4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
7 * Deadlock detection added.
8 * FIXME: one thing isn't handled yet:
9 * - mandatory locks (requires lots of changes elsewhere)
10 * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994.
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
15 * Converted file_lock_table to a linked list from an array, which eliminates
16 * the limits on how many active file locks are open.
17 * Chad Page (pageone@netcom.com), November 27, 1994
19 * Removed dependency on file descriptors. dup()'ed file descriptors now
20 * get the same locks as the original file descriptors, and a close() on
21 * any file descriptor removes ALL the locks on the file for the current
22 * process. Since locks still depend on the process id, locks are inherited
23 * after an exec() but not after a fork(). This agrees with POSIX, and both
24 * BSD and SVR4 practice.
25 * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995
27 * Scrapped free list which is redundant now that we allocate locks
28 * dynamically with kmalloc()/kfree().
29 * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
36 * FL_FLOCK locks are created with calls to flock(), through the flock()
37 * system call, which is new. Old C libraries implement flock() via fcntl()
38 * and will continue to use the old, broken implementation.
40 * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated
41 * with a file pointer (filp). As a result they can be shared by a parent
42 * process and its children after a fork(). They are removed when the last
43 * file descriptor referring to the file pointer is closed (unless explicitly
46 * FL_FLOCK locks never deadlock, an existing lock is always removed before
47 * upgrading from shared to exclusive (or vice versa). When this happens
48 * any processes blocked by the current lock are woken up and allowed to
49 * run before the new lock is applied.
50 * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995
52 * Removed some race conditions in flock_lock_file(), marked other possible
53 * races. Just grep for FIXME to see them.
54 * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996.
56 * Addressed Dmitry's concerns. Deadlock checking no longer recursive.
57 * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep
58 * once we've checked for blocking and deadlocking.
59 * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996.
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
66 * Don't allow mandatory locks on mmap()'ed files. Added simple functions to
67 * check if a file has mandatory locks, used by mmap(), open() and creat() to
68 * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
75 * Fixed deadlock condition for pathological code that mixes calls to
76 * flock() and fcntl().
77 * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996.
79 * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use
80 * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to
81 * guarantee sensible behaviour in the case where file system modules might
82 * be compiled with different options than the kernel itself.
83 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
85 * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel
86 * (Thomas.Meckel@mni.fh-giessen.de) for spotting this.
87 * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996.
89 * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK
90 * locks. Changed process synchronisation to avoid dereferencing locks that
91 * have already been freed.
92 * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996.
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
97 * Made mandatory locking a mount option. Default is not to allow mandatory
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
104 * Fixed /proc/locks interface so that we can't overrun the buffer we are handed.
105 * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997.
107 * Use slab allocator instead of kmalloc/kfree.
108 * Use generic list implementation from <linux/list.h>.
109 * Sped up posix_locks_deadlock by only considering blocked locks.
110 * Matthew Wilcox <willy@debian.org>, March, 2000.
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
117 #include <linux/capability.h>
118 #include <linux/file.h>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/module.h>
123 #include <linux/security.h>
124 #include <linux/slab.h>
125 #include <linux/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/rcupdate.h>
128 #include <linux/pid_namespace.h>
129 #include <linux/hashtable.h>
130 #include <linux/percpu.h>
131 #include <linux/lglock.h>
133 #define CREATE_TRACE_POINTS
134 #include <trace/events/filelock.h>
136 #include <asm/uaccess.h>
138 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
139 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
140 #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG))
141 #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK)
143 static bool lease_breaking(struct file_lock
*fl
)
145 return fl
->fl_flags
& (FL_UNLOCK_PENDING
| FL_DOWNGRADE_PENDING
);
148 static int target_leasetype(struct file_lock
*fl
)
150 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
152 if (fl
->fl_flags
& FL_DOWNGRADE_PENDING
)
157 int leases_enable
= 1;
158 int lease_break_time
= 45;
160 #define for_each_lock(inode, lockp) \
161 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
164 * The global file_lock_list is only used for displaying /proc/locks, so we
165 * keep a list on each CPU, with each list protected by its own spinlock via
166 * the file_lock_lglock. Note that alterations to the list also require that
167 * the relevant i_lock is held.
169 DEFINE_STATIC_LGLOCK(file_lock_lglock
);
170 static DEFINE_PER_CPU(struct hlist_head
, file_lock_list
);
173 * The blocked_hash is used to find POSIX lock loops for deadlock detection.
174 * It is protected by blocked_lock_lock.
176 * We hash locks by lockowner in order to optimize searching for the lock a
177 * particular lockowner is waiting on.
179 * FIXME: make this value scale via some heuristic? We generally will want more
180 * buckets when we have more lockowners holding locks, but that's a little
181 * difficult to determine without knowing what the workload will look like.
183 #define BLOCKED_HASH_BITS 7
184 static DEFINE_HASHTABLE(blocked_hash
, BLOCKED_HASH_BITS
);
187 * This lock protects the blocked_hash. Generally, if you're accessing it, you
188 * want to be holding this lock.
190 * In addition, it also protects the fl->fl_block list, and the fl->fl_next
191 * pointer for file_lock structures that are acting as lock requests (in
192 * contrast to those that are acting as records of acquired locks).
194 * Note that when we acquire this lock in order to change the above fields,
195 * we often hold the i_lock as well. In certain cases, when reading the fields
196 * protected by this lock, we can skip acquiring it iff we already hold the
199 * In particular, adding an entry to the fl_block list requires that you hold
200 * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting
201 * an entry from the list however only requires the file_lock_lock.
203 static DEFINE_SPINLOCK(blocked_lock_lock
);
205 static struct kmem_cache
*filelock_cache __read_mostly
;
207 static void locks_init_lock_heads(struct file_lock
*fl
)
209 INIT_HLIST_NODE(&fl
->fl_link
);
210 INIT_LIST_HEAD(&fl
->fl_block
);
211 init_waitqueue_head(&fl
->fl_wait
);
214 /* Allocate an empty lock structure. */
215 struct file_lock
*locks_alloc_lock(void)
217 struct file_lock
*fl
= kmem_cache_zalloc(filelock_cache
, GFP_KERNEL
);
220 locks_init_lock_heads(fl
);
224 EXPORT_SYMBOL_GPL(locks_alloc_lock
);
226 void locks_release_private(struct file_lock
*fl
)
229 if (fl
->fl_ops
->fl_release_private
)
230 fl
->fl_ops
->fl_release_private(fl
);
236 EXPORT_SYMBOL_GPL(locks_release_private
);
238 /* Free a lock which is not in use. */
239 void locks_free_lock(struct file_lock
*fl
)
241 BUG_ON(waitqueue_active(&fl
->fl_wait
));
242 BUG_ON(!list_empty(&fl
->fl_block
));
243 BUG_ON(!hlist_unhashed(&fl
->fl_link
));
245 locks_release_private(fl
);
246 kmem_cache_free(filelock_cache
, fl
);
248 EXPORT_SYMBOL(locks_free_lock
);
251 locks_dispose_list(struct list_head
*dispose
)
253 struct file_lock
*fl
;
255 while (!list_empty(dispose
)) {
256 fl
= list_first_entry(dispose
, struct file_lock
, fl_block
);
257 list_del_init(&fl
->fl_block
);
262 void locks_init_lock(struct file_lock
*fl
)
264 memset(fl
, 0, sizeof(struct file_lock
));
265 locks_init_lock_heads(fl
);
268 EXPORT_SYMBOL(locks_init_lock
);
270 static void locks_copy_private(struct file_lock
*new, struct file_lock
*fl
)
273 if (fl
->fl_ops
->fl_copy_lock
)
274 fl
->fl_ops
->fl_copy_lock(new, fl
);
275 new->fl_ops
= fl
->fl_ops
;
278 new->fl_lmops
= fl
->fl_lmops
;
282 * Initialize a new lock from an existing file_lock structure.
