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/security.h>
123 #include <linux/slab.h>
124 #include <linux/syscalls.h>
125 #include <linux/time.h>
126 #include <linux/rcupdate.h>
127 #include <linux/pid_namespace.h>
128 #include <linux/hashtable.h>
129 #include <linux/percpu.h>
130 #include <linux/lglock.h>
132 #define CREATE_TRACE_POINTS
133 #include <trace/events/filelock.h>
135 #include <asm/uaccess.h>
137 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
138 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
139 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
140 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
142 static bool lease_breaking(struct file_lock
*fl
)
144 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
147 static int target_leasetype(struct file_lock
*fl
)
149 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
151 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
156 int leases_enable
= 1;
157 int lease_break_time
= 45;
160 * The global file_lock_list is only used for displaying /proc/locks, so we
161 * keep a list on each CPU, with each list protected by its own spinlock via
162 * the file_lock_lglock. Note that alterations to the list also require that
163 * the relevant flc_lock is held.
165 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
166 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
169 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
170 * It is protected by blocked_lock_lock.
172 * We hash locks by lockowner in order to optimize searching for the lock a
173 * particular lockowner is waiting on.
175 * FIXME: make this value scale via some heuristic? We generally will want more
176 * buckets when we have more lockowners holding locks, but that's a little
177 * difficult to determine without knowing what the workload will look like.
179 #define BLOCKED_HASH_BITS 7
180 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
183 * This lock protects the blocked_hash. Generally, if you're accessing it, you
184 * want to be holding this lock.
186 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
187 * pointer for file_lock structures that are acting as lock requests (in
188 * contrast to those that are acting as records of acquired locks).
190 * Note that when we acquire this lock in order to change the above fields,
191 * we often hold the flc_lock as well. In certain cases, when reading the fields
192 * protected by this lock, we can skip acquiring it iff we already hold the
195 * In particular, adding an entry to the fl_block list requires that you hold
196 * both the flc_lock and the blocked_lock_lock (acquired in that order).
197 * Deleting an entry from the list however only requires the file_lock_lock.
199 static DEFINE_SPINLOCK(blocked_lock_lock
);
201 static struct kmem_cache
*flctx_cache __read_mostly
;
202 static struct kmem_cache
*filelock_cache __read_mostly
;
204 static struct file_lock_context
*
205 locks_get_lock_context(struct inode
*inode
, int type
)
207 struct file_lock_context
*ctx
;
209 /* paired with cmpxchg() below */
210 ctx
= smp_load_acquire(&inode
->i_flctx
);
211 if (likely(ctx
) || type
== F_UNLCK
)
214 ctx
= kmem_cache_alloc(flctx_cache
, GFP_KERNEL
);
218 spin_lock_init(&ctx
->flc_lock
);
219 INIT_LIST_HEAD(&ctx
->flc_flock
);
220 INIT_LIST_HEAD(&ctx
->flc_posix
);
221 INIT_LIST_HEAD(&ctx
->flc_lease
);
224 * Assign the pointer if it's not already assigned. If it is, then
225 * free the context we just allocated.
227 if (cmpxchg(&inode
->i_flctx
, NULL
, ctx
)) {
228 kmem_cache_free(flctx_cache
, ctx
);
229 ctx
= smp_load_acquire(&inode
->i_flctx
);
232 trace_locks_get_lock_context(inode
, type
, ctx
);
237 locks_dump_ctx_list(struct list_head
*list
, char *list_type
)
239 struct file_lock
*fl
;
241 list_for_each_entry(fl
, list
, fl_list
) {
242 pr_warn("%s: fl_owner=%p fl_flags=0x%x fl_type=0x%x fl_pid=%u\n", list_type
, fl
->fl_owner
, fl
->fl_flags
, fl
->fl_type
, fl
->fl_pid
);
247 locks_check_ctx_lists(struct inode
*inode
)
249 struct file_lock_context
*ctx
= inode
->i_flctx
;
251 if (unlikely(!list_empty(&ctx
->flc_flock
) ||
252 !list_empty(&ctx
->flc_posix
) ||
253 !list_empty(&ctx
->flc_lease
))) {
254 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
255 MAJOR(inode
->i_sb
->s_dev
), MINOR(inode
->i_sb
->s_dev
),
257 locks_dump_ctx_list(&ctx
->flc_flock
, "FLOCK");
258 locks_dump_ctx_list(&ctx
->flc_posix
, "POSIX");
259 locks_dump_ctx_list(&ctx
->flc_lease
, "LEASE");
264 locks_free_lock_context(struct inode
*inode
)
266 struct file_lock_context
*ctx
= inode
->i_flctx
;
269 locks_check_ctx_lists(inode
);
270 kmem_cache_free(flctx_cache
, ctx
);
274 static void locks_init_lock_heads(struct file_lock
*fl
)
276 INIT_HLIST_NODE(&fl
->fl_link
);
277 INIT_LIST_HEAD(&fl
->fl_list
);
278 INIT_LIST_HEAD(&fl
->fl_block
);
279 init_waitqueue_head(&fl
->fl_wait
);
282 /* Allocate an empty lock structure. */
283 struct file_lock
*locks_alloc_lock(void)
285 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
288 locks_init_lock_heads(fl
);
292 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
294 void locks_release_private(struct file_lock
*fl
)
297 if (fl
->fl_ops
->fl_release_private
)
298 fl
->fl_ops
->fl_release_private(fl
);
303 if (fl
->fl_lmops
->lm_put_owner
) {
304 fl
->fl_lmops
->lm_put_owner(fl
->fl_owner
);
310 EXPORT_SYMBOL_GPL(locks_release_private
);
312 /* Free a lock which is not in use. */
313 void locks_free_lock(struct file_lock
*fl
)
315 BUG_ON(waitqueue_active(&fl
->fl_wait
));
316 BUG_ON(!list_empty(&fl
->fl_list
));
317 BUG_ON(!list_empty(&fl
->fl_block
));
318 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
320 locks_release_private(fl
);
321 kmem_cache_free(filelock_cache
, fl
);
323 EXPORT_SYMBOL(locks_free_lock
);
326 locks_dispose_list(struct list_head
*dispose
)
328 struct file_lock
*fl
;
330 while (!list_empty(dispose
)) {
331 fl
= list_first_entry(dispose
, struct file_lock
, fl_list
);
332 list_del_init(&fl
->fl_list
);
337 void locks_init_lock(struct file_lock
*fl
)
339 memset(fl
, 0, sizeof(struct file_lock
));
340 locks_init_lock_heads(fl
);
343 EXPORT_SYMBOL(locks_init_lock
);
346 * Initialize a new lock from an existing file_lock structure.
