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>
131 #define CREATE_TRACE_POINTS
132 #include <trace/events/filelock.h>
134 #include <asm/uaccess.h>
136 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
137 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
138 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG|FL_LAYOUT))
139 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
141 static bool lease_breaking(struct file_lock
*fl
)
143 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
146 static int target_leasetype(struct file_lock
*fl
)
148 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
150 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
155 int leases_enable
= 1;
156 int lease_break_time
= 45;
159 * The global file_lock_list is only used for displaying /proc/locks, so we
160 * keep a list on each CPU, with each list protected by its own spinlock.
161 * Global serialization is done using file_rwsem.
163 * Note that alterations to the list also require that the relevant flc_lock is
166 struct file_lock_list_struct
{
168 struct hlist_head hlist
;
170 static DEFINE_PER_CPU(struct file_lock_list_struct
, file_lock_list
);
171 DEFINE_STATIC_PERCPU_RWSEM(file_rwsem
);
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 struct file_lock_list_struct
*fll
= this_cpu_ptr(&file_lock_list
);
597 percpu_rwsem_assert_held(&file_rwsem
);
599 spin_lock(&fll
->lock
);
600 fl
->fl_link_cpu
= smp_processor_id();
601 hlist_add_head(&fl
->fl_link
, &fll
->hlist
);
602 spin_unlock(&fll
->lock
);
605 /* Must be called with the flc_lock held! */
606 static void locks_delete_global_locks(struct file_lock
*fl
)
608 struct file_lock_list_struct
*fll
;
610 percpu_rwsem_assert_held(&file_rwsem
);
613 * Avoid taking lock if already unhashed. This is safe since this check
614 * is done while holding the flc_lock, and new insertions into the list
615 * also require that it be held.
617 if (hlist_unhashed(&fl
->fl_link
))
620 fll
= per_cpu_ptr(&file_lock_list
, fl
->fl_link_cpu
);
621 spin_lock(&fll
->lock
);
622 hlist_del_init(&fl
->fl_link
);
623 spin_unlock(&fll
->lock
);
627 posix_owner_key(struct file_lock
*fl
)
629 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
630 return fl
->fl_lmops
->lm_owner_key(fl
);
631 return (unsigned long)fl
->fl_owner
;
634 static void locks_insert_global_blocked(struct file_lock
*waiter
)
636 lockdep_assert_held(&blocked_lock_lock
);
638 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
641 static void locks_delete_global_blocked(struct file_lock
*waiter
)
643 lockdep_assert_held(&blocked_lock_lock
);
645 hash_del(&waiter
->fl_link
);
648 /* Remove waiter from blocker's block list.
649 * When blocker ends up pointing to itself then the list is empty.
651 * Must be called with blocked_lock_lock held.
653 static void __locks_delete_block(struct file_lock
*waiter
)
655 locks_delete_global_blocked(waiter
);
656 list_del_init(&waiter
->fl_block
);
657 waiter
->fl_next
= NULL
;
660 static void locks_delete_block(struct file_lock
*waiter
)
662 spin_lock(&blocked_lock_lock
);
663 __locks_delete_block(waiter
);
664 spin_unlock(&blocked_lock_lock
);
667 /* Insert waiter into blocker's block list.
668 * We use a circular list so that processes can be easily woken up in
669 * the order they blocked. The documentation doesn't require this but
670 * it seems like the reasonable thing to do.
672 * Must be called with both the flc_lock and blocked_lock_lock held. The
673 * fl_block list itself is protected by the blocked_lock_lock, but by ensuring
674 * that the flc_lock is also held on insertions we can avoid taking the
675 * blocked_lock_lock in some cases when we see that the fl_block list is empty.
677 static void __locks_insert_block(struct file_lock
*blocker
,
678 struct file_lock
*waiter
)
680 BUG_ON(!list_empty(&waiter
->fl_block
));
681 waiter
->fl_next
= blocker
;
682 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
683 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
684 locks_insert_global_blocked(waiter
);
687 /* Must be called with flc_lock held. */
688 static void locks_insert_block(struct file_lock
*blocker
,
689 struct file_lock
*waiter
)
691 spin_lock(&blocked_lock_lock
);
692 __locks_insert_block(blocker
, waiter
);
693 spin_unlock(&blocked_lock_lock
);
697 * Wake up processes blocked waiting for blocker.
699 * Must be called with the inode->flc_lock held!
701 static void locks_wake_up_blocks(struct file_lock
*blocker
)
704 * Avoid taking global lock if list is empty. This is safe since new
705 * blocked requests are only added to the list under the flc_lock, and
706 * the flc_lock is always held here. Note that removal from the fl_block
707 * list does not require the flc_lock, so we must recheck list_empty()
708 * after acquiring the blocked_lock_lock.
710 if (list_empty(&blocker
->fl_block
))
713 spin_lock(&blocked_lock_lock
);
714 while (!list_empty(&blocker
->fl_block
)) {
715 struct file_lock
*waiter
;
717 waiter
= list_first_entry(&blocker
->fl_block
,
718 struct file_lock
, fl_block
);
719 __locks_delete_block(waiter
);
720 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
721 waiter
->fl_lmops
->lm_notify(waiter
);
723 wake_up(&waiter
->fl_wait
);
725 spin_unlock(&blocked_lock_lock
);
729 locks_insert_lock_ctx(struct file_lock
*fl
, struct list_head
*before
)
731 fl
->fl_nspid
= get_pid(task_tgid(current
));
732 list_add_tail(&fl
->fl_list
, before
);
733 locks_insert_global_locks(fl
);
737 locks_unlink_lock_ctx(struct file_lock
*fl
)
739 locks_delete_global_locks(fl
);
740 list_del_init(&fl
->fl_list
);
742 put_pid(fl
->fl_nspid
);
745 locks_wake_up_blocks(fl
);
749 locks_delete_lock_ctx(struct file_lock
*fl
, struct list_head
*dispose
)
751 locks_unlink_lock_ctx(fl
);
753 list_add(&fl
->fl_list
, dispose
);
758 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
759 * checks for shared/exclusive status of overlapping locks.
761 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
763 if (sys_fl
->fl_type
== F_WRLCK
)
765 if (caller_fl
->fl_type
== F_WRLCK
)
770 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
771 * checking before calling the locks_conflict().
773 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
775 /* POSIX locks owned by the same process do not conflict with
778 if (posix_same_owner(caller_fl
, sys_fl
))
781 /* Check whether they overlap */
782 if (!locks_overlap(caller_fl
, sys_fl
))
785 return (locks_conflict(caller_fl
, sys_fl
));
788 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
789 * checking before calling the locks_conflict().
