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 inline bool is_remote_lock(struct file
*filp
)
144 return likely(!(filp
->f_path
.dentry
->d_sb
->s_flags
& MS_NOREMOTELOCK
));
147 static bool lease_breaking(struct file_lock
*fl
)
149 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
152 static int target_leasetype(struct file_lock
*fl
)
154 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
156 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
161 int leases_enable
= 1;
162 int lease_break_time
= 45;
165 * The global file_lock_list is only used for displaying /proc/locks, so we
166 * keep a list on each CPU, with each list protected by its own spinlock via
167 * the file_lock_lglock. Note that alterations to the list also require that
168 * the relevant flc_lock is held.
170 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
171 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
174 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
175 * It is protected by blocked_lock_lock.
177 * We hash locks by lockowner in order to optimize searching for the lock a
178 * particular lockowner is waiting on.
180 * FIXME: make this value scale via some heuristic? We generally will want more
181 * buckets when we have more lockowners holding locks, but that's a little
182 * difficult to determine without knowing what the workload will look like.
184 #define BLOCKED_HASH_BITS 7
185 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
188 * This lock protects the blocked_hash. Generally, if you're accessing it, you
189 * want to be holding this lock.
191 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
192 * pointer for file_lock structures that are acting as lock requests (in
193 * contrast to those that are acting as records of acquired locks).
195 * Note that when we acquire this lock in order to change the above fields,
196 * we often hold the flc_lock as well. In certain cases, when reading the fields
197 * protected by this lock, we can skip acquiring it iff we already hold the
200 * In particular, adding an entry to the fl_block list requires that you hold
201 * both the flc_lock and the blocked_lock_lock (acquired in that order).
202 * Deleting an entry from the list however only requires the file_lock_lock.
204 static DEFINE_SPINLOCK(blocked_lock_lock
);
206 static struct kmem_cache
*flctx_cache __read_mostly
;
207 static struct kmem_cache
*filelock_cache __read_mostly
;
209 static struct file_lock_context
*
210 locks_get_lock_context(struct inode
*inode
, int type
)
212 struct file_lock_context
*ctx
;
214 /* paired with cmpxchg() below */
215 ctx
= smp_load_acquire(&inode
->i_flctx
);
216 if (likely(ctx
) || type
== F_UNLCK
)
219 ctx
= kmem_cache_alloc(flctx_cache
, GFP_KERNEL
);
223 spin_lock_init(&ctx
->flc_lock
);
224 INIT_LIST_HEAD(&ctx
->flc_flock
);
225 INIT_LIST_HEAD(&ctx
->flc_posix
);
226 INIT_LIST_HEAD(&ctx
->flc_lease
);
229 * Assign the pointer if it's not already assigned. If it is, then
230 * free the context we just allocated.
232 if (cmpxchg(&inode
->i_flctx
, NULL
, ctx
)) {
233 kmem_cache_free(flctx_cache
, ctx
);
234 ctx
= smp_load_acquire(&inode
->i_flctx
);
237 trace_locks_get_lock_context(inode
, type
, ctx
);
242 locks_dump_ctx_list(struct list_head
*list
, char *list_type
)
244 struct file_lock
*fl
;
246 list_for_each_entry(fl
, list
, fl_list
) {
247 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
);
252 locks_check_ctx_lists(struct inode
*inode
)
254 struct file_lock_context
*ctx
= inode
->i_flctx
;
256 if (unlikely(!list_empty(&ctx
->flc_flock
) ||
257 !list_empty(&ctx
->flc_posix
) ||
258 !list_empty(&ctx
->flc_lease
))) {
259 pr_warn("Leaked locks on dev=0x%x:0x%x ino=0x%lx:\n",
260 MAJOR(inode
->i_sb
->s_dev
), MINOR(inode
->i_sb
->s_dev
),
262 locks_dump_ctx_list(&ctx
->flc_flock
, "FLOCK");
263 locks_dump_ctx_list(&ctx
->flc_posix
, "POSIX");
264 locks_dump_ctx_list(&ctx
->flc_lease
, "LEASE");
269 locks_free_lock_context(struct inode
*inode
)
271 struct file_lock_context
*ctx
= inode
->i_flctx
;
274 locks_check_ctx_lists(inode
);
275 kmem_cache_free(flctx_cache
, ctx
);
279 static void locks_init_lock_heads(struct file_lock
*fl
)
281 INIT_HLIST_NODE(&fl
->fl_link
);
282 INIT_LIST_HEAD(&fl
->fl_list
);
283 INIT_LIST_HEAD(&fl
->fl_block
);
284 init_waitqueue_head(&fl
->fl_wait
);
287 /* Allocate an empty lock structure. */
288 struct file_lock
*locks_alloc_lock(void)
290 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
293 locks_init_lock_heads(fl
);
297 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
299 void locks_release_private(struct file_lock
*fl
)
302 if (fl
->fl_ops
->fl_release_private
)
303 fl
->fl_ops
->fl_release_private(fl
);
308 if (fl
->fl_lmops
->lm_put_owner
) {
309 fl
->fl_lmops
->lm_put_owner(fl
->fl_owner
);
315 EXPORT_SYMBOL_GPL(locks_release_private
);
317 /* Free a lock which is not in use. */
318 void locks_free_lock(struct file_lock
*fl
)
320 BUG_ON(waitqueue_active(&fl
->fl_wait
));
321 BUG_ON(!list_empty(&fl
->fl_list
));
322 BUG_ON(!list_empty(&fl
->fl_block
));
323 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
325 locks_release_private(fl
);
326 kmem_cache_free(filelock_cache
, fl
);
328 EXPORT_SYMBOL(locks_free_lock
);
331 locks_dispose_list(struct list_head
*dispose
)
333 struct file_lock
*fl
;
335 while (!list_empty(dispose
)) {
336 fl
= list_first_entry(dispose
, struct file_lock
, fl_list
);
337 list_del_init(&fl
->fl_list
);
342 void locks_init_lock(struct file_lock
*fl
)
344 memset(fl
, 0, sizeof(struct file_lock
));
345 locks_init_lock_heads(fl
);
348 EXPORT_SYMBOL(locks_init_lock
);
351 * Initialize a new lock from an existing file_lock structure.