348 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
350 new->fl_owner
= fl
->fl_owner
;
351 new->fl_pid
= fl
->fl_pid
;
353 new->fl_flags
= fl
->fl_flags
;
354 new->fl_type
= fl
->fl_type
;
355 new->fl_start
= fl
->fl_start
;
356 new->fl_end
= fl
->fl_end
;
357 new->fl_lmops
= fl
->fl_lmops
;
361 if (fl
->fl_lmops
->lm_get_owner
)
362 fl
->fl_lmops
->lm_get_owner(fl
->fl_owner
);
365 EXPORT_SYMBOL(locks_copy_conflock
);
367 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
369 /* "new" must be a freshly-initialized lock */
370 WARN_ON_ONCE(new->fl_ops
);
372 locks_copy_conflock(new, fl
);
374 new->fl_file
= fl
->fl_file
;
375 new->fl_ops
= fl
->fl_ops
;
378 if (fl
->fl_ops
->fl_copy_lock
)
379 fl
->fl_ops
->fl_copy_lock(new, fl
);
383 EXPORT_SYMBOL(locks_copy_lock
);
385 static inline int flock_translate_cmd(int cmd
) {
387 return cmd
& (LOCK_MAND
| LOCK_RW
);
399 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
400 static struct file_lock
*
401 flock_make_lock(struct file
*filp
, unsigned int cmd
)
403 struct file_lock
*fl
;
404 int type
= flock_translate_cmd(cmd
);
407 return ERR_PTR(type
);
409 fl
= locks_alloc_lock();
411 return ERR_PTR(-ENOMEM
);
415 fl
->fl_pid
= current
->tgid
;
416 fl
->fl_flags
= FL_FLOCK
;
418 fl
->fl_end
= OFFSET_MAX
;
423 static int assign_type(struct file_lock
*fl
, long type
)
437 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
440 switch (l
->l_whence
) {
445 fl
->fl_start
= filp
->f_pos
;
448 fl
->fl_start
= i_size_read(file_inode(filp
));
453 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
455 fl
->fl_start
+= l
->l_start
;
456 if (fl
->fl_start
< 0)
459 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
460 POSIX-2001 defines it. */
462 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
464 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
466 } else if (l
->l_len
< 0) {
467 if (fl
->fl_start
+ l
->l_len
< 0)
469 fl
->fl_end
= fl
->fl_start
- 1;
470 fl
->fl_start
+= l
->l_len
;
472 fl
->fl_end
= OFFSET_MAX
;
474 fl
->fl_owner
= current
->files
;
475 fl
->fl_pid
= current
->tgid
;
477 fl
->fl_flags
= FL_POSIX
;
481 return assign_type(fl
, l
->l_type
);
484 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
487 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
490 struct flock64 ll
= {
492 .l_whence
= l
->l_whence
,
493 .l_start
= l
->l_start
,
497 return flock64_to_posix_lock(filp
, fl
, &ll
);
500 /* default lease lock manager operations */
502 lease_break_callback(struct file_lock
*fl
)
504 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
509 lease_setup(struct file_lock
*fl
, void **priv
)
511 struct file
*filp
= fl
->fl_file
;
512 struct fasync_struct
*fa
= *priv
;
515 * fasync_insert_entry() returns the old entry if any. If there was no
516 * old entry, then it used "priv" and inserted it into the fasync list.
517 * Clear the pointer to indicate that it shouldn't be freed.
519 if (!fasync_insert_entry(fa
->fa_fd
, filp
, &fl
->fl_fasync
, fa
))
522 __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
525 static const struct lock_manager_operations lease_manager_ops
= {
526 .lm_break
= lease_break_callback
,
527 .lm_change
= lease_modify
,
528 .lm_setup
= lease_setup
,
532 * Initialize a lease, use the default lock manager operations
534 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
536 if (assign_type(fl
, type
) != 0)
540 fl
->fl_pid
= current
->tgid
;
543 fl
->fl_flags
= FL_LEASE
;
545 fl
->fl_end
= OFFSET_MAX
;
547 fl
->fl_lmops
= &lease_manager_ops
;
551 /* Allocate a file_lock initialised to this type of lease */
552 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
554 struct file_lock
*fl
= locks_alloc_lock();
558 return ERR_PTR(error
);
560 error
= lease_init(filp
, type
, fl
);
563 return ERR_PTR(error
);
568 /* Check if two locks overlap each other.
570 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
572 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
573 (fl2
->fl_end
>= fl1
->fl_start
));
577 * Check whether two locks have the same owner.
579 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
581 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
582 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
583 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
584 return fl1
->fl_owner
== fl2
->fl_owner
;
587 /* Must be called with the flc_lock held! */
588 static void locks_insert_global_locks(struct file_lock
*fl
)
590 lg_local_lock(&file_lock_lglock
);
591 fl
->fl_link_cpu
= smp_processor_id();
592 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
593 lg_local_unlock(&file_lock_lglock
);
596 /* Must be called with the flc_lock held! */
597 static void locks_delete_global_locks(struct file_lock
*fl
)
600 * Avoid taking lock if already unhashed. This is safe since this check
601 * is done while holding the flc_lock, and new insertions into the list
602 * also require that it be held.
604 if (hlist_unhashed(&fl
->fl_link
))
606 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
607 hlist_del_init(&fl
->fl_link
);
608 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
612 posix_owner_key(struct file_lock
*fl
)
614 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
615 return fl
->fl_lmops
->lm_owner_key(fl
);
616 return (unsigned long)fl
->fl_owner
;
619 static void locks_insert_global_blocked(struct file_lock
*waiter
)
621 lockdep_assert_held(&blocked_lock_lock
);
623 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
626 static void locks_delete_global_blocked(struct file_lock
*waiter
)
628 lockdep_assert_held(&blocked_lock_lock
);
630 hash_del(&waiter
->fl_link
);
633 /* Remove waiter from blocker's block list.
634 * When blocker ends up pointing to itself then the list is empty.
636 * Must be called with blocked_lock_lock held.
638 static void __locks_delete_block(struct file_lock
*waiter
)
640 locks_delete_global_blocked(waiter
);
641 list_del_init(&waiter
->fl_block
);
642 waiter
->fl_next
= NULL
;
645 static void locks_delete_block(struct file_lock
*waiter
)
647 spin_lock(&blocked_lock_lock
);
648 __locks_delete_block(waiter
);
649 spin_unlock(&blocked_lock_lock
);
652 /* Insert waiter into blocker's block list.
653 * We use a circular list so that processes can be easily woken up in
654 * the order they blocked. The documentation doesn't require this but
655 * it seems like the reasonable thing to do.
657 * Must be called with both the flc_lock and blocked_lock_lock held. The
658 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
659 * that the flc_lock is also held on insertions we can avoid taking the
660 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
662 static void __locks_insert_block(struct file_lock
*blocker
,
663 struct file_lock
*waiter
)
665 BUG_ON(!list_empty(&waiter
->fl_block
));
666 waiter
->fl_next
= blocker
;
667 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
668 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
669 locks_insert_global_blocked(waiter
);
672 /* Must be called with flc_lock held. */
673 static void locks_insert_block(struct file_lock
*blocker
,
674 struct file_lock
*waiter
)
676 spin_lock(&blocked_lock_lock
);
677 __locks_insert_block(blocker
, waiter
);
678 spin_unlock(&blocked_lock_lock
);
682 * Wake up processes blocked waiting for blocker.
684 * Must be called with the inode->flc_lock held!
686 static void locks_wake_up_blocks(struct file_lock
*blocker
)
689 * Avoid taking global lock if list is empty. This is safe since new
690 * blocked requests are only added to the list under the flc_lock, and
691 * the flc_lock is always held here. Note that removal from the fl_block
692 * list does not require the flc_lock, so we must recheck list_empty()
693 * after acquiring the blocked_lock_lock.
695 if (list_empty(&blocker
->fl_block
))
698 spin_lock(&blocked_lock_lock
);
699 while (!list_empty(&blocker
->fl_block
)) {
700 struct file_lock
*waiter
;
702 waiter
= list_first_entry(&blocker
->fl_block
,
703 struct file_lock
, fl_block
);
704 __locks_delete_block(waiter
);
705 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
706 waiter
->fl_lmops
->lm_notify(waiter
);
708 wake_up(&waiter
->fl_wait
);
710 spin_unlock(&blocked_lock_lock
);
714 locks_insert_lock_ctx(struct file_lock
*fl
, struct list_head
*before
)
716 fl
->fl_nspid
= get_pid(task_tgid(current
));
717 list_add_tail(&fl
->fl_list
, before
);
718 locks_insert_global_locks(fl
);
722 locks_unlink_lock_ctx(struct file_lock
*fl
)
724 locks_delete_global_locks(fl
);
725 list_del_init(&fl
->fl_list
);
727 put_pid(fl
->fl_nspid
);
730 locks_wake_up_blocks(fl
);
734 locks_delete_lock_ctx(struct file_lock
*fl
, struct list_head
*dispose
)
736 locks_unlink_lock_ctx(fl
);
738 list_add(&fl
->fl_list
, dispose
);
743 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
744 * checks for shared/exclusive status of overlapping locks.