284 void __locks_copy_lock(struct file_lock
*new, const struct file_lock
*fl
)
286 new->fl_owner
= fl
->fl_owner
;
287 new->fl_pid
= fl
->fl_pid
;
289 new->fl_flags
= fl
->fl_flags
;
290 new->fl_type
= fl
->fl_type
;
291 new->fl_start
= fl
->fl_start
;
292 new->fl_end
= fl
->fl_end
;
294 new->fl_lmops
= NULL
;
296 EXPORT_SYMBOL(__locks_copy_lock
);
298 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
300 /* "new" must be a freshly-initialized lock */
301 WARN_ON_ONCE(new->fl_ops
);
303 __locks_copy_lock(new, fl
);
304 new->fl_file
= fl
->fl_file
;
305 new->fl_ops
= fl
->fl_ops
;
306 new->fl_lmops
= fl
->fl_lmops
;
308 locks_copy_private(new, fl
);
311 EXPORT_SYMBOL(locks_copy_lock
);
313 static inline int flock_translate_cmd(int cmd
) {
315 return cmd
& (LOCK_MAND
| LOCK_RW
);
327 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
328 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
331 struct file_lock
*fl
;
332 int type
= flock_translate_cmd(cmd
);
336 fl
= locks_alloc_lock();
342 fl
->fl_pid
= current
->tgid
;
343 fl
->fl_flags
= FL_FLOCK
;
345 fl
->fl_end
= OFFSET_MAX
;
351 static int assign_type(struct file_lock
*fl
, long type
)
365 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
368 switch (l
->l_whence
) {
373 fl
->fl_start
= filp
->f_pos
;
376 fl
->fl_start
= i_size_read(file_inode(filp
));
381 if (l
->l_start
> OFFSET_MAX
- fl
->fl_start
)
383 fl
->fl_start
+= l
->l_start
;
384 if (fl
->fl_start
< 0)
387 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
388 POSIX-2001 defines it. */
390 if (l
->l_len
- 1 > OFFSET_MAX
- fl
->fl_start
)
392 fl
->fl_end
= fl
->fl_start
+ l
->l_len
- 1;
394 } else if (l
->l_len
< 0) {
395 if (fl
->fl_start
+ l
->l_len
< 0)
397 fl
->fl_end
= fl
->fl_start
- 1;
398 fl
->fl_start
+= l
->l_len
;
400 fl
->fl_end
= OFFSET_MAX
;
402 fl
->fl_owner
= current
->files
;
403 fl
->fl_pid
= current
->tgid
;
405 fl
->fl_flags
= FL_POSIX
;
409 return assign_type(fl
, l
->l_type
);
412 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
415 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
418 struct flock64 ll
= {
420 .l_whence
= l
->l_whence
,
421 .l_start
= l
->l_start
,
425 return flock64_to_posix_lock(filp
, fl
, &ll
);
428 /* default lease lock manager operations */
429 static void lease_break_callback(struct file_lock
*fl
)
431 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
434 static const struct lock_manager_operations lease_manager_ops
= {
435 .lm_break
= lease_break_callback
,
436 .lm_change
= lease_modify
,
440 * Initialize a lease, use the default lock manager operations
442 static int lease_init(struct file
*filp
, long type
, struct file_lock
*fl
)
444 if (assign_type(fl
, type
) != 0)
447 fl
->fl_owner
= current
->files
;
448 fl
->fl_pid
= current
->tgid
;
451 fl
->fl_flags
= FL_LEASE
;
453 fl
->fl_end
= OFFSET_MAX
;
455 fl
->fl_lmops
= &lease_manager_ops
;
459 /* Allocate a file_lock initialised to this type of lease */
460 static struct file_lock
*lease_alloc(struct file
*filp
, long type
)
462 struct file_lock
*fl
= locks_alloc_lock();
466 return ERR_PTR(error
);
468 error
= lease_init(filp
, type
, fl
);
471 return ERR_PTR(error
);
476 /* Check if two locks overlap each other.
478 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
480 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
481 (fl2
->fl_end
>= fl1
->fl_start
));
485 * Check whether two locks have the same owner.
487 static int posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
489 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->lm_compare_owner
)
490 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
491 fl1
->fl_lmops
->lm_compare_owner(fl1
, fl2
);
492 return fl1
->fl_owner
== fl2
->fl_owner
;
495 /* Must be called with the i_lock held! */
496 static void locks_insert_global_locks(struct file_lock
*fl
)
498 lg_local_lock(&file_lock_lglock
);
499 fl
->fl_link_cpu
= smp_processor_id();
500 hlist_add_head(&fl
->fl_link
, this_cpu_ptr(&file_lock_list
));
501 lg_local_unlock(&file_lock_lglock
);
504 /* Must be called with the i_lock held! */
505 static void locks_delete_global_locks(struct file_lock
*fl
)
508 * Avoid taking lock if already unhashed. This is safe since this check
509 * is done while holding the i_lock, and new insertions into the list
510 * also require that it be held.
512 if (hlist_unhashed(&fl
->fl_link
))
514 lg_local_lock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
515 hlist_del_init(&fl
->fl_link
);
516 lg_local_unlock_cpu(&file_lock_lglock
, fl
->fl_link_cpu
);
520 posix_owner_key(struct file_lock
*fl
)
522 if (fl
->fl_lmops
&& fl
->fl_lmops
->lm_owner_key
)
523 return fl
->fl_lmops
->lm_owner_key(fl
);
524 return (unsigned long)fl
->fl_owner
;
527 static void locks_insert_global_blocked(struct file_lock
*waiter
)
529 hash_add(blocked_hash
, &waiter
->fl_link
, posix_owner_key(waiter
));
532 static void locks_delete_global_blocked(struct file_lock
*waiter
)
534 hash_del(&waiter
->fl_link
);
537 /* Remove waiter from blocker's block list.
538 * When blocker ends up pointing to itself then the list is empty.
540 * Must be called with blocked_lock_lock held.
542 static void __locks_delete_block(struct file_lock
*waiter
)
544 locks_delete_global_blocked(waiter
);
545 list_del_init(&waiter
->fl_block
);
546 waiter
->fl_next
= NULL
;
549 static void locks_delete_block(struct file_lock
*waiter
)
551 spin_lock(&blocked_lock_lock
);
552 __locks_delete_block(waiter
);
553 spin_unlock(&blocked_lock_lock
);
556 /* Insert waiter into blocker's block list.
557 * We use a circular list so that processes can be easily woken up in
558 * the order they blocked. The documentation doesn't require this but
559 * it seems like the reasonable thing to do.
561 * Must be called with both the i_lock and blocked_lock_lock held. The fl_block
562 * list itself is protected by the blocked_lock_lock, but by ensuring that the
563 * i_lock is also held on insertions we can avoid taking the blocked_lock_lock
564 * in some cases when we see that the fl_block list is empty.
566 static void __locks_insert_block(struct file_lock
*blocker
,
567 struct file_lock
*waiter
)
569 BUG_ON(!list_empty(&waiter
->fl_block
));
570 waiter
->fl_next
= blocker
;
571 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
572 if (IS_POSIX(blocker
) && !IS_OFDLCK(blocker
))
573 locks_insert_global_blocked(waiter
);
576 /* Must be called with i_lock held. */
577 static void locks_insert_block(struct file_lock
*blocker
,
578 struct file_lock
*waiter
)
580 spin_lock(&blocked_lock_lock
);
581 __locks_insert_block(blocker
, waiter
);
582 spin_unlock(&blocked_lock_lock
);
586 * Wake up processes blocked waiting for blocker.
588 * Must be called with the inode->i_lock held!
590 static void locks_wake_up_blocks(struct file_lock
*blocker
)
593 * Avoid taking global lock if list is empty. This is safe since new
594 * blocked requests are only added to the list under the i_lock, and
595 * the i_lock is always held here. Note that removal from the fl_block
596 * list does not require the i_lock, so we must recheck list_empty()
597 * after acquiring the blocked_lock_lock.
599 if (list_empty(&blocker
->fl_block
))
602 spin_lock(&blocked_lock_lock
);
603 while (!list_empty(&blocker
->fl_block
)) {
604 struct file_lock
*waiter
;
606 waiter
= list_first_entry(&blocker
->fl_block
,
607 struct file_lock
, fl_block
);
608 __locks_delete_block(waiter
);
609 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->lm_notify
)
610 waiter
->fl_lmops
->lm_notify(waiter
);
612 wake_up(&waiter
->fl_wait
);
614 spin_unlock(&blocked_lock_lock
);
617 /* Insert file lock fl into an inode's lock list at the position indicated
618 * by pos. At the same time add the lock to the global file lock list.
620 * Must be called with the i_lock held!
622 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
624 fl
->fl_nspid
= get_pid(task_tgid(current
));
626 /* insert into file's list */
630 locks_insert_global_locks(fl
);
634 * locks_delete_lock - Delete a lock and then free it.
635 * @thisfl_p: pointer that points to the fl_next field of the previous
636 * inode->i_flock list entry
638 * Unlink a lock from all lists and free the namespace reference, but don't
639 * free it yet. Wake up processes that are blocked waiting for this lock and
640 * notify the FS that the lock has been cleared.
642 * Must be called with the i_lock held!
644 static void locks_unlink_lock(struct file_lock
**thisfl_p
)
646 struct file_lock
*fl
= *thisfl_p
;
648 locks_delete_global_locks(fl
);
650 *thisfl_p
= fl
->fl_next
;
654 put_pid(fl
->fl_nspid
);
658 locks_wake_up_blocks(fl
);
662 * Unlink a lock from all lists and free it.
