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_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
966 struct file_lock
*fl
, *tmp
;
967 struct file_lock
*new_fl
= NULL
;
968 struct file_lock
*new_fl2
= NULL
;
969 struct file_lock
*left
= NULL
;
970 struct file_lock
*right
= NULL
;
971 struct file_lock_context
*ctx
;
976 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
978 return (request
->fl_type
== F_UNLCK
) ? 0 : -ENOMEM
;
981 * We may need two file_lock structures for this operation,
982 * so we get them in advance to avoid races.
984 * In some cases we can be sure, that no new locks will be needed
986 if (!(request
->fl_flags
& FL_ACCESS
) &&
987 (request
->fl_type
!= F_UNLCK
||
988 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
989 new_fl
= locks_alloc_lock();
990 new_fl2
= locks_alloc_lock();
993 spin_lock(&ctx
->flc_lock
);
995 * New lock request. Walk all POSIX locks and look for conflicts. If
996 * there are any, either return error or put the request on the
997 * blocker's list of waiters and the global blocked_hash.
999 if (request
->fl_type
!= F_UNLCK
) {
1000 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1001 if (!posix_locks_conflict(request
, fl
))
1004 locks_copy_conflock(conflock
, fl
);
1006 if (!(request
->fl_flags
& FL_SLEEP
))
1009 * Deadlock detection and insertion into the blocked
1010 * locks list must be done while holding the same lock!
1013 spin_lock(&blocked_lock_lock
);
1014 if (likely(!posix_locks_deadlock(request
, fl
))) {
1015 error
= FILE_LOCK_DEFERRED
;
1016 __locks_insert_block(fl
, request
);
1018 spin_unlock(&blocked_lock_lock
);
1023 /* If we're just looking for a conflict, we're done. */
1025 if (request
->fl_flags
& FL_ACCESS
)
1028 /* Find the first old lock with the same owner as the new lock */
1029 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1030 if (posix_same_owner(request
, fl
))
1034 /* Process locks with this owner. */
1035 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1036 if (!posix_same_owner(request
, fl
))
1039 /* Detect adjacent or overlapping regions (if same lock type) */
1040 if (request
->fl_type
== fl
->fl_type
) {
1041 /* In all comparisons of start vs end, use
1042 * "start - 1" rather than "end + 1". If end
1043 * is OFFSET_MAX, end + 1 will become negative.
1045 if (fl
->fl_end
< request
->fl_start
- 1)
1047 /* If the next lock in the list has entirely bigger
1048 * addresses than the new one, insert the lock here.
1050 if (fl
->fl_start
- 1 > request
->fl_end
)
1053 /* If we come here, the new and old lock are of the
1054 * same type and adjacent or overlapping. Make one
1055 * lock yielding from the lower start address of both
1056 * locks to the higher end address.
1058 if (fl
->fl_start
> request
->fl_start
)
1059 fl
->fl_start
= request
->fl_start
;
1061 request
->fl_start
= fl
->fl_start
;
1062 if (fl
->fl_end
< request
->fl_end
)
1063 fl
->fl_end
= request
->fl_end
;
1065 request
->fl_end
= fl
->fl_end
;
1067 locks_delete_lock_ctx(fl
, &dispose
);
1073 /* Processing for different lock types is a bit
1076 if (fl
->fl_end
< request
->fl_start
)
1078 if (fl
->fl_start
> request
->fl_end
)
1080 if (request
->fl_type
== F_UNLCK
)
1082 if (fl
->fl_start
< request
->fl_start
)
1084 /* If the next lock in the list has a higher end
1085 * address than the new one, insert the new one here.
1087 if (fl
->fl_end
> request
->fl_end
) {
1091 if (fl
->fl_start
>= request
->fl_start
) {
1092 /* The new lock completely replaces an old
1093 * one (This may happen several times).
1096 locks_delete_lock_ctx(fl
, &dispose
);
1100 * Replace the old lock with new_fl, and
1101 * remove the old one. It's safe to do the
1102 * insert here since we know that we won't be
1103 * using new_fl later, and that the lock is
1104 * just replacing an existing lock.
1109 locks_copy_lock(new_fl
, request
);
1112 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1113 locks_delete_lock_ctx(fl
, &dispose
);
1120 * The above code only modifies existing locks in case of merging or
1121 * replacing. If new lock(s) need to be inserted all modifications are
1122 * done below this, so it's safe yet to bail out.
1124 error
= -ENOLCK
; /* "no luck" */
1125 if (right
&& left
== right
&& !new_fl2
)
1130 if (request
->fl_type
== F_UNLCK
) {
1131 if (request
->fl_flags
& FL_EXISTS
)
1140 locks_copy_lock(new_fl
, request
);
1141 locks_insert_lock_ctx(new_fl
, &fl
->fl_list
);
1146 if (left
== right
) {
1147 /* The new lock breaks the old one in two pieces,
1148 * so we have to use the second new lock.
1152 locks_copy_lock(left
, right
);
1153 locks_insert_lock_ctx(left
, &fl
->fl_list
);
1155 right
->fl_start
= request
->fl_end
+ 1;
1156 locks_wake_up_blocks(right
);
1159 left
->fl_end
= request
->fl_start
- 1;
1160 locks_wake_up_blocks(left
);
1163 spin_unlock(&ctx
->flc_lock
);
1165 * Free any unused locks.