746 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
748 if (sys_fl
->fl_type
== F_WRLCK
)
750 if (caller_fl
->fl_type
== F_WRLCK
)
755 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
756 * checking before calling the locks_conflict().
758 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
760 /* POSIX locks owned by the same process do not conflict with
763 if (posix_same_owner(caller_fl
, sys_fl
))
766 /* Check whether they overlap */
767 if (!locks_overlap(caller_fl
, sys_fl
))
770 return (locks_conflict(caller_fl
, sys_fl
));
773 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
774 * checking before calling the locks_conflict().
776 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
778 /* FLOCK locks referring to the same filp do not conflict with
781 if (caller_fl
->fl_file
== sys_fl
->fl_file
)
783 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
786 return (locks_conflict(caller_fl
, sys_fl
));
790 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
792 struct file_lock
*cfl
;
793 struct file_lock_context
*ctx
;
794 struct inode
*inode
= file_inode(filp
);
796 ctx
= smp_load_acquire(&inode
->i_flctx
);
797 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
)) {
798 fl
->fl_type
= F_UNLCK
;
802 spin_lock(&ctx
->flc_lock
);
803 list_for_each_entry(cfl
, &ctx
->flc_posix
, fl_list
) {
804 if (posix_locks_conflict(fl
, cfl
)) {
805 locks_copy_conflock(fl
, cfl
);
807 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
811 fl
->fl_type
= F_UNLCK
;
813 spin_unlock(&ctx
->flc_lock
);
816 EXPORT_SYMBOL(posix_test_lock
);
819 * Deadlock detection:
821 * We attempt to detect deadlocks that are due purely to posix file
824 * We assume that a task can be waiting for at most one lock at a time.
825 * So for any acquired lock, the process holding that lock may be
826 * waiting on at most one other lock. That lock in turns may be held by
827 * someone waiting for at most one other lock. Given a requested lock
828 * caller_fl which is about to wait for a conflicting lock block_fl, we
829 * follow this chain of waiters to ensure we are not about to create a
832 * Since we do this before we ever put a process to sleep on a lock, we
833 * are ensured that there is never a cycle; that is what guarantees that
834 * the while() loop in posix_locks_deadlock() eventually completes.
836 * Note: the above assumption may not be true when handling lock
837 * requests from a broken NFS client. It may also fail in the presence
838 * of tasks (such as posix threads) sharing the same open file table.
839 * To handle those cases, we just bail out after a few iterations.
841 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
842 * Because the owner is not even nominally tied to a thread of
843 * execution, the deadlock detection below can't reasonably work well. Just
846 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
847 * locks that just checks for the case where two tasks are attempting to
848 * upgrade from read to write locks on the same inode.
851 #define MAX_DEADLK_ITERATIONS 10
853 /* Find a lock that the owner of the given block_fl is blocking on. */
854 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
856 struct file_lock
*fl
;
858 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
859 if (posix_same_owner(fl
, block_fl
))
865 /* Must be called with the blocked_lock_lock held! */
866 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
867 struct file_lock
*block_fl
)
871 lockdep_assert_held(&blocked_lock_lock
);
874 * This deadlock detector can't reasonably detect deadlocks with
875 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
877 if (IS_OFDLCK(caller_fl
))
880 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
881 if (i
++ > MAX_DEADLK_ITERATIONS
)
883 if (posix_same_owner(caller_fl
, block_fl
))
889 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
890 * after any leases, but before any posix locks.
892 * Note that if called with an FL_EXISTS argument, the caller may determine
893 * whether or not a lock was successfully freed by testing the return
896 static int flock_lock_inode(struct inode
*inode
, struct file_lock
*request
)
898 struct file_lock
*new_fl
= NULL
;
899 struct file_lock
*fl
;
900 struct file_lock_context
*ctx
;
905 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
907 if (request
->fl_type
!= F_UNLCK
)
909 return (request
->fl_flags
& FL_EXISTS
) ? -ENOENT
: 0;
912 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
913 new_fl
= locks_alloc_lock();
918 spin_lock(&ctx
->flc_lock
);
919 if (request
->fl_flags
& FL_ACCESS
)
922 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
923 if (request
->fl_file
!= fl
->fl_file
)
925 if (request
->fl_type
== fl
->fl_type
)
928 locks_delete_lock_ctx(fl
, &dispose
);
932 if (request
->fl_type
== F_UNLCK
) {
933 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
939 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
940 if (!flock_locks_conflict(request
, fl
))
943 if (!(request
->fl_flags
& FL_SLEEP
))
945 error
= FILE_LOCK_DEFERRED
;
946 locks_insert_block(fl
, request
);
949 if (request
->fl_flags
& FL_ACCESS
)
951 locks_copy_lock(new_fl
, request
);
952 locks_insert_lock_ctx(new_fl
, &ctx
->flc_flock
);
957 spin_unlock(&ctx
->flc_lock
);
959 locks_free_lock(new_fl
);
960 locks_dispose_list(&dispose
);
964 static int posix_lock_inode(struct inode
*inode
, struct file_lock
*request
,
965 struct file_lock
*conflock
)
967 struct file_lock
*fl
, *tmp
;
968 struct file_lock
*new_fl
= NULL
;
969 struct file_lock
*new_fl2
= NULL
;
970 struct file_lock
*left
= NULL
;
971 struct file_lock
*right
= NULL
;
972 struct file_lock_context
*ctx
;
977 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
979 return (request
->fl_type
== F_UNLCK
) ? 0 : -ENOMEM
;
982 * We may need two file_lock structures for this operation,
983 * so we get them in advance to avoid races.
985 * In some cases we can be sure, that no new locks will be needed
987 if (!(request
->fl_flags
& FL_ACCESS
) &&
988 (request
->fl_type
!= F_UNLCK
||
989 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
990 new_fl
= locks_alloc_lock();
991 new_fl2
= locks_alloc_lock();
994 spin_lock(&ctx
->flc_lock
);
996 * New lock request. Walk all POSIX locks and look for conflicts. If
997 * there are any, either return error or put the request on the
998 * blocker's list of waiters and the global blocked_hash.
1000 if (request
->fl_type
!= F_UNLCK
) {
1001 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1002 if (!posix_locks_conflict(request
, fl
))
1005 locks_copy_conflock(conflock
, fl
);
1007 if (!(request
->fl_flags
& FL_SLEEP
))
1010 * Deadlock detection and insertion into the blocked
1011 * locks list must be done while holding the same lock!
1014 spin_lock(&blocked_lock_lock
);
1015 if (likely(!posix_locks_deadlock(request
, fl
))) {
1016 error
= FILE_LOCK_DEFERRED
;
1017 __locks_insert_block(fl
, request
);
1019 spin_unlock(&blocked_lock_lock
);
1024 /* If we're just looking for a conflict, we're done. */
1026 if (request
->fl_flags
& FL_ACCESS
)
1029 /* Find the first old lock with the same owner as the new lock */
1030 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1031 if (posix_same_owner(request
, fl
))
1035 /* Process locks with this owner. */
1036 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1037 if (!posix_same_owner(request
, fl
))
1040 /* Detect adjacent or overlapping regions (if same lock type) */
1041 if (request
->fl_type
== fl
->fl_type
) {
1042 /* In all comparisons of start vs end, use
1043 * "start - 1" rather than "end + 1". If end
1044 * is OFFSET_MAX, end + 1 will become negative.
