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/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>
130 #include <asm/uaccess.h>
132 #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
133 #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
134 #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE)
136 int leases_enable = 1;
137 int lease_break_time = 45;
139 #define for_each_lock(inode, lockp) \
140 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
142 static LIST_HEAD(file_lock_list);
143 static LIST_HEAD(blocked_list);
144 static DEFINE_SPINLOCK(file_lock_lock);
147 * Protects the two list heads above, plus the inode->i_flock list
149 void lock_flocks(void)
151 spin_lock(&file_lock_lock);
153 EXPORT_SYMBOL_GPL(lock_flocks);
155 void unlock_flocks(void)
157 spin_unlock(&file_lock_lock);
159 EXPORT_SYMBOL_GPL(unlock_flocks);
161 static struct kmem_cache *filelock_cache __read_mostly;
163 /* Allocate an empty lock structure. */
164 struct file_lock *locks_alloc_lock(void)
166 return kmem_cache_alloc(filelock_cache, GFP_KERNEL);
168 EXPORT_SYMBOL_GPL(locks_alloc_lock);
170 void locks_release_private(struct file_lock *fl)
173 if (fl->fl_ops->fl_release_private)
174 fl->fl_ops->fl_release_private(fl);
178 if (fl->fl_lmops->fl_release_private)
179 fl->fl_lmops->fl_release_private(fl);
184 EXPORT_SYMBOL_GPL(locks_release_private);
186 /* Free a lock which is not in use. */
187 void locks_free_lock(struct file_lock *fl)
189 BUG_ON(waitqueue_active(&fl->fl_wait));
190 BUG_ON(!list_empty(&fl->fl_block));
191 BUG_ON(!list_empty(&fl->fl_link));
193 locks_release_private(fl);
194 kmem_cache_free(filelock_cache, fl);
196 EXPORT_SYMBOL(locks_free_lock);
198 void locks_init_lock(struct file_lock *fl)
200 INIT_LIST_HEAD(&fl->fl_link);
201 INIT_LIST_HEAD(&fl->fl_block);
202 init_waitqueue_head(&fl->fl_wait);
204 fl->fl_fasync = NULL;
211 fl->fl_start = fl->fl_end = 0;
216 EXPORT_SYMBOL(locks_init_lock);
219 * Initialises the fields of the file lock which are invariant for
222 static void init_once(void *foo)
224 struct file_lock *lock = (struct file_lock *) foo;
226 locks_init_lock(lock);
229 static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
232 if (fl->fl_ops->fl_copy_lock)
233 fl->fl_ops->fl_copy_lock(new, fl);
234 new->fl_ops = fl->fl_ops;
237 new->fl_lmops = fl->fl_lmops;
241 * Initialize a new lock from an existing file_lock structure.
243 void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
245 new->fl_owner = fl->fl_owner;
246 new->fl_pid = fl->fl_pid;
248 new->fl_flags = fl->fl_flags;
249 new->fl_type = fl->fl_type;
250 new->fl_start = fl->fl_start;
251 new->fl_end = fl->fl_end;
253 new->fl_lmops = NULL;
255 EXPORT_SYMBOL(__locks_copy_lock);
257 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
259 locks_release_private(new);
261 __locks_copy_lock(new, fl);
262 new->fl_file = fl->fl_file;
263 new->fl_ops = fl->fl_ops;
264 new->fl_lmops = fl->fl_lmops;
266 locks_copy_private(new, fl);
269 EXPORT_SYMBOL(locks_copy_lock);
271 static inline int flock_translate_cmd(int cmd) {
273 return cmd & (LOCK_MAND | LOCK_RW);
285 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
286 static int flock_make_lock(struct file *filp, struct file_lock **lock,
289 struct file_lock *fl;
290 int type = flock_translate_cmd(cmd);
294 fl = locks_alloc_lock();
299 fl->fl_pid = current->tgid;
300 fl->fl_flags = FL_FLOCK;
302 fl->fl_end = OFFSET_MAX;
308 static int assign_type(struct file_lock *fl, int type)
322 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
325 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
330 switch (l->l_whence) {
338 start = i_size_read(filp->f_path.dentry->d_inode);
344 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
345 POSIX-2001 defines it. */
349 fl->fl_end = OFFSET_MAX;
351 end = start + l->l_len - 1;
353 } else if (l->l_len < 0) {
360 fl->fl_start = start; /* we record the absolute position */
361 if (fl->fl_end < fl->fl_start)
364 fl->fl_owner = current->files;
365 fl->fl_pid = current->tgid;
367 fl->fl_flags = FL_POSIX;
371 return assign_type(fl, l->l_type);
374 #if BITS_PER_LONG == 32