1168 locks_free_lock(new_fl
);
1170 locks_free_lock(new_fl2
);
1171 locks_dispose_list(&dispose
);
1172 trace_posix_lock_inode(inode
, request
, error
);
1178 * posix_lock_file - Apply a POSIX-style lock to a file
1179 * @filp: The file to apply the lock to
1180 * @fl: The lock to be applied
1181 * @conflock: Place to return a copy of the conflicting lock, if found.
1183 * Add a POSIX style lock to a file.
1184 * We merge adjacent & overlapping locks whenever possible.
1185 * POSIX locks are sorted by owner task, then by starting address
1187 * Note that if called with an FL_EXISTS argument, the caller may determine
1188 * whether or not a lock was successfully freed by testing the return
1189 * value for -ENOENT.
1191 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1192 struct file_lock
*conflock
)
1194 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1196 EXPORT_SYMBOL(posix_lock_file
);
1199 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1200 * @inode: inode of file to which lock request should be applied
1201 * @fl: The lock to be applied
1203 * Apply a POSIX style lock request to an inode.
1205 static int posix_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1210 error
= __posix_lock_file(inode
, fl
, NULL
);
1211 if (error
!= FILE_LOCK_DEFERRED
)
1213 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1217 locks_delete_block(fl
);
1223 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1225 * locks_mandatory_locked - Check for an active lock
1226 * @file: the file to check
1228 * Searches the inode's list of locks to find any POSIX locks which conflict.
1229 * This function is called from locks_verify_locked() only.
1231 int locks_mandatory_locked(struct file
*file
)
1234 struct inode
*inode
= file_inode(file
);
1235 struct file_lock_context
*ctx
;
1236 struct file_lock
*fl
;
1238 ctx
= smp_load_acquire(&inode
->i_flctx
);
1239 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
))
1243 * Search the lock list for this inode for any POSIX locks.
1245 spin_lock(&ctx
->flc_lock
);
1247 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1248 if (fl
->fl_owner
!= current
->files
&&
1249 fl
->fl_owner
!= file
) {
1254 spin_unlock(&ctx
->flc_lock
);
1259 * locks_mandatory_area - Check for a conflicting lock
1260 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1262 * @inode: the file to check
1263 * @filp: how the file was opened (if it was)
1264 * @offset: start of area to check
1265 * @count: length of area to check
1267 * Searches the inode's list of locks to find any POSIX locks which conflict.
1268 * This function is called from rw_verify_area() and
1269 * locks_verify_truncate().
1271 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1272 struct file
*filp
, loff_t offset
,
1275 struct file_lock fl
;
1279 locks_init_lock(&fl
);
1280 fl
.fl_pid
= current
->tgid
;
1282 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1283 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1285 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1286 fl
.fl_start
= offset
;
1287 fl
.fl_end
= offset
+ count
- 1;
1292 fl
.fl_flags
&= ~FL_SLEEP
;
1293 error
= __posix_lock_file(inode
, &fl
, NULL
);
1299 fl
.fl_flags
|= FL_SLEEP
;
1300 fl
.fl_owner
= current
->files
;
1301 error
= __posix_lock_file(inode
, &fl
, NULL
);
1302 if (error
!= FILE_LOCK_DEFERRED
)
1304 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1307 * If we've been sleeping someone might have
1308 * changed the permissions behind our back.
1310 if (__mandatory_lock(inode
))
1314 locks_delete_block(&fl
);
1321 EXPORT_SYMBOL(locks_mandatory_area
);
1322 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1324 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1328 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1331 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1335 /* We already had a lease on this file; just change its type */
1336 int lease_modify(struct file_lock
*fl
, int arg
, struct list_head
*dispose
)
1338 int error
= assign_type(fl
, arg
);
1342 lease_clear_pending(fl
, arg
);
1343 locks_wake_up_blocks(fl
);
1344 if (arg
== F_UNLCK
) {
1345 struct file
*filp
= fl
->fl_file
;
1348 filp
->f_owner
.signum
= 0;
1349 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1350 if (fl
->fl_fasync
!= NULL
) {
1351 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1352 fl
->fl_fasync
= NULL
;
1354 locks_delete_lock_ctx(fl
, dispose
);
1358 EXPORT_SYMBOL(lease_modify
);
1360 static bool past_time(unsigned long then
)
1363 /* 0 is a special value meaning "this never expires": */
1365 return time_after(jiffies
, then
);
1368 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1370 struct file_lock_context
*ctx
= inode
->i_flctx
;
1371 struct file_lock
*fl
, *tmp
;
1373 lockdep_assert_held(&ctx
->flc_lock
);
1375 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1376 trace_time_out_leases(inode
, fl
);
1377 if (past_time(fl
->fl_downgrade_time
))
1378 lease_modify(fl
, F_RDLCK
, dispose
);
1379 if (past_time(fl
->fl_break_time
))
1380 lease_modify(fl
, F_UNLCK
, dispose
);
1384 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1386 if ((breaker
->fl_flags
& FL_LAYOUT
) != (lease
->fl_flags
& FL_LAYOUT
))
1388 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1390 return locks_conflict(breaker
, lease
);
1394 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1396 struct file_lock_context
*ctx
= inode
->i_flctx
;
1397 struct file_lock
*fl
;
1399 lockdep_assert_held(&ctx
->flc_lock
);
1401 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1402 if (leases_conflict(fl
, breaker
))
1409 * __break_lease - revoke all outstanding leases on file
1410 * @inode: the inode of the file to return
1411 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1413 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1416 * break_lease (inlined for speed) has checked there already is at least
1417 * some kind of lock (maybe a lease) on this file. Leases are broken on
1418 * a call to open() or truncate(). This function can sleep unless you
1419 * specified %O_NONBLOCK to your open().
