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/mandatory.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/fs.h>
120 #include <linux/init.h>
121 #include <linux/module.h>
122 #include <linux/security.h>
123 #include <linux/slab.h>
124 #include <linux/smp_lock.h>
125 #include <linux/time.h>
127 #include <asm/semaphore.h>
128 #include <asm/uaccess.h>
130 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
131 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
132 #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE)
134 int leases_enable
= 1;
135 int lease_break_time
= 45;
137 #define for_each_lock(inode, lockp) \
138 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
140 LIST_HEAD(file_lock_list
);
142 EXPORT_SYMBOL(file_lock_list
);
144 static LIST_HEAD(blocked_list
);
146 static kmem_cache_t
*filelock_cache
;
148 /* Allocate an empty lock structure. */
149 static struct file_lock
*locks_alloc_lock(void)
151 return kmem_cache_alloc(filelock_cache
, SLAB_KERNEL
);
154 /* Free a lock which is not in use. */
155 static inline void locks_free_lock(struct file_lock
*fl
)
161 if (waitqueue_active(&fl
->fl_wait
))
162 panic("Attempting to free lock with active wait queue");
164 if (!list_empty(&fl
->fl_block
))
165 panic("Attempting to free lock with active block list");
167 if (!list_empty(&fl
->fl_link
))
168 panic("Attempting to free lock on active lock list");
171 if (fl
->fl_ops
->fl_release_private
)
172 fl
->fl_ops
->fl_release_private(fl
);
177 kmem_cache_free(filelock_cache
, fl
);
180 void locks_init_lock(struct file_lock
*fl
)
182 INIT_LIST_HEAD(&fl
->fl_link
);
183 INIT_LIST_HEAD(&fl
->fl_block
);
184 init_waitqueue_head(&fl
->fl_wait
);
186 fl
->fl_fasync
= NULL
;
192 fl
->fl_start
= fl
->fl_end
= 0;
197 EXPORT_SYMBOL(locks_init_lock
);
200 * Initialises the fields of the file lock which are invariant for
203 static void init_once(void *foo
, kmem_cache_t
*cache
, unsigned long flags
)
205 struct file_lock
*lock
= (struct file_lock
*) foo
;
207 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) !=
208 SLAB_CTOR_CONSTRUCTOR
)
211 locks_init_lock(lock
);
215 * Initialize a new lock from an existing file_lock structure.
217 void locks_copy_lock(struct file_lock
*new, struct file_lock
*fl
)
219 new->fl_owner
= fl
->fl_owner
;
220 new->fl_pid
= fl
->fl_pid
;
221 new->fl_file
= fl
->fl_file
;
222 new->fl_flags
= fl
->fl_flags
;
223 new->fl_type
= fl
->fl_type
;
224 new->fl_start
= fl
->fl_start
;
225 new->fl_end
= fl
->fl_end
;
226 new->fl_ops
= fl
->fl_ops
;
227 new->fl_lmops
= fl
->fl_lmops
;
228 if (fl
->fl_ops
&& fl
->fl_ops
->fl_copy_lock
)
229 fl
->fl_ops
->fl_copy_lock(new, fl
);
232 EXPORT_SYMBOL(locks_copy_lock
);
234 static inline int flock_translate_cmd(int cmd
) {
236 return cmd
& (LOCK_MAND
| LOCK_RW
);
248 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
249 static int flock_make_lock(struct file
*filp
, struct file_lock
**lock
,
252 struct file_lock
*fl
;
253 int type
= flock_translate_cmd(cmd
);
257 fl
= locks_alloc_lock();
262 fl
->fl_pid
= current
->tgid
;
263 fl
->fl_flags
= FL_FLOCK
;
265 fl
->fl_end
= OFFSET_MAX
;
271 static int assign_type(struct file_lock
*fl
, int type
)
285 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
288 static int flock_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
293 switch (l
->l_whence
) {
301 start
= i_size_read(filp
->f_dentry
->d_inode
);
307 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
308 POSIX-2001 defines it. */
310 end
= start
+ l
->l_len
- 1;
318 if (l
->l_len
> 0 && end
< 0)
321 fl
->fl_start
= start
; /* we record the absolute position */
324 fl
->fl_end
= OFFSET_MAX
;
326 fl
->fl_owner
= current
->files
;
327 fl
->fl_pid
= current
->tgid
;
329 fl
->fl_flags
= FL_POSIX
;
333 return assign_type(fl
, l
->l_type
);
336 #if BITS_PER_LONG == 32
337 static int flock64_to_posix_lock(struct file
*filp
, struct file_lock
*fl
,
342 switch (l
->l_whence
) {
350 start
= i_size_read(filp
->f_dentry
->d_inode
);
356 if (((start
+= l
->l_start
) < 0) || (l
->l_len
< 0))
358 fl
->fl_end
= start
+ l
->l_len
- 1;
359 if (l
->l_len
> 0 && fl
->fl_end
< 0)
361 fl
->fl_start
= start
; /* we record the absolute position */
363 fl
->fl_end
= OFFSET_MAX
;
365 fl
->fl_owner
= current
->files
;
366 fl
->fl_pid
= current
->tgid
;
368 fl
->fl_flags
= FL_POSIX
;
376 fl
->fl_type
= l
->l_type
;
386 /* Allocate a file_lock initialised to this type of lease */
387 static int lease_alloc(struct file
*filp
, int type
, struct file_lock
**flp
)
389 struct file_lock
*fl
= locks_alloc_lock();
393 fl
->fl_owner
= current
->files
;
394 fl
->fl_pid
= current
->tgid
;
397 fl
->fl_flags
= FL_LEASE
;
398 if (assign_type(fl
, type
) != 0) {
403 fl
->fl_end
= OFFSET_MAX
;
411 /* Check if two locks overlap each other.
413 static inline int locks_overlap(struct file_lock
*fl1
, struct file_lock
*fl2
)
415 return ((fl1
->fl_end
>= fl2
->fl_start
) &&
416 (fl2
->fl_end
>= fl1
->fl_start
));
420 * Check whether two locks have the same owner.
423 posix_same_owner(struct file_lock
*fl1
, struct file_lock
*fl2
)
425 if (fl1
->fl_lmops
&& fl1
->fl_lmops
->fl_compare_owner
)
426 return fl2
->fl_lmops
== fl1
->fl_lmops
&&
427 fl1
->fl_lmops
->fl_compare_owner(fl1
, fl2
);
428 return fl1
->fl_owner
== fl2
->fl_owner
;
431 /* Remove waiter from blocker's block list.