664 * Must be called with i_lock held!
666 static void locks_delete_lock(struct file_lock
**thisfl_p
,
667 struct list_head
*dispose
)
669 struct file_lock
*fl
= *thisfl_p
;
671 locks_unlink_lock(thisfl_p
);
673 list_add(&fl
->fl_block
, dispose
);
678 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
679 * checks for shared/exclusive status of overlapping locks.
681 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
683 if (sys_fl
->fl_type
== F_WRLCK
)
685 if (caller_fl
->fl_type
== F_WRLCK
)
690 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
691 * checking before calling the locks_conflict().
693 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
695 /* POSIX locks owned by the same process do not conflict with
698 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
701 /* Check whether they overlap */
702 if (!locks_overlap(caller_fl
, sys_fl
))
705 return (locks_conflict(caller_fl
, sys_fl
));
708 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
709 * checking before calling the locks_conflict().
711 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
713 /* FLOCK locks referring to the same filp do not conflict with
716 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
718 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
721 return (locks_conflict(caller_fl
, sys_fl
));
725 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
727 struct file_lock
*cfl
;
728 struct inode
*inode
= file_inode(filp
);
730 spin_lock(&inode
->i_lock
);
731 for (cfl
= file_inode(filp
)->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
734 if (posix_locks_conflict(fl
, cfl
))
738 __locks_copy_lock(fl
, cfl
);
740 fl
->fl_pid
= pid_vnr(cfl
->fl_nspid
);
742 fl
->fl_type
= F_UNLCK
;
743 spin_unlock(&inode
->i_lock
);
746 EXPORT_SYMBOL(posix_test_lock
);
749 * Deadlock detection:
751 * We attempt to detect deadlocks that are due purely to posix file
754 * We assume that a task can be waiting for at most one lock at a time.
755 * So for any acquired lock, the process holding that lock may be
756 * waiting on at most one other lock. That lock in turns may be held by
757 * someone waiting for at most one other lock. Given a requested lock
758 * caller_fl which is about to wait for a conflicting lock block_fl, we
759 * follow this chain of waiters to ensure we are not about to create a
762 * Since we do this before we ever put a process to sleep on a lock, we
763 * are ensured that there is never a cycle; that is what guarantees that
764 * the while() loop in posix_locks_deadlock() eventually completes.
766 * Note: the above assumption may not be true when handling lock
767 * requests from a broken NFS client. It may also fail in the presence
768 * of tasks (such as posix threads) sharing the same open file table.
769 * To handle those cases, we just bail out after a few iterations.
771 * For FL_OFDLCK locks, the owner is the filp, not the files_struct.
772 * Because the owner is not even nominally tied to a thread of
773 * execution, the deadlock detection below can't reasonably work well. Just
776 * In principle, we could do a more limited deadlock detection on FL_OFDLCK
777 * locks that just checks for the case where two tasks are attempting to
778 * upgrade from read to write locks on the same inode.
781 #define MAX_DEADLK_ITERATIONS 10
783 /* Find a lock that the owner of the given block_fl is blocking on. */
784 static struct file_lock
*what_owner_is_waiting_for(struct file_lock
*block_fl
)
786 struct file_lock
*fl
;
788 hash_for_each_possible(blocked_hash
, fl
, fl_link
, posix_owner_key(block_fl
)) {
789 if (posix_same_owner(fl
, block_fl
))
795 /* Must be called with the blocked_lock_lock held! */
796 static int posix_locks_deadlock(struct file_lock
*caller_fl
,
797 struct file_lock
*block_fl
)
802 * This deadlock detector can't reasonably detect deadlocks with
803 * FL_OFDLCK locks, since they aren't owned by a process, per-se.
805 if (IS_OFDLCK(caller_fl
))
808 while ((block_fl
= what_owner_is_waiting_for(block_fl
))) {
809 if (i
++ > MAX_DEADLK_ITERATIONS
)
811 if (posix_same_owner(caller_fl
, block_fl
))
817 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
818 * after any leases, but before any posix locks.
820 * Note that if called with an FL_EXISTS argument, the caller may determine
821 * whether or not a lock was successfully freed by testing the return
824 static int flock_lock_file(struct file
*filp
, struct file_lock
*request
)
826 struct file_lock
*new_fl
= NULL
;
827 struct file_lock
**before
;
828 struct inode
* inode
= file_inode(filp
);
833 if (!(request
->fl_flags
& FL_ACCESS
) && (request
->fl_type
!= F_UNLCK
)) {
834 new_fl
= locks_alloc_lock();
839 spin_lock(&inode
->i_lock
);
840 if (request
->fl_flags
& FL_ACCESS
)
843 for_each_lock(inode
, before
) {
844 struct file_lock
*fl
= *before
;
849 if (filp
!= fl
->fl_file
)
851 if (request
->fl_type
== fl
->fl_type
)
854 locks_delete_lock(before
, &dispose
);
858 if (request
->fl_type
== F_UNLCK
) {
859 if ((request
->fl_flags
& FL_EXISTS
) && !found
)
865 * If a higher-priority process was blocked on the old file lock,
866 * give it the opportunity to lock the file.
869 spin_unlock(&inode
->i_lock
);
871 spin_lock(&inode
->i_lock
);
875 for_each_lock(inode
, before
) {
876 struct file_lock
*fl
= *before
;
881 if (!flock_locks_conflict(request
, fl
))
884 if (!(request
->fl_flags
& FL_SLEEP
))
886 error
= FILE_LOCK_DEFERRED
;
887 locks_insert_block(fl
, request
);
890 if (request
->fl_flags
& FL_ACCESS
)
892 locks_copy_lock(new_fl
, request
);
893 locks_insert_lock(before
, new_fl
);
898 spin_unlock(&inode
->i_lock
);
900 locks_free_lock(new_fl
);
901 locks_dispose_list(&dispose
);
905 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
, struct file_lock
*conflock
)
907 struct file_lock
*fl
;
908 struct file_lock
*new_fl
= NULL
;
909 struct file_lock
*new_fl2
= NULL
;
910 struct file_lock
*left
= NULL
;
911 struct file_lock
*right
= NULL
;
912 struct file_lock
**before
;
918 * We may need two file_lock structures for this operation,
919 * so we get them in advance to avoid races.
921 * In some cases we can be sure, that no new locks will be needed
923 if (!(request
->fl_flags
& FL_ACCESS
) &&
924 (request
->fl_type
!= F_UNLCK
||
925 request
->fl_start
!= 0 || request
->fl_end
!= OFFSET_MAX
)) {
926 new_fl
= locks_alloc_lock();
927 new_fl2
= locks_alloc_lock();
930 spin_lock(&inode
->i_lock
);
932 * New lock request. Walk all POSIX locks and look for conflicts. If
933 * there are any, either return error or put the request on the
934 * blocker's list of waiters and the global blocked_hash.
936 if (request
->fl_type
!= F_UNLCK
) {
937 for_each_lock(inode
, before
) {
941 if (!posix_locks_conflict(request
, fl
))
944 __locks_copy_lock(conflock
, fl
);
946 if (!(request
->fl_flags
& FL_SLEEP
))
949 * Deadlock detection and insertion into the blocked
950 * locks list must be done while holding the same lock!
953 spin_lock(&blocked_lock_lock
);
954 if (likely(!posix_locks_deadlock(request
, fl
))) {
955 error
= FILE_LOCK_DEFERRED
;
956 __locks_insert_block(fl
, request
);
958 spin_unlock(&blocked_lock_lock
);
963 /* If we're just looking for a conflict, we're done. */
965 if (request
->fl_flags
& FL_ACCESS
)
969 * Find the first old lock with the same owner as the new lock.
972 before
= &inode
->i_flock
;
974 /* First skip locks owned by other processes. */
975 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
976 !posix_same_owner(request
, fl
))) {
977 before
= &fl
->fl_next
;
980 /* Process locks with this owner. */
981 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
982 /* Detect adjacent or overlapping regions (if same lock type)
984 if (request
->fl_type
== fl
->fl_type
) {
985 /* In all comparisons of start vs end, use
986 * "start - 1" rather than "end + 1". If end
987 * is OFFSET_MAX, end + 1 will become negative.
989 if (fl
->fl_end
< request
->fl_start
- 1)
991 /* If the next lock in the list has entirely bigger
992 * addresses than the new one, insert the lock here.
994 if (fl
->fl_start
- 1 > request
->fl_end
)
997 /* If we come here, the new and old lock are of the
998 * same type and adjacent or overlapping. Make one
999 * lock yielding from the lower start address of both
1000 * locks to the higher end address.
