx
))
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
);
284 EXPORT_SYMBOL_GPL(locks_release_private
);
286 /* Free a lock which is not in use. */
287 void locks_free_lock(struct file_lock
*fl
)
289 BUG_ON(waitqueue_active(&fl
->fl_wait
));
290 BUG_ON(!list_empty(&fl
->fl_list
));
291 BUG_ON(!list_empty(&fl
->fl_block
));
292 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
294 locks_release_private(fl
);
295 kmem_cache_free(filelock_cache
, fl
);
297 EXPORT_SYMBOL(locks_free_lock
);
300 locks_dispose_list(struct list_head
*dispose
)
302 struct file_lock
*fl
;
304 while (!list_empty(dispose
)) {
305 fl
= list_first_entry(dispose
, struct file_lock
, fl_list
);
306 list_del_init(&fl
->fl_list
);
311 void locks_init_lock(struct file_lock
*fl
)
313 memset(fl
, 0, sizeof(struct file_lock
));
314 locks_init_lock_heads(fl
);
317 EXPORT_SYMBOL(locks_init_lock
);
320 * Initialize a new lock from an existing file_lock structure.
322 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
324 new->fl_owner
= fl
->fl_owner
;
325 new->fl_pid
= fl
->fl_pid
;
327 new->fl_flags
= fl
->fl_flags
;
328 new->fl_type
= fl
->fl_type
;
329 new->fl_start
= fl
->fl_start
;
330 new->fl_end
= fl
->fl_end
;
331 new->fl_lmops
= fl
->fl_lmops
;
335 if (fl
->fl_lmops
->lm_get_owner
)
336 fl
->fl_lmops
->lm_get_owner(new, fl
);
339 EXPORT_SYMBOL(locks_copy_conflock
);
341 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
343 /* "new" must be a freshly-initialized lock */
344 WARN_ON_ONCE(new->fl_ops
);
346 locks_copy_conflock(new, fl
);
348 new->fl_file
= fl
->fl_file
;
349 new->fl_ops
= fl
->fl_ops
;
352 if (fl
->fl_ops
->fl_copy_lock
)
353 fl
->fl_ops
->fl_copy_lock(new, fl
);
357 EXPORT_SYMBOL(locks_copy_lock
);
359 static inline int flock_translate_cmd(int cmd
) {
361 return cmd
& (LOCK_MAND
| LOCK_RW
);
373 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
374 static struct file_lock
*
375 flock_make_lock(struct file
*filp
, unsigned int cmd
)
377 struct file_lock
*fl
;
378 int type
= flock_translate_cmd(cmd
);
381 return ERR_PTR(type
);
383 fl
= locks_alloc_lock();
385 return ERR_PTR(-ENOMEM
);
389 fl
->fl_pid
= current
->tgid
;
390 fl
->fl_flags
= FL_FLOCK
;
392 fl
->fl_end
= OFFSET_MAX
;
397 static int assign_type(struct file_lock
*fl
, long type
)
411 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
414 switch (l
->l_whence
) {
419 fl
->fl_start
= filp
->f_pos
;
422 fl
->fl_start
= i_size_read(file_inode(filp
));
427 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
429 fl
->fl_start
+= l
->l_start
;
430 if (fl
->fl_start
< 0)
433 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
434 POSIX-2001 defines it. */
436 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
438 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
440 } else if (l
->l_len
< 0) {
441 if (fl
->fl_start
+ l
->l_len
< 0)
443 fl
->fl_end
= fl
->fl_start
- 1;
444 fl
->fl_start
+= l
->l_len
;
446 fl
->fl_end
= OFFSET_MAX
;
448 fl
->fl_owner
= current
->files
;
449 fl
->fl_pid
= current
->tgid
;
451 fl
->fl_flags
= FL_POSIX
;
455 return assign_type(fl
, l
->l_type
);
458 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
461 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
464 struct flock64 ll
= {
466 .l_whence
= l
->l_whence
,
467 .l_start
= l
->l_start
,
471 return flock64_to_posix_lock(filp
, fl
, &ll
);
474 /* default lease lock manager operations */
476 lease_break_callback(struct file_lock
*fl
)
478 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
483 lease_setup(struct file_lock
*fl
, void **priv
)
485 struct file
*filp
= fl
->fl_file
;
486 struct fasync_struct
*fa
= *priv
;
489 * fasync_insert_entry() returns the old entry if any. If there was no
490 * old entry, then it used "priv" and inserted it into the fasync list.
491 * Clear the pointer to indicate that it shouldn't be freed.
493 if (!fasync_insert_entry(fa
->fa_fd
, filp
, &fl
->fl_fasync
, fa
))
496 __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
499 static const struct lock_manager_operations lease_manager_ops
= {
500 .lm_break
= lease_break_callback
,
501 .lm_change
= lease_modify
,
502 .lm_setup
= lease_setup
,
506 * Initialize a lease, use the default lock manager operations
508 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
510 if (assign_type(fl
, type
) != 0)
514 fl
->fl_pid
= current
->tgid
;
517 fl
->fl_flags
= FL_LEASE
;
519 fl
->fl_end
= OFFSET_MAX
;
521 fl
->fl_lmops
= &lease_manager_ops
;
525 /* Allocate a file_lock initialised to this type of lease */
526 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
528 struct file_lock
*fl
= locks_alloc_lock();
532 return ERR_PTR(error
);
534 error
= lease_init(filp
, type
, fl
);
537 return ERR_PTR(error
);
542 /* Check if two locks overlap each other.
544 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
546 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
547 (fl2
->fl_end
>= fl1
->fl_start
));
551 * Check whether two locks have the same owner.