432 * When blocker ends up pointing to itself then the list is empty.
434 static inline void __locks_delete_block(struct file_lock
*waiter
)
436 list_del_init(&waiter
->fl_block
);
437 list_del_init(&waiter
->fl_link
);
438 waiter
->fl_next
= NULL
;
443 static void locks_delete_block(struct file_lock
*waiter
)
446 __locks_delete_block(waiter
);
450 /* Insert waiter into blocker's block list.
451 * We use a circular list so that processes can be easily woken up in
452 * the order they blocked. The documentation doesn't require this but
453 * it seems like the reasonable thing to do.
455 static void locks_insert_block(struct file_lock
*blocker
,
456 struct file_lock
*waiter
)
458 if (!list_empty(&waiter
->fl_block
)) {
459 printk(KERN_ERR
"locks_insert_block: removing duplicated lock "
460 "(pid=%d %Ld-%Ld type=%d)\n", waiter
->fl_pid
,
461 waiter
->fl_start
, waiter
->fl_end
, waiter
->fl_type
);
462 __locks_delete_block(waiter
);
464 list_add_tail(&waiter
->fl_block
, &blocker
->fl_block
);
465 waiter
->fl_next
= blocker
;
466 if (IS_POSIX(blocker
))
467 list_add(&waiter
->fl_link
, &blocked_list
);
470 /* Wake up processes blocked waiting for blocker.
471 * If told to wait then schedule the processes until the block list
472 * is empty, otherwise empty the block list ourselves.
474 static void locks_wake_up_blocks(struct file_lock
*blocker
)
476 while (!list_empty(&blocker
->fl_block
)) {
477 struct file_lock
*waiter
= list_entry(blocker
->fl_block
.next
,
478 struct file_lock
, fl_block
);
479 __locks_delete_block(waiter
);
480 if (waiter
->fl_lmops
&& waiter
->fl_lmops
->fl_notify
)
481 waiter
->fl_lmops
->fl_notify(waiter
);
483 wake_up(&waiter
->fl_wait
);
487 /* Insert file lock fl into an inode's lock list at the position indicated
488 * by pos. At the same time add the lock to the global file lock list.
490 static void locks_insert_lock(struct file_lock
**pos
, struct file_lock
*fl
)
492 list_add(&fl
->fl_link
, &file_lock_list
);
494 /* insert into file's list */
498 if (fl
->fl_ops
&& fl
->fl_ops
->fl_insert
)
499 fl
->fl_ops
->fl_insert(fl
);
503 * Delete a lock and then free it.
504 * Wake up processes that are blocked waiting for this lock,
505 * notify the FS that the lock has been cleared and
506 * finally free the lock.
508 static void locks_delete_lock(struct file_lock
**thisfl_p
)
510 struct file_lock
*fl
= *thisfl_p
;
512 *thisfl_p
= fl
->fl_next
;
514 list_del_init(&fl
->fl_link
);
516 fasync_helper(0, fl
->fl_file
, 0, &fl
->fl_fasync
);
517 if (fl
->fl_fasync
!= NULL
) {
518 printk(KERN_ERR
"locks_delete_lock: fasync == %p\n", fl
->fl_fasync
);
519 fl
->fl_fasync
= NULL
;
522 if (fl
->fl_ops
&& fl
->fl_ops
->fl_remove
)
523 fl
->fl_ops
->fl_remove(fl
);
525 locks_wake_up_blocks(fl
);
529 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
530 * checks for shared/exclusive status of overlapping locks.
532 static int locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
534 if (sys_fl
->fl_type
== F_WRLCK
)
536 if (caller_fl
->fl_type
== F_WRLCK
)
541 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
542 * checking before calling the locks_conflict().
544 static int posix_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
546 /* POSIX locks owned by the same process do not conflict with
549 if (!IS_POSIX(sys_fl
) || posix_same_owner(caller_fl
, sys_fl
))
552 /* Check whether they overlap */
553 if (!locks_overlap(caller_fl
, sys_fl
))
556 return (locks_conflict(caller_fl
, sys_fl
));
559 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
560 * checking before calling the locks_conflict().
562 static int flock_locks_conflict(struct file_lock
*caller_fl
, struct file_lock
*sys_fl
)
564 /* FLOCK locks referring to the same filp do not conflict with
567 if (!IS_FLOCK(sys_fl
) || (caller_fl
->fl_file
== sys_fl
->fl_file
))
569 if ((caller_fl
->fl_type
& LOCK_MAND
) || (sys_fl
->fl_type
& LOCK_MAND
))
572 return (locks_conflict(caller_fl
, sys_fl
));
575 static int interruptible_sleep_on_locked(wait_queue_head_t
*fl_wait
, int timeout
)
578 DECLARE_WAITQUEUE(wait
, current
);
580 __set_current_state(TASK_INTERRUPTIBLE
);
581 add_wait_queue(fl_wait
, &wait
);
585 result
= schedule_timeout(timeout
);
586 if (signal_pending(current
))
587 result
= -ERESTARTSYS
;
588 remove_wait_queue(fl_wait
, &wait
);
589 __set_current_state(TASK_RUNNING
);
593 static int locks_block_on_timeout(struct file_lock
*blocker
, struct file_lock
*waiter
, int time
)
596 locks_insert_block(blocker
, waiter
);
597 result
= interruptible_sleep_on_locked(&waiter
->fl_wait
, time
);
598 __locks_delete_block(waiter
);
603 posix_test_lock(struct file
*filp
, struct file_lock
*fl
)
605 struct file_lock
*cfl
;
608 for (cfl
= filp
->f_dentry
->d_inode
->i_flock
; cfl
; cfl
= cfl
->fl_next
) {
611 if (posix_locks_conflict(cfl
, fl
))
619 EXPORT_SYMBOL(posix_test_lock
);
621 /* This function tests for deadlock condition before putting a process to
622 * sleep. The detection scheme is no longer recursive. Recursive was neat,
623 * but dangerous - we risked stack corruption if the lock data was bad, or
624 * if the recursion was too deep for any other reason.
626 * We rely on the fact that a task can only be on one lock's wait queue
627 * at a time. When we find blocked_task on a wait queue we can re-search
628 * with blocked_task equal to that queue's owner, until either blocked_task
629 * isn't found, or blocked_task is found on a queue owned by my_task.