791 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
793 /* FLOCK locks referring to the same filp do not conflict with
796 if (caller_fl
->fl_file
== sys_fl
->fl_file
)
798 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
801 return (locks_conflict(caller_fl
, sys_fl
));
805 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
807 struct file_lock
*cfl
;
808 struct file_lock_context
*ctx
;
809 struct inode
*inode
= file_inode(filp
);
811 ctx
= smp_load_acquire(&inode
->i_flctx
);
812 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
)) {
813 fl
->fl_type
= F_UNLCK
;
817 spin_lock(&ctx
->flc_lock
);
818 list_for_each_entry(cfl
, &ctx
->flc_posix
, fl_list
) {
819 if (posix_locks_conflict(fl
, cfl
)) {
820 locks_copy_conflock(fl
, cfl
);
822 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
826 fl
->fl_type
= F_UNLCK
;
828 spin_unlock(&ctx
->flc_lock
);
831 EXPORT_SYMBOL(posix_test_lock
);
834 * Deadlock detection:
836 * We attempt to detect deadlocks that are due purely to posix file
839 * We assume that a task can be waiting for at most one lock at a time.
840 * So for any acquired lock, the process holding that lock may be
841 * waiting on at most one other lock. That lock in turns may be held by
842 * someone waiting for at most one other lock. Given a requested lock
843 * caller_fl which is about to wait for a conflicting lock block_fl, we
844 * follow this chain of waiters to ensure we are not about to create a
847 * Since we do this before we ever put a process to sleep on a lock, we
848 * are ensured that there is never a cycle; that is what guarantees that
849 * the while() loop in posix_locks_deadlock() eventually completes.
851 * Note: the above assumption may not be true when handling lock
852 * requests from a broken NFS client. It may also fail in the presence
853 * of tasks (such as posix threads) sharing the same open file table.
854 * To handle those cases, we just bail out after a few iterations.
856 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
857 * Because the owner is not even nominally tied to a thread of
858 * execution, the deadlock detection below can't reasonably work well. Just
861 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
862 * locks that just checks for the case where two tasks are attempting to
863 * upgrade from read to write locks on the same inode.
866 #define MAX_DEADLK_ITERATIONS 10
868 /* Find a lock that the owner of the given block_fl is blocking on. */
869 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
871 struct file_lock
*fl
;
873 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
874 if (posix_same_owner(fl
, block_fl
))
880 /* Must be called with the blocked_lock_lock held! */
881 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
882 struct file_lock
*block_fl
)
886 lockdep_assert_held(&blocked_lock_lock
);
889 * This deadlock detector can't reasonably detect deadlocks with
890 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
892 if (IS_OFDLCK(caller_fl
))
895 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
896 if (i
++ > MAX_DEADLK_ITERATIONS
)
898 if (posix_same_owner(caller_fl
, block_fl
))
904 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
905 * after any leases, but before any posix locks.
907 * Note that if called with an FL_EXISTS argument, the caller may determine
908 * whether or not a lock was successfully freed by testing the return
911 static int flock_lock_inode(struct inode
*inode
, struct file_lock
*request
)
913 struct file_lock
*new_fl
= NULL
;
914 struct file_lock
*fl
;
915 struct file_lock_context
*ctx
;
920 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
922 if (request
->fl_type
!= F_UNLCK
)
924 return (request
->fl_flags
& FL_EXISTS
) ? -ENOENT
: 0;
927 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
928 new_fl
= locks_alloc_lock();
933 percpu_down_read_preempt_disable(&file_rwsem
);
934 spin_lock(&ctx
->flc_lock
);
935 if (request
->fl_flags
& FL_ACCESS
)
938 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
939 if (request
->fl_file
!= fl
->fl_file
)
941 if (request
->fl_type
== fl
->fl_type
)
944 locks_delete_lock_ctx(fl
, &dispose
);
948 if (request
->fl_type
== F_UNLCK
) {
949 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
955 list_for_each_entry(fl
, &ctx
->flc_flock
, fl_list
) {
956 if (!flock_locks_conflict(request
, fl
))
959 if (!(request
->fl_flags
& FL_SLEEP
))
961 error
= FILE_LOCK_DEFERRED
;
962 locks_insert_block(fl
, request
);
965 if (request
->fl_flags
& FL_ACCESS
)
967 locks_copy_lock(new_fl
, request
);
968 locks_insert_lock_ctx(new_fl
, &ctx
->flc_flock
);
973 spin_unlock(&ctx
->flc_lock
);
974 percpu_up_read_preempt_enable(&file_rwsem
);
976 locks_free_lock(new_fl
);
977 locks_dispose_list(&dispose
);
981 static int posix_lock_inode(struct inode
*inode
, struct file_lock
*request
,
982 struct file_lock
*conflock
)
984 struct file_lock
*fl
, *tmp
;
985 struct file_lock
*new_fl
= NULL
;
986 struct file_lock
*new_fl2
= NULL
;
987 struct file_lock
*left
= NULL
;
988 struct file_lock
*right
= NULL
;
989 struct file_lock_context
*ctx
;
994 ctx
= locks_get_lock_context(inode
, request
->fl_type
);
996 return (request
->fl_type
== F_UNLCK
) ? 0 : -ENOMEM
;
999 * We may need two file_lock structures for this operation,
1000 * so we get them in advance to avoid races.
1002 * In some cases we can be sure, that no new locks will be needed
1004 if (!(request
->fl_flags
& FL_ACCESS
) &&
1005 (request
->fl_type
!= F_UNLCK
||
1006 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
1007 new_fl
= locks_alloc_lock();
1008 new_fl2
= locks_alloc_lock();
1011 percpu_down_read_preempt_disable(&file_rwsem
);
1012 spin_lock(&ctx
->flc_lock
);
1014 * New lock request. Walk all POSIX locks and look for conflicts. If
1015 * there are any, either return error or put the request on the
1016 * blocker's list of waiters and the global blocked_hash.
1018 if (request
->fl_type
!= F_UNLCK
) {
1019 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1020 if (!posix_locks_conflict(request
, fl
))
1023 locks_copy_conflock(conflock
, fl
);
1025 if (!(request
->fl_flags
& FL_SLEEP
))
1028 * Deadlock detection and insertion into the blocked
1029 * locks list must be done while holding the same lock!
1032 spin_lock(&blocked_lock_lock
);
1033 if (likely(!posix_locks_deadlock(request
, fl
))) {
1034 error
= FILE_LOCK_DEFERRED
;
1035 __locks_insert_block(fl
, request
);
1037 spin_unlock(&blocked_lock_lock
);
1042 /* If we're just looking for a conflict, we're done. */
1044 if (request
->fl_flags
& FL_ACCESS
)
1047 /* Find the first old lock with the same owner as the new lock */
1048 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1049 if (posix_same_owner(request
, fl
))
1053 /* Process locks with this owner. */
1054 list_for_each_entry_safe_from(fl
, tmp
, &ctx
->flc_posix
, fl_list
) {
1055 if (!posix_same_owner(request
, fl
))
1058 /* Detect adjacent or overlapping regions (if same lock type) */
1059 if (request
->fl_type
== fl
->fl_type
) {
1060 /* In all comparisons of start vs end, use
1061 * "start - 1" rather than "end + 1". If end
1062 * is OFFSET_MAX, end + 1 will become negative.
1064 if (fl
->fl_end
< request
->fl_start
- 1)
1066 /* If the next lock in the list has entirely bigger
1067 * addresses than the new one, insert the lock here.