553 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
555 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
556 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
557 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
558 return fl1
->fl_owner
== fl2
->fl_owner
;
561 /* Must be called with the flc_lock held! */
562 static void locks_insert_global_locks(struct file_lock
*fl
)
564 lg_local_lock(&file_lock_lglock
);
565 fl
->fl_link_cpu
= smp_processor_id();
566 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
567 lg_local_unlock(&file_lock_lglock
);
570 /* Must be called with the flc_lock held! */
571 static void locks_delete_global_locks(struct file_lock
*fl
)
574 * Avoid taking lock if already unhashed. This is safe since this check
575 * is done while holding the flc_lock, and new insertions into the list
576 * also require that it be held.
578 if (hlist_unhashed(&fl
->fl_link
))
580 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
581 hlist_del_init(&fl
->fl_link
);
582 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
586 posix_owner_key(struct file_lock
*fl
)
588 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
589 return fl
->fl_lmops
->lm_owner_key(fl
);
590 return (unsigned long)fl
->fl_owner
;
593 static void locks_insert_global_blocked(struct file_lock
*waiter
)
595 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
598 static void locks_delete_global_blocked(struct file_lock
*waiter
)
600 hash_del(&waiter
->fl_link
);
603 /* Remove waiter from blocker's block list.
604 * When blocker ends up pointing to itself then the list is empty.
606 * Must be called with blocked_lock_lock held.
608 static void __locks_delete_block(struct file_lock
*waiter
)
610 locks_delete_global_blocked(waiter
);
611 list_del_init(&waiter
->fl_block
);
612 waiter
->fl_next
= NULL
;
615 static void locks_delete_block(struct file_lock
*waiter
)
617 spin_lock(&blocked_lock_lock
);
618 __locks_delete_block(waiter
);
619 spin_unlock(&blocked_lock_lock
);
622 /* Insert waiter into blocker's block list.
623 * We use a circular list so that processes can be easily woken up in
624 * the order they blocked. The documentation doesn't require this but
625 * it seems like the reasonable thing to do.
627 * Must be called with both the flc_lock and blocked_lock_lock held. The
628 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
629 * that the flc_lock is also held on insertions we can avoid taking the
630 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
632 static void __locks_insert_block(struct file_lock
*blocker
,
633 struct file_lock
*waiter
)
635 BUG_ON(!list_empty(&waiter
->fl_block
));
636 waiter
->fl_next
= blocker
;
637 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
638 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
639 locks_insert_global_blocked(waiter
);
642 /* Must be called with flc_lock held. */
643 static void locks_insert_block(struct file_lock
*blocker
,
644 struct file_lock
*waiter
)
646 spin_lock(&blocked_lock_lock
);
647 __locks_insert_block(blocker
, waiter
);
648 spin_unlock(&blocked_lock_lock
);
652 * Wake up processes blocked waiting for blocker.
654 * Must be called with the inode->flc_lock held!
656 static void locks_wake_up_blocks(struct file_lock
*blocker
)
659 * Avoid taking global lock if list is empty. This is safe since new
660 * blocked requests are only added to the list under the flc_lock, and
661 * the flc_lock is always held here. Note that removal from the fl_block
662 * list does not require the flc_lock, so we must recheck list_empty()
663 * after acquiring the blocked_lock_lock.
665 if (list_empty(&blocker
->fl_block
))
668 spin_lock(&blocked_lock_lock
);
669 while (!list_empty(&blocker
->fl_block
)) {
670 struct file_lock
*waiter
;
672 waiter
= list_first_entry(&blocker
->fl_block
,
673 struct file_lock
, fl_block
);
674 __locks_delete_block(waiter
);
675 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
676 waiter
->fl_lmops
->lm_notify(waiter
);
678 wake_up(&waiter
->fl_wait
);
680 spin_unlock(&blocked_lock_lock
);
684 locks_insert_lock_ctx(struct file_lock
*fl
, struct list_head
*before
)
686 fl
->fl_nspid
= get_pid(task_tgid(current
));
687 list_add_tail(&fl
->fl_list
, before
);
688 locks_insert_global_locks(fl
);
692 locks_unlink_lock_ctx(struct file_lock
*fl
)
694 locks_delete_global_locks(fl
);
695 list_del_init(&fl
->fl_list
);
697 put_pid(fl
->fl_nspid
);
700 locks_wake_up_blocks(fl
);
704 locks_delete_lock_ctx(struct file_lock
*fl
, struct list_head
*dispose
)
706 locks_unlink_lock_ctx(fl
);
708 list_add(&fl
->fl_list
, dispose
);
713 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
714 * checks for shared/exclusive status of overlapping locks.
716 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
718 if (sys_fl
->fl_type
== F_WRLCK
)
720 if (caller_fl
->fl_type
== F_WRLCK
)
725 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
726 * checking before calling the locks_conflict().
728 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
730 /* POSIX locks owned by the same process do not conflict with
733 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
736 /* Check whether they overlap */
737 if (!locks_overlap(caller_fl
, sys_fl
))
740 return (locks_conflict(caller_fl
, sys_fl
));
743 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
744 * checking before calling the locks_conflict().