631 * Note: the above assumption may not be true when handling lock requests
632 * from a broken NFS client. But broken NFS clients have a lot more to
633 * worry about than proper deadlock detection anyway... --okir
635 int posix_locks_deadlock(struct file_lock
*caller_fl
,
636 struct file_lock
*block_fl
)
638 struct list_head
*tmp
;
641 if (posix_same_owner(caller_fl
, block_fl
))
643 list_for_each(tmp
, &blocked_list
) {
644 struct file_lock
*fl
= list_entry(tmp
, struct file_lock
, fl_link
);
645 if (posix_same_owner(fl
, block_fl
)) {
654 EXPORT_SYMBOL(posix_locks_deadlock
);
656 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
657 * at the head of the list, but that's secret knowledge known only to
658 * flock_lock_file and posix_lock_file.
660 static int flock_lock_file(struct file
*filp
, struct file_lock
*new_fl
)
662 struct file_lock
**before
;
663 struct inode
* inode
= filp
->f_dentry
->d_inode
;
668 for_each_lock(inode
, before
) {
669 struct file_lock
*fl
= *before
;
674 if (filp
!= fl
->fl_file
)
676 if (new_fl
->fl_type
== fl
->fl_type
)
679 locks_delete_lock(before
);
684 if (new_fl
->fl_type
== F_UNLCK
)
688 * If a higher-priority process was blocked on the old file lock,
689 * give it the opportunity to lock the file.
695 for_each_lock(inode
, before
) {
696 struct file_lock
*fl
= *before
;
701 if (!flock_locks_conflict(new_fl
, fl
))
704 if (new_fl
->fl_flags
& FL_SLEEP
) {
705 locks_insert_block(fl
, new_fl
);
709 locks_insert_lock(&inode
->i_flock
, new_fl
);
717 EXPORT_SYMBOL(posix_lock_file
);
719 static int __posix_lock_file(struct inode
*inode
, struct file_lock
*request
)
721 struct file_lock
*fl
;
722 struct file_lock
*new_fl
, *new_fl2
;
723 struct file_lock
*left
= NULL
;
724 struct file_lock
*right
= NULL
;
725 struct file_lock
**before
;
726 int error
, added
= 0;
729 * We may need two file_lock structures for this operation,
730 * so we get them in advance to avoid races.
732 new_fl
= locks_alloc_lock();
733 new_fl2
= locks_alloc_lock();
736 if (request
->fl_type
!= F_UNLCK
) {
737 for_each_lock(inode
, before
) {
738 struct file_lock
*fl
= *before
;
741 if (!posix_locks_conflict(request
, fl
))
744 if (!(request
->fl_flags
& FL_SLEEP
))
747 if (posix_locks_deadlock(request
, fl
))
750 locks_insert_block(fl
, request
);
755 /* If we're just looking for a conflict, we're done. */
757 if (request
->fl_flags
& FL_ACCESS
)
760 error
= -ENOLCK
; /* "no luck" */
761 if (!(new_fl
&& new_fl2
))
765 * We've allocated the new locks in advance, so there are no
766 * errors possible (and no blocking operations) from here on.
768 * Find the first old lock with the same owner as the new lock.
771 before
= &inode
->i_flock
;
773 /* First skip locks owned by other processes. */
774 while ((fl
= *before
) && (!IS_POSIX(fl
) ||
775 !posix_same_owner(request
, fl
))) {
776 before
= &fl
->fl_next
;
779 /* Process locks with this owner. */
780 while ((fl
= *before
) && posix_same_owner(request
, fl
)) {
781 /* Detect adjacent or overlapping regions (if same lock type)
783 if (request
->fl_type
== fl
->fl_type
) {
784 if (fl
->fl_end
< request
->fl_start
- 1)
786 /* If the next lock in the list has entirely bigger
787 * addresses than the new one, insert the lock here.
789 if (fl
->fl_start
> request
->fl_end
+ 1)
792 /* If we come here, the new and old lock are of the
793 * same type and adjacent or overlapping. Make one
794 * lock yielding from the lower start address of both
795 * locks to the higher end address.
797 if (fl
->fl_start
> request
->fl_start
)
798 fl
->fl_start
= request
->fl_start
;
800 request
->fl_start
= fl
->fl_start
;
801 if (fl
->fl_end
< request
->fl_end
)
802 fl
->fl_end
= request
->fl_end
;
804 request
->fl_end
= fl
->fl_end
;
806 locks_delete_lock(before
);
813 /* Processing for different lock types is a bit
816 if (fl
->fl_end
< request
->fl_start
)
818 if (fl
->fl_start
> request
->fl_end
)
820 if (request
->fl_type
== F_UNLCK
)
822 if (fl
->fl_start
< request
->fl_start
)
824 /* If the next lock in the list has a higher end
825 * address than the new one, insert the new one here.
827 if (fl
->fl_end
> request
->fl_end
) {
831 if (fl
->fl_start
>= request
->fl_start
) {
832 /* The new lock completely replaces an old
833 * one (This may happen several times).
836 locks_delete_lock(before
);
839 /* Replace the old lock with the new one.
840 * Wake up anybody waiting for the old one,
841 * as the change in lock type might satisfy
844 locks_wake_up_blocks(fl
);
845 fl
->fl_start
= request
->fl_start
;
846 fl
->fl_end
= request
->fl_end
;
847 fl
->fl_type
= request
->fl_type
;
848 fl
->fl_u
= request
->fl_u
;
853 /* Go on to next lock.
856 before
= &fl
->fl_next
;
861 if (request
->fl_type
== F_UNLCK
)
863 locks_copy_lock(new_fl
, request
);
864 locks_insert_lock(before
, new_fl
);
869 /* The new lock breaks the old one in two pieces,
870 * so we have to use the second new lock.
874 locks_copy_lock(left
, right
);
875 locks_insert_lock(before
, left
);
877 right
->fl_start
= request
->fl_end
+ 1;
878 locks_wake_up_blocks(right
);
881 left
->fl_end
= request
->fl_start
- 1;
882 locks_wake_up_blocks(left
);
887 * Free any unused locks.
890 locks_free_lock(new_fl
);
892 locks_free_lock(new_fl2
);
897 * posix_lock_file - Apply a POSIX-style lock to a file
898 * @filp: The file to apply the lock to
899 * @fl: The lock to be applied
901 * Add a POSIX style lock to a file.
902 * We merge adjacent & overlapping locks whenever possible.