375 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
380 switch (l->l_whence) {
388 start = i_size_read(filp->f_path.dentry->d_inode);
397 fl->fl_end = OFFSET_MAX;
399 fl->fl_end = start + l->l_len - 1;
400 } else if (l->l_len < 0) {
401 fl->fl_end = start - 1;
406 fl->fl_start = start; /* we record the absolute position */
407 if (fl->fl_end < fl->fl_start)
410 fl->fl_owner = current->files;
411 fl->fl_pid = current->tgid;
413 fl->fl_flags = FL_POSIX;
417 return assign_type(fl, l->l_type);
421 /* default lease lock manager operations */
422 static void lease_break_callback(struct file_lock *fl)
424 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
427 static void lease_release_private_callback(struct file_lock *fl)
432 f_delown(fl->fl_file);
433 fl->fl_file->f_owner.signum = 0;
436 static const struct lock_manager_operations lease_manager_ops = {
437 .fl_break = lease_break_callback,
438 .fl_release_private = lease_release_private_callback,
439 .fl_change = lease_modify,
443 * Initialize a lease, use the default lock manager operations
445 static int lease_init(struct file *filp, int type, struct file_lock *fl)
447 if (assign_type(fl, type) != 0)
450 fl->fl_owner = current->files;
451 fl->fl_pid = current->tgid;
454 fl->fl_flags = FL_LEASE;
456 fl->fl_end = OFFSET_MAX;
458 fl->fl_lmops = &lease_manager_ops;
462 /* Allocate a file_lock initialised to this type of lease */
463 static struct file_lock *lease_alloc(struct file *filp, int type)
465 struct file_lock *fl = locks_alloc_lock();
469 return ERR_PTR(error);
471 error = lease_init(filp, type, fl);
474 return ERR_PTR(error);
479 /* Check if two locks overlap each other.
481 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
483 return ((fl1->fl_end >= fl2->fl_start) &&
484 (fl2->fl_end >= fl1->fl_start));
488 * Check whether two locks have the same owner.
490 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
492 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
493 return fl2->fl_lmops == fl1->fl_lmops &&
494 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
495 return fl1->fl_owner == fl2->fl_owner;
498 /* Remove waiter from blocker's block list.
499 * When blocker ends up pointing to itself then the list is empty.
501 static void __locks_delete_block(struct file_lock *waiter)
503 list_del_init(&waiter->fl_block);
504 list_del_init(&waiter->fl_link);
505 waiter->fl_next = NULL;
510 static void locks_delete_block(struct file_lock *waiter)
513 __locks_delete_block(waiter);
517 /* Insert waiter into blocker's block list.
518 * We use a circular list so that processes can be easily woken up in
519 * the order they blocked. The documentation doesn't require this but
520 * it seems like the reasonable thing to do.
522 static void locks_insert_block(struct file_lock *blocker,
523 struct file_lock *waiter)
525 BUG_ON(!list_empty(&waiter->fl_block));
526 list_add_tail(&waiter->fl_block, &blocker->fl_block);
527 waiter->fl_next = blocker;
528 if (IS_POSIX(blocker))
529 list_add(&waiter->fl_link, &blocked_list);
532 /* Wake up processes blocked waiting for blocker.
533 * If told to wait then schedule the processes until the block list
534 * is empty, otherwise empty the block list ourselves.
536 static void locks_wake_up_blocks(struct file_lock *blocker)
538 while (!list_empty(&blocker->fl_block)) {
539 struct file_lock *waiter;
541 waiter = list_first_entry(&blocker->fl_block,
542 struct file_lock, fl_block);
543 __locks_delete_block(waiter);
544 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
545 waiter->fl_lmops->fl_notify(waiter);
547 wake_up(&waiter->fl_wait);
551 /* Insert file lock fl into an inode's lock list at the position indicated
552 * by pos. At the same time add the lock to the global file lock list.
554 static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
556 list_add(&fl->fl_link, &file_lock_list);
558 fl->fl_nspid = get_pid(task_tgid(current));
560 /* insert into file's list */
566 * Delete a lock and then free it.
567 * Wake up processes that are blocked waiting for this lock,
568 * notify the FS that the lock has been cleared and
569 * finally free the lock.