1069 if (fl
->fl_start
- 1 > request
->fl_end
)
1072 /* If we come here, the new and old lock are of the
1073 * same type and adjacent or overlapping. Make one
1074 * lock yielding from the lower start address of both
1075 * locks to the higher end address.
1077 if (fl
->fl_start
> request
->fl_start
)
1078 fl
->fl_start
= request
->fl_start
;
1080 request
->fl_start
= fl
->fl_start
;
1081 if (fl
->fl_end
< request
->fl_end
)
1082 fl
->fl_end
= request
->fl_end
;
1084 request
->fl_end
= fl
->fl_end
;
1086 locks_delete_lock_ctx(fl
, &dispose
);
1092 /* Processing for different lock types is a bit
1095 if (fl
->fl_end
< request
->fl_start
)
1097 if (fl
->fl_start
> request
->fl_end
)
1099 if (request
->fl_type
== F_UNLCK
)
1101 if (fl
->fl_start
< request
->fl_start
)
1103 /* If the next lock in the list has a higher end
1104 * address than the new one, insert the new one here.
1106 if (fl
->fl_end
> request
->fl_end
) {
1110 if (fl
->fl_start
>= request
->fl_start
) {
1111 /* The new lock completely replaces an old
1112 * one (This may happen several times).
1115 locks_delete_lock_ctx(fl
, &dispose
);
1119 * Replace the old lock with new_fl, and
1120 * remove the old one. It's safe to do the
1121 * insert here since we know that we won't be
1122 * using new_fl later, and that the lock is
1123 * just replacing an existing lock.
1128 locks_copy_lock(new_fl
, request
);
1131 locks_insert_lock_ctx(request
, &fl
->fl_list
);
1132 locks_delete_lock_ctx(fl
, &dispose
);
1139 * The above code only modifies existing locks in case of merging or
1140 * replacing. If new lock(s) need to be inserted all modifications are
1141 * done below this, so it's safe yet to bail out.
1143 error
= -ENOLCK
; /* "no luck" */
1144 if (right
&& left
== right
&& !new_fl2
)
1149 if (request
->fl_type
== F_UNLCK
) {
1150 if (request
->fl_flags
& FL_EXISTS
)
1159 locks_copy_lock(new_fl
, request
);
1160 locks_insert_lock_ctx(new_fl
, &fl
->fl_list
);
1165 if (left
== right
) {
1166 /* The new lock breaks the old one in two pieces,
1167 * so we have to use the second new lock.
1171 locks_copy_lock(left
, right
);
1172 locks_insert_lock_ctx(left
, &fl
->fl_list
);
1174 right
->fl_start
= request
->fl_end
+ 1;
1175 locks_wake_up_blocks(right
);
1178 left
->fl_end
= request
->fl_start
- 1;
1179 locks_wake_up_blocks(left
);
1182 spin_unlock(&ctx
->flc_lock
);
1183 percpu_up_read_preempt_enable(&file_rwsem
);
1185 * Free any unused locks.
1188 locks_free_lock(new_fl
);
1190 locks_free_lock(new_fl2
);
1191 locks_dispose_list(&dispose
);
1192 trace_posix_lock_inode(inode
, request
, error
);
1198 * posix_lock_file - Apply a POSIX-style lock to a file
1199 * @filp: The file to apply the lock to
1200 * @fl: The lock to be applied
1201 * @conflock: Place to return a copy of the conflicting lock, if found.
1203 * Add a POSIX style lock to a file.
1204 * We merge adjacent & overlapping locks whenever possible.
1205 * POSIX locks are sorted by owner task, then by starting address
1207 * Note that if called with an FL_EXISTS argument, the caller may determine
1208 * whether or not a lock was successfully freed by testing the return
1209 * value for -ENOENT.
1211 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1212 struct file_lock
*conflock
)
1214 return posix_lock_inode(file_inode(filp
), fl
, conflock
);
1216 EXPORT_SYMBOL(posix_lock_file
);
1219 * posix_lock_inode_wait - Apply a POSIX-style lock to a file
1220 * @inode: inode of file to which lock request should be applied
1221 * @fl: The lock to be applied
1223 * Apply a POSIX style lock request to an inode.
1225 static int posix_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1230 error
= posix_lock_inode(inode
, fl
, NULL
);
1231 if (error
!= FILE_LOCK_DEFERRED
)
1233 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1237 locks_delete_block(fl
);
1243 #ifdef CONFIG_MANDATORY_FILE_LOCKING
1245 * locks_mandatory_locked - Check for an active lock
1246 * @file: the file to check
1248 * Searches the inode's list of locks to find any POSIX locks which conflict.
1249 * This function is called from locks_verify_locked() only.
1251 int locks_mandatory_locked(struct file
*file
)
1254 struct inode
*inode
= file_inode(file
);
1255 struct file_lock_context
*ctx
;
1256 struct file_lock
*fl
;
1258 ctx
= smp_load_acquire(&inode
->i_flctx
);
1259 if (!ctx
|| list_empty_careful(&ctx
->flc_posix
))
1263 * Search the lock list for this inode for any POSIX locks.
1265 spin_lock(&ctx
->flc_lock
);
1267 list_for_each_entry(fl
, &ctx
->flc_posix
, fl_list
) {
1268 if (fl
->fl_owner
!= current
->files
&&
1269 fl
->fl_owner
!= file
) {
1274 spin_unlock(&ctx
->flc_lock
);
1279 * locks_mandatory_area - Check for a conflicting lock
1280 * @inode: the file to check
1281 * @filp: how the file was opened (if it was)
1282 * @start: first byte in the file to check
1283 * @end: lastbyte in the file to check
1284 * @type: %F_WRLCK for a write lock, else %F_RDLCK
1286 * Searches the inode's list of locks to find any POSIX locks which conflict.
1288 int locks_mandatory_area(struct inode
*inode
, struct file
*filp
, loff_t start
,
1289 loff_t end
, unsigned char type
)
1291 struct file_lock fl
;
1295 locks_init_lock(&fl
);
1296 fl
.fl_pid
= current
->tgid
;
1298 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1299 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1302 fl
.fl_start
= start
;
1308 fl
.fl_flags
&= ~FL_SLEEP
;
1309 error
= posix_lock_inode(inode
, &fl
, NULL
);
1315 fl
.fl_flags
|= FL_SLEEP
;
1316 fl
.fl_owner
= current
->files
;
1317 error
= posix_lock_inode(inode
, &fl
, NULL
);
1318 if (error
!= FILE_LOCK_DEFERRED
)
1320 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1323 * If we've been sleeping someone might have
1324 * changed the permissions behind our back.