903 * POSIX locks are sorted by owner task, then by starting address
905 int posix_lock_file(struct file
*filp
, struct file_lock
*fl
)
907 return __posix_lock_file(filp
->f_dentry
->d_inode
, fl
);
911 * posix_lock_file_wait - Apply a POSIX-style lock to a file
912 * @filp: The file to apply the lock to
913 * @fl: The lock to be applied
915 * Add a POSIX style lock to a file.
916 * We merge adjacent & overlapping locks whenever possible.
917 * POSIX locks are sorted by owner task, then by starting address
919 int posix_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
924 error
= __posix_lock_file(filp
->f_dentry
->d_inode
, fl
);
925 if ((error
!= -EAGAIN
) || !(fl
->fl_flags
& FL_SLEEP
))
927 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
931 locks_delete_block(fl
);
936 EXPORT_SYMBOL(posix_lock_file_wait
);
939 * locks_mandatory_locked - Check for an active lock
940 * @inode: the file to check
942 * Searches the inode's list of locks to find any POSIX locks which conflict.
943 * This function is called from locks_verify_locked() only.
945 int locks_mandatory_locked(struct inode
*inode
)
947 fl_owner_t owner
= current
->files
;
948 struct file_lock
*fl
;
951 * Search the lock list for this inode for any POSIX locks.
954 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
957 if (fl
->fl_owner
!= owner
)
961 return fl
? -EAGAIN
: 0;
965 * locks_mandatory_area - Check for a conflicting lock
966 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
968 * @inode: the file to check
969 * @filp: how the file was opened (if it was)
970 * @offset: start of area to check
971 * @count: length of area to check
973 * Searches the inode's list of locks to find any POSIX locks which conflict.
974 * This function is called from locks_verify_area() and
975 * locks_verify_truncate().
977 int locks_mandatory_area(int read_write
, struct inode
*inode
,
978 struct file
*filp
, loff_t offset
,
984 locks_init_lock(&fl
);
985 fl
.fl_owner
= current
->files
;
986 fl
.fl_pid
= current
->tgid
;
988 fl
.fl_flags
= FL_POSIX
| FL_ACCESS
;
989 if (filp
&& !(filp
->f_flags
& O_NONBLOCK
))
990 fl
.fl_flags
|= FL_SLEEP
;
991 fl
.fl_type
= (read_write
== FLOCK_VERIFY_WRITE
) ? F_WRLCK
: F_RDLCK
;
992 fl
.fl_start
= offset
;
993 fl
.fl_end
= offset
+ count
- 1;
996 error
= __posix_lock_file(inode
, &fl
);
997 if (error
!= -EAGAIN
)
999 if (!(fl
.fl_flags
& FL_SLEEP
))
1001 error
= wait_event_interruptible(fl
.fl_wait
, !fl
.fl_next
);
1004 * If we've been sleeping someone might have
1005 * changed the permissions behind our back.
1007 if ((inode
->i_mode
& (S_ISGID
| S_IXGRP
)) == S_ISGID
)
1011 locks_delete_block(&fl
);
1018 EXPORT_SYMBOL(locks_mandatory_area
);
1020 /* We already had a lease on this file; just change its type */
1021 static int lease_modify(struct file_lock
**before
, int arg
)
1023 struct file_lock
*fl
= *before
;
1024 int error
= assign_type(fl
, arg
);
1028 locks_wake_up_blocks(fl
);
1029 if (arg
== F_UNLCK
) {
1030 struct file
*filp
= fl
->fl_file
;
1033 filp
->f_owner
.signum
= 0;
1034 locks_delete_lock(before
);
1039 static void time_out_leases(struct inode
*inode
)
1041 struct file_lock
**before
;
1042 struct file_lock
*fl
;
1044 before
= &inode
->i_flock
;
1045 while ((fl
= *before
) && IS_LEASE(fl
) && (fl
->fl_type
& F_INPROGRESS
)) {
1046 if ((fl
->fl_break_time
== 0)
1047 || time_before(jiffies
, fl
->fl_break_time
)) {
1048 before
= &fl
->fl_next
;
1051 printk(KERN_INFO
"lease broken - owner pid = %d\n", fl
->fl_pid
);
1052 lease_modify(before
, fl
->fl_type
& ~F_INPROGRESS
);
1053 if (fl
== *before
) /* lease_modify may have freed fl */
1054 before
= &fl
->fl_next
;
1059 * __break_lease - revoke all outstanding leases on file
1060 * @inode: the inode of the file to return
1061 * @mode: the open mode (read or write)
1063 * break_lease (inlined for speed) has checked there already
1064 * is a lease on this file. Leases are broken on a call to open()
1065 * or truncate(). This function can sleep unless you
1066 * specified %O_NONBLOCK to your open().
1068 int __break_lease(struct inode
*inode
, unsigned int mode
)
1070 int error
= 0, future
;
1071 struct file_lock
*new_fl
, *flock
;
1072 struct file_lock
*fl
;
1074 unsigned long break_time
;
1075 int i_have_this_lease
= 0;
1077 alloc_err
= lease_alloc(NULL
, mode
& FMODE_WRITE
? F_WRLCK
: F_RDLCK
,
1082 time_out_leases(inode
);
1084 flock
= inode
->i_flock
;
1085 if ((flock
== NULL
) || !IS_LEASE(flock
))
1088 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
)
1089 if (fl
->fl_owner
== current
->files
)
1090 i_have_this_lease
= 1;
1092 if (mode
& FMODE_WRITE
) {
1093 /* If we want write access, we have to revoke any lease. */
1094 future
= F_UNLCK
| F_INPROGRESS
;
1095 } else if (flock
->fl_type
& F_INPROGRESS
) {
1096 /* If the lease is already being broken, we just leave it */
1097 future
= flock
->fl_type
;
1098 } else if (flock
->fl_type
& F_WRLCK
) {
1099 /* Downgrade the exclusive lease to a read-only lease. */
1100 future
= F_RDLCK
| F_INPROGRESS
;
1102 /* the existing lease was read-only, so we can read too. */
1106 if (alloc_err
&& !i_have_this_lease
&& ((mode
& O_NONBLOCK
) == 0)) {
1112 if (lease_break_time
> 0) {
1113 break_time
= jiffies
+ lease_break_time
* HZ
;
1114 if (break_time
== 0)
1115 break_time
++; /* so that 0 means no break time */
1118 for (fl
= flock
; fl
&& IS_LEASE(fl
); fl
= fl
->fl_next
) {
1119 if (fl
->fl_type
!= future
) {
1120 fl
->fl_type
= future
;
1121 fl
->fl_break_time
= break_time
;
1122 kill_fasync(&fl
->fl_fasync
, SIGIO
, POLL_MSG
);
1126 if (i_have_this_lease
|| (mode
& O_NONBLOCK
)) {
1127 error
= -EWOULDBLOCK
;
1132 break_time
= flock
->fl_break_time
;
1133 if (break_time
!= 0) {
1134 break_time
-= jiffies
;
1135 if (break_time
== 0)
1138 error
= locks_block_on_timeout(flock
, new_fl
, break_time
);
1141 time_out_leases(inode
);
1142 /* Wait for the next lease that has not been broken yet */
1143 for (flock
= inode
->i_flock
; flock
&& IS_LEASE(flock
);
1144 flock
= flock
->fl_next
) {
1145 if (flock
->fl_type
& F_INPROGRESS
)
1154 locks_free_lock(new_fl
);
1158 EXPORT_SYMBOL(__break_lease
);
1163 * @time: pointer to a timespec which will contain the last modified time
1165 * This is to force NFS clients to flush their caches for files with
1166 * exclusive leases. The justification is that if someone has an
1167 * exclusive lease, then they could be modifiying it.