571 static void locks_delete_lock(struct file_lock **thisfl_p)
573 struct file_lock *fl = *thisfl_p;
575 *thisfl_p = fl->fl_next;
577 list_del_init(&fl->fl_link);
579 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
580 if (fl->fl_fasync != NULL) {
581 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
582 fl->fl_fasync = NULL;
586 put_pid(fl->fl_nspid);
590 locks_wake_up_blocks(fl);
594 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
595 * checks for shared/exclusive status of overlapping locks.
597 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
599 if (sys_fl->fl_type == F_WRLCK)
601 if (caller_fl->fl_type == F_WRLCK)
606 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
607 * checking before calling the locks_conflict().
609 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
611 /* POSIX locks owned by the same process do not conflict with
614 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
617 /* Check whether they overlap */
618 if (!locks_overlap(caller_fl, sys_fl))
621 return (locks_conflict(caller_fl, sys_fl));
624 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
625 * checking before calling the locks_conflict().
627 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
629 /* FLOCK locks referring to the same filp do not conflict with
632 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
634 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
637 return (locks_conflict(caller_fl, sys_fl));
641 posix_test_lock(struct file *filp, struct file_lock *fl)
643 struct file_lock *cfl;
646 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
649 if (posix_locks_conflict(fl, cfl))
653 __locks_copy_lock(fl, cfl);
655 fl->fl_pid = pid_vnr(cfl->fl_nspid);
657 fl->fl_type = F_UNLCK;
661 EXPORT_SYMBOL(posix_test_lock);
664 * Deadlock detection:
666 * We attempt to detect deadlocks that are due purely to posix file
669 * We assume that a task can be waiting for at most one lock at a time.
670 * So for any acquired lock, the process holding that lock may be
671 * waiting on at most one other lock. That lock in turns may be held by
672 * someone waiting for at most one other lock. Given a requested lock
673 * caller_fl which is about to wait for a conflicting lock block_fl, we
674 * follow this chain of waiters to ensure we are not about to create a
677 * Since we do this before we ever put a process to sleep on a lock, we
678 * are ensured that there is never a cycle; that is what guarantees that
679 * the while() loop in posix_locks_deadlock() eventually completes.
681 * Note: the above assumption may not be true when handling lock
682 * requests from a broken NFS client. It may also fail in the presence
683 * of tasks (such as posix threads) sharing the same open file table.
685 * To handle those cases, we just bail out after a few iterations.
688 #define MAX_DEADLK_ITERATIONS 10
690 /* Find a lock that the owner of the given block_fl is blocking on. */
691 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
693 struct file_lock *fl;
695 list_for_each_entry(fl, &blocked_list, fl_link) {
696 if (posix_same_owner(fl, block_fl))
702 static int posix_locks_deadlock(struct file_lock *caller_fl,
703 struct file_lock *block_fl)
707 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
708 if (i++ > MAX_DEADLK_ITERATIONS)
710 if (posix_same_owner(caller_fl, block_fl))
716 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
717 * after any leases, but before any posix locks.
719 * Note that if called with an FL_EXISTS argument, the caller may determine
720 * whether or not a lock was successfully freed by testing the return
723 static int flock_lock_file(struct file *filp, struct file_lock *request)
725 struct file_lock *new_fl = NULL;
726 struct file_lock **before;
727 struct inode * inode = filp->f_path.dentry->d_inode;
731 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
732 new_fl = locks_alloc_lock();
738 if (request->fl_flags & FL_ACCESS)
741 for_each_lock(inode, before) {
742 struct file_lock *fl = *before;
747 if (filp != fl->fl_file)
749 if (request->fl_type == fl->fl_type)
752 locks_delete_lock(before);
756 if (request->fl_type == F_UNLCK) {
757 if ((request->fl_flags & FL_EXISTS) && !found)
763 * If a higher-priority process was blocked on the old file lock,
764 * give it the opportunity to lock the file.
773 for_each_lock(inode, before) {
774 struct file_lock *fl = *before;
779 if (!flock_locks_conflict(request, fl))
782 if (!(request->fl_flags & FL_SLEEP))
784 error = FILE_LOCK_DEFERRED;
785 locks_insert_block(fl, request);
788 if (request->fl_flags & FL_ACCESS)
790 locks_copy_lock(new_fl, request);
791 locks_insert_lock(before, new_fl);
798 locks_free_lock(new_fl);
802 static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
804 struct file_lock *fl;
805 struct file_lock *new_fl = NULL;
806 struct file_lock *new_fl2 = NULL;
807 struct file_lock *left = NULL;
808 struct file_lock *right = NULL;
809 struct file_lock **before;
810 int error, added = 0;
813 * We may need two file_lock structures for this operation,
814 * so we get them in advance to avoid races.