1046 if (fl
->fl_end
< request
->fl_start
- 1)
1048 /* If the next lock in the list has entirely bigger
1049 * addresses than the new one, insert the lock here.
1051 if (fl
->fl_start
- 1 > request
->fl_end
)
1054 /* If we come here, the new and old lock are of the
1055 * same type and adjacent or overlapping. Make one
1056 * lock yielding from the lower start address of both
1057 * locks to the higher end address.
1059 if (fl
->fl_start
> request
->fl_start
)
1060 fl
->fl_start
= request
->fl_start
;
1062 request
->fl_start
= fl
->fl_start
;
1063 if (fl
->fl_end
< request
->fl_end
)
1064 fl
->fl_end
= request
->fl_end
;
1066 request
->fl_end
= fl
->fl_end
;
1068 locks_delete_lock_ctx(fl
, &dispose
);
1074 /* Processing for different lock types is a bit
1077 if (fl
->fl_end
< request
->fl_start
)
1079 if (fl
->fl_start
> request
->fl_end
)
1081 if (request
->fl_type
== F_UNLCK
)
1083 if (fl
->fl_start
< request
->fl_start
)
1085 /* If the next lock in the list has a higher end
1086 * address than the new one, insert the new one here.
1088 if (fl
->fl_end
> request
->fl_end
) {
1092 if (fl
->fl_start
>= request
->fl_start
) {
1093 /* The new lock completely replaces an old
1094 * one (This may happen several times).
1097 locks_delete_lock_ctx(fl
, &dispose
);
1101 * Replace the old lock with new_fl, and
1102 * remove the old one. It's safe to do the
1103 * insert here since we know that we won't be
1104 * using new_fl later, and that the lock is
1105 * just replacing an existing lock.
1110 locks_copy_lock(new_fl
, request
);
1113 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1114 locks_delete_lock_ctx(fl
, &dispose
);
1121 * The above code only modifies existing locks in case of merging or
1122 * replacing. If new lock(s) need to be inserted all modifications are
1123 * done below this, so it's safe yet to bail out.
1125 error
= -ENOLCK
; /* "no luck" */
1126 if (right
&& left
== right
&& !new_fl2
)
1131 if (request
->fl_type
== F_UNLCK
) {
1132 if (request
->fl_flags
& FL_EXISTS
)
1141 locks_copy_lock(new_fl
, request
);
1142 locks_insert_lock_ctx(new_fl
, &fl
->fl_list
);
1147 if (left
== right
) {
1148 /* The new lock breaks the old one in two pieces,
1149 * so we have to use the second new lock.
1153 locks_copy_lock(left
, right
);
1154 locks_insert_lock_ctx(left
, &fl
->fl_list
);
1156 right
->fl_start
= request
->fl_end
+ 1;
1157 locks_wake_up_blocks(right
);
1160 left
->fl_end
= request
->fl_start
- 1;
1161 locks_wake_up_blocks(left
);
1164 spin_unlock(&ctx
->flc_lock
);
1166 * Free any unused locks.
1169 locks_free_lock(new_fl
);
1171 locks_free_lock(new_fl2
);
1172 locks_dispose_list(&dispose
);
1173 trace_posix_lock_inode(inode
, request
, error
);
1179 * posix_lock_file - Apply a POSIX-style lock to a file
1180 * @filp: The file to apply the lock to
1181 * @fl: The lock to be applied
1182 * @conflock: Place to return a copy of the conflicting lock, if found.
1184 * Add a POSIX style lock to a file.
1185 * We merge adjacent & overlapping locks whenever possible.
1186 * POSIX locks are sorted by owner task, then by starting address
1188 * Note that if called with an FL_EXISTS argument, the caller may determine
1189 * whether or not a lock was successfully freed by testing the return
1190 * value for -ENOENT.
1192 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1193 struct file_lock
*conflock
)
1195 return posix_lock_inode(file_inode(filp
), fl
, conflock
);
1197 EXPORT_SYMBOL(posix_lock_file
);
1200 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1201 * @inode: inode of file to which lock request should be applied
1202 * @fl: The lock to be applied
1204 * Apply a POSIX style lock request to an inode.
1206 static int posix_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1211 error
= posix_lock_inode(inode
, fl
, NULL
);
1212 if (error
!= FILE_LOCK_DEFERRED
)
1214 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1218 locks_delete_block(fl
);
1224 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1226 * locks_mandatory_locked - Check for an active lock
1227 * @file: the file to check
1229 * Searches the inode's list of locks to find any POSIX locks which conflict.
1230 * This function is called from locks_verify_locked() only.
1232 int locks_mandatory_locked(struct file
*file
)
1235 struct inode
*inode
= file_inode(file
);
1236 struct file_lock_context
*ctx
;
1237 struct file_lock
*fl
;
1239 ctx
= smp_load_acquire(&inode
->i_flctx
);
1240 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
))
1244 * Search the lock list for this inode for any POSIX locks.
1246 spin_lock(&ctx
->flc_lock
);
1248 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1249 if (fl
->fl_owner
!= current
->files
&&
1250 fl
->fl_owner
!= file
) {
1255 spin_unlock(&ctx
->flc_lock
);
1260 * locks_mandatory_area - Check for a conflicting lock
1261 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1263 * @inode: the file to check
1264 * @filp: how the file was opened (if it was)
1265 * @offset: start of area to check
1266 * @count: length of area to check
1268 * Searches the inode's list of locks to find any POSIX locks which conflict.
1269 * This function is called from rw_verify_area() and
1270 * locks_verify_truncate().
1272 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1273 struct file
*filp
, loff_t offset
,
1276 struct file_lock fl
;
1280 locks_init_lock(&fl
);
1281 fl
.fl_pid
= current
->tgid
;
1283 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1284 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1286 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1287 fl
.fl_start
= offset
;
1288 fl
.fl_end
= offset
+ count
- 1;
1293 fl
.fl_flags
&= ~FL_SLEEP
;
1294 error
= posix_lock_inode(inode
, &fl
, NULL
);
1300 fl
.fl_flags
|= FL_SLEEP
;
1301 fl
.fl_owner
= current
->files
;
1302 error
= posix_lock_inode(inode
, &fl
, NULL
);
1303 if (error
!= FILE_LOCK_DEFERRED
)
1305 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1308 * If we've been sleeping someone might have
1309 * changed the permissions behind our back.
1311 if (__mandatory_lock(inode
))
1315 locks_delete_block(&fl
);
1322 EXPORT_SYMBOL(locks_mandatory_area
);
1323 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1325 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1329 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1332 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1336 /* We already had a lease on this file; just change its type */
1337 int lease_modify(struct file_lock
*fl
, int arg
, struct list_head
*dispose
)
1339 int error
= assign_type(fl
, arg
);
1343 lease_clear_pending(fl
, arg
);
1344 locks_wake_up_blocks(fl
);
1345 if (arg
== F_UNLCK
) {
1346 struct file
*filp
= fl
->fl_file
;
1349 filp
->f_owner
.signum
= 0;
1350 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1351 if (fl
->fl_fasync
!= NULL
) {
1352 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1353 fl
->fl_fasync
= NULL
;
1355 locks_delete_lock_ctx(fl
, dispose
);
1359 EXPORT_SYMBOL(lease_modify
);
1361 static bool past_time(unsigned long then
)
1364 /* 0 is a special value meaning "this never expires": */
1366 return time_after(jiffies
, then
);
1369 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1371 struct file_lock_context
*ctx
= inode
->i_flctx
;
1372 struct file_lock
*fl
, *tmp
;
1374 lockdep_assert_held(&ctx
->flc_lock
);
1376 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1377 trace_time_out_leases(inode
, fl
);
1378 if (past_time(fl
->fl_downgrade_time
))
1379 lease_modify(fl
, F_RDLCK
, dispose
);
1380 if (past_time(fl
->fl_break_time
))
1381 lease_modify(fl
, F_UNLCK
, dispose
);
1385 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1387 if ((breaker
->fl_flags
& FL_LAYOUT
) != (lease
->fl_flags
& FL_LAYOUT
))
1389 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1391 return locks_conflict(breaker
, lease
);
1395 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1397 struct file_lock_context
*ctx
= inode
->i_flctx
;
1398 struct file_lock
*fl
;
1400 lockdep_assert_held(&ctx
->flc_lock
);
1402 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1403 if (leases_conflict(fl
, breaker
))
1410 * __break_lease - revoke all outstanding leases on file
1411 * @inode: the inode of the file to return
1412 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1414 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1417 * break_lease (inlined for speed) has checked there already is at least
1418 * some kind of lock (maybe a lease) on this file. Leases are broken on
1419 * a call to open() or truncate(). This function can sleep unless you
1420 * specified %O_NONBLOCK to your open().