1421 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1424 struct file_lock_context
*ctx
;
1425 struct file_lock
*new_fl
, *fl
, *tmp
;
1426 unsigned long break_time
;
1427 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1430 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1432 return PTR_ERR(new_fl
);
1433 new_fl
->fl_flags
= type
;
1435 /* typically we will check that ctx is non-NULL before calling */
1436 ctx
= smp_load_acquire(&inode
->i_flctx
);
1442 spin_lock(&ctx
->flc_lock
);
1444 time_out_leases(inode
, &dispose
);
1446 if (!any_leases_conflict(inode
, new_fl
))
1450 if (lease_break_time
> 0) {
1451 break_time
= jiffies
+ lease_break_time
* HZ
;
1452 if (break_time
== 0)
1453 break_time
++; /* so that 0 means no break time */
1456 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1457 if (!leases_conflict(fl
, new_fl
))
1460 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1462 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1463 fl
->fl_break_time
= break_time
;
1465 if (lease_breaking(fl
))
1467 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1468 fl
->fl_downgrade_time
= break_time
;
1470 if (fl
->fl_lmops
->lm_break(fl
))
1471 locks_delete_lock_ctx(fl
, &dispose
);
1474 if (list_empty(&ctx
->flc_lease
))
1477 if (mode
& O_NONBLOCK
) {
1478 trace_break_lease_noblock(inode
, new_fl
);
1479 error
= -EWOULDBLOCK
;
1484 fl
= list_first_entry(&ctx
->flc_lease
, struct file_lock
, fl_list
);
1485 break_time
= fl
->fl_break_time
;
1486 if (break_time
!= 0)
1487 break_time
-= jiffies
;
1488 if (break_time
== 0)
1490 locks_insert_block(fl
, new_fl
);
1491 trace_break_lease_block(inode
, new_fl
);
1492 spin_unlock(&ctx
->flc_lock
);
1493 locks_dispose_list(&dispose
);
1494 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1495 !new_fl
->fl_next
, break_time
);
1496 spin_lock(&ctx
->flc_lock
);
1497 trace_break_lease_unblock(inode
, new_fl
);
1498 locks_delete_block(new_fl
);
1501 * Wait for the next conflicting lease that has not been
1505 time_out_leases(inode
, &dispose
);
1506 if (any_leases_conflict(inode
, new_fl
))
1511 spin_unlock(&ctx
->flc_lock
);
1512 locks_dispose_list(&dispose
);
1513 locks_free_lock(new_fl
);
1517 EXPORT_SYMBOL(__break_lease
);
1520 * lease_get_mtime - get the last modified time of an inode
1522 * @time: pointer to a timespec which will contain the last modified time
1524 * This is to force NFS clients to flush their caches for files with
1525 * exclusive leases. The justification is that if someone has an
1526 * exclusive lease, then they could be modifying it.
1528 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1530 bool has_lease
= false;
1531 struct file_lock_context
*ctx
;
1532 struct file_lock
*fl
;
1534 ctx
= smp_load_acquire(&inode
->i_flctx
);
1535 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1536 spin_lock(&ctx
->flc_lock
);
1537 fl
= list_first_entry_or_null(&ctx
->flc_lease
,
1538 struct file_lock
, fl_list
);
1539 if (fl
&& (fl
->fl_type
== F_WRLCK
))
1541 spin_unlock(&ctx
->flc_lock
);
1545 *time
= current_fs_time(inode
->i_sb
);
1547 *time
= inode
->i_mtime
;
1550 EXPORT_SYMBOL(lease_get_mtime
);
1553 * fcntl_getlease - Enquire what lease is currently active
1556 * The value returned by this function will be one of
1557 * (if no lease break is pending):
1559 * %F_RDLCK to indicate a shared lease is held.
1561 * %F_WRLCK to indicate an exclusive lease is held.
1563 * %F_UNLCK to indicate no lease is held.
1565 * (if a lease break is pending):
1567 * %F_RDLCK to indicate an exclusive lease needs to be
1568 * changed to a shared lease (or removed).
1570 * %F_UNLCK to indicate the lease needs to be removed.
1572 * XXX: sfr & willy disagree over whether F_INPROGRESS
1573 * should be returned to userspace.