746 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
748 /* FLOCK locks referring to the same filp do not conflict with
751 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
753 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
756 return (locks_conflict(caller_fl
, sys_fl
));
760 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
762 struct file_lock
*cfl
;
763 struct file_lock_context
*ctx
;
764 struct inode
*inode
= file_inode(filp
);
766 ctx
= inode
->i_flctx
;
767 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
)) {
768 fl
->fl_type
= F_UNLCK
;
772 spin_lock(&ctx
->flc_lock
);
773 list_for_each_entry(cfl
, &ctx
->flc_posix
, fl_list
) {
774 if (posix_locks_conflict(fl
, cfl
)) {
775 locks_copy_conflock(fl
, cfl
);
777 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
781 fl
->fl_type
= F_UNLCK
;
783 spin_unlock(&ctx
->flc_lock
);
786 EXPORT_SYMBOL(posix_test_lock
);
789 * Deadlock detection:
791 * We attempt to detect deadlocks that are due purely to posix file
794 * We assume that a task can be waiting for at most one lock at a time.
795 * So for any acquired lock, the process holding that lock may be
796 * waiting on at most one other lock. That lock in turns may be held by
797 * someone waiting for at most one other lock. Given a requested lock
798 * caller_fl which is about to wait for a conflicting lock block_fl, we
799 * follow this chain of waiters to ensure we are not about to create a
802 * Since we do this before we ever put a process to sleep on a lock, we
803 * are ensured that there is never a cycle; that is what guarantees that
804 * the while() loop in posix_locks_deadlock() eventually completes.
806 * Note: the above assumption may not be true when handling lock
807 * requests from a broken NFS client. It may also fail in the presence
808 * of tasks (such as posix threads) sharing the same open file table.
809 * To handle those cases, we just bail out after a few iterations.
811 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
812 * Because the owner is not even nominally tied to a thread of
813 * execution, the deadlock detection below can't reasonably work well. Just
816 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
817 * locks that just checks for the case where two tasks are attempting to
818 * upgrade from read to write locks on the same inode.
821 #define MAX_DEADLK_ITERATIONS 10
823 /* Find a lock that the owner of the given block_fl is blocking on. */
824 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
826 struct file_lock
*fl
;
828 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
829 if (posix_same_owner(fl
, block_fl
))
835 /* Must be called with the blocked_lock_lock held! */
836 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
837 struct file_lock
*block_fl
)
842 * This deadlock detector can't reasonably detect deadlocks with
843 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
845 if (IS_OFDLCK(caller_fl
))
848 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
849 if (i
++ > MAX_DEADLK_ITERATIONS
)
851 if (posix_same_owner(caller_fl
, block_fl
))
857 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
858 * after any leases, but before any posix locks.
860 * Note that if called with an FL_EXISTS argument, the caller may determine
861 * whether or not a lock was successfully freed by testing the return
864 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
866 struct file_lock
*new_fl
= NULL
;
867 struct file_lock
*fl
;
868 struct file_lock_context
*ctx
;
869 struct inode
*inode
= file_inode(filp
);
874 ctx
= locks_get_lock_context(inode
);
878 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
879 new_fl
= locks_alloc_lock();
884 spin_lock(&ctx
->flc_lock
);
885 if (request
->fl_flags
& FL_ACCESS
)
888 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
889 if (filp
!= fl
->fl_file
)
891 if (request
->fl_type
== fl
->fl_type
)
894 locks_delete_lock_ctx(fl
, &dispose
);
898 if (request
->fl_type
== F_UNLCK
) {
899 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
905 * If a higher-priority process was blocked on the old file lock,
906 * give it the opportunity to lock the file.
909 spin_unlock(&ctx
->flc_lock
);
911 spin_lock(&ctx
->flc_lock
);
915 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
916 if (!flock_locks_conflict(request
, fl
))
919 if (!(request
->fl_flags
& FL_SLEEP
))
921 error
= FILE_LOCK_DEFERRED
;
922 locks_insert_block(fl
, request
);
925 if (request
->fl_flags
& FL_ACCESS
)
927 locks_copy_lock(new_fl
, request
);
928 locks_insert_lock_ctx(new_fl
, &ctx
->flc_flock
);
933 spin_unlock(&ctx
->flc_lock
);
935 locks_free_lock(new_fl
);
936 locks_dispose_list(&dispose
);
940 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
942 struct file_lock
*fl
, *tmp
;
943 struct file_lock
*new_fl
= NULL
;
944 struct file_lock
*new_fl2
= NULL
;
945 struct file_lock
*left
= NULL
;
946 struct file_lock
*right
= NULL
;
947 struct file_lock_context
*ctx
;
952 ctx
= locks_get_lock_context(inode
);
957 * We may need two file_lock structures for this operation,
958 * so we get them in advance to avoid races.
960 * In some cases we can be sure, that no new locks will be needed
962 if (!(request
->fl_flags
& FL_ACCESS
) &&
963 (request
->fl_type
!= F_UNLCK
||
964 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
965 new_fl
= locks_alloc_lock();
966 new_fl2
= locks_alloc_lock();
969 spin_lock(&ctx
->flc_lock
);
971 * New lock request. Walk all POSIX locks and look for conflicts. If
972 * there are any, either return error or put the request on the
973 * blocker's list of waiters and the global blocked_hash.
975 if (request
->fl_type
!= F_UNLCK
) {
976 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
979 if (!posix_locks_conflict(request
, fl
))
982 locks_copy_conflock(conflock
, fl
);
984 if (!(request
->fl_flags
& FL_SLEEP
))
987 * Deadlock detection and insertion into the blocked
988 * locks list must be done while holding the same lock!