353 void locks_copy_conflock(struct file_lock
*new, struct file_lock
*fl
)
355 new->fl_owner
= fl
->fl_owner
;
356 new->fl_pid
= fl
->fl_pid
;
358 new->fl_flags
= fl
->fl_flags
;
359 new->fl_type
= fl
->fl_type
;
360 new->fl_start
= fl
->fl_start
;
361 new->fl_end
= fl
->fl_end
;
362 new->fl_lmops
= fl
->fl_lmops
;
366 if (fl
->fl_lmops
->lm_get_owner
)
367 fl
->fl_lmops
->lm_get_owner(fl
->fl_owner
);
370 EXPORT_SYMBOL(locks_copy_conflock
);
372 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
374 /* "new" must be a freshly-initialized lock */
375 WARN_ON_ONCE(new->fl_ops
);
377 locks_copy_conflock(new, fl
);
379 new->fl_file
= fl
->fl_file
;
380 new->fl_ops
= fl
->fl_ops
;
383 if (fl
->fl_ops
->fl_copy_lock
)
384 fl
->fl_ops
->fl_copy_lock(new, fl
);
388 EXPORT_SYMBOL(locks_copy_lock
);
390 static inline int flock_translate_cmd(int cmd
) {
392 return cmd
& (LOCK_MAND
| LOCK_RW
);
404 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
405 static struct file_lock
*
406 flock_make_lock(struct file
*filp
, unsigned int cmd
)
408 struct file_lock
*fl
;
409 int type
= flock_translate_cmd(cmd
);
412 return ERR_PTR(type
);
414 fl
= locks_alloc_lock();
416 return ERR_PTR(-ENOMEM
);
420 fl
->fl_pid
= current
->tgid
;
421 fl
->fl_flags
= FL_FLOCK
;
423 fl
->fl_end
= OFFSET_MAX
;
428 static int assign_type(struct file_lock
*fl
, long type
)
442 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
445 switch (l
->l_whence
) {
450 fl
->fl_start
= filp
->f_pos
;
453 fl
->fl_start
= i_size_read(file_inode(filp
));
458 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
460 fl
->fl_start
+= l
->l_start
;
461 if (fl
->fl_start
< 0)
464 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
465 POSIX-2001 defines it. */
467 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
469 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
471 } else if (l
->l_len
< 0) {
472 if (fl
->fl_start
+ l
->l_len
< 0)
474 fl
->fl_end
= fl
->fl_start
- 1;
475 fl
->fl_start
+= l
->l_len
;
477 fl
->fl_end
= OFFSET_MAX
;
479 fl
->fl_owner
= current
->files
;
480 fl
->fl_pid
= current
->tgid
;
482 fl
->fl_flags
= FL_POSIX
;
486 return assign_type(fl
, l
->l_type
);
489 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
492 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
495 struct flock64 ll
= {
497 .l_whence
= l
->l_whence
,
498 .l_start
= l
->l_start
,
502 return flock64_to_posix_lock(filp
, fl
, &ll
);
505 /* default lease lock manager operations */
507 lease_break_callback(struct file_lock
*fl
)
509 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
514 lease_setup(struct file_lock
*fl
, void **priv
)
516 struct file
*filp
= fl
->fl_file
;
517 struct fasync_struct
*fa
= *priv
;
520 * fasync_insert_entry() returns the old entry if any. If there was no
521 * old entry, then it used "priv" and inserted it into the fasync list.
522 * Clear the pointer to indicate that it shouldn't be freed.
524 if (!fasync_insert_entry(fa
->fa_fd
, filp
, &fl
->fl_fasync
, fa
))
527 __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
530 static const struct lock_manager_operations lease_manager_ops
= {
531 .lm_break
= lease_break_callback
,
532 .lm_change
= lease_modify
,
533 .lm_setup
= lease_setup
,
537 * Initialize a lease, use the default lock manager operations
539 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
541 if (assign_type(fl
, type
) != 0)
545 fl
->fl_pid
= current
->tgid
;
548 fl
->fl_flags
= FL_LEASE
;
550 fl
->fl_end
= OFFSET_MAX
;
552 fl
->fl_lmops
= &lease_manager_ops
;
556 /* Allocate a file_lock initialised to this type of lease */
557 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
559 struct file_lock
*fl
= locks_alloc_lock();
563 return ERR_PTR(error
);
565 error
= lease_init(filp
, type
, fl
);
568 return ERR_PTR(error
);
573 /* Check if two locks overlap each other.
575 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
577 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
578 (fl2
->fl_end
>= fl1
->fl_start
));
582 * Check whether two locks have the same owner.
584 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
586 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
587 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
588 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
589 return fl1
->fl_owner
== fl2
->fl_owner
;
592 /* Must be called with the flc_lock held! */
593 static void locks_insert_global_locks(struct file_lock
*fl
)
595 lg_local_lock(&file_lock_lglock
);
596 fl
->fl_link_cpu
= smp_processor_id();
597 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
598 lg_local_unlock(&file_lock_lglock
);
601 /* Must be called with the flc_lock held! */
602 static void locks_delete_global_locks(struct file_lock
*fl
)
605 * Avoid taking lock if already unhashed. This is safe since this check
606 * is done while holding the flc_lock, and new insertions into the list
607 * also require that it be held.
609 if (hlist_unhashed(&fl
->fl_link
))
611 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
612 hlist_del_init(&fl
->fl_link
);
613 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
617 posix_owner_key(struct file_lock
*fl
)
619 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
620 return fl
->fl_lmops
->lm_owner_key(fl
);
621 return (unsigned long)fl
->fl_owner
;
624 static void locks_insert_global_blocked(struct file_lock
*waiter
)
626 lockdep_assert_held(&blocked_lock_lock
);
628 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
631 static void locks_delete_global_blocked(struct file_lock
*waiter
)
633 lockdep_assert_held(&blocked_lock_lock
);
635 hash_del(&waiter
->fl_link
);
638 /* Remove waiter from blocker's block list.
639 * When blocker ends up pointing to itself then the list is empty.
641 * Must be called with blocked_lock_lock held.
643 static void __locks_delete_block(struct file_lock
*waiter
)
645 locks_delete_global_blocked(waiter
);
646 list_del_init(&waiter
->fl_block
);
647 waiter
->fl_next
= NULL
;
650 static void locks_delete_block(struct file_lock
*waiter
)
652 spin_lock(&blocked_lock_lock
);
653 __locks_delete_block(waiter
);
654 spin_unlock(&blocked_lock_lock
);
657 /* Insert waiter into blocker's block list.
658 * We use a circular list so that processes can be easily woken up in
659 * the order they blocked. The documentation doesn't require this but
660 * it seems like the reasonable thing to do.
662 * Must be called with both the flc_lock and blocked_lock_lock held. The
663 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
664 * that the flc_lock is also held on insertions we can avoid taking the
665 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
667 static void __locks_insert_block(struct file_lock
*blocker
,
668 struct file_lock
*waiter
)
670 BUG_ON(!list_empty(&waiter
->fl_block
));
671 waiter
->fl_next
= blocker
;
672 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
673 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
674 locks_insert_global_blocked(waiter
);
677 /* Must be called with flc_lock held. */
678 static void locks_insert_block(struct file_lock
*blocker
,
679 struct file_lock
*waiter
)
681 spin_lock(&blocked_lock_lock
);
682 __locks_insert_block(blocker
, waiter
);
683 spin_unlock(&blocked_lock_lock
);
687 * Wake up processes blocked waiting for blocker.
689 * Must be called with the inode->flc_lock held!
691 static void locks_wake_up_blocks(struct file_lock
*blocker
)
694 * Avoid taking global lock if list is empty. This is safe since new
695 * blocked requests are only added to the list under the flc_lock, and
696 * the flc_lock is always held here. Note that removal from the fl_block
697 * list does not require the flc_lock, so we must recheck list_empty()
698 * after acquiring the blocked_lock_lock.
700 if (list_empty(&blocker
->fl_block
))
703 spin_lock(&blocked_lock_lock
);
704 while (!list_empty(&blocker
->fl_block
)) {
705 struct file_lock
*waiter
;
707 waiter
= list_first_entry(&blocker
->fl_block
,
708 struct file_lock
, fl_block
);
709 __locks_delete_block(waiter
);
710 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
711 waiter
->fl_lmops
->lm_notify(waiter
);
713 wake_up(&waiter
->fl_wait
);
715 spin_unlock(&blocked_lock_lock
);
719 locks_insert_lock_ctx(struct file_lock
*fl
, struct list_head
*before
)
721 fl
->fl_nspid
= get_pid(task_tgid(current
));
722 list_add_tail(&fl
->fl_list
, before
);
723 locks_insert_global_locks(fl
);
727 locks_unlink_lock_ctx(struct file_lock
*fl
)
729 locks_delete_global_locks(fl
);
730 list_del_init(&fl
->fl_list
);
732 put_pid(fl
->fl_nspid
);
735 locks_wake_up_blocks(fl
);
739 locks_delete_lock_ctx(struct file_lock
*fl
, struct list_head
*dispose
)
741 locks_unlink_lock_ctx(fl
);
743 list_add(&fl
->fl_list
, dispose
);
748 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
749 * checks for shared/exclusive status of overlapping locks.
751 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
753 if (sys_fl
->fl_type
== F_WRLCK
)
755 if (caller_fl
->fl_type
== F_WRLCK
)
760 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
761 * checking before calling the locks_conflict().