1575 int fcntl_getlease(struct file
*filp
)
1577 struct file_lock
*fl
;
1578 struct inode
*inode
= file_inode(filp
);
1579 struct file_lock_context
*ctx
;
1583 ctx
= smp_load_acquire(&inode
->i_flctx
);
1584 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1585 spin_lock(&ctx
->flc_lock
);
1586 time_out_leases(file_inode(filp
), &dispose
);
1587 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1588 if (fl
->fl_file
!= filp
)
1590 type
= target_leasetype(fl
);
1593 spin_unlock(&ctx
->flc_lock
);
1594 locks_dispose_list(&dispose
);
1600 * check_conflicting_open - see if the given dentry points to a file that has
1601 * an existing open that would conflict with the
1603 * @dentry: dentry to check
1604 * @arg: type of lease that we're trying to acquire
1605 * @flags: current lock flags
1607 * Check to see if there's an existing open fd on this file that would
1608 * conflict with the lease we're trying to set.
1611 check_conflicting_open(const struct dentry
*dentry
, const long arg
, int flags
)
1614 struct inode
*inode
= dentry
->d_inode
;
1616 if (flags
& FL_LAYOUT
)
1619 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1622 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1623 (atomic_read(&inode
->i_count
) > 1)))
1630 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1632 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1633 struct dentry
*dentry
= filp
->f_path
.dentry
;
1634 struct inode
*inode
= dentry
->d_inode
;
1635 struct file_lock_context
*ctx
;
1636 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1641 trace_generic_add_lease(inode
, lease
);
1643 /* Note that arg is never F_UNLCK here */
1644 ctx
= locks_get_lock_context(inode
, arg
);
1649 * In the delegation case we need mutual exclusion with
1650 * a number of operations that take the i_mutex. We trylock
1651 * because delegations are an optional optimization, and if
1652 * there's some chance of a conflict--we'd rather not
1653 * bother, maybe that's a sign this just isn't a good file to
1654 * hand out a delegation on.
1656 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1659 if (is_deleg
&& arg
== F_WRLCK
) {
1660 /* Write delegations are not currently supported: */
1661 mutex_unlock(&inode
->i_mutex
);
1666 spin_lock(&ctx
->flc_lock
);
1667 time_out_leases(inode
, &dispose
);
1668 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1673 * At this point, we know that if there is an exclusive
1674 * lease on this file, then we hold it on this filp
1675 * (otherwise our open of this file would have blocked).
1676 * And if we are trying to acquire an exclusive lease,
1677 * then the file is not open by anyone (including us)
1678 * except for this filp.
1681 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1682 if (fl
->fl_file
== filp
&&
1683 fl
->fl_owner
== lease
->fl_owner
) {
1689 * No exclusive leases if someone else has a lease on
1695 * Modifying our existing lease is OK, but no getting a
1696 * new lease if someone else is opening for write:
1698 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1702 if (my_fl
!= NULL
) {
1704 error
= lease
->fl_lmops
->lm_change(lease
, arg
, &dispose
);
1714 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1716 * The check in break_lease() is lockless. It's possible for another
1717 * open to race in after we did the earlier check for a conflicting
1718 * open but before the lease was inserted. Check again for a
1719 * conflicting open and cancel the lease if there is one.
1721 * We also add a barrier here to ensure that the insertion of the lock
1722 * precedes these checks.
1725 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1727 locks_unlink_lock_ctx(lease
);
1732 if (lease
->fl_lmops
->lm_setup
)
1733 lease
->fl_lmops
->lm_setup(lease
, priv
);
1735 spin_unlock(&ctx
->flc_lock
);
1736 locks_dispose_list(&dispose
);
1738 mutex_unlock(&inode
->i_mutex
);
1739 if (!error
&& !my_fl
)
1744 static int generic_delete_lease(struct file
*filp
, void *owner
)
1746 int error
= -EAGAIN
;
1747 struct file_lock
*fl
, *victim
= NULL
;
1748 struct inode
*inode
= file_inode(filp
);
1749 struct file_lock_context
*ctx
;
1752 ctx
= smp_load_acquire(&inode
->i_flctx
);
1754 trace_generic_delete_lease(inode
, NULL
);
1758 spin_lock(&ctx
->flc_lock
);
1759 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1760 if (fl
->fl_file
== filp
&&
1761 fl
->fl_owner
== owner
) {
1766 trace_generic_delete_lease(inode
, victim
);
1768 error
= fl
->fl_lmops
->lm_change(victim
, F_UNLCK
, &dispose
);
1769 spin_unlock(&ctx
->flc_lock
);
1770 locks_dispose_list(&dispose
);
1775 * generic_setlease - sets a lease on an open file
1776 * @filp: file pointer
1777 * @arg: type of lease to obtain
1778 * @flp: input - file_lock to use, output - file_lock inserted
1779 * @priv: private data for lm_setup (may be NULL if lm_setup
1780 * doesn't require it)
1782 * The (input) flp->fl_lmops->lm_break function is required
1785 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1788 struct inode
*inode
= file_inode(filp
);
1791 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1793 if (!S_ISREG(inode
->i_mode
))
1795 error
= security_file_lock(filp
, arg
);
1801 return generic_delete_lease(filp
, *priv
);
1804 if (!(*flp
)->fl_lmops
->lm_break
) {
1809 return generic_add_lease(filp
, arg
, flp
, priv
);
1814 EXPORT_SYMBOL(generic_setlease
);
1817 * vfs_setlease - sets a lease on an open file
1818 * @filp: file pointer
1819 * @arg: type of lease to obtain
1820 * @lease: file_lock to use when adding a lease
1821 * @priv: private info for lm_setup when adding a lease (may be
1822 * NULL if lm_setup doesn't require it)
1824 * Call this to establish a lease on the file. The "lease" argument is not
1825 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1826 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1827 * if not, this function will return -ENOLCK (and generate a scary-looking
1830 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1831 * may be NULL if the lm_setup operation doesn't require it.