991 spin_lock(&blocked_lock_lock
);
992 if (likely(!posix_locks_deadlock(request
, fl
))) {
993 error
= FILE_LOCK_DEFERRED
;
994 __locks_insert_block(fl
, request
);
996 spin_unlock(&blocked_lock_lock
);
1001 /* If we're just looking for a conflict, we're done. */
1003 if (request
->fl_flags
& FL_ACCESS
)
1006 /* Find the first old lock with the same owner as the new lock */
1007 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1008 if (posix_same_owner(request
, fl
))
1012 /* Process locks with this owner. */
1013 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1014 if (!posix_same_owner(request
, fl
))
1017 /* Detect adjacent or overlapping regions (if same lock type) */
1018 if (request
->fl_type
== fl
->fl_type
) {
1019 /* In all comparisons of start vs end, use
1020 * "start - 1" rather than "end + 1". If end
1021 * is OFFSET_MAX, end + 1 will become negative.
1023 if (fl
->fl_end
< request
->fl_start
- 1)
1025 /* If the next lock in the list has entirely bigger
1026 * addresses than the new one, insert the lock here.
1028 if (fl
->fl_start
- 1 > request
->fl_end
)
1031 /* If we come here, the new and old lock are of the
1032 * same type and adjacent or overlapping. Make one
1033 * lock yielding from the lower start address of both
1034 * locks to the higher end address.
1036 if (fl
->fl_start
> request
->fl_start
)
1037 fl
->fl_start
= request
->fl_start
;
1039 request
->fl_start
= fl
->fl_start
;
1040 if (fl
->fl_end
< request
->fl_end
)
1041 fl
->fl_end
= request
->fl_end
;
1043 request
->fl_end
= fl
->fl_end
;
1045 locks_delete_lock_ctx(fl
, &dispose
);
1051 /* Processing for different lock types is a bit
1054 if (fl
->fl_end
< request
->fl_start
)
1056 if (fl
->fl_start
> request
->fl_end
)
1058 if (request
->fl_type
== F_UNLCK
)
1060 if (fl
->fl_start
< request
->fl_start
)
1062 /* If the next lock in the list has a higher end
1063 * address than the new one, insert the new one here.
1065 if (fl
->fl_end
> request
->fl_end
) {
1069 if (fl
->fl_start
>= request
->fl_start
) {
1070 /* The new lock completely replaces an old
1071 * one (This may happen several times).
1074 locks_delete_lock_ctx(fl
, &dispose
);
1078 * Replace the old lock with new_fl, and
1079 * remove the old one. It's safe to do the
1080 * insert here since we know that we won't be
1081 * using new_fl later, and that the lock is
1082 * just replacing an existing lock.
1087 locks_copy_lock(new_fl
, request
);
1090 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1091 locks_delete_lock_ctx(fl
, &dispose
);
1098 * The above code only modifies existing locks in case of merging or
1099 * replacing. If new lock(s) need to be inserted all modifications are
1100 * done below this, so it's safe yet to bail out.
1102 error
= -ENOLCK
; /* "no luck" */
1103 if (right
&& left
== right
&& !new_fl2
)
1108 if (request
->fl_type
== F_UNLCK
) {
1109 if (request
->fl_flags
& FL_EXISTS
)
1118 locks_copy_lock(new_fl
, request
);
1119 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
**before
, int arg
, struct list_head
*dispose
)
1314 struct file_lock
*fl
= *before
;
1315 int error
= assign_type(fl
, arg
);
1319 lease_clear_pending(fl
, arg
);
1320 locks_wake_up_blocks(fl
);
1321 if (arg
== F_UNLCK
) {
1322 struct file
*filp
= fl
->fl_file
;
1325 filp
->f_owner
.signum
= 0;
1326 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1327 if (fl
->fl_fasync
!= NULL
) {
1328 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1329 fl
->fl_fasync
= NULL
;
1331 locks_delete_lock_ctx(fl
, dispose
);
1335 EXPORT_SYMBOL(lease_modify
);
1337 static bool past_time(unsigned long then
)
1340 /* 0 is a special value meaning "this never expires": */
1342 return time_after(jiffies
, then
);
1345 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1347 struct file_lock_context
*ctx
= inode
->i_flctx
;
1348 struct file_lock
*fl
, *tmp
;
1350 lockdep_assert_held(&ctx
->flc_lock
);
1352 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1353 trace_time_out_leases(inode
, fl
);
1354 if (past_time(fl
->fl_downgrade_time
))
1355 lease_modify(&fl
, F_RDLCK
, dispose
);
1356 if (past_time(fl
->fl_break_time
))
1357 lease_modify(&fl
, F_UNLCK
, dispose
);
1361 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1363 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1365 return locks_conflict(breaker
, lease
);
1369 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1371 struct file_lock_context
*ctx
= inode
->i_flctx
;
1372 struct file_lock
*fl
;
1374 lockdep_assert_held(&ctx
->flc_lock
);
1376 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1377 if (leases_conflict(fl
, breaker
))
1384 * __break_lease - revoke all outstanding leases on file
1385 * @inode: the inode of the file to return
1386 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1388 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1391 * break_lease (inlined for speed) has checked there already is at least
1392 * some kind of lock (maybe a lease) on this file. Leases are broken on
1393 * a call to open() or truncate(). This function can sleep unless you
1394 * specified %O_NONBLOCK to your open().
1396 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1399 struct file_lock
*new_fl
;
1400 struct file_lock_context
*ctx
= inode
->i_flctx
;
1401 struct file_lock
*fl
;
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(fl
, &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
)
1588 struct inode
*inode
= dentry
->d_inode
;
1590 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1593 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1594 (atomic_read(&inode
->i_count
) > 1)))
1601 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1603 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1604 struct dentry
*dentry
= filp
->f_path
.dentry
;
1605 struct inode
*inode
= dentry
->d_inode
;
1606 struct file_lock_context
*ctx
;
1607 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1612 trace_generic_add_lease(inode
, lease
);
1614 ctx
= locks_get_lock_context(inode
);
1619 * In the delegation case we need mutual exclusion with
1620 * a number of operations that take the i_mutex. We trylock
1621 * because delegations are an optional optimization, and if
1622 * there's some chance of a conflict--we'd rather not
1623 * bother, maybe that's a sign this just isn't a good file to
1624 * hand out a delegation on.