763 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
765 /* POSIX locks owned by the same process do not conflict with
768 if (posix_same_owner(caller_fl
, sys_fl
))
771 /* Check whether they overlap */
772 if (!locks_overlap(caller_fl
, sys_fl
))
775 return (locks_conflict(caller_fl
, sys_fl
));
778 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
779 * checking before calling the locks_conflict().
781 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
783 /* FLOCK locks referring to the same filp do not conflict with
786 if (caller_fl
->fl_file
== sys_fl
->fl_file
)
788 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
791 return (locks_conflict(caller_fl
, sys_fl
));
795 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
797 struct file_lock
*cfl
;
798 struct file_lock_context
*ctx
;
799 struct inode
*inode
= locks_inode(filp
);
801 ctx
= smp_load_acquire(&inode
->i_flctx
);
802 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
)) {
803 fl
->fl_type
= F_UNLCK
;
807 spin_lock(&ctx
->flc_lock
);
808 list_for_each_entry(cfl
, &ctx
->flc_posix
, fl_list
) {
809 if (posix_locks_conflict(fl
, cfl
)) {
810 locks_copy_conflock(fl
, cfl
);
812 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
816 fl
->fl_type
= F_UNLCK
;
818 spin_unlock(&ctx
->flc_lock
);
821 EXPORT_SYMBOL(posix_test_lock
);
824 * Deadlock detection:
826 * We attempt to detect deadlocks that are due purely to posix file
829 * We assume that a task can be waiting for at most one lock at a time.
830 * So for any acquired lock, the process holding that lock may be
831 * waiting on at most one other lock. That lock in turns may be held by
832 * someone waiting for at most one other lock. Given a requested lock
833 * caller_fl which is about to wait for a conflicting lock block_fl, we
834 * follow this chain of waiters to ensure we are not about to create a
837 * Since we do this before we ever put a process to sleep on a lock, we
838 * are ensured that there is never a cycle; that is what guarantees that
839 * the while() loop in posix_locks_deadlock() eventually completes.
841 * Note: the above assumption may not be true when handling lock
842 * requests from a broken NFS client. It may also fail in the presence
843 * of tasks (such as posix threads) sharing the same open file table.
844 * To handle those cases, we just bail out after a few iterations.
846 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
847 * Because the owner is not even nominally tied to a thread of
848 * execution, the deadlock detection below can't reasonably work well. Just
851 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
852 * locks that just checks for the case where two tasks are attempting to
853 * upgrade from read to write locks on the same inode.
856 #define MAX_DEADLK_ITERATIONS 10
858 /* Find a lock that the owner of the given block_fl is blocking on. */
859 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
861 struct file_lock
*fl
;
863 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
864 if (posix_same_owner(fl
, block_fl
))
870 /* Must be called with the blocked_lock_lock held! */
871 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
872 struct file_lock
*block_fl
)
876 lockdep_assert_held(&blocked_lock_lock
);
879 * This deadlock detector can't reasonably detect deadlocks with
880 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
882 if (IS_OFDLCK(caller_fl
))
885 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
886 if (i
++ > MAX_DEADLK_ITERATIONS
)
888 if (posix_same_owner(caller_fl
, block_fl
))
894 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
895 * after any leases, but before any posix locks.
897 * Note that if called with an FL_EXISTS argument, the caller may determine
898 * whether or not a lock was successfully freed by testing the return
901 static int flock_lock_inode(struct inode
*inode
, struct file_lock
*request
)
903 struct file_lock
*new_fl
= NULL
;
904 struct file_lock
*fl
;
905 struct file_lock_context
*ctx
;
910 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
912 if (request
->fl_type
!= F_UNLCK
)
914 return (request
->fl_flags
& FL_EXISTS
) ? -ENOENT
: 0;
917 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
918 new_fl
= locks_alloc_lock();
923 spin_lock(&ctx
->flc_lock
);
924 if (request
->fl_flags
& FL_ACCESS
)
927 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
928 if (request
->fl_file
!= fl
->fl_file
)
930 if (request
->fl_type
== fl
->fl_type
)
933 locks_delete_lock_ctx(fl
, &dispose
);
937 if (request
->fl_type
== F_UNLCK
) {
938 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
944 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
945 if (!flock_locks_conflict(request
, fl
))
948 if (!(request
->fl_flags
& FL_SLEEP
))
950 error
= FILE_LOCK_DEFERRED
;
951 locks_insert_block(fl
, request
);
954 if (request
->fl_flags
& FL_ACCESS
)
956 locks_copy_lock(new_fl
, request
);
957 locks_insert_lock_ctx(new_fl
, &ctx
->flc_flock
);
962 spin_unlock(&ctx
->flc_lock
);
964 locks_free_lock(new_fl
);
965 locks_dispose_list(&dispose
);
969 static int posix_lock_inode(struct inode
*inode
, struct file_lock
*request
,
970 struct file_lock
*conflock
)
972 struct file_lock
*fl
, *tmp
;
973 struct file_lock
*new_fl
= NULL
;
974 struct file_lock
*new_fl2
= NULL
;
975 struct file_lock
*left
= NULL
;
976 struct file_lock
*right
= NULL
;
977 struct file_lock_context
*ctx
;
982 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
984 return (request
->fl_type
== F_UNLCK
) ? 0 : -ENOMEM
;
987 * We may need two file_lock structures for this operation,
988 * so we get them in advance to avoid races.
990 * In some cases we can be sure, that no new locks will be needed
992 if (!(request
->fl_flags
& FL_ACCESS
) &&
993 (request
->fl_type
!= F_UNLCK
||
994 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
995 new_fl
= locks_alloc_lock();
996 new_fl2
= locks_alloc_lock();
999 spin_lock(&ctx
->flc_lock
);
1001 * New lock request. Walk all POSIX locks and look for conflicts. If
1002 * there are any, either return error or put the request on the
1003 * blocker's list of waiters and the global blocked_hash.
1005 if (request
->fl_type
!= F_UNLCK
) {
1006 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1007 if (!posix_locks_conflict(request
, fl
))
1010 locks_copy_conflock(conflock
, fl
);
1012 if (!(request
->fl_flags
& FL_SLEEP
))
1015 * Deadlock detection and insertion into the blocked
1016 * locks list must be done while holding the same lock!
1019 spin_lock(&blocked_lock_lock
);
1020 if (likely(!posix_locks_deadlock(request
, fl
))) {
1021 error
= FILE_LOCK_DEFERRED
;
1022 __locks_insert_block(fl
, request
);
1024 spin_unlock(&blocked_lock_lock
);
1029 /* If we're just looking for a conflict, we're done. */
1031 if (request
->fl_flags
& FL_ACCESS
)
1034 /* Find the first old lock with the same owner as the new lock */
1035 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1036 if (posix_same_owner(request
, fl
))
1040 /* Process locks with this owner. */
1041 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1042 if (!posix_same_owner(request
, fl
))
1045 /* Detect adjacent or overlapping regions (if same lock type) */
1046 if (request
->fl_type
== fl
->fl_type
) {
1047 /* In all comparisons of start vs end, use
1048 * "start - 1" rather than "end + 1". If end
1049 * is OFFSET_MAX, end + 1 will become negative.
1051 if (fl
->fl_end
< request
->fl_start
- 1)
1053 /* If the next lock in the list has entirely bigger
1054 * addresses than the new one, insert the lock here.
1056 if (fl
->fl_start
- 1 > request
->fl_end
)
1059 /* If we come here, the new and old lock are of the
1060 * same type and adjacent or overlapping. Make one
1061 * lock yielding from the lower start address of both
1062 * locks to the higher end address.