1834 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1836 if (filp
->f_op
->setlease
)
1837 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1839 return generic_setlease(filp
, arg
, lease
, priv
);
1841 EXPORT_SYMBOL_GPL(vfs_setlease
);
1843 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1845 struct file_lock
*fl
;
1846 struct fasync_struct
*new;
1849 fl
= lease_alloc(filp
, arg
);
1853 new = fasync_alloc();
1855 locks_free_lock(fl
);
1860 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1862 locks_free_lock(fl
);
1869 * fcntl_setlease - sets a lease on an open file
1870 * @fd: open file descriptor
1871 * @filp: file pointer
1872 * @arg: type of lease to obtain
1874 * Call this fcntl to establish a lease on the file.
1875 * Note that you also need to call %F_SETSIG to
1876 * receive a signal when the lease is broken.
1878 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1881 return vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&filp
);
1882 return do_fcntl_add_lease(fd
, filp
, arg
);
1886 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1887 * @inode: inode of the file to apply to
1888 * @fl: The lock to be applied
1890 * Apply a FLOCK style lock request to an inode.
1892 static int flock_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1897 error
= flock_lock_inode(inode
, fl
);
1898 if (error
!= FILE_LOCK_DEFERRED
)
1900 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1904 locks_delete_block(fl
);
1911 * locks_lock_inode_wait - Apply a lock to an inode
1912 * @inode: inode of the file to apply to
1913 * @fl: The lock to be applied
1915 * Apply a POSIX or FLOCK style lock request to an inode.
1917 int locks_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1920 switch (fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)) {
1922 res
= posix_lock_inode_wait(inode
, fl
);
1925 res
= flock_lock_inode_wait(inode
, fl
);
1932 EXPORT_SYMBOL(locks_lock_inode_wait
);
1935 * sys_flock: - flock() system call.
1936 * @fd: the file descriptor to lock.
1937 * @cmd: the type of lock to apply.
1939 * Apply a %FL_FLOCK style lock to an open file descriptor.
1940 * The @cmd can be one of
1942 * %LOCK_SH -- a shared lock.
1944 * %LOCK_EX -- an exclusive lock.
1946 * %LOCK_UN -- remove an existing lock.
1948 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1950 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1951 * processes read and write access respectively.
1953 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1955 struct fd f
= fdget(fd
);
1956 struct file_lock
*lock
;
1957 int can_sleep
, unlock
;
1964 can_sleep
= !(cmd
& LOCK_NB
);
1966 unlock
= (cmd
== LOCK_UN
);
1968 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1969 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1972 lock
= flock_make_lock(f
.file
, cmd
);
1974 error
= PTR_ERR(lock
);
1979 lock
->fl_flags
|= FL_SLEEP
;
1981 error
= security_file_lock(f
.file
, lock
->fl_type
);
1985 if (f
.file
->f_op
->flock
)
1986 error
= f
.file
->f_op
->flock(f
.file
,
1987 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1990 error
= locks_lock_file_wait(f
.file
, lock
);
1993 locks_free_lock(lock
);
2002 * vfs_test_lock - test file byte range lock
2003 * @filp: The file to test lock for
2004 * @fl: The lock to test; also used to hold result
2006 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2007 * setting conf->fl_type to something other than F_UNLCK.
2009 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
2011 if (filp
->f_op
->lock
)
2012 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
2013 posix_test_lock(filp
, fl
);
2016 EXPORT_SYMBOL_GPL(vfs_test_lock
);
2018 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
2020 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2021 #if BITS_PER_LONG == 32
2023 * Make sure we can represent the posix lock via
2024 * legacy 32bit flock.
2026 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
2028 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
2031 flock
->l_start
= fl
->fl_start
;
2032 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2033 fl
->fl_end
- fl
->fl_start
+ 1;
2034 flock
->l_whence
= 0;
2035 flock
->l_type
= fl
->fl_type
;
2039 #if BITS_PER_LONG == 32
2040 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
2042 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2043 flock
->l_start
= fl
->fl_start
;
2044 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2045 fl
->fl_end
- fl
->fl_start
+ 1;
2046 flock
->l_whence
= 0;
2047 flock
->l_type
= fl
->fl_type
;
2051 /* Report the first existing lock that would conflict with l.
2052 * This implements the F_GETLK command of fcntl().