1326 if (__mandatory_lock(inode
))
1330 locks_delete_block(&fl
);
1337 EXPORT_SYMBOL(locks_mandatory_area
);
1338 #endif /* CONFIG_MANDATORY_FILE_LOCKING */
1340 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1344 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1347 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1351 /* We already had a lease on this file; just change its type */
1352 int lease_modify(struct file_lock
*fl
, int arg
, struct list_head
*dispose
)
1354 int error
= assign_type(fl
, arg
);
1358 lease_clear_pending(fl
, arg
);
1359 locks_wake_up_blocks(fl
);
1360 if (arg
== F_UNLCK
) {
1361 struct file
*filp
= fl
->fl_file
;
1364 filp
->f_owner
.signum
= 0;
1365 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1366 if (fl
->fl_fasync
!= NULL
) {
1367 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1368 fl
->fl_fasync
= NULL
;
1370 locks_delete_lock_ctx(fl
, dispose
);
1374 EXPORT_SYMBOL(lease_modify
);
1376 static bool past_time(unsigned long then
)
1379 /* 0 is a special value meaning "this never expires": */
1381 return time_after(jiffies
, then
);
1384 static void time_out_leases(struct inode
*inode
, struct list_head
*dispose
)
1386 struct file_lock_context
*ctx
= inode
->i_flctx
;
1387 struct file_lock
*fl
, *tmp
;
1389 lockdep_assert_held(&ctx
->flc_lock
);
1391 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1392 trace_time_out_leases(inode
, fl
);
1393 if (past_time(fl
->fl_downgrade_time
))
1394 lease_modify(fl
, F_RDLCK
, dispose
);
1395 if (past_time(fl
->fl_break_time
))
1396 lease_modify(fl
, F_UNLCK
, dispose
);
1400 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1402 if ((breaker
->fl_flags
& FL_LAYOUT
) != (lease
->fl_flags
& FL_LAYOUT
))
1404 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1406 return locks_conflict(breaker
, lease
);
1410 any_leases_conflict(struct inode
*inode
, struct file_lock
*breaker
)
1412 struct file_lock_context
*ctx
= inode
->i_flctx
;
1413 struct file_lock
*fl
;
1415 lockdep_assert_held(&ctx
->flc_lock
);
1417 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1418 if (leases_conflict(fl
, breaker
))
1425 * __break_lease - revoke all outstanding leases on file
1426 * @inode: the inode of the file to return
1427 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1429 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1432 * break_lease (inlined for speed) has checked there already is at least
1433 * some kind of lock (maybe a lease) on this file. Leases are broken on
1434 * a call to open() or truncate(). This function can sleep unless you
1435 * specified %O_NONBLOCK to your open().
1437 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1440 struct file_lock_context
*ctx
;
1441 struct file_lock
*new_fl
, *fl
, *tmp
;
1442 unsigned long break_time
;
1443 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1446 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1448 return PTR_ERR(new_fl
);
1449 new_fl
->fl_flags
= type
;
1451 /* typically we will check that ctx is non-NULL before calling */
1452 ctx
= smp_load_acquire(&inode
->i_flctx
);
1458 percpu_down_read_preempt_disable(&file_rwsem
);
1459 spin_lock(&ctx
->flc_lock
);
1461 time_out_leases(inode
, &dispose
);
1463 if (!any_leases_conflict(inode
, new_fl
))
1467 if (lease_break_time
> 0) {
1468 break_time
= jiffies
+ lease_break_time
* HZ
;
1469 if (break_time
== 0)
1470 break_time
++; /* so that 0 means no break time */
1473 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
) {
1474 if (!leases_conflict(fl
, new_fl
))
1477 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1479 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1480 fl
->fl_break_time
= break_time
;
1482 if (lease_breaking(fl
))
1484 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1485 fl
->fl_downgrade_time
= break_time
;
1487 if (fl
->fl_lmops
->lm_break(fl
))
1488 locks_delete_lock_ctx(fl
, &dispose
);
1491 if (list_empty(&ctx
->flc_lease
))
1494 if (mode
& O_NONBLOCK
) {
1495 trace_break_lease_noblock(inode
, new_fl
);
1496 error
= -EWOULDBLOCK
;
1501 fl
= list_first_entry(&ctx
->flc_lease
, struct file_lock
, fl_list
);
1502 break_time
= fl
->fl_break_time
;
1503 if (break_time
!= 0)
1504 break_time
-= jiffies
;
1505 if (break_time
== 0)
1507 locks_insert_block(fl
, new_fl
);
1508 trace_break_lease_block(inode
, new_fl
);
1509 spin_unlock(&ctx
->flc_lock
);
1510 percpu_up_read_preempt_enable(&file_rwsem
);
1512 locks_dispose_list(&dispose
);
1513 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1514 !new_fl
->fl_next
, break_time
);
1516 percpu_down_read_preempt_disable(&file_rwsem
);
1517 spin_lock(&ctx
->flc_lock
);
1518 trace_break_lease_unblock(inode
, new_fl
);
1519 locks_delete_block(new_fl
);
1522 * Wait for the next conflicting lease that has not been
1526 time_out_leases(inode
, &dispose
);
1527 if (any_leases_conflict(inode
, new_fl
))
1532 spin_unlock(&ctx
->flc_lock
);
1533 percpu_up_read_preempt_enable(&file_rwsem
);
1534 locks_dispose_list(&dispose
);
1535 locks_free_lock(new_fl
);
1539 EXPORT_SYMBOL(__break_lease
);
1542 * lease_get_mtime - get the last modified time of an inode
1544 * @time: pointer to a timespec which will contain the last modified time
1546 * This is to force NFS clients to flush their caches for files with
1547 * exclusive leases. The justification is that if someone has an
1548 * exclusive lease, then they could be modifying it.
1550 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1552 bool has_lease
= false;
1553 struct file_lock_context
*ctx
;
1554 struct file_lock
*fl
;
1556 ctx
= smp_load_acquire(&inode
->i_flctx
);
1557 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1558 spin_lock(&ctx
->flc_lock
);
1559 fl
= list_first_entry_or_null(&ctx
->flc_lease
,
1560 struct file_lock
, fl_list
);
1561 if (fl
&& (fl
->fl_type
== F_WRLCK
))
1563 spin_unlock(&ctx
->flc_lock
);
1567 *time
= current_fs_time(inode
->i_sb
);
1569 *time
= inode
->i_mtime
;
1572 EXPORT_SYMBOL(lease_get_mtime
);
1575 * fcntl_getlease - Enquire what lease is currently active
1578 * The value returned by this function will be one of
1579 * (if no lease break is pending):
1581 * %F_RDLCK to indicate a shared lease is held.
1583 * %F_WRLCK to indicate an exclusive lease is held.
1585 * %F_UNLCK to indicate no lease is held.
1587 * (if a lease break is pending):
1589 * %F_RDLCK to indicate an exclusive lease needs to be
1590 * changed to a shared lease (or removed).
1592 * %F_UNLCK to indicate the lease needs to be removed.
1594 * XXX: sfr & willy disagree over whether F_INPROGRESS
1595 * should be returned to userspace.