1169 void lease_get_mtime(struct inode
*inode
, struct timespec
*time
)
1171 struct file_lock
*flock
= inode
->i_flock
;
1172 if (flock
&& IS_LEASE(flock
) && (flock
->fl_type
& F_WRLCK
))
1173 *time
= CURRENT_TIME
;
1175 *time
= inode
->i_mtime
;
1178 EXPORT_SYMBOL(lease_get_mtime
);
1181 * fcntl_getlease - Enquire what lease is currently active
1184 * The value returned by this function will be one of
1185 * (if no lease break is pending):
1187 * %F_RDLCK to indicate a shared lease is held.
1189 * %F_WRLCK to indicate an exclusive lease is held.
1191 * %F_UNLCK to indicate no lease is held.
1193 * (if a lease break is pending):
1195 * %F_RDLCK to indicate an exclusive lease needs to be
1196 * changed to a shared lease (or removed).
1198 * %F_UNLCK to indicate the lease needs to be removed.
1200 * XXX: sfr & willy disagree over whether F_INPROGRESS
1201 * should be returned to userspace.
1203 int fcntl_getlease(struct file
*filp
)
1205 struct file_lock
*fl
;
1209 time_out_leases(filp
->f_dentry
->d_inode
);
1210 for (fl
= filp
->f_dentry
->d_inode
->i_flock
; fl
&& IS_LEASE(fl
);
1212 if (fl
->fl_file
== filp
) {
1213 type
= fl
->fl_type
& ~F_INPROGRESS
;
1222 * fcntl_setlease - sets a lease on an open file
1223 * @fd: open file descriptor
1224 * @filp: file pointer
1225 * @arg: type of lease to obtain
1227 * Call this fcntl to establish a lease on the file.
1228 * Note that you also need to call %F_SETSIG to
1229 * receive a signal when the lease is broken.
1231 int fcntl_setlease(unsigned int fd
, struct file
*filp
, long arg
)
1233 struct file_lock
*fl
, **before
, **my_before
= NULL
;
1234 struct dentry
*dentry
;
1235 struct inode
*inode
;
1236 int error
, rdlease_count
= 0, wrlease_count
= 0;
1238 dentry
= filp
->f_dentry
;
1239 inode
= dentry
->d_inode
;
1241 if ((current
->fsuid
!= inode
->i_uid
) && !capable(CAP_LEASE
))
1243 if (!S_ISREG(inode
->i_mode
))
1245 error
= security_file_lock(filp
, arg
);
1251 time_out_leases(inode
);
1254 * FIXME: What about F_RDLCK and files open for writing?
1257 if ((arg
== F_WRLCK
)
1258 && ((atomic_read(&dentry
->d_count
) > 1)
1259 || (atomic_read(&inode
->i_count
) > 1)))
1263 * At this point, we know that if there is an exclusive
1264 * lease on this file, then we hold it on this filp
1265 * (otherwise our open of this file would have blocked).
1266 * And if we are trying to acquire an exclusive lease,
1267 * then the file is not open by anyone (including us)
1268 * except for this filp.
1270 for (before
= &inode
->i_flock
;
1271 ((fl
= *before
) != NULL
) && IS_LEASE(fl
);
1272 before
= &fl
->fl_next
) {
1273 if (fl
->fl_file
== filp
)
1275 else if (fl
->fl_type
== (F_INPROGRESS
| F_UNLCK
))
1277 * Someone is in the process of opening this
1278 * file for writing so we may not take an
1279 * exclusive lease on it.
1286 if ((arg
== F_RDLCK
&& (wrlease_count
> 0)) ||
1287 (arg
== F_WRLCK
&& ((rdlease_count
+ wrlease_count
) > 0)))
1290 if (my_before
!= NULL
) {
1291 error
= lease_modify(my_before
, arg
);
1303 error
= lease_alloc(filp
, arg
, &fl
);
1307 error
= fasync_helper(fd
, filp
, 1, &fl
->fl_fasync
);
1309 locks_free_lock(fl
);
1313 locks_insert_lock(before
, fl
);
1315 error
= f_setown(filp
, current
->pid
, 0);
1322 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1323 * @filp: The file to apply the lock to
1324 * @fl: The lock to be applied
1326 * Add a FLOCK style lock to a file.
1328 int flock_lock_file_wait(struct file
*filp
, struct file_lock
*fl
)
1333 error
= flock_lock_file(filp
, fl
);
1334 if ((error
!= -EAGAIN
) || !(fl
->fl_flags
& FL_SLEEP
))
1336 error
= wait_event_interruptible(fl
->fl_wait
, !fl
->fl_next
);
1340 locks_delete_block(fl
);
1346 EXPORT_SYMBOL(flock_lock_file_wait
);
1349 * sys_flock: - flock() system call.
1350 * @fd: the file descriptor to lock.
1351 * @cmd: the type of lock to apply.
1353 * Apply a %FL_FLOCK style lock to an open file descriptor.
1354 * The @cmd can be one of
1356 * %LOCK_SH -- a shared lock.
1358 * %LOCK_EX -- an exclusive lock.
1360 * %LOCK_UN -- remove an existing lock.
1362 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1364 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1365 * processes read and write access respectively.