1626 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1629 if (is_deleg
&& arg
== F_WRLCK
) {
1630 /* Write delegations are not currently supported: */
1631 mutex_unlock(&inode
->i_mutex
);
1636 spin_lock(&ctx
->flc_lock
);
1637 time_out_leases(inode
, &dispose
);
1638 error
= check_conflicting_open(dentry
, arg
);
1643 * At this point, we know that if there is an exclusive
1644 * lease on this file, then we hold it on this filp
1645 * (otherwise our open of this file would have blocked).
1646 * And if we are trying to acquire an exclusive lease,
1647 * then the file is not open by anyone (including us)
1648 * except for this filp.
1651 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1652 if (fl
->fl_file
== filp
) {
1658 * No exclusive leases if someone else has a lease on
1664 * Modifying our existing lease is OK, but no getting a
1665 * new lease if someone else is opening for write:
1667 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1671 if (my_fl
!= NULL
) {
1672 error
= lease
->fl_lmops
->lm_change(&my_fl
, arg
, &dispose
);
1682 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1684 * The check in break_lease() is lockless. It's possible for another
1685 * open to race in after we did the earlier check for a conflicting
1686 * open but before the lease was inserted. Check again for a
1687 * conflicting open and cancel the lease if there is one.
1689 * We also add a barrier here to ensure that the insertion of the lock
1690 * precedes these checks.
1693 error
= check_conflicting_open(dentry
, arg
);
1695 locks_unlink_lock_ctx(lease
);
1700 if (lease
->fl_lmops
->lm_setup
)
1701 lease
->fl_lmops
->lm_setup(lease
, priv
);
1703 spin_unlock(&ctx
->flc_lock
);
1704 locks_dispose_list(&dispose
);
1706 mutex_unlock(&inode
->i_mutex
);
1707 if (!error
&& !my_fl
)
1712 static int generic_delete_lease(struct file
*filp
)
1714 int error
= -EAGAIN
;
1715 struct file_lock
*fl
, *victim
= NULL
;
1716 struct dentry
*dentry
= filp
->f_path
.dentry
;
1717 struct inode
*inode
= dentry
->d_inode
;
1718 struct file_lock_context
*ctx
= inode
->i_flctx
;
1722 trace_generic_delete_lease(inode
, NULL
);
1726 spin_lock(&ctx
->flc_lock
);
1727 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1728 if (fl
->fl_file
== filp
) {
1733 trace_generic_delete_lease(inode
, fl
);
1735 error
= fl
->fl_lmops
->lm_change(&victim
, F_UNLCK
, &dispose
);
1736 spin_unlock(&ctx
->flc_lock
);
1737 locks_dispose_list(&dispose
);
1742 * generic_setlease - sets a lease on an open file
1743 * @filp: file pointer
1744 * @arg: type of lease to obtain
1745 * @flp: input - file_lock to use, output - file_lock inserted
1746 * @priv: private data for lm_setup (may be NULL if lm_setup
1747 * doesn't require it)
1749 * The (input) flp->fl_lmops->lm_break function is required
1752 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1755 struct dentry
*dentry
= filp
->f_path
.dentry
;
1756 struct inode
*inode
= dentry
->d_inode
;
1759 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1761 if (!S_ISREG(inode
->i_mode
))
1763 error
= security_file_lock(filp
, arg
);
1769 return generic_delete_lease(filp
);
1772 if (!(*flp
)->fl_lmops
->lm_break
) {
1776 return generic_add_lease(filp
, arg
, flp
, priv
);
1781 EXPORT_SYMBOL(generic_setlease
);
1784 * vfs_setlease - sets a lease on an open file
1785 * @filp: file pointer
1786 * @arg: type of lease to obtain
1787 * @lease: file_lock to use when adding a lease
1788 * @priv: private info for lm_setup when adding a lease (may be
1789 * NULL if lm_setup doesn't require it)
1791 * Call this to establish a lease on the file. The "lease" argument is not
1792 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1793 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1794 * if not, this function will return -ENOLCK (and generate a scary-looking
1797 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1798 * may be NULL if the lm_setup operation doesn't require it.
1801 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1803 if (filp
->f_op
->setlease
)
1804 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1806 return generic_setlease(filp
, arg
, lease
, priv
);
1808 EXPORT_SYMBOL_GPL(vfs_setlease
);
1810 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1812 struct file_lock
*fl
;
1813 struct fasync_struct
*new;
1816 fl
= lease_alloc(filp
, arg
);
1820 new = fasync_alloc();
1822 locks_free_lock(fl
);
1827 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1829 locks_free_lock(fl
);
1836 * fcntl_setlease - sets a lease on an open file
1837 * @fd: open file descriptor
1838 * @filp: file pointer
1839 * @arg: type of lease to obtain
1841 * Call this fcntl to establish a lease on the file.
1842 * Note that you also need to call %F_SETSIG to
1843 * receive a signal when the lease is broken.
1845 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1848 return vfs_setlease(filp
, F_UNLCK
, NULL
, NULL
);
1849 return do_fcntl_add_lease(fd
, filp
, arg
);
1853 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1854 * @filp: The file to apply the lock to
1855 * @fl: The lock to be applied
1857 * Add a FLOCK style lock to a file.