1064 if (fl
->fl_start
> request
->fl_start
)
1065 fl
->fl_start
= request
->fl_start
;
1067 request
->fl_start
= fl
->fl_start
;
1068 if (fl
->fl_end
< request
->fl_end
)
1069 fl
->fl_end
= request
->fl_end
;
1071 request
->fl_end
= fl
->fl_end
;
1073 locks_delete_lock_ctx(fl
, &dispose
);
1079 /* Processing for different lock types is a bit
1082 if (fl
->fl_end
< request
->fl_start
)
1084 if (fl
->fl_start
> request
->fl_end
)
1086 if (request
->fl_type
== F_UNLCK
)
1088 if (fl
->fl_start
< request
->fl_start
)
1090 /* If the next lock in the list has a higher end
1091 * address than the new one, insert the new one here.
1093 if (fl
->fl_end
> request
->fl_end
) {
1097 if (fl
->fl_start
>= request
->fl_start
) {
1098 /* The new lock completely replaces an old
1099 * one (This may happen several times).
1102 locks_delete_lock_ctx(fl
, &dispose
);
1106 * Replace the old lock with new_fl, and
1107 * remove the old one. It's safe to do the
1108 * insert here since we know that we won't be
1109 * using new_fl later, and that the lock is
1110 * just replacing an existing lock.
1115 locks_copy_lock(new_fl
, request
);
1118 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1119 locks_delete_lock_ctx(fl
, &dispose
);
1126 * The above code only modifies existing locks in case of merging or
1127 * replacing. If new lock(s) need to be inserted all modifications are
1128 * done below this, so it's safe yet to bail out.
1130 error
= -ENOLCK
; /* "no luck" */
1131 if (right
&& left
== right
&& !new_fl2
)
1136 if (request
->fl_type
== F_UNLCK
) {
1137 if (request
->fl_flags
& FL_EXISTS
)
1146 locks_copy_lock(new_fl
, request
);
1147 locks_insert_lock_ctx(new_fl
, &fl
->fl_list
);
1152 if (left
== right
) {
1153 /* The new lock breaks the old one in two pieces,
1154 * so we have to use the second new lock.
1158 locks_copy_lock(left
, right
);
1159 locks_insert_lock_ctx(left
, &fl
->fl_list
);
1161 right
->fl_start
= request
->fl_end
+ 1;
1162 locks_wake_up_blocks(right
);
1165 left
->fl_end
= request
->fl_start
- 1;
1166 locks_wake_up_blocks(left
);
1169 spin_unlock(&ctx
->flc_lock
);
1171 * Free any unused locks.
1174 locks_free_lock(new_fl
);
1176 locks_free_lock(new_fl2
);
1177 locks_dispose_list(&dispose
);
1178 trace_posix_lock_inode(inode
, request
, error
);
1184 * posix_lock_file - Apply a POSIX-style lock to a file
1185 * @filp: The file to apply the lock to
1186 * @fl: The lock to be applied
1187 * @conflock: Place to return a copy of the conflicting lock, if found.
1189 * Add a POSIX style lock to a file.
1190 * We merge adjacent & overlapping locks whenever possible.
1191 * POSIX locks are sorted by owner task, then by starting address
1193 * Note that if called with an FL_EXISTS argument, the caller may determine
1194 * whether or not a lock was successfully freed by testing the return
1195 * value for -ENOENT.
1197 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1198 struct file_lock
*conflock
)
1200 return posix_lock_inode(locks_inode(filp
), fl
, conflock
);
1202 EXPORT_SYMBOL(posix_lock_file
);
1205 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1206 * @inode: inode of file to which lock request should be applied
1207 * @fl: The lock to be applied
1209 * Apply a POSIX style lock request to an inode.
1211 static int posix_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1216 error
= posix_lock_inode(inode
, fl
, NULL
);
1217 if (error
!= FILE_LOCK_DEFERRED
)
1219 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1223 locks_delete_block(fl
);
1229 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1231 * locks_mandatory_locked - Check for an active lock
1232 * @file: the file to check
1234 * Searches the inode's list of locks to find any POSIX locks which conflict.
1235 * This function is called from locks_verify_locked() only.
1237 int locks_mandatory_locked(struct file
*file
)
1240 struct inode
*inode
= locks_inode(file
);
1241 struct file_lock_context
*ctx
;
1242 struct file_lock
*fl
;
1244 ctx
= smp_load_acquire(&inode
->i_flctx
);
1245 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
))
1249 * Search the lock list for this inode for any POSIX locks.
1251 spin_lock(&ctx
->flc_lock
);
1253 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1254 if (fl
->fl_owner
!= current
->files
&&
1255 fl
->fl_owner
!= file
) {
1260 spin_unlock(&ctx
->flc_lock
);
1265 * locks_mandatory_area - Check for a conflicting lock
1266 * @inode: the file to check
1267 * @filp: how the file was opened (if it was)
1268 * @start: first byte in the file to check
1269 * @end: lastbyte in the file to check
1270 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1272 * Searches the inode's list of locks to find any POSIX locks which conflict.
1274 int locks_mandatory_area(struct inode
*inode
, struct file
*filp
, loff_t start
,
1275 loff_t end
, unsigned char type
)
1277 struct file_lock fl
;
1281 locks_init_lock(&fl
);
1282 fl
.fl_pid
= current
->tgid
;
1284 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1285 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1288 fl
.fl_start
= start
;
1294 fl
.fl_flags
&= ~FL_SLEEP
;
1295 error
= posix_lock_inode(inode
, &fl
, NULL
);
1301 fl
.fl_flags
|= FL_SLEEP
;
1302 fl
.fl_owner
= current
->files
;
1303 error
= posix_lock_inode(inode
, &fl
, NULL
);
1304 if (error
!= FILE_LOCK_DEFERRED
)
1306 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1309 * If we've been sleeping someone might have
1310 * changed the permissions behind our back.
1312 if (__mandatory_lock(inode
))
1316 locks_delete_block(&fl
);
1323 EXPORT_SYMBOL(locks_mandatory_area
);
1324 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1326 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1330 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1333 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1337 /* We already had a lease on this file; just change its type */
1338 int lease_modify(struct file_lock
*fl
, int arg
, struct list_head
*dispose
)
1340 int error
= assign_type(fl
, arg
);
1344 lease_clear_pending(fl
, arg
);
1345 locks_wake_up_blocks(fl
);
1346 if (arg
== F_UNLCK
) {
1347 struct file
*filp
= fl
->fl_file
;
1350 filp
->f_owner
.signum
= 0;
1351 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1352 if (fl
->fl_fasync
!= NULL
) {
1353 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1354 fl
->fl_fasync
= NULL
;
1356 locks_delete_lock_ctx(fl
, dispose
);
1360 EXPORT_SYMBOL(lease_modify
);
1362 static bool past_time(unsigned long then
)
1365 /* 0 is a special value meaning "this never expires": */
1367 return time_after(jiffies
, then
);
1370 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1372 struct file_lock_context
*ctx
= inode
->i_flctx
;
1373 struct file_lock
*fl
, *tmp
;
1375 lockdep_assert_held(&ctx
->flc_lock
);
1377 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1378 trace_time_out_leases(inode
, fl
);
1379 if (past_time(fl
->fl_downgrade_time
))
1380 lease_modify(fl
, F_RDLCK
, dispose
);
1381 if (past_time(fl
->fl_break_time
))
1382 lease_modify(fl
, F_UNLCK
, dispose
);
1386 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1388 if ((breaker
->fl_flags
& FL_LAYOUT
) != (lease
->fl_flags
& FL_LAYOUT
))
1390 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1392 return locks_conflict(breaker
, lease
);
1396 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1398 struct file_lock_context
*ctx
= inode
->i_flctx
;
1399 struct file_lock
*fl
;
1401 lockdep_assert_held(&ctx
->flc_lock
);
1403 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1404 if (leases_conflict(fl
, breaker
))
1411 * __break_lease - revoke all outstanding leases on file
1412 * @inode: the inode of the file to return
1413 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1415 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1418 * break_lease (inlined for speed) has checked there already is at least
1419 * some kind of lock (maybe a lease) on this file. Leases are broken on
1420 * a call to open() or truncate(). This function can sleep unless you
1421 * specified %O_NONBLOCK to your open().