816 * In some cases we can be sure, that no new locks will be needed
818 if (!(request->fl_flags & FL_ACCESS) &&
819 (request->fl_type != F_UNLCK ||
820 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
821 new_fl = locks_alloc_lock();
822 new_fl2 = locks_alloc_lock();
826 if (request->fl_type != F_UNLCK) {
827 for_each_lock(inode, before) {
831 if (!posix_locks_conflict(request, fl))
834 __locks_copy_lock(conflock, fl);
836 if (!(request->fl_flags & FL_SLEEP))
839 if (posix_locks_deadlock(request, fl))
841 error = FILE_LOCK_DEFERRED;
842 locks_insert_block(fl, request);
847 /* If we're just looking for a conflict, we're done. */
849 if (request->fl_flags & FL_ACCESS)
853 * Find the first old lock with the same owner as the new lock.
856 before = &inode->i_flock;
858 /* First skip locks owned by other processes. */
859 while ((fl = *before) && (!IS_POSIX(fl) ||
860 !posix_same_owner(request, fl))) {
861 before = &fl->fl_next;
864 /* Process locks with this owner. */
865 while ((fl = *before) && posix_same_owner(request, fl)) {
866 /* Detect adjacent or overlapping regions (if same lock type)
868 if (request->fl_type == fl->fl_type) {
869 /* In all comparisons of start vs end, use
870 * "start - 1" rather than "end + 1". If end
871 * is OFFSET_MAX, end + 1 will become negative.
873 if (fl->fl_end < request->fl_start - 1)
875 /* If the next lock in the list has entirely bigger
876 * addresses than the new one, insert the lock here.
878 if (fl->fl_start - 1 > request->fl_end)
881 /* If we come here, the new and old lock are of the
882 * same type and adjacent or overlapping. Make one
883 * lock yielding from the lower start address of both
884 * locks to the higher end address.
886 if (fl->fl_start > request->fl_start)
887 fl->fl_start = request->fl_start;
889 request->fl_start = fl->fl_start;
890 if (fl->fl_end < request->fl_end)
891 fl->fl_end = request->fl_end;
893 request->fl_end = fl->fl_end;
895 locks_delete_lock(before);
902 /* Processing for different lock types is a bit
905 if (fl->fl_end < request->fl_start)
907 if (fl->fl_start > request->fl_end)
909 if (request->fl_type == F_UNLCK)
911 if (fl->fl_start < request->fl_start)
913 /* If the next lock in the list has a higher end
914 * address than the new one, insert the new one here.
916 if (fl->fl_end > request->fl_end) {
920 if (fl->fl_start >= request->fl_start) {
921 /* The new lock completely replaces an old
922 * one (This may happen several times).
925 locks_delete_lock(before);
928 /* Replace the old lock with the new one.
929 * Wake up anybody waiting for the old one,
930 * as the change in lock type might satisfy
933 locks_wake_up_blocks(fl);
934 fl->fl_start = request->fl_start;
935 fl->fl_end = request->fl_end;
936 fl->fl_type = request->fl_type;
937 locks_release_private(fl);
938 locks_copy_private(fl, request);
943 /* Go on to next lock.
946 before = &fl->fl_next;
950 * The above code only modifies existing locks in case of
951 * merging or replacing. If new lock(s) need to be inserted
952 * all modifications are done bellow this, so it's safe yet to
955 error = -ENOLCK; /* "no luck" */
956 if (right && left == right && !new_fl2)
961 if (request->fl_type == F_UNLCK) {
962 if (request->fl_flags & FL_EXISTS)
971 locks_copy_lock(new_fl, request);
972 locks_insert_lock(before, new_fl);
977 /* The new lock breaks the old one in two pieces,
978 * so we have to use the second new lock.
982 locks_copy_lock(left, right);
983 locks_insert_lock(before, left);
985 right->fl_start = request->fl_end + 1;
986 locks_wake_up_blocks(right);
989 left->fl_end = request->fl_start - 1;
990 locks_wake_up_blocks(left);
995 * Free any unused locks.
998 locks_free_lock(new_fl);
1000 locks_free_lock(new_fl2);
1005 * posix_lock_file - Apply a POSIX-style lock to a file
1006 * @filp: The file to apply the lock to
1007 * @fl: The lock to be applied
1008 * @conflock: Place to return a copy of the conflicting lock, if found.
1010 * Add a POSIX style lock to a file.
1011 * We merge adjacent & overlapping locks whenever possible.