1002 if (fl
->fl_start
> request
->fl_start
)
1003 fl
->fl_start
= request
->fl_start
;
1005 request
->fl_start
= fl
->fl_start
;
1006 if (fl
->fl_end
< request
->fl_end
)
1007 fl
->fl_end
= request
->fl_end
;
1009 request
->fl_end
= fl
->fl_end
;
1011 locks_delete_lock(before
, &dispose
);
1018 /* Processing for different lock types is a bit
1021 if (fl
->fl_end
< request
->fl_start
)
1023 if (fl
->fl_start
> request
->fl_end
)
1025 if (request
->fl_type
== F_UNLCK
)
1027 if (fl
->fl_start
< request
->fl_start
)
1029 /* If the next lock in the list has a higher end
1030 * address than the new one, insert the new one here.
1032 if (fl
->fl_end
> request
->fl_end
) {
1036 if (fl
->fl_start
>= request
->fl_start
) {
1037 /* The new lock completely replaces an old
1038 * one (This may happen several times).
1041 locks_delete_lock(before
, &dispose
);
1045 * Replace the old lock with new_fl, and
1046 * remove the old one. It's safe to do the
1047 * insert here since we know that we won't be
1048 * using new_fl later, and that the lock is
1049 * just replacing an existing lock.
1054 locks_copy_lock(new_fl
, request
);
1057 locks_delete_lock(before
, &dispose
);
1058 locks_insert_lock(before
, request
);
1062 /* Go on to next lock.
1065 before
= &fl
->fl_next
;
1069 * The above code only modifies existing locks in case of merging or
1070 * replacing. If new lock(s) need to be inserted all modifications are
1071 * done below this, so it's safe yet to bail out.
1073 error
= -ENOLCK
; /* "no luck" */
1074 if (right
&& left
== right
&& !new_fl2
)
1079 if (request
->fl_type
== F_UNLCK
) {
1080 if (request
->fl_flags
& FL_EXISTS
)
1089 locks_copy_lock(new_fl
, request
);
1090 locks_insert_lock(before
, new_fl
);
1094 if (left
== right
) {
1095 /* The new lock breaks the old one in two pieces,
1096 * so we have to use the second new lock.
1100 locks_copy_lock(left
, right
);
1101 locks_insert_lock(before
, left
);
1103 right
->fl_start
= request
->fl_end
+ 1;
1104 locks_wake_up_blocks(right
);
1107 left
->fl_end
= request
->fl_start
- 1;
1108 locks_wake_up_blocks(left
);
1111 spin_unlock(&inode
->i_lock
);
1113 * Free any unused locks.
1116 locks_free_lock(new_fl
);
1118 locks_free_lock(new_fl2
);
1119 locks_dispose_list(&dispose
);
1124 * posix_lock_file - Apply a POSIX-style lock to a file
1125 * @filp: The file to apply the lock to
1126 * @fl: The lock to be applied
1127 * @conflock: Place to return a copy of the conflicting lock, if found.
1129 * Add a POSIX style lock to a file.
1130 * We merge adjacent & overlapping locks whenever possible.
1131 * POSIX locks are sorted by owner task, then by starting address
1133 * Note that if called with an FL_EXISTS argument, the caller may determine
1134 * whether or not a lock was successfully freed by testing the return
1135 * value for -ENOENT.
1137 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
,
1138 struct file_lock
*conflock
)
1140 return __posix_lock_file(file_inode(filp
), fl
, conflock
);
1142 EXPORT_SYMBOL(posix_lock_file
);
1145 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1146 * @filp: The file to apply the lock to
1147 * @fl: The lock to be applied
1149 * Add a POSIX style lock to a file.
1150 * We merge adjacent & overlapping locks whenever possible.
1151 * POSIX locks are sorted by owner task, then by starting address
1153 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1158 error
= posix_lock_file(filp
, fl
, NULL
);
1159 if (error
!= FILE_LOCK_DEFERRED
)
1161 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1165 locks_delete_block(fl
);
1170 EXPORT_SYMBOL(posix_lock_file_wait
);
1173 * locks_mandatory_locked - Check for an active lock
1174 * @file: the file to check
1176 * Searches the inode's list of locks to find any POSIX locks which conflict.
1177 * This function is called from locks_verify_locked() only.
1179 int locks_mandatory_locked(struct file
*file
)
1181 struct inode
*inode
= file_inode(file
);
1182 struct file_lock
*fl
;
1185 * Search the lock list for this inode for any POSIX locks.
1187 spin_lock(&inode
->i_lock
);
1188 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1191 if (fl
->fl_owner
!= current
->files
&&
1192 fl
->fl_owner
!= file
)
1195 spin_unlock(&inode
->i_lock
);
1196 return fl
? -EAGAIN
: 0;
1200 * locks_mandatory_area - Check for a conflicting lock
1201 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1203 * @inode: the file to check
1204 * @filp: how the file was opened (if it was)
1205 * @offset: start of area to check
1206 * @count: length of area to check
1208 * Searches the inode's list of locks to find any POSIX locks which conflict.
1209 * This function is called from rw_verify_area() and
1210 * locks_verify_truncate().
1212 int locks_mandatory_area(int read_write
, struct inode
*inode
,
1213 struct file
*filp
, loff_t offset
,
1216 struct file_lock fl
;
1220 locks_init_lock(&fl
);
1221 fl
.fl_pid
= current
->tgid
;
1223 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
1224 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
1226 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
1227 fl
.fl_start
= offset
;
1228 fl
.fl_end
= offset
+ count
- 1;
1233 fl
.fl_flags
&= ~FL_SLEEP
;
1234 error
= __posix_lock_file(inode
, &fl
, NULL
);
1240 fl
.fl_flags
|= FL_SLEEP
;
1241 fl
.fl_owner
= current
->files
;
1242 error
= __posix_lock_file(inode
, &fl
, NULL
);
1243 if (error
!= FILE_LOCK_DEFERRED
)
1245 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1248 * If we've been sleeping someone might have
1249 * changed the permissions behind our back.
1251 if (__mandatory_lock(inode
))
1255 locks_delete_block(&fl
);
1262 EXPORT_SYMBOL(locks_mandatory_area
);
1264 static void lease_clear_pending(struct file_lock
*fl
, int arg
)
1268 fl
->fl_flags
&= ~FL_UNLOCK_PENDING
;
1271 fl
->fl_flags
&= ~FL_DOWNGRADE_PENDING
;
1275 /* We already had a lease on this file; just change its type */
1276 int lease_modify(struct file_lock
**before
, int arg
)
1278 struct file_lock
*fl
= *before
;
1279 int error
= assign_type(fl
, arg
);
1283 lease_clear_pending(fl
, arg
);
1284 locks_wake_up_blocks(fl
);
1285 if (arg
== F_UNLCK
) {
1286 struct file
*filp
= fl
->fl_file
;
1289 filp
->f_owner
.signum
= 0;
1290 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
1291 if (fl
->fl_fasync
!= NULL
) {
1292 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
1293 fl
->fl_fasync
= NULL
;
1295 locks_delete_lock(before
, NULL
);
1300 EXPORT_SYMBOL(lease_modify
);
1302 static bool past_time(unsigned long then
)
1305 /* 0 is a special value meaning "this never expires": */
1307 return time_after(jiffies
, then
);
1310 static void time_out_leases(struct inode
*inode
)
1312 struct file_lock
**before
;
1313 struct file_lock
*fl
;
1315 before
= &inode
->i_flock
;
1316 while ((fl
= *before
) && IS_LEASE(fl
) && lease_breaking(fl
)) {
1317 trace_time_out_leases(inode
, fl
);
1318 if (past_time(fl
->fl_downgrade_time
))
1319 lease_modify(before
, F_RDLCK
);
1320 if (past_time(fl
->fl_break_time
))
1321 lease_modify(before
, F_UNLCK
);
1322 if (fl
== *before
) /* lease_modify may have freed fl */
1323 before
= &fl
->fl_next
;
1327 static bool leases_conflict(struct file_lock
*lease
, struct file_lock
*breaker
)
1329 if ((breaker
->fl_flags
& FL_DELEG
) && (lease
->fl_flags
& FL_LEASE
))
1331 return locks_conflict(breaker
, lease
);
1335 * __break_lease - revoke all outstanding leases on file
1336 * @inode: the inode of the file to return
1337 * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR:
1339 * @type: FL_LEASE: break leases and delegations; FL_DELEG: break
1342 * break_lease (inlined for speed) has checked there already is at least
1343 * some kind of lock (maybe a lease) on this file. Leases are broken on
1344 * a call to open() or truncate(). This function can sleep unless you
1345 * specified %O_NONBLOCK to your open().