2054 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2056 struct file_lock file_lock
;
2061 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2064 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2067 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2071 if (cmd
== F_OFD_GETLK
) {
2073 if (flock
.l_pid
!= 0)
2077 file_lock
.fl_flags
|= FL_OFDLCK
;
2078 file_lock
.fl_owner
= filp
;
2081 error
= vfs_test_lock(filp
, &file_lock
);
2085 flock
.l_type
= file_lock
.fl_type
;
2086 if (file_lock
.fl_type
!= F_UNLCK
) {
2087 error
= posix_lock_to_flock(&flock
, &file_lock
);
2092 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2095 locks_release_private(&file_lock
);
2101 * vfs_lock_file - file byte range lock
2102 * @filp: The file to apply the lock to
2103 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2104 * @fl: The lock to be applied
2105 * @conf: Place to return a copy of the conflicting lock, if found.
2107 * A caller that doesn't care about the conflicting lock may pass NULL
2108 * as the final argument.
2110 * If the filesystem defines a private ->lock() method, then @conf will
2111 * be left unchanged; so a caller that cares should initialize it to
2112 * some acceptable default.
2114 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2115 * locks, the ->lock() interface may return asynchronously, before the lock has
2116 * been granted or denied by the underlying filesystem, if (and only if)
2117 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2118 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2119 * the request is for a blocking lock. When ->lock() does return asynchronously,
2120 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2121 * request completes.
2122 * If the request is for non-blocking lock the file system should return
2123 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2124 * with the result. If the request timed out the callback routine will return a
2125 * nonzero return code and the file system should release the lock. The file
2126 * system is also responsible to keep a corresponding posix lock when it
2127 * grants a lock so the VFS can find out which locks are locally held and do
2128 * the correct lock cleanup when required.
2129 * The underlying filesystem must not drop the kernel lock or call
2130 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2133 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2135 if (filp
->f_op
->lock
)
2136 return filp
->f_op
->lock(filp
, cmd
, fl
);
2138 return posix_lock_file(filp
, fl
, conf
);
2140 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2142 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2143 struct file_lock
*fl
)
2147 error
= security_file_lock(filp
, fl
->fl_type
);
2152 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2153 if (error
!= FILE_LOCK_DEFERRED
)
2155 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2159 locks_delete_block(fl
);
2166 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2168 check_fmode_for_setlk(struct file_lock
*fl
)
2170 switch (fl
->fl_type
) {
2172 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2176 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2182 /* Apply the lock described by l to an open file descriptor.
2183 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2185 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2186 struct flock __user
*l
)
2188 struct file_lock
*file_lock
= locks_alloc_lock();
2190 struct inode
*inode
;
2194 if (file_lock
== NULL
)
2197 inode
= file_inode(filp
);
2200 * This might block, so we do it before checking the inode.
2203 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2206 /* Don't allow mandatory locks on files that may be memory mapped
2209 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2214 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2218 error
= check_fmode_for_setlk(file_lock
);
2223 * If the cmd is requesting file-private locks, then set the
2224 * FL_OFDLCK flag and override the owner.
2229 if (flock
.l_pid
!= 0)
2233 file_lock
->fl_flags
|= FL_OFDLCK
;
2234 file_lock
->fl_owner
= filp
;
2238 if (flock
.l_pid
!= 0)
2242 file_lock
->fl_flags
|= FL_OFDLCK
;
2243 file_lock
->fl_owner
= filp
;
2246 file_lock
->fl_flags
|= FL_SLEEP
;
2249 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2252 * Attempt to detect a close/fcntl race and recover by releasing the
2253 * lock that was just acquired. There is no need to do that when we're
2254 * unlocking though, or for OFD locks.
2256 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2257 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2259 * We need that spin_lock here - it prevents reordering between
2260 * update of i_flctx->flc_posix and check for it done in
2261 * close(). rcu_read_lock() wouldn't do.
2263 spin_lock(¤t
->files
->file_lock
);
2265 spin_unlock(¤t
->files
->file_lock
);
2267 file_lock
->fl_type
= F_UNLCK
;
2268 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2269 WARN_ON_ONCE(error
);
2274 trace_fcntl_setlk(inode
, file_lock
, error
);
2275 locks_free_lock(file_lock
);
2279 #if BITS_PER_LONG == 32
2280 /* Report the first existing lock that would conflict with l.
2281 * This implements the F_GETLK command of fcntl().
2283 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2285 struct file_lock file_lock
;
2286 struct flock64 flock
;
2290 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2293 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2296 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2300 if (cmd
== F_OFD_GETLK
) {
2302 if (flock
.l_pid
!= 0)
2306 file_lock
.fl_flags
|= FL_OFDLCK
;
2307 file_lock
.fl_owner
= filp
;
2310 error
= vfs_test_lock(filp
, &file_lock
);
2314 flock
.l_type
= file_lock
.fl_type
;
2315 if (file_lock
.fl_type
!= F_UNLCK
)
2316 posix_lock_to_flock64(&flock
, &file_lock
);
2319 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2322 locks_release_private(&file_lock
);
2327 /* Apply the lock described by l to an open file descriptor.
2328 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2330 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2331 struct flock64 __user
*l
)
2333 struct file_lock
*file_lock
= locks_alloc_lock();
2334 struct flock64 flock
;
2335 struct inode
*inode
;
2339 if (file_lock
== NULL
)
2343 * This might block, so we do it before checking the inode.