1597 int fcntl_getlease(struct file
*filp
)
1599 struct file_lock
*fl
;
1600 struct inode
*inode
= file_inode(filp
);
1601 struct file_lock_context
*ctx
;
1605 ctx
= smp_load_acquire(&inode
->i_flctx
);
1606 if (ctx
&& !list_empty_careful(&ctx
->flc_lease
)) {
1607 spin_lock(&ctx
->flc_lock
);
1608 time_out_leases(file_inode(filp
), &dispose
);
1609 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1610 if (fl
->fl_file
!= filp
)
1612 type
= target_leasetype(fl
);
1615 spin_unlock(&ctx
->flc_lock
);
1616 locks_dispose_list(&dispose
);
1622 * check_conflicting_open - see if the given dentry points to a file that has
1623 * an existing open that would conflict with the
1625 * @dentry: dentry to check
1626 * @arg: type of lease that we're trying to acquire
1627 * @flags: current lock flags
1629 * Check to see if there's an existing open fd on this file that would
1630 * conflict with the lease we're trying to set.
1633 check_conflicting_open(const struct dentry
*dentry
, const long arg
, int flags
)
1636 struct inode
*inode
= dentry
->d_inode
;
1638 if (flags
& FL_LAYOUT
)
1641 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1644 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1645 (atomic_read(&inode
->i_count
) > 1)))
1652 generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
, void **priv
)
1654 struct file_lock
*fl
, *my_fl
= NULL
, *lease
;
1655 struct dentry
*dentry
= filp
->f_path
.dentry
;
1656 struct inode
*inode
= file_inode(filp
);
1657 struct file_lock_context
*ctx
;
1658 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1663 trace_generic_add_lease(inode
, lease
);
1665 /* Note that arg is never F_UNLCK here */
1666 ctx
= locks_get_lock_context(inode
, arg
);
1671 * In the delegation case we need mutual exclusion with
1672 * a number of operations that take the i_mutex. We trylock
1673 * because delegations are an optional optimization, and if
1674 * there's some chance of a conflict--we'd rather not
1675 * bother, maybe that's a sign this just isn't a good file to
1676 * hand out a delegation on.
1678 if (is_deleg
&& !inode_trylock(inode
))
1681 if (is_deleg
&& arg
== F_WRLCK
) {
1682 /* Write delegations are not currently supported: */
1683 inode_unlock(inode
);
1688 percpu_down_read_preempt_disable(&file_rwsem
);
1689 spin_lock(&ctx
->flc_lock
);
1690 time_out_leases(inode
, &dispose
);
1691 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1696 * At this point, we know that if there is an exclusive
1697 * lease on this file, then we hold it on this filp
1698 * (otherwise our open of this file would have blocked).
1699 * And if we are trying to acquire an exclusive lease,
1700 * then the file is not open by anyone (including us)
1701 * except for this filp.
1704 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1705 if (fl
->fl_file
== filp
&&
1706 fl
->fl_owner
== lease
->fl_owner
) {
1712 * No exclusive leases if someone else has a lease on
1718 * Modifying our existing lease is OK, but no getting a
1719 * new lease if someone else is opening for write:
1721 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1725 if (my_fl
!= NULL
) {
1727 error
= lease
->fl_lmops
->lm_change(lease
, arg
, &dispose
);
1737 locks_insert_lock_ctx(lease
, &ctx
->flc_lease
);
1739 * The check in break_lease() is lockless. It's possible for another
1740 * open to race in after we did the earlier check for a conflicting
1741 * open but before the lease was inserted. Check again for a
1742 * conflicting open and cancel the lease if there is one.
1744 * We also add a barrier here to ensure that the insertion of the lock
1745 * precedes these checks.
1748 error
= check_conflicting_open(dentry
, arg
, lease
->fl_flags
);
1750 locks_unlink_lock_ctx(lease
);
1755 if (lease
->fl_lmops
->lm_setup
)
1756 lease
->fl_lmops
->lm_setup(lease
, priv
);
1758 spin_unlock(&ctx
->flc_lock
);
1759 percpu_up_read_preempt_enable(&file_rwsem
);
1760 locks_dispose_list(&dispose
);
1762 inode_unlock(inode
);
1763 if (!error
&& !my_fl
)
1768 static int generic_delete_lease(struct file
*filp
, void *owner
)
1770 int error
= -EAGAIN
;
1771 struct file_lock
*fl
, *victim
= NULL
;
1772 struct inode
*inode
= file_inode(filp
);
1773 struct file_lock_context
*ctx
;
1776 ctx
= smp_load_acquire(&inode
->i_flctx
);
1778 trace_generic_delete_lease(inode
, NULL
);
1782 percpu_down_read_preempt_disable(&file_rwsem
);
1783 spin_lock(&ctx
->flc_lock
);
1784 list_for_each_entry(fl
, &ctx
->flc_lease
, fl_list
) {
1785 if (fl
->fl_file
== filp
&&
1786 fl
->fl_owner
== owner
) {
1791 trace_generic_delete_lease(inode
, victim
);
1793 error
= fl
->fl_lmops
->lm_change(victim
, F_UNLCK
, &dispose
);
1794 spin_unlock(&ctx
->flc_lock
);
1795 percpu_up_read_preempt_enable(&file_rwsem
);
1796 locks_dispose_list(&dispose
);
1801 * generic_setlease - sets a lease on an open file
1802 * @filp: file pointer
1803 * @arg: type of lease to obtain
1804 * @flp: input - file_lock to use, output - file_lock inserted
1805 * @priv: private data for lm_setup (may be NULL if lm_setup
1806 * doesn't require it)
1808 * The (input) flp->fl_lmops->lm_break function is required
1811 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
,
1814 struct inode
*inode
= file_inode(filp
);
1817 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1819 if (!S_ISREG(inode
->i_mode
))
1821 error
= security_file_lock(filp
, arg
);
1827 return generic_delete_lease(filp
, *priv
);
1830 if (!(*flp
)->fl_lmops
->lm_break
) {
1835 return generic_add_lease(filp
, arg
, flp
, priv
);
1840 EXPORT_SYMBOL(generic_setlease
);
1843 * vfs_setlease - sets a lease on an open file
1844 * @filp: file pointer
1845 * @arg: type of lease to obtain
1846 * @lease: file_lock to use when adding a lease
1847 * @priv: private info for lm_setup when adding a lease (may be
1848 * NULL if lm_setup doesn't require it)
1850 * Call this to establish a lease on the file. The "lease" argument is not
1851 * used for F_UNLCK requests and may be NULL. For commands that set or alter
1852 * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set;
1853 * if not, this function will return -ENOLCK (and generate a scary-looking
1856 * The "priv" pointer is passed directly to the lm_setup function as-is. It
1857 * may be NULL if the lm_setup operation doesn't require it.
1860 vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
, void **priv
)
1862 if (filp
->f_op
->setlease
)
1863 return filp
->f_op
->setlease(filp
, arg
, lease
, priv
);
1865 return generic_setlease(filp
, arg
, lease
, priv
);
1867 EXPORT_SYMBOL_GPL(vfs_setlease
);
1869 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1871 struct file_lock
*fl
;
1872 struct fasync_struct
*new;
1875 fl
= lease_alloc(filp
, arg
);
1879 new = fasync_alloc();
1881 locks_free_lock(fl
);
1886 error
= vfs_setlease(filp
, arg
, &fl
, (void **)&new);
1888 locks_free_lock(fl
);
1895 * fcntl_setlease - sets a lease on an open file
1896 * @fd: open file descriptor
1897 * @filp: file pointer
1898 * @arg: type of lease to obtain
1900 * Call this fcntl to establish a lease on the file.