1859 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1864 error
= flock_lock_file(filp
, fl
);
1865 if (error
!= FILE_LOCK_DEFERRED
)
1867 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1871 locks_delete_block(fl
);
1877 EXPORT_SYMBOL(flock_lock_file_wait
);
1880 * sys_flock: - flock() system call.
1881 * @fd: the file descriptor to lock.
1882 * @cmd: the type of lock to apply.
1884 * Apply a %FL_FLOCK style lock to an open file descriptor.
1885 * The @cmd can be one of
1887 * %LOCK_SH -- a shared lock.
1889 * %LOCK_EX -- an exclusive lock.
1891 * %LOCK_UN -- remove an existing lock.
1893 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1895 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1896 * processes read and write access respectively.
1898 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1900 struct fd f
= fdget(fd
);
1901 struct file_lock
*lock
;
1902 int can_sleep
, unlock
;
1909 can_sleep
= !(cmd
& LOCK_NB
);
1911 unlock
= (cmd
== LOCK_UN
);
1913 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1914 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1917 lock
= flock_make_lock(f
.file
, cmd
);
1919 error
= PTR_ERR(lock
);
1924 lock
->fl_flags
|= FL_SLEEP
;
1926 error
= security_file_lock(f
.file
, lock
->fl_type
);
1930 if (f
.file
->f_op
->flock
)
1931 error
= f
.file
->f_op
->flock(f
.file
,
1932 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1935 error
= flock_lock_file_wait(f
.file
, lock
);
1938 locks_free_lock(lock
);
1947 * vfs_test_lock - test file byte range lock
1948 * @filp: The file to test lock for
1949 * @fl: The lock to test; also used to hold result
1951 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1952 * setting conf->fl_type to something other than F_UNLCK.
1954 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1956 if (filp
->f_op
->lock
)
1957 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1958 posix_test_lock(filp
, fl
);
1961 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1963 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1965 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1966 #if BITS_PER_LONG == 32
1968 * Make sure we can represent the posix lock via
1969 * legacy 32bit flock.
1971 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1973 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1976 flock
->l_start
= fl
->fl_start
;
1977 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1978 fl
->fl_end
- fl
->fl_start
+ 1;
1979 flock
->l_whence
= 0;
1980 flock
->l_type
= fl
->fl_type
;
1984 #if BITS_PER_LONG == 32
1985 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1987 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1988 flock
->l_start
= fl
->fl_start
;
1989 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1990 fl
->fl_end
- fl
->fl_start
+ 1;
1991 flock
->l_whence
= 0;
1992 flock
->l_type
= fl
->fl_type
;
1996 /* Report the first existing lock that would conflict with l.
1997 * This implements the F_GETLK command of fcntl().
1999 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2001 struct file_lock file_lock
;
2006 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2009 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2012 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2016 if (cmd
== F_OFD_GETLK
) {
2018 if (flock
.l_pid
!= 0)
2022 file_lock
.fl_flags
|= FL_OFDLCK
;
2023 file_lock
.fl_owner
= filp
;
2026 error
= vfs_test_lock(filp
, &file_lock
);
2030 flock
.l_type
= file_lock
.fl_type
;
2031 if (file_lock
.fl_type
!= F_UNLCK
) {
2032 error
= posix_lock_to_flock(&flock
, &file_lock
);
2037 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2040 locks_release_private(&file_lock
);
2046 * vfs_lock_file - file byte range lock
2047 * @filp: The file to apply the lock to
2048 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2049 * @fl: The lock to be applied
2050 * @conf: Place to return a copy of the conflicting lock, if found.
2052 * A caller that doesn't care about the conflicting lock may pass NULL
2053 * as the final argument.
2055 * If the filesystem defines a private ->lock() method, then @conf will
2056 * be left unchanged; so a caller that cares should initialize it to
2057 * some acceptable default.
2059 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2060 * locks, the ->lock() interface may return asynchronously, before the lock has
2061 * been granted or denied by the underlying filesystem, if (and only if)
2062 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2063 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2064 * the request is for a blocking lock. When ->lock() does return asynchronously,
2065 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2066 * request completes.
2067 * If the request is for non-blocking lock the file system should return
2068 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2069 * with the result. If the request timed out the callback routine will return a
2070 * nonzero return code and the file system should release the lock. The file
2071 * system is also responsible to keep a corresponding posix lock when it
2072 * grants a lock so the VFS can find out which locks are locally held and do
2073 * the correct lock cleanup when required.
2074 * The underlying filesystem must not drop the kernel lock or call
2075 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2078 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2080 if (filp
->f_op
->lock
)
2081 return filp
->f_op
->lock(filp
, cmd
, fl
);
2083 return posix_lock_file(filp
, fl
, conf
);
2085 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2087 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2088 struct file_lock
*fl
)
2092 error
= security_file_lock(filp
, fl
->fl_type
);
2097 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2098 if (error
!= FILE_LOCK_DEFERRED
)
2100 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2104 locks_delete_block(fl
);
2111 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2113 check_fmode_for_setlk(struct file_lock
*fl
)
2115 switch (fl
->fl_type
) {
2117 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2121 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2127 /* Apply the lock described by l to an open file descriptor.
2128 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2130 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2131 struct flock __user
*l
)
2133 struct file_lock
*file_lock
= locks_alloc_lock();
2135 struct inode
*inode
;
2139 if (file_lock
== NULL
)
2143 * This might block, so we do it before checking the inode.