1367 asmlinkage
long sys_flock(unsigned int fd
, unsigned int cmd
)
1370 struct file_lock
*lock
;
1371 int can_sleep
, unlock
;
1379 can_sleep
= !(cmd
& LOCK_NB
);
1381 unlock
= (cmd
== LOCK_UN
);
1383 if (!unlock
&& !(cmd
& LOCK_MAND
) && !(filp
->f_mode
& 3))
1386 error
= flock_make_lock(filp
, &lock
, cmd
);
1390 lock
->fl_flags
|= FL_SLEEP
;
1392 error
= security_file_lock(filp
, cmd
);
1396 if (filp
->f_op
&& filp
->f_op
->flock
)
1397 error
= filp
->f_op
->flock(filp
,
1398 (can_sleep
) ? F_SETLKW
: F_SETLK
,
1401 error
= flock_lock_file_wait(filp
, lock
);
1404 if (list_empty(&lock
->fl_link
)) {
1405 locks_free_lock(lock
);
1414 /* Report the first existing lock that would conflict with l.
1415 * This implements the F_GETLK command of fcntl().
1417 int fcntl_getlk(struct file
*filp
, struct flock __user
*l
)
1419 struct file_lock
*fl
, file_lock
;
1424 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1427 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1430 error
= flock_to_posix_lock(filp
, &file_lock
, &flock
);
1434 if (filp
->f_op
&& filp
->f_op
->lock
) {
1435 error
= filp
->f_op
->lock(filp
, F_GETLK
, &file_lock
);
1438 else if (error
== LOCK_USE_CLNT
)
1439 /* Bypass for NFS with no locking - 2.0.36 compat */
1440 fl
= posix_test_lock(filp
, &file_lock
);
1442 fl
= (file_lock
.fl_type
== F_UNLCK
? NULL
: &file_lock
);
1444 fl
= posix_test_lock(filp
, &file_lock
);
1447 flock
.l_type
= F_UNLCK
;
1449 flock
.l_pid
= fl
->fl_pid
;
1450 #if BITS_PER_LONG == 32
1452 * Make sure we can represent the posix lock via
1453 * legacy 32bit flock.
1456 if (fl
->fl_start
> OFFT_OFFSET_MAX
)
1458 if ((fl
->fl_end
!= OFFSET_MAX
)
1459 && (fl
->fl_end
> OFFT_OFFSET_MAX
))
1462 flock
.l_start
= fl
->fl_start
;
1463 flock
.l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1464 fl
->fl_end
- fl
->fl_start
+ 1;
1466 flock
.l_type
= fl
->fl_type
;
1469 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1475 /* Apply the lock described by l to an open file descriptor.
1476 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1478 int fcntl_setlk(struct file
*filp
, unsigned int cmd
, struct flock __user
*l
)
1480 struct file_lock
*file_lock
= locks_alloc_lock();
1482 struct inode
*inode
;
1485 if (file_lock
== NULL
)
1489 * This might block, so we do it before checking the inode.
1492 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1495 inode
= filp
->f_dentry
->d_inode
;
1497 /* Don't allow mandatory locks on files that may be memory mapped
1500 if (IS_MANDLOCK(inode
) &&
1501 (inode
->i_mode
& (S_ISGID
| S_IXGRP
)) == S_ISGID
&&
1502 mapping_writably_mapped(filp
->f_mapping
)) {
1507 error
= flock_to_posix_lock(filp
, file_lock
, &flock
);
1510 if (cmd
== F_SETLKW
) {
1511 file_lock
->fl_flags
|= FL_SLEEP
;
1515 switch (flock
.l_type
) {
1517 if (!(filp
->f_mode
& FMODE_READ
))
1521 if (!(filp
->f_mode
& FMODE_WRITE
))
1531 error
= security_file_lock(filp
, file_lock
->fl_type
);
1535 if (filp
->f_op
&& filp
->f_op
->lock
!= NULL
) {
1536 error
= filp
->f_op
->lock(filp
, cmd
, file_lock
);
1541 error
= __posix_lock_file(inode
, file_lock
);
1542 if ((error
!= -EAGAIN
) || (cmd
== F_SETLK
))
1544 error
= wait_event_interruptible(file_lock
->fl_wait
,
1545 !file_lock
->fl_next
);
1549 locks_delete_block(file_lock
);
1554 locks_free_lock(file_lock
);
1558 #if BITS_PER_LONG == 32
1559 /* Report the first existing lock that would conflict with l.
1560 * This implements the F_GETLK command of fcntl().
1562 int fcntl_getlk64(struct file
*filp
, struct flock64 __user
*l
)
1564 struct file_lock
*fl
, file_lock
;
1565 struct flock64 flock
;
1569 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1572 if ((flock
.l_type
!= F_RDLCK
) && (flock
.l_type
!= F_WRLCK
))
1575 error
= flock64_to_posix_lock(filp
, &file_lock
, &flock
);
1579 if (filp
->f_op
&& filp
->f_op
->lock
) {
1580 error
= filp
->f_op
->lock(filp
, F_GETLK
, &file_lock
);
1583 else if (error
== LOCK_USE_CLNT
)
1584 /* Bypass for NFS with no locking - 2.0.36 compat */
1585 fl
= posix_test_lock(filp
, &file_lock
);
1587 fl
= (file_lock
.fl_type
== F_UNLCK
? NULL
: &file_lock
);
1589 fl
= posix_test_lock(filp
, &file_lock
);
1592 flock
.l_type
= F_UNLCK
;
1594 flock
.l_pid
= fl
->fl_pid
;
1595 flock
.l_start
= fl
->fl_start
;
1596 flock
.l_len
= fl
->fl_end
== OFFSET_MAX
? 0 :
1597 fl
->fl_end
- fl
->fl_start
+ 1;
1599 flock
.l_type
= fl
->fl_type
;
1602 if (!copy_to_user(l
, &flock
, sizeof(flock
)))
1609 /* Apply the lock described by l to an open file descriptor.
1610 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1612 int fcntl_setlk64(struct file
*filp
, unsigned int cmd
, struct flock64 __user
*l
)
1614 struct file_lock
*file_lock
= locks_alloc_lock();
1615 struct flock64 flock
;
1616 struct inode
*inode
;
1619 if (file_lock
== NULL
)
1623 * This might block, so we do it before checking the inode.