1901 * Note that you also need to call %F_SETSIG to
1902 * receive a signal when the lease is broken.
1904 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1907 return vfs_setlease(filp
, F_UNLCK
, NULL
, (void **)&filp
);
1908 return do_fcntl_add_lease(fd
, filp
, arg
);
1912 * flock_lock_inode_wait - Apply a FLOCK-style lock to a file
1913 * @inode: inode of the file to apply to
1914 * @fl: The lock to be applied
1916 * Apply a FLOCK style lock request to an inode.
1918 static int flock_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1923 error
= flock_lock_inode(inode
, fl
);
1924 if (error
!= FILE_LOCK_DEFERRED
)
1926 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1930 locks_delete_block(fl
);
1937 * locks_lock_inode_wait - Apply a lock to an inode
1938 * @inode: inode of the file to apply to
1939 * @fl: The lock to be applied
1941 * Apply a POSIX or FLOCK style lock request to an inode.
1943 int locks_lock_inode_wait(struct inode
*inode
, struct file_lock
*fl
)
1946 switch (fl
->fl_flags
& (FL_POSIX
|FL_FLOCK
)) {
1948 res
= posix_lock_inode_wait(inode
, fl
);
1951 res
= flock_lock_inode_wait(inode
, fl
);
1958 EXPORT_SYMBOL(locks_lock_inode_wait
);
1961 * sys_flock: - flock() system call.
1962 * @fd: the file descriptor to lock.
1963 * @cmd: the type of lock to apply.
1965 * Apply a %FL_FLOCK style lock to an open file descriptor.
1966 * The @cmd can be one of
1968 * %LOCK_SH -- a shared lock.
1970 * %LOCK_EX -- an exclusive lock.
1972 * %LOCK_UN -- remove an existing lock.
1974 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1976 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1977 * processes read and write access respectively.
1979 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1981 struct fd f
= fdget(fd
);
1982 struct file_lock
*lock
;
1983 int can_sleep
, unlock
;
1990 can_sleep
= !(cmd
& LOCK_NB
);
1992 unlock
= (cmd
== LOCK_UN
);
1994 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1995 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1998 lock
= flock_make_lock(f
.file
, cmd
);
2000 error
= PTR_ERR(lock
);
2005 lock
->fl_flags
|= FL_SLEEP
;
2007 error
= security_file_lock(f
.file
, lock
->fl_type
);
2011 if (f
.file
->f_op
->flock
)
2012 error
= f
.file
->f_op
->flock(f
.file
,
2013 (can_sleep
) ? F_SETLKW
: F_SETLK
,
2016 error
= locks_lock_file_wait(f
.file
, lock
);
2019 locks_free_lock(lock
);
2028 * vfs_test_lock - test file byte range lock
2029 * @filp: The file to test lock for
2030 * @fl: The lock to test; also used to hold result
2032 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
2033 * setting conf->fl_type to something other than F_UNLCK.
2035 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
2037 if (filp
->f_op
->lock
)
2038 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
2039 posix_test_lock(filp
, fl
);
2042 EXPORT_SYMBOL_GPL(vfs_test_lock
);
2044 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
2046 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2047 #if BITS_PER_LONG == 32
2049 * Make sure we can represent the posix lock via
2050 * legacy 32bit flock.
2052 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
2054 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
2057 flock
->l_start
= fl
->fl_start
;
2058 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2059 fl
->fl_end
- fl
->fl_start
+ 1;
2060 flock
->l_whence
= 0;
2061 flock
->l_type
= fl
->fl_type
;
2065 #if BITS_PER_LONG == 32
2066 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
2068 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
2069 flock
->l_start
= fl
->fl_start
;
2070 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
2071 fl
->fl_end
- fl
->fl_start
+ 1;
2072 flock
->l_whence
= 0;
2073 flock
->l_type
= fl
->fl_type
;
2077 /* Report the first existing lock that would conflict with l.
2078 * This implements the F_GETLK command of fcntl().
2080 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
2082 struct file_lock file_lock
;
2087 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2090 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2093 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
2097 if (cmd
== F_OFD_GETLK
) {
2099 if (flock
.l_pid
!= 0)
2103 file_lock
.fl_flags
|= FL_OFDLCK
;
2104 file_lock
.fl_owner
= filp
;
2107 error
= vfs_test_lock(filp
, &file_lock
);
2111 flock
.l_type
= file_lock
.fl_type
;
2112 if (file_lock
.fl_type
!= F_UNLCK
) {
2113 error
= posix_lock_to_flock(&flock
, &file_lock
);
2118 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2121 locks_release_private(&file_lock
);
2127 * vfs_lock_file - file byte range lock
2128 * @filp: The file to apply the lock to
2129 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
2130 * @fl: The lock to be applied
2131 * @conf: Place to return a copy of the conflicting lock, if found.
2133 * A caller that doesn't care about the conflicting lock may pass NULL
2134 * as the final argument.
2136 * If the filesystem defines a private ->lock() method, then @conf will
2137 * be left unchanged; so a caller that cares should initialize it to
2138 * some acceptable default.
2140 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2141 * locks, the ->lock() interface may return asynchronously, before the lock has
2142 * been granted or denied by the underlying filesystem, if (and only if)
2143 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2144 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2145 * the request is for a blocking lock. When ->lock() does return asynchronously,
2146 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2147 * request completes.
2148 * If the request is for non-blocking lock the file system should return
2149 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2150 * with the result. If the request timed out the callback routine will return a
2151 * nonzero return code and the file system should release the lock. The file
2152 * system is also responsible to keep a corresponding posix lock when it
2153 * grants a lock so the VFS can find out which locks are locally held and do
2154 * the correct lock cleanup when required.
2155 * The underlying filesystem must not drop the kernel lock or call
2156 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2159 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2161 if (filp
->f_op
->lock
)
2162 return filp
->f_op
->lock(filp
, cmd
, fl
);
2164 return posix_lock_file(filp
, fl
, conf
);
2166 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2168 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2169 struct file_lock
*fl
)
2173 error
= security_file_lock(filp
, fl
->fl_type
);
2178 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2179 if (error
!= FILE_LOCK_DEFERRED
)
2181 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2185 locks_delete_block(fl
);
2192 /* Ensure that fl->fl_file has compatible f_mode for F_SETLK calls */
2194 check_fmode_for_setlk(struct file_lock
*fl
)
2196 switch (fl
->fl_type
) {
2198 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2202 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2208 /* Apply the lock described by l to an open file descriptor.
2209 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2211 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2212 struct flock __user
*l
)
2214 struct file_lock
*file_lock
= locks_alloc_lock();
2216 struct inode
*inode
;
2220 if (file_lock
== NULL
)
2223 inode
= file_inode(filp
);
2226 * This might block, so we do it before checking the inode.
2229 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2232 /* Don't allow mandatory locks on files that may be memory mapped
2235 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2240 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2244 error
= check_fmode_for_setlk(file_lock
);
2249 * If the cmd is requesting file-private locks, then set the
2250 * FL_OFDLCK flag and override the owner.