2146 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2149 inode
= file_inode(filp
);
2151 /* Don't allow mandatory locks on files that may be memory mapped
2154 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2160 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2164 error
= check_fmode_for_setlk(file_lock
);
2169 * If the cmd is requesting file-private locks, then set the
2170 * FL_OFDLCK flag and override the owner.
2175 if (flock
.l_pid
!= 0)
2179 file_lock
->fl_flags
|= FL_OFDLCK
;
2180 file_lock
->fl_owner
= filp
;
2184 if (flock
.l_pid
!= 0)
2188 file_lock
->fl_flags
|= FL_OFDLCK
;
2189 file_lock
->fl_owner
= filp
;
2192 file_lock
->fl_flags
|= FL_SLEEP
;
2195 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2198 * Attempt to detect a close/fcntl race and recover by
2199 * releasing the lock that was just acquired.
2202 * we need that spin_lock here - it prevents reordering between
2203 * update of inode->i_flock and check for it done in close().
2204 * rcu_read_lock() wouldn't do.
2206 spin_lock(¤t
->files
->file_lock
);
2208 spin_unlock(¤t
->files
->file_lock
);
2209 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2210 flock
.l_type
= F_UNLCK
;
2215 locks_free_lock(file_lock
);
2219 #if BITS_PER_LONG == 32
2220 /* Report the first existing lock that would conflict with l.
2221 * This implements the F_GETLK command of fcntl().
2223 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2225 struct file_lock file_lock
;
2226 struct flock64 flock
;
2230 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2233 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2236 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2240 if (cmd
== F_OFD_GETLK
) {
2242 if (flock
.l_pid
!= 0)
2246 file_lock
.fl_flags
|= FL_OFDLCK
;
2247 file_lock
.fl_owner
= filp
;
2250 error
= vfs_test_lock(filp
, &file_lock
);
2254 flock
.l_type
= file_lock
.fl_type
;
2255 if (file_lock
.fl_type
!= F_UNLCK
)
2256 posix_lock_to_flock64(&flock
, &file_lock
);
2259 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2262 locks_release_private(&file_lock
);
2267 /* Apply the lock described by l to an open file descriptor.
2268 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2270 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2271 struct flock64 __user
*l
)
2273 struct file_lock
*file_lock
= locks_alloc_lock();
2274 struct flock64 flock
;
2275 struct inode
*inode
;
2279 if (file_lock
== NULL
)
2283 * This might block, so we do it before checking the inode.
2286 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2289 inode
= file_inode(filp
);
2291 /* Don't allow mandatory locks on files that may be memory mapped
2294 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2300 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2304 error
= check_fmode_for_setlk(file_lock
);
2309 * If the cmd is requesting file-private locks, then set the
2310 * FL_OFDLCK flag and override the owner.
2315 if (flock
.l_pid
!= 0)
2319 file_lock
->fl_flags
|= FL_OFDLCK
;
2320 file_lock
->fl_owner
= filp
;
2324 if (flock
.l_pid
!= 0)
2328 file_lock
->fl_flags
|= FL_OFDLCK
;
2329 file_lock
->fl_owner
= filp
;
2332 file_lock
->fl_flags
|= FL_SLEEP
;
2335 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2338 * Attempt to detect a close/fcntl race and recover by
2339 * releasing the lock that was just acquired.
2341 spin_lock(¤t
->files
->file_lock
);
2343 spin_unlock(¤t
->files
->file_lock
);
2344 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2345 flock
.l_type
= F_UNLCK
;
2350 locks_free_lock(file_lock
);
2353 #endif /* BITS_PER_LONG == 32 */
2356 * This function is called when the file is being removed
2357 * from the task's fd array. POSIX locks belonging to this task
2358 * are deleted at this time.
2360 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2362 struct file_lock lock
;
2363 struct file_lock_context
*ctx
= file_inode(filp
)->i_flctx
;
2366 * If there are no locks held on this file, we don't need to call
2367 * posix_lock_file(). Another process could be setting a lock on this
2368 * file at the same time, but we wouldn't remove that lock anyway.
2370 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2373 lock
.fl_type
= F_UNLCK
;
2374 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2376 lock
.fl_end
= OFFSET_MAX
;
2377 lock
.fl_owner
= owner
;
2378 lock
.fl_pid
= current
->tgid
;
2379 lock
.fl_file
= filp
;
2381 lock
.fl_lmops
= NULL
;
2383 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2385 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2386 lock
.fl_ops
->fl_release_private(&lock
);
2389 EXPORT_SYMBOL(locks_remove_posix
);
2392 locks_remove_flock(struct file
*filp
)
2394 struct file_lock fl
= {
2396 .fl_pid
= current
->tgid
,
2398 .fl_flags
= FL_FLOCK
,
2400 .fl_end
= OFFSET_MAX
,
2402 struct file_lock_context
*flctx
= file_inode(filp
)->i_flctx
;
2404 if (!flctx
|| list_empty(&flctx
->flc_flock
))
2407 if (filp
->f_op
->flock
)
2408 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2410 flock_lock_file(filp
, &fl
);
2412 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2413 fl
.fl_ops
->fl_release_private(&fl
);
2417 locks_remove_lease(struct file
*filp
)
2419 struct inode
*inode
= file_inode(filp
);
2420 struct file_lock_context
*ctx
= inode
->i_flctx
;
2421 struct file_lock
*fl
, *tmp
;
2424 if (!ctx
|| list_empty(&ctx
->flc_lease
))
2427 spin_lock(&ctx
->flc_lock
);
2428 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2429 lease_modify(&fl
, F_UNLCK
, &dispose
);
2430 spin_unlock(&ctx
->flc_lock
);
2431 locks_dispose_list(&dispose
);
2435 * This function is called on the last close of an open file.