1012 * POSIX locks are sorted by owner task, then by starting address
1014 * Note that if called with an FL_EXISTS argument, the caller may determine
1015 * whether or not a lock was successfully freed by testing the return
1016 * value for -ENOENT.
1018 int posix_lock_file(struct file *filp, struct file_lock *fl,
1019 struct file_lock *conflock)
1021 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
1023 EXPORT_SYMBOL(posix_lock_file);
1026 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1027 * @filp: The file to apply the lock to
1028 * @fl: The lock to be applied
1030 * Add a POSIX style lock to a file.
1031 * We merge adjacent & overlapping locks whenever possible.
1032 * POSIX locks are sorted by owner task, then by starting address
1034 int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1039 error = posix_lock_file(filp, fl, NULL);
1040 if (error != FILE_LOCK_DEFERRED)
1042 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1046 locks_delete_block(fl);
1051 EXPORT_SYMBOL(posix_lock_file_wait);
1054 * locks_mandatory_locked - Check for an active lock
1055 * @inode: the file to check
1057 * Searches the inode's list of locks to find any POSIX locks which conflict.
1058 * This function is called from locks_verify_locked() only.
1060 int locks_mandatory_locked(struct inode *inode)
1062 fl_owner_t owner = current->files;
1063 struct file_lock *fl;
1066 * Search the lock list for this inode for any POSIX locks.
1069 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1072 if (fl->fl_owner != owner)
1076 return fl ? -EAGAIN : 0;
1080 * locks_mandatory_area - Check for a conflicting lock
1081 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1083 * @inode: the file to check
1084 * @filp: how the file was opened (if it was)
1085 * @offset: start of area to check
1086 * @count: length of area to check
1088 * Searches the inode's list of locks to find any POSIX locks which conflict.
1089 * This function is called from rw_verify_area() and
1090 * locks_verify_truncate().
1092 int locks_mandatory_area(int read_write, struct inode *inode,
1093 struct file *filp, loff_t offset,
1096 struct file_lock fl;
1099 locks_init_lock(&fl);
1100 fl.fl_owner = current->files;
1101 fl.fl_pid = current->tgid;
1103 fl.fl_flags = FL_POSIX | FL_ACCESS;
1104 if (filp && !(filp->f_flags & O_NONBLOCK))
1105 fl.fl_flags |= FL_SLEEP;
1106 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1107 fl.fl_start = offset;
1108 fl.fl_end = offset + count - 1;
1111 error = __posix_lock_file(inode, &fl, NULL);
1112 if (error != FILE_LOCK_DEFERRED)
1114 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1117 * If we've been sleeping someone might have
1118 * changed the permissions behind our back.
1120 if (__mandatory_lock(inode))
1124 locks_delete_block(&fl);
1131 EXPORT_SYMBOL(locks_mandatory_area);
1133 /* We already had a lease on this file; just change its type */
1134 int lease_modify(struct file_lock **before, int arg)
1136 struct file_lock *fl = *before;
1137 int error = assign_type(fl, arg);
1141 locks_wake_up_blocks(fl);
1143 locks_delete_lock(before);
1147 EXPORT_SYMBOL(lease_modify);
1149 static void time_out_leases(struct inode *inode)
1151 struct file_lock **before;
1152 struct file_lock *fl;
1154 before = &inode->i_flock;
1155 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1156 if ((fl->fl_break_time == 0)
1157 || time_before(jiffies, fl->fl_break_time)) {
1158 before = &fl->fl_next;
1161 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1162 if (fl == *before) /* lease_modify may have freed fl */
1163 before = &fl->fl_next;
1168 * __break_lease - revoke all outstanding leases on file
1169 * @inode: the inode of the file to return
1170 * @mode: the open mode (read or write)
1172 * break_lease (inlined for speed) has checked there already is at least
1173 * some kind of lock (maybe a lease) on this file. Leases are broken on
1174 * a call to open() or truncate(). This function can sleep unless you
1175 * specified %O_NONBLOCK to your open().