2255 if (flock
.l_pid
!= 0)
2259 file_lock
->fl_flags
|= FL_OFDLCK
;
2260 file_lock
->fl_owner
= filp
;
2264 if (flock
.l_pid
!= 0)
2268 file_lock
->fl_flags
|= FL_OFDLCK
;
2269 file_lock
->fl_owner
= filp
;
2272 file_lock
->fl_flags
|= FL_SLEEP
;
2275 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2278 * Attempt to detect a close/fcntl race and recover by releasing the
2279 * lock that was just acquired. There is no need to do that when we're
2280 * unlocking though, or for OFD locks.
2282 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2283 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2285 * We need that spin_lock here - it prevents reordering between
2286 * update of i_flctx->flc_posix and check for it done in
2287 * close(). rcu_read_lock() wouldn't do.
2289 spin_lock(¤t
->files
->file_lock
);
2291 spin_unlock(¤t
->files
->file_lock
);
2293 file_lock
->fl_type
= F_UNLCK
;
2294 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2295 WARN_ON_ONCE(error
);
2300 trace_fcntl_setlk(inode
, file_lock
, error
);
2301 locks_free_lock(file_lock
);
2305 #if BITS_PER_LONG == 32
2306 /* Report the first existing lock that would conflict with l.
2307 * This implements the F_GETLK command of fcntl().
2309 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2311 struct file_lock file_lock
;
2312 struct flock64 flock
;
2316 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2319 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2322 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2326 if (cmd
== F_OFD_GETLK
) {
2328 if (flock
.l_pid
!= 0)
2332 file_lock
.fl_flags
|= FL_OFDLCK
;
2333 file_lock
.fl_owner
= filp
;
2336 error
= vfs_test_lock(filp
, &file_lock
);
2340 flock
.l_type
= file_lock
.fl_type
;
2341 if (file_lock
.fl_type
!= F_UNLCK
)
2342 posix_lock_to_flock64(&flock
, &file_lock
);
2345 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2348 locks_release_private(&file_lock
);
2353 /* Apply the lock described by l to an open file descriptor.
2354 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2356 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2357 struct flock64 __user
*l
)
2359 struct file_lock
*file_lock
= locks_alloc_lock();
2360 struct flock64 flock
;
2361 struct inode
*inode
;
2365 if (file_lock
== NULL
)
2369 * This might block, so we do it before checking the inode.
2372 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2375 inode
= file_inode(filp
);
2377 /* Don't allow mandatory locks on files that may be memory mapped
2380 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2385 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2389 error
= check_fmode_for_setlk(file_lock
);
2394 * If the cmd is requesting file-private locks, then set the
2395 * FL_OFDLCK flag and override the owner.
2400 if (flock
.l_pid
!= 0)
2404 file_lock
->fl_flags
|= FL_OFDLCK
;
2405 file_lock
->fl_owner
= filp
;
2409 if (flock
.l_pid
!= 0)
2413 file_lock
->fl_flags
|= FL_OFDLCK
;
2414 file_lock
->fl_owner
= filp
;
2417 file_lock
->fl_flags
|= FL_SLEEP
;
2420 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2423 * Attempt to detect a close/fcntl race and recover by releasing the
2424 * lock that was just acquired. There is no need to do that when we're
2425 * unlocking though, or for OFD locks.
2427 if (!error
&& file_lock
->fl_type
!= F_UNLCK
&&
2428 !(file_lock
->fl_flags
& FL_OFDLCK
)) {
2430 * We need that spin_lock here - it prevents reordering between
2431 * update of i_flctx->flc_posix and check for it done in
2432 * close(). rcu_read_lock() wouldn't do.
2434 spin_lock(¤t
->files
->file_lock
);
2436 spin_unlock(¤t
->files
->file_lock
);
2438 file_lock
->fl_type
= F_UNLCK
;
2439 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2440 WARN_ON_ONCE(error
);
2445 locks_free_lock(file_lock
);
2448 #endif /* BITS_PER_LONG == 32 */
2451 * This function is called when the file is being removed
2452 * from the task's fd array. POSIX locks belonging to this task
2453 * are deleted at this time.
2455 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2458 struct file_lock lock
;
2459 struct file_lock_context
*ctx
;
2462 * If there are no locks held on this file, we don't need to call
2463 * posix_lock_file(). Another process could be setting a lock on this
2464 * file at the same time, but we wouldn't remove that lock anyway.
2466 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2467 if (!ctx
|| list_empty(&ctx
->flc_posix
))
2470 lock
.fl_type
= F_UNLCK
;
2471 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2473 lock
.fl_end
= OFFSET_MAX
;
2474 lock
.fl_owner
= owner
;
2475 lock
.fl_pid
= current
->tgid
;
2476 lock
.fl_file
= filp
;
2478 lock
.fl_lmops
= NULL
;
2480 error
= vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2482 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2483 lock
.fl_ops
->fl_release_private(&lock
);
2484 trace_locks_remove_posix(file_inode(filp
), &lock
, error
);
2487 EXPORT_SYMBOL(locks_remove_posix
);
2489 /* The i_flctx must be valid when calling into here */
2491 locks_remove_flock(struct file
*filp
, struct file_lock_context
*flctx
)
2493 struct file_lock fl
= {
2495 .fl_pid
= current
->tgid
,
2497 .fl_flags
= FL_FLOCK
,
2499 .fl_end
= OFFSET_MAX
,
2501 struct inode
*inode
= file_inode(filp
);
2503 if (list_empty(&flctx
->flc_flock
))
2506 if (filp
->f_op
->flock
)
2507 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2509 flock_lock_inode(inode
, &fl
);
2511 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2512 fl
.fl_ops
->fl_release_private(&fl
);
2515 /* The i_flctx must be valid when calling into here */
2517 locks_remove_lease(struct file
*filp
, struct file_lock_context
*ctx
)
2519 struct file_lock
*fl
, *tmp
;
2522 if (list_empty(&ctx
->flc_lease
))
2525 spin_lock(&ctx
->flc_lock
);
2526 list_for_each_entry_safe(fl
, tmp
, &ctx
->flc_lease
, fl_list
)
2527 if (filp
== fl
->fl_file
)
2528 lease_modify(fl
, F_UNLCK
, &dispose
);
2529 spin_unlock(&ctx
->flc_lock
);
2530 locks_dispose_list(&dispose
);
2534 * This function is called on the last close of an open file.
2536 void locks_remove_file(struct file
*filp
)
2538 struct file_lock_context
*ctx
;
2540 ctx
= smp_load_acquire(&file_inode(filp
)->i_flctx
);
2544 /* remove any OFD locks */
2545 locks_remove_posix(filp
, filp
);
2547 /* remove flock locks */
2548 locks_remove_flock(filp
, ctx
);
2550 /* remove any leases */
2551 locks_remove_lease(filp
, ctx
);
2555 * posix_unblock_lock - stop waiting for a file lock
2556 * @waiter: the lock which was waiting
2558 * lockd needs to block waiting for locks.