2437 void locks_remove_file(struct file
*filp
)
2439 /* remove any OFD locks */
2440 locks_remove_posix(filp
, filp
);
2442 /* remove flock locks */
2443 locks_remove_flock(filp
);
2445 /* remove any leases */
2446 locks_remove_lease(filp
);
2450 * posix_unblock_lock - stop waiting for a file lock
2451 * @waiter: the lock which was waiting
2453 * lockd needs to block waiting for locks.
2456 posix_unblock_lock(struct file_lock
*waiter
)
2460 spin_lock(&blocked_lock_lock
);
2461 if (waiter
->fl_next
)
2462 __locks_delete_block(waiter
);
2465 spin_unlock(&blocked_lock_lock
);
2468 EXPORT_SYMBOL(posix_unblock_lock
);
2471 * vfs_cancel_lock - file byte range unblock lock
2472 * @filp: The file to apply the unblock to
2473 * @fl: The lock to be unblocked
2475 * Used by lock managers to cancel blocked requests
2477 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2479 if (filp
->f_op
->lock
)
2480 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2484 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2486 #ifdef CONFIG_PROC_FS
2487 #include <linux/proc_fs.h>
2488 #include <linux/seq_file.h>
2490 struct locks_iterator
{
2495 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2496 loff_t id
, char *pfx
)
2498 struct inode
*inode
= NULL
;
2499 unsigned int fl_pid
;
2502 fl_pid
= pid_vnr(fl
->fl_nspid
);
2504 fl_pid
= fl
->fl_pid
;
2506 if (fl
->fl_file
!= NULL
)
2507 inode
= file_inode(fl
->fl_file
);
2509 seq_printf(f
, "%lld:%s ", id
, pfx
);
2511 if (fl
->fl_flags
& FL_ACCESS
)
2512 seq_puts(f
, "ACCESS");
2513 else if (IS_OFDLCK(fl
))
2514 seq_puts(f
, "OFDLCK");
2516 seq_puts(f
, "POSIX ");
2518 seq_printf(f
, " %s ",
2519 (inode
== NULL
) ? "*NOINODE*" :
2520 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2521 } else if (IS_FLOCK(fl
)) {
2522 if (fl
->fl_type
& LOCK_MAND
) {
2523 seq_puts(f
, "FLOCK MSNFS ");
2525 seq_puts(f
, "FLOCK ADVISORY ");
2527 } else if (IS_LEASE(fl
)) {
2528 if (fl
->fl_flags
& FL_DELEG
)
2529 seq_puts(f
, "DELEG ");
2531 seq_puts(f
, "LEASE ");
2533 if (lease_breaking(fl
))
2534 seq_puts(f
, "BREAKING ");
2535 else if (fl
->fl_file
)
2536 seq_puts(f
, "ACTIVE ");
2538 seq_puts(f
, "BREAKER ");
2540 seq_puts(f
, "UNKNOWN UNKNOWN ");
2542 if (fl
->fl_type
& LOCK_MAND
) {
2543 seq_printf(f
, "%s ",
2544 (fl
->fl_type
& LOCK_READ
)
2545 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2546 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2548 seq_printf(f
, "%s ",
2549 (lease_breaking(fl
))
2550 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2551 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2554 #ifdef WE_CAN_BREAK_LSLK_NOW
2555 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2556 inode
->i_sb
->s_id
, inode
->i_ino
);
2558 /* userspace relies on this representation of dev_t ;-( */
2559 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2560 MAJOR(inode
->i_sb
->s_dev
),
2561 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2564 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2567 if (fl
->fl_end
== OFFSET_MAX
)
2568 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2570 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2572 seq_puts(f
, "0 EOF\n");
2576 static int locks_show(struct seq_file
*f
, void *v
)
2578 struct locks_iterator
*iter
= f
->private;
2579 struct file_lock
*fl
, *bfl
;
2581 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2583 lock_get_status(f
, fl
, iter
->li_pos
, "");
2585 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2586 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2591 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2592 __acquires(&blocked_lock_lock
)
2594 struct locks_iterator
*iter
= f
->private;
2596 iter
->li_pos
= *pos
+ 1;
2597 lg_global_lock(&file_lock_lglock
);
2598 spin_lock(&blocked_lock_lock
);
2599 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2602 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2604 struct locks_iterator
*iter
= f
->private;
2607 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2610 static void locks_stop(struct seq_file
*f
, void *v
)
2611 __releases(&blocked_lock_lock
)
2613 spin_unlock(&blocked_lock_lock
);
2614 lg_global_unlock(&file_lock_lglock
);
2617 static const struct seq_operations locks_seq_operations
= {
2618 .start
= locks_start
,
2624 static int locks_open(struct inode
*inode
, struct file
*filp
)
2626 return seq_open_private(filp
, &locks_seq_operations
,
2627 sizeof(struct locks_iterator
));
2630 static const struct file_operations proc_locks_operations
= {
2633 .llseek
= seq_lseek
,
2634 .release
= seq_release_private
,
2637 static int __init
proc_locks_init(void)
2639 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2642 module_init(proc_locks_init
);
2645 static int __init
filelock_init(void)
2649 flctx_cache
= kmem_cache_create("file_lock_ctx",
2650 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2652 filelock_cache
= kmem_cache_create("file_lock_cache",
2653 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2655 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2657 for_each_possible_cpu(i
)
2658 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2663 core_initcall(filelock_init
);