1347 int __break_lease(struct inode
*inode
, unsigned int mode
, unsigned int type
)
1350 struct file_lock
*new_fl
, *flock
;
1351 struct file_lock
*fl
;
1352 unsigned long break_time
;
1353 int i_have_this_lease
= 0;
1354 bool lease_conflict
= false;
1355 int want_write
= (mode
& O_ACCMODE
) != O_RDONLY
;
1357 new_fl
= lease_alloc(NULL
, want_write
? F_WRLCK
: F_RDLCK
);
1359 return PTR_ERR(new_fl
);
1360 new_fl
->fl_flags
= type
;
1362 spin_lock(&inode
->i_lock
);
1364 time_out_leases(inode
);
1366 flock
= inode
->i_flock
;
1367 if ((flock
== NULL
) || !IS_LEASE(flock
))
1370 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1371 if (leases_conflict(fl
, new_fl
)) {
1372 lease_conflict
= true;
1373 if (fl
->fl_owner
== current
->files
)
1374 i_have_this_lease
= 1;
1377 if (!lease_conflict
)
1381 if (lease_break_time
> 0) {
1382 break_time
= jiffies
+ lease_break_time
* HZ
;
1383 if (break_time
== 0)
1384 break_time
++; /* so that 0 means no break time */
1387 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1388 if (!leases_conflict(fl
, new_fl
))
1391 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1393 fl
->fl_flags
|= FL_UNLOCK_PENDING
;
1394 fl
->fl_break_time
= break_time
;
1396 if (lease_breaking(flock
))
1398 fl
->fl_flags
|= FL_DOWNGRADE_PENDING
;
1399 fl
->fl_downgrade_time
= break_time
;
1401 fl
->fl_lmops
->lm_break(fl
);
1404 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1405 trace_break_lease_noblock(inode
, new_fl
);
1406 error
= -EWOULDBLOCK
;
1411 break_time
= flock
->fl_break_time
;
1412 if (break_time
!= 0)
1413 break_time
-= jiffies
;
1414 if (break_time
== 0)
1416 locks_insert_block(flock
, new_fl
);
1417 trace_break_lease_block(inode
, new_fl
);
1418 spin_unlock(&inode
->i_lock
);
1419 error
= wait_event_interruptible_timeout(new_fl
->fl_wait
,
1420 !new_fl
->fl_next
, break_time
);
1421 spin_lock(&inode
->i_lock
);
1422 trace_break_lease_unblock(inode
, new_fl
);
1423 locks_delete_block(new_fl
);
1426 time_out_leases(inode
);
1428 * Wait for the next conflicting lease that has not been
1431 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1432 flock
= flock
->fl_next
) {
1433 if (leases_conflict(new_fl
, flock
))
1440 spin_unlock(&inode
->i_lock
);
1441 locks_free_lock(new_fl
);
1445 EXPORT_SYMBOL(__break_lease
);
1448 * lease_get_mtime - get the last modified time of an inode
1450 * @time: pointer to a timespec which will contain the last modified time
1452 * This is to force NFS clients to flush their caches for files with
1453 * exclusive leases. The justification is that if someone has an
1454 * exclusive lease, then they could be modifying it.
1456 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1458 struct file_lock
*flock
= inode
->i_flock
;
1459 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
== F_WRLCK
))
1460 *time
= current_fs_time(inode
->i_sb
);
1462 *time
= inode
->i_mtime
;
1465 EXPORT_SYMBOL(lease_get_mtime
);
1468 * fcntl_getlease - Enquire what lease is currently active
1471 * The value returned by this function will be one of
1472 * (if no lease break is pending):
1474 * %F_RDLCK to indicate a shared lease is held.
1476 * %F_WRLCK to indicate an exclusive lease is held.
1478 * %F_UNLCK to indicate no lease is held.
1480 * (if a lease break is pending):
1482 * %F_RDLCK to indicate an exclusive lease needs to be
1483 * changed to a shared lease (or removed).
1485 * %F_UNLCK to indicate the lease needs to be removed.
1487 * XXX: sfr & willy disagree over whether F_INPROGRESS
1488 * should be returned to userspace.
1490 int fcntl_getlease(struct file
*filp
)
1492 struct file_lock
*fl
;
1493 struct inode
*inode
= file_inode(filp
);
1496 spin_lock(&inode
->i_lock
);
1497 time_out_leases(file_inode(filp
));
1498 for (fl
= file_inode(filp
)->i_flock
; fl
&& IS_LEASE(fl
);
1500 if (fl
->fl_file
== filp
) {
1501 type
= target_leasetype(fl
);
1505 spin_unlock(&inode
->i_lock
);
1510 * check_conflicting_open - see if the given dentry points to a file that has
1511 * an existing open that would conflict with the
1513 * @dentry: dentry to check
1514 * @arg: type of lease that we're trying to acquire
1516 * Check to see if there's an existing open fd on this file that would
1517 * conflict with the lease we're trying to set.
1520 check_conflicting_open(const struct dentry
*dentry
, const long arg
)
1523 struct inode
*inode
= dentry
->d_inode
;
1525 if ((arg
== F_RDLCK
) && (atomic_read(&inode
->i_writecount
) > 0))
1528 if ((arg
== F_WRLCK
) && ((d_count(dentry
) > 1) ||
1529 (atomic_read(&inode
->i_count
) > 1)))
1535 static int generic_add_lease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1537 struct file_lock
*fl
, **before
, **my_before
= NULL
, *lease
;
1538 struct dentry
*dentry
= filp
->f_path
.dentry
;
1539 struct inode
*inode
= dentry
->d_inode
;
1540 bool is_deleg
= (*flp
)->fl_flags
& FL_DELEG
;
1544 trace_generic_add_lease(inode
, lease
);
1547 * In the delegation case we need mutual exclusion with
1548 * a number of operations that take the i_mutex. We trylock
1549 * because delegations are an optional optimization, and if
1550 * there's some chance of a conflict--we'd rather not
1551 * bother, maybe that's a sign this just isn't a good file to
1552 * hand out a delegation on.
1554 if (is_deleg
&& !mutex_trylock(&inode
->i_mutex
))
1557 if (is_deleg
&& arg
== F_WRLCK
) {
1558 /* Write delegations are not currently supported: */
1559 mutex_unlock(&inode
->i_mutex
);
1564 error
= check_conflicting_open(dentry
, arg
);
1569 * At this point, we know that if there is an exclusive
1570 * lease on this file, then we hold it on this filp
1571 * (otherwise our open of this file would have blocked).
1572 * And if we are trying to acquire an exclusive lease,
1573 * then the file is not open by anyone (including us)
1574 * except for this filp.
1577 for (before
= &inode
->i_flock
;
1578 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1579 before
= &fl
->fl_next
) {
1580 if (fl
->fl_file
== filp
) {
1585 * No exclusive leases if someone else has a lease on
1591 * Modifying our existing lease is OK, but no getting a
1592 * new lease if someone else is opening for write:
1594 if (fl
->fl_flags
& FL_UNLOCK_PENDING
)
1598 if (my_before
!= NULL
) {
1599 error
= lease
->fl_lmops
->lm_change(my_before
, arg
);
1609 locks_insert_lock(before
, lease
);
1611 * The check in break_lease() is lockless. It's possible for another
1612 * open to race in after we did the earlier check for a conflicting
1613 * open but before the lease was inserted. Check again for a
1614 * conflicting open and cancel the lease if there is one.
1616 * We also add a barrier here to ensure that the insertion of the lock
1617 * precedes these checks.
1620 error
= check_conflicting_open(dentry
, arg
);
1622 locks_unlink_lock(before
);
1625 mutex_unlock(&inode
->i_mutex
);
1629 static int generic_delete_lease(struct file
*filp
, struct file_lock
**flp
)
1631 struct file_lock
*fl
, **before
;
1632 struct dentry
*dentry
= filp
->f_path
.dentry
;
1633 struct inode
*inode
= dentry
->d_inode
;
1635 trace_generic_delete_lease(inode
, *flp
);
1637 for (before
= &inode
->i_flock
;
1638 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1639 before
= &fl
->fl_next
) {
1640 if (fl
->fl_file
!= filp
)
1642 return (*flp
)->fl_lmops
->lm_change(before
, F_UNLCK
);
1648 * generic_setlease - sets a lease on an open file
1649 * @filp: file pointer
1650 * @arg: type of lease to obtain
1651 * @flp: input - file_lock to use, output - file_lock inserted
1653 * The (input) flp->fl_lmops->lm_break function is required
1656 * Called with inode->i_lock held.