2346 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2349 inode
= file_inode(filp
);
2351 /* Don't allow mandatory locks on files that may be memory mapped
2354 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2359 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2363 error
= check_fmode_for_setlk(file_lock
);
2368 * If the cmd is requesting file-private locks, then set the
2369 * FL_OFDLCK flag and override the owner.
2374 if (flock
.l_pid
!= 0)
2378 file_lock
->fl_flags
|= FL_OFDLCK
;
2379 file_lock
->fl_owner
= filp
;
2383 if (flock
.l_pid
!= 0)
2387 file_lock
->fl_flags
|= FL_OFDLCK
;
2388 file_lock
->fl_owner
= filp
;
2391 file_lock
->fl_flags
|= FL_SLEEP
;
2394 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2397 * Attempt to detect a close/fcntl race and recover by releasing the
2398 * lock that was just acquired. There is no need to do that when we're
2399 * unlocking though, or for OFD locks.
2401 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2402 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2404 * We need that spin_lock here - it prevents reordering between
2405 * update of i_flctx->flc_posix and check for it done in
2406 * close(). rcu_read_lock() wouldn't do.
2408 spin_lock(¤t
->files
->file_lock
);
2410 spin_unlock(¤t
->files
->file_lock
);
2412 file_lock
->fl_type
= F_UNLCK
;
2413 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2414 WARN_ON_ONCE(error
);
2419 locks_free_lock(file_lock
);
2422 #endif /* BITS_PER_LONG == 32 */
2425 * This function is called when the file is being removed
2426 * from the task's fd array. POSIX locks belonging to this task
2427 * are deleted at this time.
2429 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2432 struct file_lock lock
;
2433 struct file_lock_context
*ctx
;
2436 * If there are no locks held on this file, we don't need to call
2437 * posix_lock_file(). Another process could be setting a lock on this
2438 * file at the same time, but we wouldn't remove that lock anyway.
2440 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2441 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2444 lock
.fl_type
= F_UNLCK
;
2445 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2447 lock
.fl_end
= OFFSET_MAX
;
2448 lock
.fl_owner
= owner
;
2449 lock
.fl_pid
= current
->tgid
;
2450 lock
.fl_file
= filp
;
2452 lock
.fl_lmops
= NULL
;
2454 error
= vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2456 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2457 lock
.fl_ops
->fl_release_private(&lock
);
2458 trace_locks_remove_posix(file_inode(filp
), &lock
, error
);
2461 EXPORT_SYMBOL(locks_remove_posix
);
2463 /* The i_flctx must be valid when calling into here */
2465 locks_remove_flock(struct file
*filp
, struct file_lock_context
*flctx
)
2467 struct file_lock fl
= {
2469 .fl_pid
= current
->tgid
,
2471 .fl_flags
= FL_FLOCK
,
2473 .fl_end
= OFFSET_MAX
,
2475 struct inode
*inode
= file_inode(filp
);
2477 if (list_empty(&flctx
->flc_flock
))
2480 if (filp
->f_op
->flock
)
2481 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2483 flock_lock_inode(inode
, &fl
);
2485 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2486 fl
.fl_ops
->fl_release_private(&fl
);
2489 /* The i_flctx must be valid when calling into here */
2491 locks_remove_lease(struct file
*filp
, struct file_lock_context
*ctx
)
2493 struct file_lock
*fl
, *tmp
;
2496 if (list_empty(&ctx
->flc_lease
))
2499 spin_lock(&ctx
->flc_lock
);
2500 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2501 if (filp
== fl
->fl_file
)
2502 lease_modify(fl
, F_UNLCK
, &dispose
);
2503 spin_unlock(&ctx
->flc_lock
);
2504 locks_dispose_list(&dispose
);
2508 * This function is called on the last close of an open file.
2510 void locks_remove_file(struct file
*filp
)
2512 struct file_lock_context
*ctx
;
2514 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2518 /* remove any OFD locks */
2519 locks_remove_posix(filp
, filp
);
2521 /* remove flock locks */
2522 locks_remove_flock(filp
, ctx
);
2524 /* remove any leases */
2525 locks_remove_lease(filp
, ctx
);
2529 * posix_unblock_lock - stop waiting for a file lock
2530 * @waiter: the lock which was waiting
2532 * lockd needs to block waiting for locks.