1626 if (copy_from_user(&flock
, l
, sizeof(flock
)))
1629 inode
= filp
->f_dentry
->d_inode
;
1631 /* Don't allow mandatory locks on files that may be memory mapped
1634 if (IS_MANDLOCK(inode
) &&
1635 (inode
->i_mode
& (S_ISGID
| S_IXGRP
)) == S_ISGID
&&
1636 mapping_writably_mapped(filp
->f_mapping
)) {
1641 error
= flock64_to_posix_lock(filp
, file_lock
, &flock
);
1644 if (cmd
== F_SETLKW64
) {
1645 file_lock
->fl_flags
|= FL_SLEEP
;
1649 switch (flock
.l_type
) {
1651 if (!(filp
->f_mode
& FMODE_READ
))
1655 if (!(filp
->f_mode
& FMODE_WRITE
))
1665 error
= security_file_lock(filp
, file_lock
->fl_type
);
1669 if (filp
->f_op
&& filp
->f_op
->lock
!= NULL
) {
1670 error
= filp
->f_op
->lock(filp
, cmd
, file_lock
);
1675 error
= __posix_lock_file(inode
, file_lock
);
1676 if ((error
!= -EAGAIN
) || (cmd
== F_SETLK64
))
1678 error
= wait_event_interruptible(file_lock
->fl_wait
,
1679 !file_lock
->fl_next
);
1683 locks_delete_block(file_lock
);
1688 locks_free_lock(file_lock
);
1691 #endif /* BITS_PER_LONG == 32 */
1694 * This function is called when the file is being removed
1695 * from the task's fd array. POSIX locks belonging to this task
1696 * are deleted at this time.
1698 void locks_remove_posix(struct file
*filp
, fl_owner_t owner
)
1700 struct file_lock lock
, **before
;
1703 * If there are no locks held on this file, we don't need to call
1704 * posix_lock_file(). Another process could be setting a lock on this
1705 * file at the same time, but we wouldn't remove that lock anyway.
1707 before
= &filp
->f_dentry
->d_inode
->i_flock
;
1708 if (*before
== NULL
)
1711 lock
.fl_type
= F_UNLCK
;
1712 lock
.fl_flags
= FL_POSIX
;
1714 lock
.fl_end
= OFFSET_MAX
;
1715 lock
.fl_owner
= owner
;
1716 lock
.fl_pid
= current
->tgid
;
1717 lock
.fl_file
= filp
;
1719 lock
.fl_lmops
= NULL
;
1721 if (filp
->f_op
&& filp
->f_op
->lock
!= NULL
) {
1722 filp
->f_op
->lock(filp
, F_SETLK
, &lock
);
1726 /* Can't use posix_lock_file here; we need to remove it no matter
1727 * which pid we have.
1730 while (*before
!= NULL
) {
1731 struct file_lock
*fl
= *before
;
1732 if (IS_POSIX(fl
) && posix_same_owner(fl
, &lock
)) {
1733 locks_delete_lock(before
);
1736 before
= &fl
->fl_next
;
1740 if (lock
.fl_ops
&& lock
.fl_ops
->fl_release_private
)
1741 lock
.fl_ops
->fl_release_private(&lock
);
1744 EXPORT_SYMBOL(locks_remove_posix
);
1747 * This function is called on the last close of an open file.
1749 void locks_remove_flock(struct file
*filp
)
1751 struct inode
* inode
= filp
->f_dentry
->d_inode
;
1752 struct file_lock
*fl
;
1753 struct file_lock
**before
;
1755 if (!inode
->i_flock
)
1758 if (filp
->f_op
&& filp
->f_op
->flock
) {
1759 struct file_lock fl
= { .fl_flags
= FL_FLOCK
,
1760 .fl_type
= F_UNLCK
};
1761 filp
->f_op
->flock(filp
, F_SETLKW
, &fl
);
1765 before
= &inode
->i_flock
;
1767 while ((fl
= *before
) != NULL
) {
1768 if (fl
->fl_file
== filp
) {
1770 * We might have a POSIX lock that was created at the same time
1771 * the filp was closed for the last time. Just remove that too,
1772 * regardless of ownership, since nobody can own it.
1774 if (IS_FLOCK(fl
) || IS_POSIX(fl
)) {
1775 locks_delete_lock(before
);
1779 lease_modify(before
, F_UNLCK
);
1785 before
= &fl
->fl_next
;
1791 * posix_block_lock - blocks waiting for a file lock
1792 * @blocker: the lock which is blocking
1793 * @waiter: the lock which conflicts and has to wait
1795 * lockd needs to block waiting for locks.
1798 posix_block_lock(struct file_lock
*blocker
, struct file_lock
*waiter
)
1800 locks_insert_block(blocker
, waiter
);
1803 EXPORT_SYMBOL(posix_block_lock
);
1806 * posix_unblock_lock - stop waiting for a file lock
1807 * @filp: how the file was opened
1808 * @waiter: the lock which was waiting
1810 * lockd needs to block waiting for locks.
1813 posix_unblock_lock(struct file
*filp
, struct file_lock
*waiter
)
1816 * A remote machine may cancel the lock request after it's been
1817 * granted locally. If that happens, we need to delete the lock.
1820 if (waiter
->fl_next
) {
1821 __locks_delete_block(waiter
);
1825 waiter
->fl_type
= F_UNLCK
;
1826 posix_lock_file(filp
, waiter
);
1830 EXPORT_SYMBOL(posix_unblock_lock
);
1832 static void lock_get_status(char* out
, struct file_lock
*fl
, int id
, char *pfx
)
1834 struct inode
*inode
= NULL
;
1836 if (fl
->fl_file
!= NULL
)
1837 inode
= fl
->fl_file
->f_dentry
->d_inode
;
1839 out
+= sprintf(out
, "%d:%s ", id
, pfx
);
1841 out
+= sprintf(out
, "%6s %s ",
1842 (fl
->fl_flags
& FL_ACCESS
) ? "ACCESS" : "POSIX ",
1843 (inode
== NULL
) ? "*NOINODE*" :
1844 (IS_MANDLOCK(inode
) &&
1845 (inode
->i_mode
& (S_IXGRP
| S_ISGID
)) == S_ISGID
) ?