1658 int generic_setlease(struct file
*filp
, long arg
, struct file_lock
**flp
)
1660 struct dentry
*dentry
= filp
->f_path
.dentry
;
1661 struct inode
*inode
= dentry
->d_inode
;
1664 if ((!uid_eq(current_fsuid(), inode
->i_uid
)) && !capable(CAP_LEASE
))
1666 if (!S_ISREG(inode
->i_mode
))
1668 error
= security_file_lock(filp
, arg
);
1672 time_out_leases(inode
);
1674 BUG_ON(!(*flp
)->fl_lmops
->lm_break
);
1678 return generic_delete_lease(filp
, flp
);
1681 return generic_add_lease(filp
, arg
, flp
);
1686 EXPORT_SYMBOL(generic_setlease
);
1688 static int __vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1690 if (filp
->f_op
->setlease
)
1691 return filp
->f_op
->setlease(filp
, arg
, lease
);
1693 return generic_setlease(filp
, arg
, lease
);
1697 * vfs_setlease - sets a lease on an open file
1698 * @filp: file pointer
1699 * @arg: type of lease to obtain
1700 * @lease: file_lock to use
1702 * Call this to establish a lease on the file.
1703 * The (*lease)->fl_lmops->lm_break operation must be set; if not,
1704 * break_lease will oops!
1706 * This will call the filesystem's setlease file method, if
1707 * defined. Note that there is no getlease method; instead, the
1708 * filesystem setlease method should call back to setlease() to
1709 * add a lease to the inode's lease list, where fcntl_getlease() can
1710 * find it. Since fcntl_getlease() only reports whether the current
1711 * task holds a lease, a cluster filesystem need only do this for
1712 * leases held by processes on this node.
1714 * There is also no break_lease method; filesystems that
1715 * handle their own leases should break leases themselves from the
1716 * filesystem's open, create, and (on truncate) setattr methods.
1718 * Warning: the only current setlease methods exist only to disable
1719 * leases in certain cases. More vfs changes may be required to
1720 * allow a full filesystem lease implementation.
1723 int vfs_setlease(struct file
*filp
, long arg
, struct file_lock
**lease
)
1725 struct inode
*inode
= file_inode(filp
);
1728 spin_lock(&inode
->i_lock
);
1729 error
= __vfs_setlease(filp
, arg
, lease
);
1730 spin_unlock(&inode
->i_lock
);
1734 EXPORT_SYMBOL_GPL(vfs_setlease
);
1736 static int do_fcntl_delete_lease(struct file
*filp
)
1738 struct file_lock fl
, *flp
= &fl
;
1740 lease_init(filp
, F_UNLCK
, flp
);
1742 return vfs_setlease(filp
, F_UNLCK
, &flp
);
1745 static int do_fcntl_add_lease(unsigned int fd
, struct file
*filp
, long arg
)
1747 struct file_lock
*fl
, *ret
;
1748 struct inode
*inode
= file_inode(filp
);
1749 struct fasync_struct
*new;
1752 fl
= lease_alloc(filp
, arg
);
1756 new = fasync_alloc();
1758 locks_free_lock(fl
);
1762 spin_lock(&inode
->i_lock
);
1763 error
= __vfs_setlease(filp
, arg
, &ret
);
1770 * fasync_insert_entry() returns the old entry if any.
1771 * If there was no old entry, then it used 'new' and
1772 * inserted it into the fasync list. Clear new so that
1773 * we don't release it here.
1775 if (!fasync_insert_entry(fd
, filp
, &ret
->fl_fasync
, new))
1778 error
= __f_setown(filp
, task_pid(current
), PIDTYPE_PID
, 0);
1780 spin_unlock(&inode
->i_lock
);
1782 locks_free_lock(fl
);
1789 * fcntl_setlease - sets a lease on an open file
1790 * @fd: open file descriptor
1791 * @filp: file pointer
1792 * @arg: type of lease to obtain
1794 * Call this fcntl to establish a lease on the file.
1795 * Note that you also need to call %F_SETSIG to
1796 * receive a signal when the lease is broken.
1798 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1801 return do_fcntl_delete_lease(filp
);
1802 return do_fcntl_add_lease(fd
, filp
, arg
);
1806 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1807 * @filp: The file to apply the lock to
1808 * @fl: The lock to be applied
1810 * Add a FLOCK style lock to a file.
1812 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1817 error
= flock_lock_file(filp
, fl
);
1818 if (error
!= FILE_LOCK_DEFERRED
)
1820 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1824 locks_delete_block(fl
);
1830 EXPORT_SYMBOL(flock_lock_file_wait
);
1833 * sys_flock: - flock() system call.
1834 * @fd: the file descriptor to lock.
1835 * @cmd: the type of lock to apply.
1837 * Apply a %FL_FLOCK style lock to an open file descriptor.
1838 * The @cmd can be one of
1840 * %LOCK_SH -- a shared lock.
1842 * %LOCK_EX -- an exclusive lock.
1844 * %LOCK_UN -- remove an existing lock.
1846 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1848 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1849 * processes read and write access respectively.
1851 SYSCALL_DEFINE2(flock
, unsigned int, fd
, unsigned int, cmd
)
1853 struct fd f
= fdget(fd
);
1854 struct file_lock
*lock
;
1855 int can_sleep
, unlock
;
1862 can_sleep
= !(cmd
& LOCK_NB
);
1864 unlock
= (cmd
== LOCK_UN
);
1866 if (!unlock
&& !(cmd
& LOCK_MAND
) &&
1867 !(f
.file
->f_mode
& (FMODE_READ
|FMODE_WRITE
)))
1870 error
= flock_make_lock(f
.file
, &lock
, cmd
);
1874 lock
->fl_flags
|= FL_SLEEP
;
1876 error
= security_file_lock(f
.file
, lock
->fl_type
);
1880 if (f
.file
->f_op
->flock
)
1881 error
= f
.file
->f_op
->flock(f
.file
,
1882 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1885 error
= flock_lock_file_wait(f
.file
, lock
);
1888 locks_free_lock(lock
);
1897 * vfs_test_lock - test file byte range lock
1898 * @filp: The file to test lock for
1899 * @fl: The lock to test; also used to hold result
1901 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1902 * setting conf->fl_type to something other than F_UNLCK.
1904 int vfs_test_lock(struct file
*filp
, struct file_lock
*fl
)
1906 if (filp
->f_op
->lock
)
1907 return filp
->f_op
->lock(filp
, F_GETLK
, fl
);
1908 posix_test_lock(filp
, fl
);
1911 EXPORT_SYMBOL_GPL(vfs_test_lock
);
1913 static int posix_lock_to_flock(struct flock
*flock
, struct file_lock
*fl
)
1915 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1916 #if BITS_PER_LONG == 32
1918 * Make sure we can represent the posix lock via
1919 * legacy 32bit flock.
1921 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1923 if (fl
->fl_end
!= OFFSET_MAX
&& fl
->fl_end
> OFFT_OFFSET_MAX
)
1926 flock
->l_start
= fl
->fl_start
;
1927 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1928 fl
->fl_end
- fl
->fl_start
+ 1;
1929 flock
->l_whence
= 0;
1930 flock
->l_type
= fl
->fl_type
;
1934 #if BITS_PER_LONG == 32
1935 static void posix_lock_to_flock64(struct flock64
*flock
, struct file_lock
*fl
)
1937 flock
->l_pid
= IS_OFDLCK(fl
) ? -1 : fl
->fl_pid
;
1938 flock
->l_start
= fl
->fl_start
;
1939 flock
->l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1940 fl
->fl_end
- fl
->fl_start
+ 1;
1941 flock
->l_whence
= 0;
1942 flock
->l_type
= fl
->fl_type
;
1946 /* Report the first existing lock that would conflict with l.
1947 * This implements the F_GETLK command of fcntl().
1949 int fcntl_getlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
1951 struct file_lock file_lock
;
1956 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1959 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1962 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1966 if (cmd
== F_OFD_GETLK
) {
1968 if (flock
.l_pid
!= 0)
1972 file_lock
.fl_flags
|= FL_OFDLCK
;
1973 file_lock
.fl_owner
= filp
;
1976 error
= vfs_test_lock(filp
, &file_lock
);
1980 flock
.l_type
= file_lock
.fl_type
;
1981 if (file_lock
.fl_type
!= F_UNLCK
) {
1982 error
= posix_lock_to_flock(&flock
, &file_lock
);
1987 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1994 * vfs_lock_file - file byte range lock
1995 * @filp: The file to apply the lock to
1996 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1997 * @fl: The lock to be applied
1998 * @conf: Place to return a copy of the conflicting lock, if found.
2000 * A caller that doesn't care about the conflicting lock may pass NULL
2001 * as the final argument.
2003 * If the filesystem defines a private ->lock() method, then @conf will
2004 * be left unchanged; so a caller that cares should initialize it to
2005 * some acceptable default.