2535 posix_unblock_lock(struct file_lock
*waiter
)
2539 spin_lock(&blocked_lock_lock
);
2540 if (waiter
->fl_next
)
2541 __locks_delete_block(waiter
);
2544 spin_unlock(&blocked_lock_lock
);
2547 EXPORT_SYMBOL(posix_unblock_lock
);
2550 * vfs_cancel_lock - file byte range unblock lock
2551 * @filp: The file to apply the unblock to
2552 * @fl: The lock to be unblocked
2554 * Used by lock managers to cancel blocked requests
2556 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2558 if (filp
->f_op
->lock
)
2559 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2563 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2565 #ifdef CONFIG_PROC_FS
2566 #include <linux/proc_fs.h>
2567 #include <linux/seq_file.h>
2569 struct locks_iterator
{
2574 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2575 loff_t id
, char *pfx
)
2577 struct inode
*inode
= NULL
;
2578 unsigned int fl_pid
;
2581 fl_pid
= pid_vnr(fl
->fl_nspid
);
2583 fl_pid
= fl
->fl_pid
;
2585 if (fl
->fl_file
!= NULL
)
2586 inode
= file_inode(fl
->fl_file
);
2588 seq_printf(f
, "%lld:%s ", id
, pfx
);
2590 if (fl
->fl_flags
& FL_ACCESS
)
2591 seq_puts(f
, "ACCESS");
2592 else if (IS_OFDLCK(fl
))
2593 seq_puts(f
, "OFDLCK");
2595 seq_puts(f
, "POSIX ");
2597 seq_printf(f
, " %s ",
2598 (inode
== NULL
) ? "*NOINODE*" :
2599 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2600 } else if (IS_FLOCK(fl
)) {
2601 if (fl
->fl_type
& LOCK_MAND
) {
2602 seq_puts(f
, "FLOCK MSNFS ");
2604 seq_puts(f
, "FLOCK ADVISORY ");
2606 } else if (IS_LEASE(fl
)) {
2607 if (fl
->fl_flags
& FL_DELEG
)
2608 seq_puts(f
, "DELEG ");
2610 seq_puts(f
, "LEASE ");
2612 if (lease_breaking(fl
))
2613 seq_puts(f
, "BREAKING ");
2614 else if (fl
->fl_file
)
2615 seq_puts(f
, "ACTIVE ");
2617 seq_puts(f
, "BREAKER ");
2619 seq_puts(f
, "UNKNOWN UNKNOWN ");
2621 if (fl
->fl_type
& LOCK_MAND
) {
2622 seq_printf(f
, "%s ",
2623 (fl
->fl_type
& LOCK_READ
)
2624 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2625 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2627 seq_printf(f
, "%s ",
2628 (lease_breaking(fl
))
2629 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2630 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2633 /* userspace relies on this representation of dev_t */
2634 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2635 MAJOR(inode
->i_sb
->s_dev
),
2636 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2638 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2641 if (fl
->fl_end
== OFFSET_MAX
)
2642 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2644 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2646 seq_puts(f
, "0 EOF\n");
2650 static int locks_show(struct seq_file
*f
, void *v
)
2652 struct locks_iterator
*iter
= f
->private;
2653 struct file_lock
*fl
, *bfl
;
2655 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2657 lock_get_status(f
, fl
, iter
->li_pos
, "");
2659 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2660 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2665 static void __show_fd_locks(struct seq_file
*f
,
2666 struct list_head
*head
, int *id
,
2667 struct file
*filp
, struct files_struct
*files
)
2669 struct file_lock
*fl
;
2671 list_for_each_entry(fl
, head
, fl_list
) {
2673 if (filp
!= fl
->fl_file
)
2675 if (fl
->fl_owner
!= files
&&
2676 fl
->fl_owner
!= filp
)
2680 seq_puts(f
, "lock:\t");
2681 lock_get_status(f
, fl
, *id
, "");
2685 void show_fd_locks(struct seq_file
*f
,
2686 struct file
*filp
, struct files_struct
*files
)
2688 struct inode
*inode
= file_inode(filp
);
2689 struct file_lock_context
*ctx
;
2692 ctx
= smp_load_acquire(&inode
->i_flctx
);
2696 spin_lock(&ctx
->flc_lock
);
2697 __show_fd_locks(f
, &ctx
->flc_flock
, &id
, filp
, files
);
2698 __show_fd_locks(f
, &ctx
->flc_posix
, &id
, filp
, files
);
2699 __show_fd_locks(f
, &ctx
->flc_lease
, &id
, filp
, files
);
2700 spin_unlock(&ctx
->flc_lock
);
2703 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2704 __acquires(&blocked_lock_lock
)
2706 struct locks_iterator
*iter
= f
->private;
2708 iter
->li_pos
= *pos
+ 1;
2709 lg_global_lock(&file_lock_lglock
);
2710 spin_lock(&blocked_lock_lock
);
2711 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2714 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2716 struct locks_iterator
*iter
= f
->private;
2719 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2722 static void locks_stop(struct seq_file
*f
, void *v
)
2723 __releases(&blocked_lock_lock
)
2725 spin_unlock(&blocked_lock_lock
);
2726 lg_global_unlock(&file_lock_lglock
);
2729 static const struct seq_operations locks_seq_operations
= {
2730 .start
= locks_start
,
2736 static int locks_open(struct inode
*inode
, struct file
*filp
)
2738 return seq_open_private(filp
, &locks_seq_operations
,
2739 sizeof(struct locks_iterator
));
2742 static const struct file_operations proc_locks_operations
= {
2745 .llseek
= seq_lseek
,
2746 .release
= seq_release_private
,
2749 static int __init
proc_locks_init(void)
2751 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2754 fs_initcall(proc_locks_init
);
2757 static int __init
filelock_init(void)
2761 flctx_cache
= kmem_cache_create("file_lock_ctx",
2762 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2764 filelock_cache
= kmem_cache_create("file_lock_cache",
2765 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2767 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2769 for_each_possible_cpu(i
)
2770 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2775 core_initcall(filelock_init
);