1846 "MANDATORY" : "ADVISORY ");
1847 } else if (IS_FLOCK(fl
)) {
1848 if (fl
->fl_type
& LOCK_MAND
) {
1849 out
+= sprintf(out
, "FLOCK MSNFS ");
1851 out
+= sprintf(out
, "FLOCK ADVISORY ");
1853 } else if (IS_LEASE(fl
)) {
1854 out
+= sprintf(out
, "LEASE ");
1855 if (fl
->fl_type
& F_INPROGRESS
)
1856 out
+= sprintf(out
, "BREAKING ");
1857 else if (fl
->fl_file
)
1858 out
+= sprintf(out
, "ACTIVE ");
1860 out
+= sprintf(out
, "BREAKER ");
1862 out
+= sprintf(out
, "UNKNOWN UNKNOWN ");
1864 if (fl
->fl_type
& LOCK_MAND
) {
1865 out
+= sprintf(out
, "%s ",
1866 (fl
->fl_type
& LOCK_READ
)
1867 ? (fl
->fl_type
& LOCK_WRITE
) ? "RW " : "READ "
1868 : (fl
->fl_type
& LOCK_WRITE
) ? "WRITE" : "NONE ");
1870 out
+= sprintf(out
, "%s ",
1871 (fl
->fl_type
& F_INPROGRESS
)
1872 ? (fl
->fl_type
& F_UNLCK
) ? "UNLCK" : "READ "
1873 : (fl
->fl_type
& F_WRLCK
) ? "WRITE" : "READ ");
1876 #ifdef WE_CAN_BREAK_LSLK_NOW
1877 out
+= sprintf(out
, "%d %s:%ld ", fl
->fl_pid
,
1878 inode
->i_sb
->s_id
, inode
->i_ino
);
1880 /* userspace relies on this representation of dev_t ;-( */
1881 out
+= sprintf(out
, "%d %02x:%02x:%ld ", fl
->fl_pid
,
1882 MAJOR(inode
->i_sb
->s_dev
),
1883 MINOR(inode
->i_sb
->s_dev
), inode
->i_ino
);
1886 out
+= sprintf(out
, "%d <none>:0 ", fl
->fl_pid
);
1889 if (fl
->fl_end
== OFFSET_MAX
)
1890 out
+= sprintf(out
, "%Ld EOF\n", fl
->fl_start
);
1892 out
+= sprintf(out
, "%Ld %Ld\n", fl
->fl_start
,
1895 out
+= sprintf(out
, "0 EOF\n");
1899 static void move_lock_status(char **p
, off_t
* pos
, off_t offset
)
1903 if(*pos
>= offset
) {
1904 /* the complete line is valid */
1909 if(*pos
+len
> offset
) {
1910 /* use the second part of the line */
1911 int i
= offset
-*pos
;
1912 memmove(*p
,*p
+i
,len
-i
);
1917 /* discard the complete line */
1922 * get_locks_status - reports lock usage in /proc/locks
1923 * @buffer: address in userspace to write into
1925 * @offset: how far we are through the buffer
1926 * @length: how much to read
1929 int get_locks_status(char *buffer
, char **start
, off_t offset
, int length
)
1931 struct list_head
*tmp
;
1937 list_for_each(tmp
, &file_lock_list
) {
1938 struct list_head
*btmp
;
1939 struct file_lock
*fl
= list_entry(tmp
, struct file_lock
, fl_link
);
1940 lock_get_status(q
, fl
, ++i
, "");
1941 move_lock_status(&q
, &pos
, offset
);
1943 if(pos
>= offset
+length
)
1946 list_for_each(btmp
, &fl
->fl_block
) {
1947 struct file_lock
*bfl
= list_entry(btmp
,
1948 struct file_lock
, fl_block
);
1949 lock_get_status(q
, bfl
, i
, " ->");
1950 move_lock_status(&q
, &pos
, offset
);
1952 if(pos
>= offset
+length
)
1959 if(q
-buffer
< length
)
1965 * lock_may_read - checks that the region is free of locks
1966 * @inode: the inode that is being read
1967 * @start: the first byte to read
1968 * @len: the number of bytes to read
1970 * Emulates Windows locking requirements. Whole-file
1971 * mandatory locks (share modes) can prohibit a read and
1972 * byte-range POSIX locks can prohibit a read if they overlap.
1974 * N.B. this function is only ever called
1975 * from knfsd and ownership of locks is never checked.
1977 int lock_may_read(struct inode
*inode
, loff_t start
, unsigned long len
)
1979 struct file_lock
*fl
;
1982 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
1984 if (fl
->fl_type
== F_RDLCK
)
1986 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
1988 } else if (IS_FLOCK(fl
)) {
1989 if (!(fl
->fl_type
& LOCK_MAND
))
1991 if (fl
->fl_type
& LOCK_READ
)
2002 EXPORT_SYMBOL(lock_may_read
);
2005 * lock_may_write - checks that the region is free of locks
2006 * @inode: the inode that is being written
2007 * @start: the first byte to write
2008 * @len: the number of bytes to write
2010 * Emulates Windows locking requirements. Whole-file
2011 * mandatory locks (share modes) can prohibit a write and
2012 * byte-range POSIX locks can prohibit a write if they overlap.
2014 * N.B. this function is only ever called
2015 * from knfsd and ownership of locks is never checked.
2017 int lock_may_write(struct inode
*inode
, loff_t start
, unsigned long len
)
2019 struct file_lock
*fl
;
2022 for (fl
= inode
->i_flock
; fl
!= NULL
; fl
= fl
->fl_next
) {
2024 if ((fl
->fl_end
< start
) || (fl
->fl_start
> (start
+ len
)))
2026 } else if (IS_FLOCK(fl
)) {
2027 if (!(fl
->fl_type
& LOCK_MAND
))
2029 if (fl
->fl_type
& LOCK_WRITE
)
2040 EXPORT_SYMBOL(lock_may_write
);
2042 static inline void __steal_locks(struct file
*file
, fl_owner_t from
)
2044 struct inode
*inode
= file
->f_dentry
->d_inode
;
2045 struct file_lock
*fl
= inode
->i_flock
;
2048 if (fl
->fl_file
== file
&& fl
->fl_owner
== from
)
2049 fl
->fl_owner
= current
->files
;
2054 /* When getting ready for executing a binary, we make sure that current
2055 * has a files_struct on its own. Before dropping the old files_struct,
2056 * we take over ownership of all locks for all file descriptors we own.
2057 * Note that we may accidentally steal a lock for a file that a sibling
2058 * has created since the unshare_files() call.
2060 void steal_locks(fl_owner_t from
)
2062 struct files_struct
*files
= current
->files
;
2073 if (i
>= files
->max_fdset
|| i
>= files
->max_fds
)
2075 set
= files
->open_fds
->fds_bits
[j
++];
2078 struct file
*file
= files
->fd
[i
];
2080 __steal_locks(file
, from
);
2088 EXPORT_SYMBOL(steal_locks
);
2090 static int __init
filelock_init(void)
2092 filelock_cache
= kmem_cache_create("file_lock_cache",
2093 sizeof(struct file_lock
), 0, SLAB_PANIC
,
2098 core_initcall(filelock_init
);