2007 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
2008 * locks, the ->lock() interface may return asynchronously, before the lock has
2009 * been granted or denied by the underlying filesystem, if (and only if)
2010 * lm_grant is set. Callers expecting ->lock() to return asynchronously
2011 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
2012 * the request is for a blocking lock. When ->lock() does return asynchronously,
2013 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
2014 * request completes.
2015 * If the request is for non-blocking lock the file system should return
2016 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
2017 * with the result. If the request timed out the callback routine will return a
2018 * nonzero return code and the file system should release the lock. The file
2019 * system is also responsible to keep a corresponding posix lock when it
2020 * grants a lock so the VFS can find out which locks are locally held and do
2021 * the correct lock cleanup when required.
2022 * The underlying filesystem must not drop the kernel lock or call
2023 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
2026 int vfs_lock_file(struct file
*filp
, unsigned int cmd
, struct file_lock
*fl
, struct file_lock
*conf
)
2028 if (filp
->f_op
->lock
)
2029 return filp
->f_op
->lock(filp
, cmd
, fl
);
2031 return posix_lock_file(filp
, fl
, conf
);
2033 EXPORT_SYMBOL_GPL(vfs_lock_file
);
2035 static int do_lock_file_wait(struct file
*filp
, unsigned int cmd
,
2036 struct file_lock
*fl
)
2040 error
= security_file_lock(filp
, fl
->fl_type
);
2045 error
= vfs_lock_file(filp
, cmd
, fl
, NULL
);
2046 if (error
!= FILE_LOCK_DEFERRED
)
2048 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
2052 locks_delete_block(fl
);
2059 /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */
2061 check_fmode_for_setlk(struct file_lock
*fl
)
2063 switch (fl
->fl_type
) {
2065 if (!(fl
->fl_file
->f_mode
& FMODE_READ
))
2069 if (!(fl
->fl_file
->f_mode
& FMODE_WRITE
))
2075 /* Apply the lock described by l to an open file descriptor.
2076 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2078 int fcntl_setlk(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2079 struct flock __user
*l
)
2081 struct file_lock
*file_lock
= locks_alloc_lock();
2083 struct inode
*inode
;
2087 if (file_lock
== NULL
)
2091 * This might block, so we do it before checking the inode.
2094 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2097 inode
= file_inode(filp
);
2099 /* Don't allow mandatory locks on files that may be memory mapped
2102 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2108 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
2112 error
= check_fmode_for_setlk(file_lock
);
2117 * If the cmd is requesting file-private locks, then set the
2118 * FL_OFDLCK flag and override the owner.
2123 if (flock
.l_pid
!= 0)
2127 file_lock
->fl_flags
|= FL_OFDLCK
;
2128 file_lock
->fl_owner
= filp
;
2132 if (flock
.l_pid
!= 0)
2136 file_lock
->fl_flags
|= FL_OFDLCK
;
2137 file_lock
->fl_owner
= filp
;
2140 file_lock
->fl_flags
|= FL_SLEEP
;
2143 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2146 * Attempt to detect a close/fcntl race and recover by
2147 * releasing the lock that was just acquired.
2150 * we need that spin_lock here - it prevents reordering between
2151 * update of inode->i_flock and check for it done in close().
2152 * rcu_read_lock() wouldn't do.
2154 spin_lock(¤t
->files
->file_lock
);
2156 spin_unlock(¤t
->files
->file_lock
);
2157 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2158 flock
.l_type
= F_UNLCK
;
2163 locks_free_lock(file_lock
);
2167 #if BITS_PER_LONG == 32
2168 /* Report the first existing lock that would conflict with l.
2169 * This implements the F_GETLK command of fcntl().
2171 int fcntl_getlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
2173 struct file_lock file_lock
;
2174 struct flock64 flock
;
2178 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2181 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
2184 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
2188 if (cmd
== F_OFD_GETLK
) {
2190 if (flock
.l_pid
!= 0)
2194 file_lock
.fl_flags
|= FL_OFDLCK
;
2195 file_lock
.fl_owner
= filp
;
2198 error
= vfs_test_lock(filp
, &file_lock
);
2202 flock
.l_type
= file_lock
.fl_type
;
2203 if (file_lock
.fl_type
!= F_UNLCK
)
2204 posix_lock_to_flock64(&flock
, &file_lock
);
2207 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
2214 /* Apply the lock described by l to an open file descriptor.
2215 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
2217 int fcntl_setlk64(unsigned int fd
, struct file
*filp
, unsigned int cmd
,
2218 struct flock64 __user
*l
)
2220 struct file_lock
*file_lock
= locks_alloc_lock();
2221 struct flock64 flock
;
2222 struct inode
*inode
;
2226 if (file_lock
== NULL
)
2230 * This might block, so we do it before checking the inode.
2233 if (copy_from_user(&flock
, l
, sizeof(flock
)))
2236 inode
= file_inode(filp
);
2238 /* Don't allow mandatory locks on files that may be memory mapped
2241 if (mandatory_lock(inode
) && mapping_writably_mapped(filp
->f_mapping
)) {
2247 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
2251 error
= check_fmode_for_setlk(file_lock
);
2256 * If the cmd is requesting file-private locks, then set the
2257 * FL_OFDLCK flag and override the owner.
2262 if (flock
.l_pid
!= 0)
2266 file_lock
->fl_flags
|= FL_OFDLCK
;
2267 file_lock
->fl_owner
= filp
;
2271 if (flock
.l_pid
!= 0)
2275 file_lock
->fl_flags
|= FL_OFDLCK
;
2276 file_lock
->fl_owner
= filp
;
2279 file_lock
->fl_flags
|= FL_SLEEP
;
2282 error
= do_lock_file_wait(filp
, cmd
, file_lock
);
2285 * Attempt to detect a close/fcntl race and recover by
2286 * releasing the lock that was just acquired.
2288 spin_lock(¤t
->files
->file_lock
);
2290 spin_unlock(¤t
->files
->file_lock
);
2291 if (!error
&& f
!= filp
&& flock
.l_type
!= F_UNLCK
) {
2292 flock
.l_type
= F_UNLCK
;
2297 locks_free_lock(file_lock
);
2300 #endif /* BITS_PER_LONG == 32 */
2303 * This function is called when the file is being removed
2304 * from the task's fd array. POSIX locks belonging to this task
2305 * are deleted at this time.
2307 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
2309 struct file_lock lock
;
2312 * If there are no locks held on this file, we don't need to call
2313 * posix_lock_file(). Another process could be setting a lock on this
2314 * file at the same time, but we wouldn't remove that lock anyway.
2316 if (!file_inode(filp
)->i_flock
)
2319 lock
.fl_type
= F_UNLCK
;
2320 lock
.fl_flags
= FL_POSIX
| FL_CLOSE
;
2322 lock
.fl_end
= OFFSET_MAX
;
2323 lock
.fl_owner
= owner
;
2324 lock
.fl_pid
= current
->tgid
;
2325 lock
.fl_file
= filp
;
2327 lock
.fl_lmops
= NULL
;
2329 vfs_lock_file(filp
, F_SETLK
, &lock
, NULL
);
2331 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
2332 lock
.fl_ops
->fl_release_private(&lock
);
2335 EXPORT_SYMBOL(locks_remove_posix
);
2338 * This function is called on the last close of an open file.
2340 void locks_remove_file(struct file
*filp
)
2342 struct inode
* inode
= file_inode(filp
);
2343 struct file_lock
*fl
;
2344 struct file_lock
**before
;
2347 if (!inode
->i_flock
)
2350 locks_remove_posix(filp
, filp
);
2352 if (filp
->f_op
->flock
) {
2353 struct file_lock fl
= {
2355 .fl_pid
= current
->tgid
,
2357 .fl_flags
= FL_FLOCK
,
2359 .fl_end
= OFFSET_MAX
,
2361 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
2362 if (fl
.fl_ops
&& fl
.fl_ops
->fl_release_private
)
2363 fl
.fl_ops
->fl_release_private(&fl
);
2366 spin_lock(&inode
->i_lock
);
2367 before
= &inode
->i_flock
;
2369 while ((fl
= *before
) != NULL
) {
2370 if (fl
->fl_file
== filp
) {
2372 lease_modify(before
, F_UNLCK
);
2377 * There's a leftover lock on the list of a type that
2378 * we didn't expect to see. Most likely a classic
2379 * POSIX lock that ended up not getting released
2380 * properly, or that raced onto the list somehow. Log
2381 * some info about it and then just remove it from
2385 "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n",
2386 MAJOR(inode
->i_sb
->s_dev
),
2387 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
,
2388 fl
->fl_type
, fl
->fl_flags
,
2389 fl
->fl_start
, fl
->fl_end
);
2391 locks_delete_lock(before
, &dispose
);
2394 before
= &fl
->fl_next
;
2396 spin_unlock(&inode
->i_lock
);
2397 locks_dispose_list(&dispose
);
2401 * posix_unblock_lock - stop waiting for a file lock
2402 * @waiter: the lock which was waiting
2404 * lockd needs to block waiting for locks.