1423 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1426 struct file_lock_context
*ctx
;
1427 struct file_lock
*new_fl
, *fl
, *tmp
;
1428 unsigned long break_time
;
1429 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1432 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1434 return PTR_ERR(new_fl
);
1435 new_fl
->fl_flags
= type
;
1437 /* typically we will check that ctx is non-NULL before calling */
1438 ctx
= smp_load_acquire(&inode
->i_flctx
);
1444 spin_lock(&ctx
->flc_lock
);
1446 time_out_leases(inode
, &dispose
);
1448 if (!any_leases_conflict(inode
, new_fl
))
1452 if (lease_break_time
> 0) {
1453 break_time
= jiffies
+ lease_break_time
* HZ
;
1454 if (break_time
== 0)
1455 break_time
++; /* so that 0 means no break time */
1458 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1459 if (!leases_conflict(fl
, new_fl
))
1462 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1464 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1465 fl
->fl_break_time
= break_time
;
1467 if (lease_breaking(fl
))
1469 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1470 fl
->fl_downgrade_time
= break_time
;
1472 if (fl
->fl_lmops
->lm_break(fl
))
1473 locks_delete_lock_ctx(fl
, &dispose
);
1476 if (list_empty(&ctx
->flc_lease
))
1479 if (mode
& O_NONBLOCK
) {
1480 trace_break_lease_noblock(inode
, new_fl
);
1481 error
= -EWOULDBLOCK
;
1486 fl
= list_first_entry(&ctx
->flc_lease
, struct file_lock
, fl_list
);
1487 break_time
= fl
->fl_break_time
;
1488 if (break_time
!= 0)
1489 break_time
-= jiffies
;
1490 if (break_time
== 0)
1492 locks_insert_block(fl
, new_fl
);
1493 trace_break_lease_block(inode
, new_fl
);
1494 spin_unlock(&ctx
->flc_lock
);
1495 locks_dispose_list(&dispose
);
1496 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1497 !new_fl
->fl_next
, break_time
);
1498 spin_lock(&ctx
->flc_lock
);
1499 trace_break_lease_unblock(inode
, new_fl
);
1500 locks_delete_block(new_fl
);
1503 * Wait for the next conflicting lease that has not been
1507 time_out_leases(inode
, &dispose
);
1508 if (any_leases_conflict(inode
, new_fl
))
1513 spin_unlock(&ctx
->flc_lock
);
1514 locks_dispose_list(&dispose
);
1515 locks_free_lock(new_fl
);
1519 EXPORT_SYMBOL(__break_lease
);
1522 * lease_get_mtime - get the last modified time of an inode
1524 * @time: pointer to a timespec which will contain the last modified time
1526 * This is to force NFS clients to flush their caches for files with
1527 * exclusive leases. The justification is that if someone has an
1528 * exclusive lease, then they could be modifying it.
1530 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1532 bool has_lease
= false;
1533 struct file_lock_context
*ctx
;
1534 struct file_lock
*fl
;
1536 ctx
= smp_load_acquire(&inode
->i_flctx
);
1537 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1538 spin_lock(&ctx
->flc_lock
);
1539 fl
= list_first_entry_or_null(&ctx
->flc_lease
,
1540 struct file_lock
, fl_list
);
1541 if (fl
&& (fl
->fl_type
== F_WRLCK
))
1543 spin_unlock(&ctx
->flc_lock
);
1547 *time
= current_fs_time(inode
->i_sb
);
1549 *time
= inode
->i_mtime
;
1552 EXPORT_SYMBOL(lease_get_mtime
);
1555 * fcntl_getlease - Enquire what lease is currently active
1558 * The value returned by this function will be one of
1559 * (if no lease break is pending):
1561 * %F_RDLCK to indicate a shared lease is held.
1563 * %F_WRLCK to indicate an exclusive lease is held.
1565 * %F_UNLCK to indicate no lease is held.
1567 * (if a lease break is pending):
1569 * %F_RDLCK to indicate an exclusive lease needs to be
1570 * changed to a shared lease (or removed).
1572 * %F_UNLCK to indicate the lease needs to be removed.
1574 * XXX: sfr & willy disagree over whether F_INPROGRESS
1575 * should be returned to userspace.
1577 int fcntl_getlease(struct file
*filp
)
1579 struct file_lock
*fl
;
1580 struct inode
*inode
= locks_inode(filp
);
1581 struct file_lock_context
*ctx
;
1585 ctx
= smp_load_acquire(&inode
->i_flctx
);
1586 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1587 spin_lock(&ctx
->flc_lock
);
1588 time_out_leases(inode
, &dispose
);
1589 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1590 if (fl
->fl_file
!= filp
)
1592 type
= target_leasetype(fl
);
1595 spin_unlock(&ctx
->flc_lock
);
1596 locks_dispose_list(&dispose
);
1602 * check_conflicting_open - see if the given dentry points to a file that has
1603 * an existing open that would conflict with the
1605 * @dentry: dentry to check
1606 * @arg: type of lease that we're trying to acquire
1607 * @flags: current lock flags
1609 * Check to see if there's an existing open fd on this file that would
1610 * conflict with the lease we're trying to set.
1613 check_conflicting_open(const struct dentry
*dentry
, const long arg
, int flags
)
1616 struct inode
*inode
= dentry
->d_inode
;
1618 if (flags
& FL_LAYOUT
)
1621 if ((arg
== F_RDLCK
) &&
1622 (atomic_read(&d_real_inode(dentry
)->i_writecount
) > 0))
1625 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1626 (atomic_read(&inode
->i_count
) > 1)))
1633 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1635 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1636 struct dentry
*dentry
= filp
->f_path
.dentry
;
1637 struct inode
*inode
= dentry
->d_inode
;
1638 struct file_lock_context
*ctx
;
1639 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1644 trace_generic_add_lease(inode
, lease
);
1646 /* Note that arg is never F_UNLCK here */
1647 ctx
= locks_get_lock_context(inode
, arg
);
1652 * In the delegation case we need mutual exclusion with
1653 * a number of operations that take the i_mutex. We trylock
1654 * because delegations are an optional optimization, and if
1655 * there's some chance of a conflict--we'd rather not
1656 * bother, maybe that's a sign this just isn't a good file to
1657 * hand out a delegation on.
1659 if (is_deleg
&& !inode_trylock(inode
))
1662 if (is_deleg
&& arg
== F_WRLCK
) {
1663 /* Write delegations are not currently supported: */
1664 inode_unlock(inode
);
1669 spin_lock(&ctx
->flc_lock
);
1670 time_out_leases(inode
, &dispose
);
1671 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1676 * At this point, we know that if there is an exclusive
1677 * lease on this file, then we hold it on this filp
1678 * (otherwise our open of this file would have blocked).
1679 * And if we are trying to acquire an exclusive lease,
1680 * then the file is not open by anyone (including us)
1681 * except for this filp.
1684 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1685 if (fl
->fl_file
== filp
&&
1686 fl
->fl_owner
== lease
->fl_owner
) {
1692 * No exclusive leases if someone else has a lease on
1698 * Modifying our existing lease is OK, but no getting a
1699 * new lease if someone else is opening for write:
1701 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1705 if (my_fl
!= NULL
) {
1707 error
= lease
->fl_lmops
->lm_change(lease
, arg
, &dispose
);
1717 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1719 * The check in break_lease() is lockless. It's possible for another
1720 * open to race in after we did the earlier check for a conflicting
1721 * open but before the lease was inserted. Check again for a
1722 * conflicting open and cancel the lease if there is one.
1724 * We also add a barrier here to ensure that the insertion of the lock
1725 * precedes these checks.