1422 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1425 struct file_lock_context
*ctx
;
1426 struct file_lock
*new_fl
, *fl
, *tmp
;
1427 unsigned long break_time
;
1428 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1431 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1433 return PTR_ERR(new_fl
);
1434 new_fl
->fl_flags
= type
;
1436 /* typically we will check that ctx is non-NULL before calling */
1437 ctx
= smp_load_acquire(&inode
->i_flctx
);
1443 spin_lock(&ctx
->flc_lock
);
1445 time_out_leases(inode
, &dispose
);
1447 if (!any_leases_conflict(inode
, new_fl
))
1451 if (lease_break_time
> 0) {
1452 break_time
= jiffies
+ lease_break_time
* HZ
;
1453 if (break_time
== 0)
1454 break_time
++; /* so that 0 means no break time */
1457 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1458 if (!leases_conflict(fl
, new_fl
))
1461 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1463 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1464 fl
->fl_break_time
= break_time
;
1466 if (lease_breaking(fl
))
1468 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1469 fl
->fl_downgrade_time
= break_time
;
1471 if (fl
->fl_lmops
->lm_break(fl
))
1472 locks_delete_lock_ctx(fl
, &dispose
);
1475 if (list_empty(&ctx
->flc_lease
))
1478 if (mode
& O_NONBLOCK
) {
1479 trace_break_lease_noblock(inode
, new_fl
);
1480 error
= -EWOULDBLOCK
;
1485 fl
= list_first_entry(&ctx
->flc_lease
, struct file_lock
, fl_list
);
1486 break_time
= fl
->fl_break_time
;
1487 if (break_time
!= 0)
1488 break_time
-= jiffies
;
1489 if (break_time
== 0)
1491 locks_insert_block(fl
, new_fl
);
1492 trace_break_lease_block(inode
, new_fl
);
1493 spin_unlock(&ctx
->flc_lock
);
1494 locks_dispose_list(&dispose
);
1495 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1496 !new_fl
->fl_next
, break_time
);
1497 spin_lock(&ctx
->flc_lock
);
1498 trace_break_lease_unblock(inode
, new_fl
);
1499 locks_delete_block(new_fl
);
1502 * Wait for the next conflicting lease that has not been
1506 time_out_leases(inode
, &dispose
);
1507 if (any_leases_conflict(inode
, new_fl
))
1512 spin_unlock(&ctx
->flc_lock
);
1513 locks_dispose_list(&dispose
);
1514 locks_free_lock(new_fl
);
1518 EXPORT_SYMBOL(__break_lease
);
1521 * lease_get_mtime - get the last modified time of an inode
1523 * @time: pointer to a timespec which will contain the last modified time
1525 * This is to force NFS clients to flush their caches for files with
1526 * exclusive leases. The justification is that if someone has an
1527 * exclusive lease, then they could be modifying it.
1529 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1531 bool has_lease
= false;
1532 struct file_lock_context
*ctx
;
1533 struct file_lock
*fl
;
1535 ctx
= smp_load_acquire(&inode
->i_flctx
);
1536 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1537 spin_lock(&ctx
->flc_lock
);
1538 fl
= list_first_entry_or_null(&ctx
->flc_lease
,
1539 struct file_lock
, fl_list
);
1540 if (fl
&& (fl
->fl_type
== F_WRLCK
))
1542 spin_unlock(&ctx
->flc_lock
);
1546 *time
= current_fs_time(inode
->i_sb
);
1548 *time
= inode
->i_mtime
;
1551 EXPORT_SYMBOL(lease_get_mtime
);
1554 * fcntl_getlease - Enquire what lease is currently active
1557 * The value returned by this function will be one of
1558 * (if no lease break is pending):
1560 * %F_RDLCK to indicate a shared lease is held.
1562 * %F_WRLCK to indicate an exclusive lease is held.
1564 * %F_UNLCK to indicate no lease is held.
1566 * (if a lease break is pending):
1568 * %F_RDLCK to indicate an exclusive lease needs to be
1569 * changed to a shared lease (or removed).
1571 * %F_UNLCK to indicate the lease needs to be removed.
1573 * XXX: sfr & willy disagree over whether F_INPROGRESS
1574 * should be returned to userspace.
1576 int fcntl_getlease(struct file
*filp
)
1578 struct file_lock
*fl
;
1579 struct inode
*inode
= file_inode(filp
);
1580 struct file_lock_context
*ctx
;
1584 ctx
= smp_load_acquire(&inode
->i_flctx
);
1585 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1586 spin_lock(&ctx
->flc_lock
);
1587 time_out_leases(file_inode(filp
), &dispose
);
1588 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1589 if (fl
->fl_file
!= filp
)
1591 type
= target_leasetype(fl
);
1594 spin_unlock(&ctx
->flc_lock
);
1595 locks_dispose_list(&dispose
);
1601 * check_conflicting_open - see if the given dentry points to a file that has
1602 * an existing open that would conflict with the
1604 * @dentry: dentry to check
1605 * @arg: type of lease that we're trying to acquire
1606 * @flags: current lock flags
1608 * Check to see if there's an existing open fd on this file that would
1609 * conflict with the lease we're trying to set.
1612 check_conflicting_open(const struct dentry
*dentry
, const long arg
, int flags
)
1615 struct inode
*inode
= dentry
->d_inode
;
1617 if (flags
& FL_LAYOUT
)
1620 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1623 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1624 (atomic_read(&inode
->i_count
) > 1)))
1631 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1633 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1634 struct dentry
*dentry
= filp
->f_path
.dentry
;
1635 struct inode
*inode
= dentry
->d_inode
;
1636 struct file_lock_context
*ctx
;
1637 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1642 trace_generic_add_lease(inode
, lease
);
1644 /* Note that arg is never F_UNLCK here */
1645 ctx
= locks_get_lock_context(inode
, arg
);
1650 * In the delegation case we need mutual exclusion with
1651 * a number of operations that take the i_mutex. We trylock
1652 * because delegations are an optional optimization, and if
1653 * there's some chance of a conflict--we'd rather not
1654 * bother, maybe that's a sign this just isn't a good file to
1655 * hand out a delegation on.
1657 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1660 if (is_deleg
&& arg
== F_WRLCK
) {
1661 /* Write delegations are not currently supported: */
1662 mutex_unlock(&inode
->i_mutex
);
1667 spin_lock(&ctx
->flc_lock
);
1668 time_out_leases(inode
, &dispose
);
1669 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1674 * At this point, we know that if there is an exclusive
1675 * lease on this file, then we hold it on this filp
1676 * (otherwise our open of this file would have blocked).
1677 * And if we are trying to acquire an exclusive lease,
1678 * then the file is not open by anyone (including us)
1679 * except for this filp.