2407 posix_unblock_lock(struct file_lock
*waiter
)
2411 spin_lock(&blocked_lock_lock
);
2412 if (waiter
->fl_next
)
2413 __locks_delete_block(waiter
);
2416 spin_unlock(&blocked_lock_lock
);
2419 EXPORT_SYMBOL(posix_unblock_lock
);
2422 * vfs_cancel_lock - file byte range unblock lock
2423 * @filp: The file to apply the unblock to
2424 * @fl: The lock to be unblocked
2426 * Used by lock managers to cancel blocked requests
2428 int vfs_cancel_lock(struct file
*filp
, struct file_lock
*fl
)
2430 if (filp
->f_op
->lock
)
2431 return filp
->f_op
->lock(filp
, F_CANCELLK
, fl
);
2435 EXPORT_SYMBOL_GPL(vfs_cancel_lock
);
2437 #ifdef CONFIG_PROC_FS
2438 #include <linux/proc_fs.h>
2439 #include <linux/seq_file.h>
2441 struct locks_iterator
{
2446 static void lock_get_status(struct seq_file
*f
, struct file_lock
*fl
,
2447 loff_t id
, char *pfx
)
2449 struct inode
*inode
= NULL
;
2450 unsigned int fl_pid
;
2453 fl_pid
= pid_vnr(fl
->fl_nspid
);
2455 fl_pid
= fl
->fl_pid
;
2457 if (fl
->fl_file
!= NULL
)
2458 inode
= file_inode(fl
->fl_file
);
2460 seq_printf(f
, "%lld:%s ", id
, pfx
);
2462 if (fl
->fl_flags
& FL_ACCESS
)
2463 seq_puts(f
, "ACCESS");
2464 else if (IS_OFDLCK(fl
))
2465 seq_puts(f
, "OFDLCK");
2467 seq_puts(f
, "POSIX ");
2469 seq_printf(f
, " %s ",
2470 (inode
== NULL
) ? "*NOINODE*" :
2471 mandatory_lock(inode
) ? "MANDATORY" : "ADVISORY ");
2472 } else if (IS_FLOCK(fl
)) {
2473 if (fl
->fl_type
& LOCK_MAND
) {
2474 seq_puts(f
, "FLOCK MSNFS ");
2476 seq_puts(f
, "FLOCK ADVISORY ");
2478 } else if (IS_LEASE(fl
)) {
2479 if (fl
->fl_flags
& FL_DELEG
)
2480 seq_puts(f
, "DELEG ");
2482 seq_puts(f
, "LEASE ");
2484 if (lease_breaking(fl
))
2485 seq_puts(f
, "BREAKING ");
2486 else if (fl
->fl_file
)
2487 seq_puts(f
, "ACTIVE ");
2489 seq_puts(f
, "BREAKER ");
2491 seq_puts(f
, "UNKNOWN UNKNOWN ");
2493 if (fl
->fl_type
& LOCK_MAND
) {
2494 seq_printf(f
, "%s ",
2495 (fl
->fl_type
& LOCK_READ
)
2496 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
2497 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
2499 seq_printf(f
, "%s ",
2500 (lease_breaking(fl
))
2501 ? (fl
->fl_type
== F_UNLCK
) ? "UNLCK" : "READ "
2502 : (fl
->fl_type
== F_WRLCK
) ? "WRITE" : "READ ");
2505 #ifdef WE_CAN_BREAK_LSLK_NOW
2506 seq_printf(f
, "%d %s:%ld ", fl_pid
,
2507 inode
->i_sb
->s_id
, inode
->i_ino
);
2509 /* userspace relies on this representation of dev_t ;-( */
2510 seq_printf(f
, "%d %02x:%02x:%ld ", fl_pid
,
2511 MAJOR(inode
->i_sb
->s_dev
),
2512 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
2515 seq_printf(f
, "%d <none>:0 ", fl_pid
);
2518 if (fl
->fl_end
== OFFSET_MAX
)
2519 seq_printf(f
, "%Ld EOF\n", fl
->fl_start
);
2521 seq_printf(f
, "%Ld %Ld\n", fl
->fl_start
, fl
->fl_end
);
2523 seq_puts(f
, "0 EOF\n");
2527 static int locks_show(struct seq_file
*f
, void *v
)
2529 struct locks_iterator
*iter
= f
->private;
2530 struct file_lock
*fl
, *bfl
;
2532 fl
= hlist_entry(v
, struct file_lock
, fl_link
);
2534 lock_get_status(f
, fl
, iter
->li_pos
, "");
2536 list_for_each_entry(bfl
, &fl
->fl_block
, fl_block
)
2537 lock_get_status(f
, bfl
, iter
->li_pos
, " ->");
2542 static void *locks_start(struct seq_file
*f
, loff_t
*pos
)
2543 __acquires(&blocked_lock_lock
)
2545 struct locks_iterator
*iter
= f
->private;
2547 iter
->li_pos
= *pos
+ 1;
2548 lg_global_lock(&file_lock_lglock
);
2549 spin_lock(&blocked_lock_lock
);
2550 return seq_hlist_start_percpu(&file_lock_list
, &iter
->li_cpu
, *pos
);
2553 static void *locks_next(struct seq_file
*f
, void *v
, loff_t
*pos
)
2555 struct locks_iterator
*iter
= f
->private;
2558 return seq_hlist_next_percpu(v
, &file_lock_list
, &iter
->li_cpu
, pos
);
2561 static void locks_stop(struct seq_file
*f
, void *v
)
2562 __releases(&blocked_lock_lock
)
2564 spin_unlock(&blocked_lock_lock
);
2565 lg_global_unlock(&file_lock_lglock
);
2568 static const struct seq_operations locks_seq_operations
= {
2569 .start
= locks_start
,
2575 static int locks_open(struct inode
*inode
, struct file
*filp
)
2577 return seq_open_private(filp
, &locks_seq_operations
,
2578 sizeof(struct locks_iterator
));
2581 static const struct file_operations proc_locks_operations
= {
2584 .llseek
= seq_lseek
,
2585 .release
= seq_release_private
,
2588 static int __init
proc_locks_init(void)
2590 proc_create("locks", 0, NULL
, &proc_locks_operations
);
2593 module_init(proc_locks_init
);
2597 * lock_may_read - checks that the region is free of locks
2598 * @inode: the inode that is being read
2599 * @start: the first byte to read
2600 * @len: the number of bytes to read
2602 * Emulates Windows locking requirements. Whole-file
2603 * mandatory locks (share modes) can prohibit a read and
2604 * byte-range POSIX locks can prohibit a read if they overlap.
2606 * N.B. this function is only ever called
2607 * from knfsd and ownership of locks is never checked.
2609 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
2611 struct file_lock
*fl
;
2614 spin_lock(&inode
->i_lock
);
2615 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2617 if (fl
->fl_type
== F_RDLCK
)
2619 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2621 } else if (IS_FLOCK(fl
)) {
2622 if (!(fl
->fl_type
& LOCK_MAND
))
2624 if (fl
->fl_type
& LOCK_READ
)
2631 spin_unlock(&inode
->i_lock
);
2635 EXPORT_SYMBOL(lock_may_read
);
2638 * lock_may_write - checks that the region is free of locks
2639 * @inode: the inode that is being written
2640 * @start: the first byte to write
2641 * @len: the number of bytes to write
2643 * Emulates Windows locking requirements. Whole-file
2644 * mandatory locks (share modes) can prohibit a write and
2645 * byte-range POSIX locks can prohibit a write if they overlap.
2647 * N.B. this function is only ever called
2648 * from knfsd and ownership of locks is never checked.
2650 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2652 struct file_lock
*fl
;
2655 spin_lock(&inode
->i_lock
);
2656 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2658 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2660 } else if (IS_FLOCK(fl
)) {
2661 if (!(fl
->fl_type
& LOCK_MAND
))
2663 if (fl
->fl_type
& LOCK_WRITE
)
2670 spin_unlock(&inode
->i_lock
);
2674 EXPORT_SYMBOL(lock_may_write
);
2676 static int __init
filelock_init(void)
2680 filelock_cache
= kmem_cache_create("file_lock_cache",
2681 sizeof(struct file_lock
), 0, SLAB_PANIC
, NULL
);
2683 lg_lock_init(&file_lock_lglock
, "file_lock_lglock");
2685 for_each_possible_cpu(i
)
2686 INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list
, i
));
2691 core_initcall(filelock_init
);