1728 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1730 locks_unlink_lock_ctx(lease
);
1735 if (lease
->fl_lmops
->lm_setup
)
1736 lease
->fl_lmops
->lm_setup(lease
, priv
);
1738 spin_unlock(&ctx
->flc_lock
);
1739 locks_dispose_list(&dispose
);
1741 inode_unlock(inode
);
1742 if (!error
&& !my_fl
)
1747 static int generic_delete_lease(struct file
*filp
, void *owner
)
1749 int error
= -EAGAIN
;
1750 struct file_lock
*fl
, *victim
= NULL
;
1751 struct inode
*inode
= locks_inode(filp
);
1752 struct file_lock_context
*ctx
;
1755 ctx
= smp_load_acquire(&inode
->i_flctx
);
1757 trace_generic_delete_lease(inode
, NULL
);
1761 spin_lock(&ctx
->flc_lock
);
1762 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1763 if (fl
->fl_file
== filp
&&
1764 fl
->fl_owner
== owner
) {
1769 trace_generic_delete_lease(inode
, victim
);
1771 error
= fl
->fl_lmops
->lm_change(victim
, F_UNLCK
, &dispose
);
1772 spin_unlock(&ctx
->flc_lock
);
1773 locks_dispose_list(&dispose
);
1778 * generic_setlease - sets a lease on an open file
1779 * @filp: file pointer
1780 * @arg: type of lease to obtain
1781 * @flp: input - file_lock to use, output - file_lock inserted
1782 * @priv: private data for lm_setup (may be NULL if lm_setup
1783 * doesn't require it)
1785 * The (input) flp->fl_lmops->lm_break function is required
1788 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1791 struct inode
*inode
= locks_inode(filp
);
1794 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1796 if (!S_ISREG(inode
->i_mode
))
1798 error
= security_file_lock(filp
, arg
);
1804 return generic_delete_lease(filp
, *priv
);
1807 if (!(*flp
)->fl_lmops
->lm_break
) {
1812 return generic_add_lease(filp
, arg
, flp
, priv
);
1817 EXPORT_SYMBOL(generic_setlease
);
1820 * vfs_setlease - sets a lease on an open file
1821 * @filp: file pointer
1822 * @arg: type of lease to obtain
1823 * @lease: file_lock to use when adding a lease
1824 * @priv: private info for lm_setup when adding a lease (may be
1825 * NULL if lm_setup doesn't require it)
1827 * Call this to establish a lease on the file. The "lease" argument is not
1828 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1829 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1830 * if not, this function will return -ENOLCK (and generate a scary-looking
1833 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1834 * may be NULL if the lm_setup operation doesn't require it.
1837 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1839 if (filp
->f_op
->setlease
&& is_remote_lock(filp
))
1840 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1842 return generic_setlease(filp
, arg
, lease
, priv
);
1844 EXPORT_SYMBOL_GPL(vfs_setlease
);
1846 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1848 struct file_lock
*fl
;
1849 struct fasync_struct
*new;
1852 fl
= lease_alloc(filp
, arg
);
1856 new = fasync_alloc();
1858 locks_free_lock(fl
);
1863 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1865 locks_free_lock(fl
);
1872 * fcntl_setlease - sets a lease on an open file
1873 * @fd: open file descriptor
1874 * @filp: file pointer
1875 * @arg: type of lease to obtain
1877 * Call this fcntl to establish a lease on the file.
1878 * Note that you also need to call %F_SETSIG to
1879 * receive a signal when the lease is broken.
1881 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1884 return vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&filp
);
1885 return do_fcntl_add_lease(fd
, filp
, arg
);
1889 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1890 * @inode: inode of the file to apply to
1891 * @fl: The lock to be applied
1893 * Apply a FLOCK style lock request to an inode.
1895 static int flock_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1900 error
= flock_lock_inode(inode
, fl
);
1901 if (error
!= FILE_LOCK_DEFERRED
)
1903 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1907 locks_delete_block(fl
);
1914 * locks_lock_inode_wait - Apply a lock to an inode
1915 * @inode: inode of the file to apply to
1916 * @fl: The lock to be applied
1918 * Apply a POSIX or FLOCK style lock request to an inode.
1920 int locks_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1923 switch (fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)) {
1925 res
= posix_lock_inode_wait(inode
, fl
);
1928 res
= flock_lock_inode_wait(inode
, fl
);
1935 EXPORT_SYMBOL(locks_lock_inode_wait
);
1938 * sys_flock: - flock() system call.
1939 * @fd: the file descriptor to lock.
1940 * @cmd: the type of lock to apply.
1942 * Apply a %FL_FLOCK style lock to an open file descriptor.
1943 * The @cmd can be one of
1945 * %LOCK_SH -- a shared lock.
1947 * %LOCK_EX -- an exclusive lock.
1949 * %LOCK_UN -- remove an existing lock.
1951 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1953 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1954 * processes read and write access respectively.
1956 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1958 struct fd f
= fdget(fd
);
1959 struct file_lock
*lock
;
1960 int can_sleep
, unlock
;
1967 can_sleep
= !(cmd
& LOCK_NB
);
1969 unlock
= (cmd
== LOCK_UN
);
1971 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1972 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1975 lock
= flock_make_lock(f
.file
, cmd
);
1977 error
= PTR_ERR(lock
);
1982 lock
->fl_flags
|= FL_SLEEP
;
1984 error
= security_file_lock(f
.file
, lock
->fl_type
);
1988 if (f
.file
->f_op
->flock
&& is_remote_lock(f
.file
))
1989 error
= f
.file
->f_op
->flock(f
.file
,
1990 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1993 error
= locks_lock_file_wait(f
.file
, lock
);
1996 locks_free_lock(lock
);
2005 * vfs_test_lock - test file byte range lock
2006 * @filp: The file to test lock for
2007 * @fl: The lock to test; also used to hold result
2009 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2010 * setting conf->fl_type to something other than F_UNLCK.
2012 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
2014 if (filp
->f_op
->lock
&& is_remote_lock(filp
))
2015 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
2016 posix_test_lock(filp
, fl
);
2019 EXPORT_SYMBOL_GPL(vfs_test_lock
);
2021 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
2023 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2024 #if BITS_PER_LONG == 32
2026 * Make sure we can represent the posix lock via
2027 * legacy 32bit flock.
2029 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
2031 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
2034 flock
->l_start
= fl
->fl_start
;
2035 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2036 fl
->fl_end
- fl
->fl_start
+ 1;
2037 flock
->l_whence
= 0;
2038 flock
->l_type
= fl
->fl_type
;
2042 #if BITS_PER_LONG == 32
2043 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
2045 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2046 flock
->l_start
= fl
->fl_start
;
2047 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2048 fl
->fl_end
- fl
->fl_start
+ 1;
2049 flock
->l_whence
= 0;
2050 flock
->l_type
= fl
->fl_type
;
2054 /* Report the first existing lock that would conflict with l.
2055 * This implements the F_GETLK command of fcntl().
2057 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2059 struct file_lock file_lock
;
2064 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2067 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2070 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2074 if (cmd
== F_OFD_GETLK
) {
2076 if (flock
.l_pid
!= 0)
2080 file_lock
.fl_flags
|= FL_OFDLCK
;
2081 file_lock
.fl_owner
= filp
;
2084 error
= vfs_test_lock(filp
, &file_lock
);
2088 flock
.l_type
= file_lock
.fl_type
;
2089 if (file_lock
.fl_type
!= F_UNLCK
) {
2090 error
= posix_lock_to_flock(&flock
, &file_lock
);
2095 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2098 locks_release_private(&file_lock
);
2104 * vfs_lock_file - file byte range lock
2105 * @filp: The file to apply the lock to
2106 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2107 * @fl: The lock to be applied
2108 * @conf: Place to return a copy of the conflicting lock, if found.