1682 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1683 if (fl
->fl_file
== filp
&&
1684 fl
->fl_owner
== lease
->fl_owner
) {
1690 * No exclusive leases if someone else has a lease on
1696 * Modifying our existing lease is OK, but no getting a
1697 * new lease if someone else is opening for write:
1699 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1703 if (my_fl
!= NULL
) {
1705 error
= lease
->fl_lmops
->lm_change(lease
, arg
, &dispose
);
1715 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1717 * The check in break_lease() is lockless. It's possible for another
1718 * open to race in after we did the earlier check for a conflicting
1719 * open but before the lease was inserted. Check again for a
1720 * conflicting open and cancel the lease if there is one.
1722 * We also add a barrier here to ensure that the insertion of the lock
1723 * precedes these checks.
1726 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1728 locks_unlink_lock_ctx(lease
);
1733 if (lease
->fl_lmops
->lm_setup
)
1734 lease
->fl_lmops
->lm_setup(lease
, priv
);
1736 spin_unlock(&ctx
->flc_lock
);
1737 locks_dispose_list(&dispose
);
1739 mutex_unlock(&inode
->i_mutex
);
1740 if (!error
&& !my_fl
)
1745 static int generic_delete_lease(struct file
*filp
, void *owner
)
1747 int error
= -EAGAIN
;
1748 struct file_lock
*fl
, *victim
= NULL
;
1749 struct inode
*inode
= file_inode(filp
);
1750 struct file_lock_context
*ctx
;
1753 ctx
= smp_load_acquire(&inode
->i_flctx
);
1755 trace_generic_delete_lease(inode
, NULL
);
1759 spin_lock(&ctx
->flc_lock
);
1760 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1761 if (fl
->fl_file
== filp
&&
1762 fl
->fl_owner
== owner
) {
1767 trace_generic_delete_lease(inode
, victim
);
1769 error
= fl
->fl_lmops
->lm_change(victim
, F_UNLCK
, &dispose
);
1770 spin_unlock(&ctx
->flc_lock
);
1771 locks_dispose_list(&dispose
);
1776 * generic_setlease - sets a lease on an open file
1777 * @filp: file pointer
1778 * @arg: type of lease to obtain
1779 * @flp: input - file_lock to use, output - file_lock inserted
1780 * @priv: private data for lm_setup (may be NULL if lm_setup
1781 * doesn't require it)
1783 * The (input) flp->fl_lmops->lm_break function is required
1786 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1789 struct inode
*inode
= file_inode(filp
);
1792 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1794 if (!S_ISREG(inode
->i_mode
))
1796 error
= security_file_lock(filp
, arg
);
1802 return generic_delete_lease(filp
, *priv
);
1805 if (!(*flp
)->fl_lmops
->lm_break
) {
1810 return generic_add_lease(filp
, arg
, flp
, priv
);
1815 EXPORT_SYMBOL(generic_setlease
);
1818 * vfs_setlease - sets a lease on an open file
1819 * @filp: file pointer
1820 * @arg: type of lease to obtain
1821 * @lease: file_lock to use when adding a lease
1822 * @priv: private info for lm_setup when adding a lease (may be
1823 * NULL if lm_setup doesn't require it)
1825 * Call this to establish a lease on the file. The "lease" argument is not
1826 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1827 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1828 * if not, this function will return -ENOLCK (and generate a scary-looking
1831 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1832 * may be NULL if the lm_setup operation doesn't require it.
1835 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1837 if (filp
->f_op
->setlease
)
1838 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1840 return generic_setlease(filp
, arg
, lease
, priv
);
1842 EXPORT_SYMBOL_GPL(vfs_setlease
);
1844 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1846 struct file_lock
*fl
;
1847 struct fasync_struct
*new;
1850 fl
= lease_alloc(filp
, arg
);
1854 new = fasync_alloc();
1856 locks_free_lock(fl
);
1861 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1863 locks_free_lock(fl
);
1870 * fcntl_setlease - sets a lease on an open file
1871 * @fd: open file descriptor
1872 * @filp: file pointer
1873 * @arg: type of lease to obtain
1875 * Call this fcntl to establish a lease on the file.
1876 * Note that you also need to call %F_SETSIG to
1877 * receive a signal when the lease is broken.
1879 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1882 return vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&filp
);
1883 return do_fcntl_add_lease(fd
, filp
, arg
);
1887 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1888 * @inode: inode of the file to apply to
1889 * @fl: The lock to be applied
1891 * Apply a FLOCK style lock request to an inode.
1893 static int flock_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1898 error
= flock_lock_inode(inode
, fl
);
1899 if (error
!= FILE_LOCK_DEFERRED
)
1901 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1905 locks_delete_block(fl
);
1912 * locks_lock_inode_wait - Apply a lock to an inode
1913 * @inode: inode of the file to apply to
1914 * @fl: The lock to be applied
1916 * Apply a POSIX or FLOCK style lock request to an inode.
1918 int locks_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1921 switch (fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)) {
1923 res
= posix_lock_inode_wait(inode
, fl
);
1926 res
= flock_lock_inode_wait(inode
, fl
);
1933 EXPORT_SYMBOL(locks_lock_inode_wait
);
1936 * sys_flock: - flock() system call.
1937 * @fd: the file descriptor to lock.
1938 * @cmd: the type of lock to apply.
1940 * Apply a %FL_FLOCK style lock to an open file descriptor.
1941 * The @cmd can be one of
1943 * %LOCK_SH -- a shared lock.
1945 * %LOCK_EX -- an exclusive lock.
1947 * %LOCK_UN -- remove an existing lock.
1949 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1951 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1952 * processes read and write access respectively.
1954 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1956 struct fd f
= fdget(fd
);
1957 struct file_lock
*lock
;
1958 int can_sleep
, unlock
;
1965 can_sleep
= !(cmd
& LOCK_NB
);
1967 unlock
= (cmd
== LOCK_UN
);
1969 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1970 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1973 lock
= flock_make_lock(f
.file
, cmd
);
1975 error
= PTR_ERR(lock
);
1980 lock
->fl_flags
|= FL_SLEEP
;
1982 error
= security_file_lock(f
.file
, lock
->fl_type
);
1986 if (f
.file
->f_op
->flock
)
1987 error
= f
.file
->f_op
->flock(f
.file
,
1988 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1991 error
= locks_lock_file_wait(f
.file
, lock
);
1994 locks_free_lock(lock
);
2003 * vfs_test_lock - test file byte range lock
2004 * @filp: The file to test lock for
2005 * @fl: The lock to test; also used to hold result
2007 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2008 * setting conf->fl_type to something other than F_UNLCK.
2010 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
2012 if (filp
->f_op
->lock
)
2013 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
2014 posix_test_lock(filp
, fl
);
2017 EXPORT_SYMBOL_GPL(vfs_test_lock
);
2019 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
2021 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2022 #if BITS_PER_LONG == 32
2024 * Make sure we can represent the posix lock via
2025 * legacy 32bit flock.
2027 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
2029 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
2032 flock
->l_start
= fl
->fl_start
;
2033 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2034 fl
->fl_end
- fl
->fl_start
+ 1;
2035 flock
->l_whence
= 0;
2036 flock
->l_type
= fl
->fl_type
;
2040 #if BITS_PER_LONG == 32
2041 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
2043 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2044 flock
->l_start
= fl
->fl_start
;
2045 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2046 fl
->fl_end
- fl
->fl_start
+ 1;
2047 flock
->l_whence
= 0;
2048 flock
->l_type
= fl
->fl_type
;
2052 /* Report the first existing lock that would conflict with l.
2053 * This implements the F_GETLK command of fcntl().