2561 posix_unblock_lock(struct file_lock
*waiter
)
2565 spin_lock(&blocked_lock_lock
);
2566 if (waiter
->fl_next
)
2567 __locks_delete_block(waiter
);
2570 spin_unlock(&blocked_lock_lock
);
2573 EXPORT_SYMBOL(posix_unblock_lock
);
2576 * vfs_cancel_lock - file byte range unblock lock
2577 * @filp: The file to apply the unblock to
2578 * @fl: The lock to be unblocked
2580 * Used by lock managers to cancel blocked requests
2582 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2584 if (filp
->f_op
->lock
)
2585 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2589 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2591 #ifdef CONFIG_PROC_FS
2592 #include <linux/proc_fs.h>
2593 #include <linux/seq_file.h>
2595 struct locks_iterator
{
2600 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2601 loff_t id
, char *pfx
)
2603 struct inode
*inode
= NULL
;
2604 unsigned int fl_pid
;
2607 struct pid_namespace
*proc_pidns
= file_inode(f
->file
)->i_sb
->s_fs_info
;
2609 /* Don't let fl_pid change based on who is reading the file */
2610 fl_pid
= pid_nr_ns(fl
->fl_nspid
, proc_pidns
);
2613 * If there isn't a fl_pid don't display who is waiting on
2614 * the lock if we are called from locks_show, or if we are
2615 * called from __show_fd_info - skip lock entirely
2620 fl_pid
= fl
->fl_pid
;
2622 if (fl
->fl_file
!= NULL
)
2623 inode
= file_inode(fl
->fl_file
);
2625 seq_printf(f
, "%lld:%s ", id
, pfx
);
2627 if (fl
->fl_flags
& FL_ACCESS
)
2628 seq_puts(f
, "ACCESS");
2629 else if (IS_OFDLCK(fl
))
2630 seq_puts(f
, "OFDLCK");
2632 seq_puts(f
, "POSIX ");
2634 seq_printf(f
, " %s ",
2635 (inode
== NULL
) ? "*NOINODE*" :
2636 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2637 } else if (IS_FLOCK(fl
)) {
2638 if (fl
->fl_type
& LOCK_MAND
) {
2639 seq_puts(f
, "FLOCK MSNFS ");
2641 seq_puts(f
, "FLOCK ADVISORY ");
2643 } else if (IS_LEASE(fl
)) {
2644 if (fl
->fl_flags
& FL_DELEG
)
2645 seq_puts(f
, "DELEG ");
2647 seq_puts(f
, "LEASE ");
2649 if (lease_breaking(fl
))
2650 seq_puts(f
, "BREAKING ");
2651 else if (fl
->fl_file
)
2652 seq_puts(f
, "ACTIVE ");
2654 seq_puts(f
, "BREAKER ");
2656 seq_puts(f
, "UNKNOWN UNKNOWN ");
2658 if (fl
->fl_type
& LOCK_MAND
) {
2659 seq_printf(f
, "%s ",
2660 (fl
->fl_type
& LOCK_READ
)
2661 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2662 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2664 seq_printf(f
, "%s ",
2665 (lease_breaking(fl
))
2666 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2667 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2670 /* userspace relies on this representation of dev_t */
2671 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2672 MAJOR(inode
->i_sb
->s_dev
),
2673 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2675 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2678 if (fl
->fl_end
== OFFSET_MAX
)
2679 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2681 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2683 seq_puts(f
, "0 EOF\n");
2687 static int locks_show(struct seq_file
*f
, void *v
)
2689 struct locks_iterator
*iter
= f
->private;
2690 struct file_lock
*fl
, *bfl
;
2691 struct pid_namespace
*proc_pidns
= file_inode(f
->file
)->i_sb
->s_fs_info
;
2693 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2695 if (fl
->fl_nspid
&& !pid_nr_ns(fl
->fl_nspid
, proc_pidns
))
2698 lock_get_status(f
, fl
, iter
->li_pos
, "");
2700 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2701 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2706 static void __show_fd_locks(struct seq_file
*f
,
2707 struct list_head
*head
, int *id
,
2708 struct file
*filp
, struct files_struct
*files
)
2710 struct file_lock
*fl
;
2712 list_for_each_entry(fl
, head
, fl_list
) {
2714 if (filp
!= fl
->fl_file
)
2716 if (fl
->fl_owner
!= files
&&
2717 fl
->fl_owner
!= filp
)
2721 seq_puts(f
, "lock:\t");
2722 lock_get_status(f
, fl
, *id
, "");
2726 void show_fd_locks(struct seq_file
*f
,
2727 struct file
*filp
, struct files_struct
*files
)
2729 struct inode
*inode
= file_inode(filp
);
2730 struct file_lock_context
*ctx
;
2733 ctx
= smp_load_acquire(&inode
->i_flctx
);
2737 spin_lock(&ctx
->flc_lock
);
2738 __show_fd_locks(f
, &ctx
->flc_flock
, &id
, filp
, files
);
2739 __show_fd_locks(f
, &ctx
->flc_posix
, &id
, filp
, files
);
2740 __show_fd_locks(f
, &ctx
->flc_lease
, &id
, filp
, files
);
2741 spin_unlock(&ctx
->flc_lock
);
2744 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2745 __acquires(&blocked_lock_lock
)
2747 struct locks_iterator
*iter
= f
->private;
2749 iter
->li_pos
= *pos
+ 1;
2750 percpu_down_write(&file_rwsem
);
2751 spin_lock(&blocked_lock_lock
);
2752 return seq_hlist_start_percpu(&file_lock_list
.hlist
, &iter
->li_cpu
, *pos
);
2755 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2757 struct locks_iterator
*iter
= f
->private;
2760 return seq_hlist_next_percpu(v
, &file_lock_list
.hlist
, &iter
->li_cpu
, pos
);
2763 static void locks_stop(struct seq_file
*f
, void *v
)
2764 __releases(&blocked_lock_lock
)
2766 spin_unlock(&blocked_lock_lock
);
2767 percpu_up_write(&file_rwsem
);
2770 static const struct seq_operations locks_seq_operations
= {
2771 .start
= locks_start
,
2777 static int locks_open(struct inode
*inode
, struct file
*filp
)
2779 return seq_open_private(filp
, &locks_seq_operations
,
2780 sizeof(struct locks_iterator
));
2783 static const struct file_operations proc_locks_operations
= {
2786 .llseek
= seq_lseek
,
2787 .release
= seq_release_private
,
2790 static int __init
proc_locks_init(void)
2792 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2795 fs_initcall(proc_locks_init
);
2798 static int __init
filelock_init(void)
2802 flctx_cache
= kmem_cache_create("file_lock_ctx",
2803 sizeof(struct file_lock_context
), 0, SLAB_PANIC
, NULL
);
2805 filelock_cache
= kmem_cache_create("file_lock_cache",
2806 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2809 for_each_possible_cpu(i
) {
2810 struct file_lock_list_struct
*fll
= per_cpu_ptr(&file_lock_list
, i
);
2812 spin_lock_init(&fll
->lock
);
2813 INIT_HLIST_HEAD(&fll
->hlist
);
2819 core_initcall(filelock_init
);