2110 * A caller that doesn't care about the conflicting lock may pass NULL
2111 * as the final argument.
2113 * If the filesystem defines a private ->lock() method, then @conf will
2114 * be left unchanged; so a caller that cares should initialize it to
2115 * some acceptable default.
2117 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2118 * locks, the ->lock() interface may return asynchronously, before the lock has
2119 * been granted or denied by the underlying filesystem, if (and only if)
2120 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2121 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2122 * the request is for a blocking lock. When ->lock() does return asynchronously,
2123 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2124 * request completes.
2125 * If the request is for non-blocking lock the file system should return
2126 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2127 * with the result. If the request timed out the callback routine will return a
2128 * nonzero return code and the file system should release the lock. The file
2129 * system is also responsible to keep a corresponding posix lock when it
2130 * grants a lock so the VFS can find out which locks are locally held and do
2131 * the correct lock cleanup when required.
2132 * The underlying filesystem must not drop the kernel lock or call
2133 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2136 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2138 if (filp
->f_op
->lock
&& is_remote_lock(filp
))
2139 return filp
->f_op
->lock(filp
, cmd
, fl
);
2141 return posix_lock_file(filp
, fl
, conf
);
2143 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2145 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2146 struct file_lock
*fl
)
2150 error
= security_file_lock(filp
, fl
->fl_type
);
2155 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2156 if (error
!= FILE_LOCK_DEFERRED
)
2158 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2162 locks_delete_block(fl
);
2169 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2171 check_fmode_for_setlk(struct file_lock
*fl
)
2173 switch (fl
->fl_type
) {
2175 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2179 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2185 /* Apply the lock described by l to an open file descriptor.
2186 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2188 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2189 struct flock __user
*l
)
2191 struct file_lock
*file_lock
= locks_alloc_lock();
2193 struct inode
*inode
;
2197 if (file_lock
== NULL
)
2200 inode
= locks_inode(filp
);
2203 * This might block, so we do it before checking the inode.
2206 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2209 /* Don't allow mandatory locks on files that may be memory mapped
2212 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2217 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2221 error
= check_fmode_for_setlk(file_lock
);
2226 * If the cmd is requesting file-private locks, then set the
2227 * FL_OFDLCK flag and override the owner.
2232 if (flock
.l_pid
!= 0)
2236 file_lock
->fl_flags
|= FL_OFDLCK
;
2237 file_lock
->fl_owner
= filp
;
2241 if (flock
.l_pid
!= 0)
2245 file_lock
->fl_flags
|= FL_OFDLCK
;
2246 file_lock
->fl_owner
= filp
;
2249 file_lock
->fl_flags
|= FL_SLEEP
;
2252 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2255 * Attempt to detect a close/fcntl race and recover by releasing the
2256 * lock that was just acquired. There is no need to do that when we're
2257 * unlocking though, or for OFD locks.
2259 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2260 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2262 * We need that spin_lock here - it prevents reordering between
2263 * update of i_flctx->flc_posix and check for it done in
2264 * close(). rcu_read_lock() wouldn't do.
2266 spin_lock(¤t
->files
->file_lock
);
2268 spin_unlock(¤t
->files
->file_lock
);
2270 file_lock
->fl_type
= F_UNLCK
;
2271 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2272 WARN_ON_ONCE(error
);
2277 trace_fcntl_setlk(inode
, file_lock
, error
);
2278 locks_free_lock(file_lock
);
2282 #if BITS_PER_LONG == 32
2283 /* Report the first existing lock that would conflict with l.
2284 * This implements the F_GETLK command of fcntl().
2286 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2288 struct file_lock file_lock
;
2289 struct flock64 flock
;
2293 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2296 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2299 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2303 if (cmd
== F_OFD_GETLK
) {
2305 if (flock
.l_pid
!= 0)
2309 file_lock
.fl_flags
|= FL_OFDLCK
;
2310 file_lock
.fl_owner
= filp
;
2313 error
= vfs_test_lock(filp
, &file_lock
);
2317 flock
.l_type
= file_lock
.fl_type
;
2318 if (file_lock
.fl_type
!= F_UNLCK
)
2319 posix_lock_to_flock64(&flock
, &file_lock
);
2322 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2325 locks_release_private(&file_lock
);
2330 /* Apply the lock described by l to an open file descriptor.
2331 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2333 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2334 struct flock64 __user
*l
)
2336 struct file_lock
*file_lock
= locks_alloc_lock();
2337 struct flock64 flock
;
2338 struct inode
*inode
;
2342 if (file_lock
== NULL
)
2346 * This might block, so we do it before checking the inode.
2349 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2352 inode
= locks_inode(filp
);
2354 /* Don't allow mandatory locks on files that may be memory mapped
2357 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2362 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2366 error
= check_fmode_for_setlk(file_lock
);
2371 * If the cmd is requesting file-private locks, then set the
2372 * FL_OFDLCK flag and override the owner.
2377 if (flock
.l_pid
!= 0)
2381 file_lock
->fl_flags
|= FL_OFDLCK
;
2382 file_lock
->fl_owner
= filp
;
2386 if (flock
.l_pid
!= 0)
2390 file_lock
->fl_flags
|= FL_OFDLCK
;
2391 file_lock
->fl_owner
= filp
;
2394 file_lock
->fl_flags
|= FL_SLEEP
;
2397 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2400 * Attempt to detect a close/fcntl race and recover by releasing the
2401 * lock that was just acquired. There is no need to do that when we're
2402 * unlocking though, or for OFD locks.
2404 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2405 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2407 * We need that spin_lock here - it prevents reordering between
2408 * update of i_flctx->flc_posix and check for it done in
2409 * close(). rcu_read_lock() wouldn't do.
2411 spin_lock(¤t
->files
->file_lock
);
2413 spin_unlock(¤t
->files
->file_lock
);
2415 file_lock
->fl_type
= F_UNLCK
;
2416 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2417 WARN_ON_ONCE(error
);
2422 locks_free_lock(file_lock
);
2425 #endif /* BITS_PER_LONG == 32 */
2428 * This function is called when the file is being removed
2429 * from the task's fd array. POSIX locks belonging to this task
2430 * are deleted at this time.
2432 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2435 struct inode
*inode
= locks_inode(filp
);
2436 struct file_lock lock
;
2437 struct file_lock_context
*ctx
;
2440 * If there are no locks held on this file, we don't need to call
2441 * posix_lock_file(). Another process could be setting a lock on this
2442 * file at the same time, but we wouldn't remove that lock anyway.
2444 ctx
= smp_load_acquire(&inode
->i_flctx
);
2445 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2448 lock
.fl_type
= F_UNLCK
;
2449 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2451 lock
.fl_end
= OFFSET_MAX
;
2452 lock
.fl_owner
= owner
;
2453 lock
.fl_pid
= current
->tgid
;
2454 lock
.fl_file
= filp
;
2456 lock
.fl_lmops
= NULL
;
2458 error
= vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2460 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2461 lock
.fl_ops
->fl_release_private(&lock
);
2462 trace_locks_remove_posix(inode
, &lock
, error
);
2465 EXPORT_SYMBOL(locks_remove_posix
);
2467 /* The i_flctx must be valid when calling into here */
2469 locks_remove_flock(struct file
*filp
, struct file_lock_context
*flctx
)
2471 struct file_lock fl
= {
2473 .fl_pid
= current
->tgid
,
2475 .fl_flags
= FL_FLOCK
,
2477 .fl_end
= OFFSET_MAX
,
2479 struct inode
*inode
= locks_inode(filp
);
2481 if (list_empty(&flctx
->flc_flock
))
2484 if (filp
->f_op
->flock
&& is_remote_lock(filp
))
2485 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2487 flock_lock_inode(inode
, &fl
);
2489 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2490 fl
.fl_ops
->fl_release_private(&fl
);
2493 /* The i_flctx must be valid when calling into here */
2495 locks_remove_lease(struct file
*filp
, struct file_lock_context
*ctx
)
2497 struct file_lock
*fl
, *tmp
;
2500 if (list_empty(&ctx
->flc_lease
))
2503 spin_lock(&ctx
->flc_lock
);
2504 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2505 if (filp
== fl
->fl_file
)
2506 lease_modify(fl
, F_UNLCK
, &dispose
);
2507 spin_unlock(&ctx
->flc_lock
);
2508 locks_dispose_list(&dispose
);
2512 * This function is called on the last close of an open file.