2055 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2057 struct file_lock file_lock
;
2062 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2065 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2068 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2072 if (cmd
== F_OFD_GETLK
) {
2074 if (flock
.l_pid
!= 0)
2078 file_lock
.fl_flags
|= FL_OFDLCK
;
2079 file_lock
.fl_owner
= filp
;
2082 error
= vfs_test_lock(filp
, &file_lock
);
2086 flock
.l_type
= file_lock
.fl_type
;
2087 if (file_lock
.fl_type
!= F_UNLCK
) {
2088 error
= posix_lock_to_flock(&flock
, &file_lock
);
2093 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2096 locks_release_private(&file_lock
);
2102 * vfs_lock_file - file byte range lock
2103 * @filp: The file to apply the lock to
2104 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2105 * @fl: The lock to be applied
2106 * @conf: Place to return a copy of the conflicting lock, if found.
2108 * A caller that doesn't care about the conflicting lock may pass NULL
2109 * as the final argument.
2111 * If the filesystem defines a private ->lock() method, then @conf will
2112 * be left unchanged; so a caller that cares should initialize it to
2113 * some acceptable default.
2115 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2116 * locks, the ->lock() interface may return asynchronously, before the lock has
2117 * been granted or denied by the underlying filesystem, if (and only if)
2118 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2119 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2120 * the request is for a blocking lock. When ->lock() does return asynchronously,
2121 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2122 * request completes.
2123 * If the request is for non-blocking lock the file system should return
2124 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2125 * with the result. If the request timed out the callback routine will return a
2126 * nonzero return code and the file system should release the lock. The file
2127 * system is also responsible to keep a corresponding posix lock when it
2128 * grants a lock so the VFS can find out which locks are locally held and do
2129 * the correct lock cleanup when required.
2130 * The underlying filesystem must not drop the kernel lock or call
2131 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2134 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2136 if (filp
->f_op
->lock
)
2137 return filp
->f_op
->lock(filp
, cmd
, fl
);
2139 return posix_lock_file(filp
, fl
, conf
);
2141 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2143 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2144 struct file_lock
*fl
)
2148 error
= security_file_lock(filp
, fl
->fl_type
);
2153 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2154 if (error
!= FILE_LOCK_DEFERRED
)
2156 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2160 locks_delete_block(fl
);
2167 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2169 check_fmode_for_setlk(struct file_lock
*fl
)
2171 switch (fl
->fl_type
) {
2173 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2177 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2183 /* Apply the lock described by l to an open file descriptor.
2184 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2186 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2187 struct flock __user
*l
)
2189 struct file_lock
*file_lock
= locks_alloc_lock();
2191 struct inode
*inode
;
2195 if (file_lock
== NULL
)
2198 inode
= file_inode(filp
);
2201 * This might block, so we do it before checking the inode.
2204 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2207 /* Don't allow mandatory locks on files that may be memory mapped
2210 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2215 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2219 error
= check_fmode_for_setlk(file_lock
);
2224 * If the cmd is requesting file-private locks, then set the
2225 * FL_OFDLCK flag and override the owner.
2230 if (flock
.l_pid
!= 0)
2234 file_lock
->fl_flags
|= FL_OFDLCK
;
2235 file_lock
->fl_owner
= filp
;
2239 if (flock
.l_pid
!= 0)
2243 file_lock
->fl_flags
|= FL_OFDLCK
;
2244 file_lock
->fl_owner
= filp
;
2247 file_lock
->fl_flags
|= FL_SLEEP
;
2250 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2253 * Attempt to detect a close/fcntl race and recover by releasing the
2254 * lock that was just acquired. There is no need to do that when we're
2255 * unlocking though, or for OFD locks.
2257 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2258 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2260 * We need that spin_lock here - it prevents reordering between
2261 * update of i_flctx->flc_posix and check for it done in
2262 * close(). rcu_read_lock() wouldn't do.
2264 spin_lock(¤t
->files
->file_lock
);
2266 spin_unlock(¤t
->files
->file_lock
);
2268 file_lock
->fl_type
= F_UNLCK
;
2269 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2270 WARN_ON_ONCE(error
);
2275 trace_fcntl_setlk(inode
, file_lock
, error
);
2276 locks_free_lock(file_lock
);
2280 #if BITS_PER_LONG == 32
2281 /* Report the first existing lock that would conflict with l.
2282 * This implements the F_GETLK command of fcntl().
2284 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2286 struct file_lock file_lock
;
2287 struct flock64 flock
;
2291 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2294 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2297 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2301 if (cmd
== F_OFD_GETLK
) {
2303 if (flock
.l_pid
!= 0)
2307 file_lock
.fl_flags
|= FL_OFDLCK
;
2308 file_lock
.fl_owner
= filp
;
2311 error
= vfs_test_lock(filp
, &file_lock
);
2315 flock
.l_type
= file_lock
.fl_type
;
2316 if (file_lock
.fl_type
!= F_UNLCK
)
2317 posix_lock_to_flock64(&flock
, &file_lock
);
2320 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2323 locks_release_private(&file_lock
);
2328 /* Apply the lock described by l to an open file descriptor.
2329 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2331 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2332 struct flock64 __user
*l
)
2334 struct file_lock
*file_lock
= locks_alloc_lock();
2335 struct flock64 flock
;
2336 struct inode
*inode
;
2340 if (file_lock
== NULL
)
2344 * This might block, so we do it before checking the inode.
2347 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2350 inode
= file_inode(filp
);
2352 /* Don't allow mandatory locks on files that may be memory mapped
2355 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2360 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2364 error
= check_fmode_for_setlk(file_lock
);
2369 * If the cmd is requesting file-private locks, then set the
2370 * FL_OFDLCK flag and override the owner.
2375 if (flock
.l_pid
!= 0)
2379 file_lock
->fl_flags
|= FL_OFDLCK
;
2380 file_lock
->fl_owner
= filp
;
2384 if (flock
.l_pid
!= 0)
2388 file_lock
->fl_flags
|= FL_OFDLCK
;
2389 file_lock
->fl_owner
= filp
;
2392 file_lock
->fl_flags
|= FL_SLEEP
;
2395 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2398 * Attempt to detect a close/fcntl race and recover by releasing the
2399 * lock that was just acquired. There is no need to do that when we're
2400 * unlocking though, or for OFD locks.
2402 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2403 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2405 * We need that spin_lock here - it prevents reordering between
2406 * update of i_flctx->flc_posix and check for it done in
2407 * close(). rcu_read_lock() wouldn't do.
2409 spin_lock(¤t
->files
->file_lock
);
2411 spin_unlock(¤t
->files
->file_lock
);
2413 file_lock
->fl_type
= F_UNLCK
;
2414 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2415 WARN_ON_ONCE(error
);
2420 locks_free_lock(file_lock
);
2423 #endif /* BITS_PER_LONG == 32 */
2426 * This function is called when the file is being removed
2427 * from the task's fd array. POSIX locks belonging to this task
2428 * are deleted at this time.
2430 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2433 struct file_lock lock
;
2434 struct file_lock_context
*ctx
;
2437 * If there are no locks held on this file, we don't need to call
2438 * posix_lock_file(). Another process could be setting a lock on this
2439 * file at the same time, but we wouldn't remove that lock anyway.