1177 int __break_lease(struct inode *inode, unsigned int mode)
1179 int error = 0, future;
1180 struct file_lock *new_fl, *flock;
1181 struct file_lock *fl;
1182 unsigned long break_time;
1183 int i_have_this_lease = 0;
1184 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1186 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1190 time_out_leases(inode);
1192 flock = inode->i_flock;
1193 if ((flock == NULL) || !IS_LEASE(flock))
1196 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1197 if (fl->fl_owner == current->files)
1198 i_have_this_lease = 1;
1201 /* If we want write access, we have to revoke any lease. */
1202 future = F_UNLCK | F_INPROGRESS;
1203 } else if (flock->fl_type & F_INPROGRESS) {
1204 /* If the lease is already being broken, we just leave it */
1205 future = flock->fl_type;
1206 } else if (flock->fl_type & F_WRLCK) {
1207 /* Downgrade the exclusive lease to a read-only lease. */
1208 future = F_RDLCK | F_INPROGRESS;
1210 /* the existing lease was read-only, so we can read too. */
1214 if (IS_ERR(new_fl) && !i_have_this_lease
1215 && ((mode & O_NONBLOCK) == 0)) {
1216 error = PTR_ERR(new_fl);
1221 if (lease_break_time > 0) {
1222 break_time = jiffies + lease_break_time * HZ;
1223 if (break_time == 0)
1224 break_time++; /* so that 0 means no break time */
1227 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1228 if (fl->fl_type != future) {
1229 fl->fl_type = future;
1230 fl->fl_break_time = break_time;
1231 /* lease must have lmops break callback */
1232 fl->fl_lmops->fl_break(fl);
1236 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1237 error = -EWOULDBLOCK;
1242 break_time = flock->fl_break_time;
1243 if (break_time != 0) {
1244 break_time -= jiffies;
1245 if (break_time == 0)
1248 locks_insert_block(flock, new_fl);
1250 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1251 !new_fl->fl_next, break_time);
1253 __locks_delete_block(new_fl);
1256 time_out_leases(inode);
1257 /* Wait for the next lease that has not been broken yet */
1258 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1259 flock = flock->fl_next) {
1260 if (flock->fl_type & F_INPROGRESS)
1268 if (!IS_ERR(new_fl))
1269 locks_free_lock(new_fl);
1273 EXPORT_SYMBOL(__break_lease);
1276 * lease_get_mtime - get the last modified time of an inode
1278 * @time: pointer to a timespec which will contain the last modified time
1280 * This is to force NFS clients to flush their caches for files with
1281 * exclusive leases. The justification is that if someone has an
1282 * exclusive lease, then they could be modifying it.
1284 void lease_get_mtime(struct inode *inode, struct timespec *time)
1286 struct file_lock *flock = inode->i_flock;
1287 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1288 *time = current_fs_time(inode->i_sb);
1290 *time = inode->i_mtime;
1293 EXPORT_SYMBOL(lease_get_mtime);
1296 * fcntl_getlease - Enquire what lease is currently active
1299 * The value returned by this function will be one of
1300 * (if no lease break is pending):
1302 * %F_RDLCK to indicate a shared lease is held.
1304 * %F_WRLCK to indicate an exclusive lease is held.
1306 * %F_UNLCK to indicate no lease is held.
1308 * (if a lease break is pending):
1310 * %F_RDLCK to indicate an exclusive lease needs to be
1311 * changed to a shared lease (or removed).
1313 * %F_UNLCK to indicate the lease needs to be removed.
1315 * XXX: sfr & willy disagree over whether F_INPROGRESS
1316 * should be returned to userspace.
1318 int fcntl_getlease(struct file *filp)
1320 struct file_lock *fl;
1324 time_out_leases(filp->f_path.dentry->d_inode);
1325 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
1327 if (fl->fl_file == filp) {
1328 type = fl->fl_type & ~F_INPROGRESS;
1337 * generic_setlease - sets a lease on an open file
1338 * @filp: file pointer
1339 * @arg: type of lease to obtain
1340 * @flp: input - file_lock to use, output - file_lock inserted
1342 * The (input) flp->fl_lmops->fl_break function is required
1345 * Called with file_lock_lock held.
1347 int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
1349 struct file_lock *fl, **before, **my_before = NULL, *lease;
1350 struct dentry *dentry = filp->f_path.dentry;
1351 struct inode *inode = dentry->d_inode;
1352 int error, rdlease_count = 0, wrlease_count = 0;
1357 if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE))
1360 if (!S_ISREG(inode->i_mode))
1362 error = security_file_lock(filp, arg);
1366 time_out_leases(inode);
1368 BUG_ON(!(*flp)->fl_lmops->fl_break);
1370 if (arg != F_UNLCK) {
1372 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1374 if ((arg == F_WRLCK)
1375 && ((dentry->d_count > 1)
1376 || (atomic_read(&inode->i_count) > 1)))
1381 * At this point, we know that if there is an exclusive
1382 * lease on this file, then we hold it on this filp
1383 * (otherwise our open of this file would have blocked).
1384 * And if we are trying to acquire an exclusive lease,
1385 * then the file is not open by anyone (including us)
1386 * except for this filp.