2514 void locks_remove_file(struct file
*filp
)
2516 struct file_lock_context
*ctx
;
2518 ctx
= smp_load_acquire(&locks_inode(filp
)->i_flctx
);
2522 /* remove any OFD locks */
2523 locks_remove_posix(filp
, filp
);
2525 /* remove flock locks */
2526 locks_remove_flock(filp
, ctx
);
2528 /* remove any leases */
2529 locks_remove_lease(filp
, ctx
);
2533 * posix_unblock_lock - stop waiting for a file lock
2534 * @waiter: the lock which was waiting
2536 * lockd needs to block waiting for locks.
2539 posix_unblock_lock(struct file_lock
*waiter
)
2543 spin_lock(&blocked_lock_lock
);
2544 if (waiter
->fl_next
)
2545 __locks_delete_block(waiter
);
2548 spin_unlock(&blocked_lock_lock
);
2551 EXPORT_SYMBOL(posix_unblock_lock
);
2554 * vfs_cancel_lock - file byte range unblock lock
2555 * @filp: The file to apply the unblock to
2556 * @fl: The lock to be unblocked
2558 * Used by lock managers to cancel blocked requests
2560 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2562 if (filp
->f_op
->lock
&& is_remote_lock(filp
))
2563 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2567 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2569 #ifdef CONFIG_PROC_FS
2570 #include <linux/proc_fs.h>
2571 #include <linux/seq_file.h>
2573 struct locks_iterator
{
2578 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2579 loff_t id
, char *pfx
)
2581 struct inode
*inode
= NULL
;
2582 unsigned int fl_pid
;
2585 struct pid_namespace
*proc_pidns
= file_inode(f
->file
)->i_sb
->s_fs_info
;
2587 /* Don't let fl_pid change based on who is reading the file */
2588 fl_pid
= pid_nr_ns(fl
->fl_nspid
, proc_pidns
);
2591 * If there isn't a fl_pid don't display who is waiting on
2592 * the lock if we are called from locks_show, or if we are
2593 * called from __show_fd_info - skip lock entirely
2598 fl_pid
= fl
->fl_pid
;
2600 if (fl
->fl_file
!= NULL
)
2601 inode
= locks_inode(fl
->fl_file
);
2603 seq_printf(f
, "%lld:%s ", id
, pfx
);
2605 if (fl
->fl_flags
& FL_ACCESS
)
2606 seq_puts(f
, "ACCESS");
2607 else if (IS_OFDLCK(fl
))
2608 seq_puts(f
, "OFDLCK");
2610 seq_puts(f
, "POSIX ");
2612 seq_printf(f
, " %s ",
2613 (inode
== NULL
) ? "*NOINODE*" :
2614 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2615 } else if (IS_FLOCK(fl
)) {
2616 if (fl
->fl_type
& LOCK_MAND
) {
2617 seq_puts(f
, "FLOCK MSNFS ");
2619 seq_puts(f
, "FLOCK ADVISORY ");
2621 } else if (IS_LEASE(fl
)) {
2622 if (fl
->fl_flags
& FL_DELEG
)
2623 seq_puts(f
, "DELEG ");
2625 seq_puts(f
, "LEASE ");
2627 if (lease_breaking(fl
))
2628 seq_puts(f
, "BREAKING ");
2629 else if (fl
->fl_file
)
2630 seq_puts(f
, "ACTIVE ");
2632 seq_puts(f
, "BREAKER ");
2634 seq_puts(f
, "UNKNOWN UNKNOWN ");
2636 if (fl
->fl_type
& LOCK_MAND
) {
2637 seq_printf(f
, "%s ",
2638 (fl
->fl_type
& LOCK_READ
)
2639 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2640 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2642 seq_printf(f
, "%s ",
2643 (lease_breaking(fl
))
2644 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2645 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2648 /* userspace relies on this representation of dev_t */
2649 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2650 MAJOR(inode
->i_sb
->s_dev
),
2651 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2653 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2656 if (fl
->fl_end
== OFFSET_MAX
)
2657 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2659 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2661 seq_puts(f
, "0 EOF\n");
2665 static int locks_show(struct seq_file
*f
, void *v
)
2667 struct locks_iterator
*iter
= f
->private;
2668 struct file_lock
*fl
, *bfl
;
2669 struct pid_namespace
*proc_pidns
= file_inode(f
->file
)->i_sb
->s_fs_info
;
2671 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2673 if (fl
->fl_nspid
&& !pid_nr_ns(fl
->fl_nspid
, proc_pidns
))
2676 lock_get_status(f
, fl
, iter
->li_pos
, "");
2678 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2679 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2684 static void __show_fd_locks(struct seq_file
*f
,
2685 struct list_head
*head
, int *id
,
2686 struct file
*filp
, struct files_struct
*files
)
2688 struct file_lock
*fl
;
2690 list_for_each_entry(fl
, head
, fl_list
) {
2692 if (filp
!= fl
->fl_file
)
2694 if (fl
->fl_owner
!= files
&&
2695 fl
->fl_owner
!= filp
)
2699 seq_puts(f
, "lock:\t");
2700 lock_get_status(f
, fl
, *id
, "");
2704 void show_fd_locks(struct seq_file
*f
,
2705 struct file
*filp
, struct files_struct
*files
)
2707 struct inode
*inode
= locks_inode(filp
);
2708 struct file_lock_context
*ctx
;
2711 ctx
= smp_load_acquire(&inode
->i_flctx
);
2715 spin_lock(&ctx
->flc_lock
);
2716 __show_fd_locks(f
, &ctx
->flc_flock
, &id
, filp
, files
);
2717 __show_fd_locks(f
, &ctx
->flc_posix
, &id
, filp
, files
);
2718 __show_fd_locks(f
, &ctx
->flc_lease
, &id
, filp
, files
);
2719 spin_unlock(&ctx
->flc_lock
);
2722 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2723 __acquires(&blocked_lock_lock
)
2725 struct locks_iterator
*iter
= f
->private;
2727 iter
->li_pos
= *pos
+ 1;
2728 lg_global_lock(&file_lock_lglock
);
2729 spin_lock(&blocked_lock_lock
);
2730 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2733 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2735 struct locks_iterator
*iter
= f
->private;
2738 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2741 static void locks_stop(struct seq_file
*f
, void *v
)
2742 __releases(&blocked_lock_lock
)
2744 spin_unlock(&blocked_lock_lock
);
2745 lg_global_unlock(&file_lock_lglock
);
2748 static const struct seq_operations locks_seq_operations
= {
2749 .start
= locks_start
,
2755 static int locks_open(struct inode
*inode
, struct file
*filp
)
2757 return seq_open_private(filp
, &locks_seq_operations
,
2758 sizeof(struct locks_iterator
));
2761 static const struct file_operations proc_locks_operations
= {
2764 .llseek
= seq_lseek
,
2765 .release
= seq_release_private
,
2768 static int __init
proc_locks_init(void)
2770 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2773 fs_initcall(proc_locks_init
);
2776 static int __init
filelock_init(void)
2780 flctx_cache
= kmem_cache_create("file_lock_ctx",
2781 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2783 filelock_cache
= kmem_cache_create("file_lock_cache",
2784 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2786 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2788 for_each_possible_cpu(i
)
2789 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2794 core_initcall(filelock_init
);