2441 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2442 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2445 lock
.fl_type
= F_UNLCK
;
2446 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2448 lock
.fl_end
= OFFSET_MAX
;
2449 lock
.fl_owner
= owner
;
2450 lock
.fl_pid
= current
->tgid
;
2451 lock
.fl_file
= filp
;
2453 lock
.fl_lmops
= NULL
;
2455 error
= vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2457 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2458 lock
.fl_ops
->fl_release_private(&lock
);
2459 trace_locks_remove_posix(file_inode(filp
), &lock
, error
);
2462 EXPORT_SYMBOL(locks_remove_posix
);
2464 /* The i_flctx must be valid when calling into here */
2466 locks_remove_flock(struct file
*filp
, struct file_lock_context
*flctx
)
2468 struct file_lock fl
= {
2470 .fl_pid
= current
->tgid
,
2472 .fl_flags
= FL_FLOCK
,
2474 .fl_end
= OFFSET_MAX
,
2476 struct inode
*inode
= file_inode(filp
);
2478 if (list_empty(&flctx
->flc_flock
))
2481 if (filp
->f_op
->flock
)
2482 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2484 flock_lock_inode(inode
, &fl
);
2486 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2487 fl
.fl_ops
->fl_release_private(&fl
);
2490 /* The i_flctx must be valid when calling into here */
2492 locks_remove_lease(struct file
*filp
, struct file_lock_context
*ctx
)
2494 struct file_lock
*fl
, *tmp
;
2497 if (list_empty(&ctx
->flc_lease
))
2500 spin_lock(&ctx
->flc_lock
);
2501 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2502 if (filp
== fl
->fl_file
)
2503 lease_modify(fl
, F_UNLCK
, &dispose
);
2504 spin_unlock(&ctx
->flc_lock
);
2505 locks_dispose_list(&dispose
);
2509 * This function is called on the last close of an open file.
2511 void locks_remove_file(struct file
*filp
)
2513 struct file_lock_context
*ctx
;
2515 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2519 /* remove any OFD locks */
2520 locks_remove_posix(filp
, filp
);
2522 /* remove flock locks */
2523 locks_remove_flock(filp
, ctx
);
2525 /* remove any leases */
2526 locks_remove_lease(filp
, ctx
);
2530 * posix_unblock_lock - stop waiting for a file lock
2531 * @waiter: the lock which was waiting
2533 * lockd needs to block waiting for locks.
2536 posix_unblock_lock(struct file_lock
*waiter
)
2540 spin_lock(&blocked_lock_lock
);
2541 if (waiter
->fl_next
)
2542 __locks_delete_block(waiter
);
2545 spin_unlock(&blocked_lock_lock
);
2548 EXPORT_SYMBOL(posix_unblock_lock
);
2551 * vfs_cancel_lock - file byte range unblock lock
2552 * @filp: The file to apply the unblock to
2553 * @fl: The lock to be unblocked
2555 * Used by lock managers to cancel blocked requests
2557 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2559 if (filp
->f_op
->lock
)
2560 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2564 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2566 #ifdef CONFIG_PROC_FS
2567 #include <linux/proc_fs.h>
2568 #include <linux/seq_file.h>
2570 struct locks_iterator
{
2575 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2576 loff_t id
, char *pfx
)
2578 struct inode
*inode
= NULL
;
2579 unsigned int fl_pid
;
2582 fl_pid
= pid_vnr(fl
->fl_nspid
);
2584 fl_pid
= fl
->fl_pid
;
2586 if (fl
->fl_file
!= NULL
)
2587 inode
= file_inode(fl
->fl_file
);
2589 seq_printf(f
, "%lld:%s ", id
, pfx
);
2591 if (fl
->fl_flags
& FL_ACCESS
)
2592 seq_puts(f
, "ACCESS");
2593 else if (IS_OFDLCK(fl
))
2594 seq_puts(f
, "OFDLCK");
2596 seq_puts(f
, "POSIX ");
2598 seq_printf(f
, " %s ",
2599 (inode
== NULL
) ? "*NOINODE*" :
2600 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2601 } else if (IS_FLOCK(fl
)) {
2602 if (fl
->fl_type
& LOCK_MAND
) {
2603 seq_puts(f
, "FLOCK MSNFS ");
2605 seq_puts(f
, "FLOCK ADVISORY ");
2607 } else if (IS_LEASE(fl
)) {
2608 if (fl
->fl_flags
& FL_DELEG
)
2609 seq_puts(f
, "DELEG ");
2611 seq_puts(f
, "LEASE ");
2613 if (lease_breaking(fl
))
2614 seq_puts(f
, "BREAKING ");
2615 else if (fl
->fl_file
)
2616 seq_puts(f
, "ACTIVE ");
2618 seq_puts(f
, "BREAKER ");
2620 seq_puts(f
, "UNKNOWN UNKNOWN ");
2622 if (fl
->fl_type
& LOCK_MAND
) {
2623 seq_printf(f
, "%s ",
2624 (fl
->fl_type
& LOCK_READ
)
2625 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2626 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2628 seq_printf(f
, "%s ",
2629 (lease_breaking(fl
))
2630 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2631 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2634 /* userspace relies on this representation of dev_t */
2635 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2636 MAJOR(inode
->i_sb
->s_dev
),
2637 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2639 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2642 if (fl
->fl_end
== OFFSET_MAX
)
2643 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2645 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2647 seq_puts(f
, "0 EOF\n");
2651 static int locks_show(struct seq_file
*f
, void *v
)
2653 struct locks_iterator
*iter
= f
->private;
2654 struct file_lock
*fl
, *bfl
;
2656 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2658 lock_get_status(f
, fl
, iter
->li_pos
, "");
2660 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2661 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2666 static void __show_fd_locks(struct seq_file
*f
,
2667 struct list_head
*head
, int *id
,
2668 struct file
*filp
, struct files_struct
*files
)
2670 struct file_lock
*fl
;
2672 list_for_each_entry(fl
, head
, fl_list
) {
2674 if (filp
!= fl
->fl_file
)
2676 if (fl
->fl_owner
!= files
&&
2677 fl
->fl_owner
!= filp
)
2681 seq_puts(f
, "lock:\t");
2682 lock_get_status(f
, fl
, *id
, "");
2686 void show_fd_locks(struct seq_file
*f
,
2687 struct file
*filp
, struct files_struct
*files
)
2689 struct inode
*inode
= file_inode(filp
);
2690 struct file_lock_context
*ctx
;
2693 ctx
= smp_load_acquire(&inode
->i_flctx
);
2697 spin_lock(&ctx
->flc_lock
);
2698 __show_fd_locks(f
, &ctx
->flc_flock
, &id
, filp
, files
);
2699 __show_fd_locks(f
, &ctx
->flc_posix
, &id
, filp
, files
);
2700 __show_fd_locks(f
, &ctx
->flc_lease
, &id
, filp
, files
);
2701 spin_unlock(&ctx
->flc_lock
);
2704 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2705 __acquires(&blocked_lock_lock
)
2707 struct locks_iterator
*iter
= f
->private;
2709 iter
->li_pos
= *pos
+ 1;
2710 lg_global_lock(&file_lock_lglock
);
2711 spin_lock(&blocked_lock_lock
);
2712 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2715 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2717 struct locks_iterator
*iter
= f
->private;
2720 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2723 static void locks_stop(struct seq_file
*f
, void *v
)
2724 __releases(&blocked_lock_lock
)
2726 spin_unlock(&blocked_lock_lock
);
2727 lg_global_unlock(&file_lock_lglock
);
2730 static const struct seq_operations locks_seq_operations
= {
2731 .start
= locks_start
,
2737 static int locks_open(struct inode
*inode
, struct file
*filp
)
2739 return seq_open_private(filp
, &locks_seq_operations
,
2740 sizeof(struct locks_iterator
));
2743 static const struct file_operations proc_locks_operations
= {
2746 .llseek
= seq_lseek
,
2747 .release
= seq_release_private
,
2750 static int __init
proc_locks_init(void)
2752 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2755 fs_initcall(proc_locks_init
);
2758 static int __init
filelock_init(void)
2762 flctx_cache
= kmem_cache_create("file_lock_ctx",
2763 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2765 filelock_cache
= kmem_cache_create("file_lock_cache",
2766 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2768 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2770 for_each_possible_cpu(i
)
2771 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2776 core_initcall(filelock_init
);