1388 for (before = &inode->i_flock;
1389 ((fl = *before) != NULL) && IS_LEASE(fl);
1390 before = &fl->fl_next) {
1391 if (fl->fl_file == filp)
1393 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1395 * Someone is in the process of opening this
1396 * file for writing so we may not take an
1397 * exclusive lease on it.
1405 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1406 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1409 if (my_before != NULL) {
1410 error = lease->fl_lmops->fl_change(my_before, arg);
1423 locks_insert_lock(before, lease);
1429 EXPORT_SYMBOL(generic_setlease);
1431 static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
1433 if (filp->f_op && filp->f_op->setlease)
1434 return filp->f_op->setlease(filp, arg, lease);
1436 return generic_setlease(filp, arg, lease);
1440 * vfs_setlease - sets a lease on an open file
1441 * @filp: file pointer
1442 * @arg: type of lease to obtain
1443 * @lease: file_lock to use
1445 * Call this to establish a lease on the file.
1446 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1447 * break_lease will oops!
1449 * This will call the filesystem's setlease file method, if
1450 * defined. Note that there is no getlease method; instead, the
1451 * filesystem setlease method should call back to setlease() to
1452 * add a lease to the inode's lease list, where fcntl_getlease() can
1453 * find it. Since fcntl_getlease() only reports whether the current
1454 * task holds a lease, a cluster filesystem need only do this for
1455 * leases held by processes on this node.
1457 * There is also no break_lease method; filesystems that
1458 * handle their own leases should break leases themselves from the
1459 * filesystem's open, create, and (on truncate) setattr methods.
1461 * Warning: the only current setlease methods exist only to disable
1462 * leases in certain cases. More vfs changes may be required to
1463 * allow a full filesystem lease implementation.
1466 int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
1471 error = __vfs_setlease(filp, arg, lease);
1476 EXPORT_SYMBOL_GPL(vfs_setlease);
1478 static int do_fcntl_delete_lease(struct file *filp)
1480 struct file_lock fl, *flp = &fl;
1482 lease_init(filp, F_UNLCK, flp);
1484 return vfs_setlease(filp, F_UNLCK, &flp);
1487 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1489 struct file_lock *fl, *ret;
1490 struct fasync_struct *new;
1493 fl = lease_alloc(filp, arg);
1497 new = fasync_alloc();
1499 locks_free_lock(fl);
1504 error = __vfs_setlease(filp, arg, &ret);
1507 locks_free_lock(fl);
1508 goto out_free_fasync;
1511 locks_free_lock(fl);
1514 * fasync_insert_entry() returns the old entry if any.
1515 * If there was no old entry, then it used 'new' and
1516 * inserted it into the fasync list. Clear new so that
1517 * we don't release it here.
1519 if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new))
1522 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
1532 * fcntl_setlease - sets a lease on an open file
1533 * @fd: open file descriptor
1534 * @filp: file pointer
1535 * @arg: type of lease to obtain
1537 * Call this fcntl to establish a lease on the file.
1538 * Note that you also need to call %F_SETSIG to
1539 * receive a signal when the lease is broken.
1541 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1544 return do_fcntl_delete_lease(filp);
1545 return do_fcntl_add_lease(fd, filp, arg);
1549 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1550 * @filp: The file to apply the lock to
1551 * @fl: The lock to be applied
1553 * Add a FLOCK style lock to a file.
1555 int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1560 error = flock_lock_file(filp, fl);
1561 if (error != FILE_LOCK_DEFERRED)
1563 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1567 locks_delete_block(fl);
1573 EXPORT_SYMBOL(flock_lock_file_wait);
1576 * sys_flock: - flock() system call.
1577 * @fd: the file descriptor to lock.
1578 * @cmd: the type of lock to apply.
1580 * Apply a %FL_FLOCK style lock to an open file descriptor.
1581 * The @cmd can be one of
1583 * %LOCK_SH -- a shared lock.
1585 * %LOCK_EX -- an exclusive lock.
1587 * %LOCK_UN -- remove an existing lock.
1589 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1591 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1592 * processes read and write access respectively.
1594 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1597 struct file_lock *lock;
1598 int can_sleep, unlock;
1606 can_sleep = !(cmd & LOCK_NB);
1608 unlock = (cmd == LOCK_UN);
1610 if (!unlock && !(cmd & LOCK_MAND) &&
1611 !(filp->f_mode & (FMODE_READ|FMODE_WRITE)))
1614 error = flock_make_lock(filp, &lock, cmd);
1618 lock->fl_flags |= FL_SLEEP;
1620 error = security_file_lock(filp, lock->fl_type);
1624 if (filp->f_op && filp->f_op->flock)
1625 error = filp->f_op->flock(filp,
1626 (can_sleep) ? F_SETLKW : F_SETLK,
1629 error = flock_lock_file_wait(filp, lock);
1632 locks_free_lock(lock);
1641 * vfs_test_lock - test file byte range lock
1642 * @filp: The file to test lock for
1643 * @fl: The lock to test; also used to hold result
1645 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1646 * setting conf->fl_type to something other than F_UNLCK.
1648 int vfs_test_lock(struct file *filp, struct file_lock *fl)
1650 if (filp->f_op && filp->f_op->lock)
1651 return filp->f_op->lock(filp, F_GETLK, fl);
1652 posix_test_lock(filp, fl);
1655 EXPORT_SYMBOL_GPL(vfs_test_lock);
1657 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1659 flock->l_pid = fl->fl_pid;
1660 #if BITS_PER_LONG == 32
1662 * Make sure we can represent the posix lock via
1663 * legacy 32bit flock.
1665 if (fl->fl_start > OFFT_OFFSET_MAX)
1667 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1670 flock->l_start = fl->fl_start;
1671 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1672 fl->fl_end - fl->fl_start + 1;
1673 flock->l_whence = 0;
1674 flock->l_type = fl->fl_type;
1678 #if BITS_PER_LONG == 32
1679 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1681 flock->l_pid = fl->fl_pid;
1682 flock->l_start = fl->fl_start;
1683 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1684 fl->fl_end - fl->fl_start + 1;
1685 flock->l_whence = 0;
1686 flock->l_type = fl->fl_type;
1690 /* Report the first existing lock that would conflict with l.
1691 * This implements the F_GETLK command of fcntl().
1693 int fcntl_getlk(struct file *filp, struct flock __user *l)
1695 struct file_lock file_lock;
1700 if (copy_from_user(&flock, l, sizeof(flock)))
1703 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1706 error = flock_to_posix_lock(filp, &file_lock, &flock);
1710 error = vfs_test_lock(filp, &file_lock);
1714 flock.l_type = file_lock.fl_type;
1715 if (file_lock.fl_type != F_UNLCK) {
1716 error = posix_lock_to_flock(&flock, &file_lock);
1721 if (!copy_to_user(l, &flock, sizeof(flock)))
1728 * vfs_lock_file - file byte range lock
1729 * @filp: The file to apply the lock to
1730 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1731 * @fl: The lock to be applied
1732 * @conf: Place to return a copy of the conflicting lock, if found.
1734 * A caller that doesn't care about the conflicting lock may pass NULL
1735 * as the final argument.
1737 * If the filesystem defines a private ->lock() method, then @conf will
1738 * be left unchanged; so a caller that cares should initialize it to
1739 * some acceptable default.
1741 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1742 * locks, the ->lock() interface may return asynchronously, before the lock has
1743 * been granted or denied by the underlying filesystem, if (and only if)
1744 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1745 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1746 * the request is for a blocking lock. When ->lock() does return asynchronously,
1747 * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock
1748 * request completes.
1749 * If the request is for non-blocking lock the file system should return
1750 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1751 * with the result. If the request timed out the callback routine will return a
1752 * nonzero return code and the file system should release the lock. The file
1753 * system is also responsible to keep a corresponding posix lock when it
1754 * grants a lock so the VFS can find out which locks are locally held and do
1755 * the correct lock cleanup when required.
1756 * The underlying filesystem must not drop the kernel lock or call
1757 * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED
1760 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
1762 if (filp->f_op && filp->f_op->lock)
1763 return filp->f_op->lock(filp, cmd, fl);
1765 return posix_lock_file(filp, fl, conf);
1767 EXPORT_SYMBOL_GPL(vfs_lock_file);
1769 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
1770 struct file_lock *fl)
1774 error = security_file_lock(filp, fl->fl_type);
1779 error = vfs_lock_file(filp, cmd, fl, NULL);
1780 if (error != FILE_LOCK_DEFERRED)
1782 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1786 locks_delete_block(fl);
1793 /* Apply the lock described by l to an open file descriptor.
1794 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1796 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1797 struct flock __user *l)
1799 struct file_lock *file_lock = locks_alloc_lock();
1801 struct inode *inode;
1805 if (file_lock == NULL)
1809 * This might block, so we do it before checking the inode.
1812 if (copy_from_user(&flock, l, sizeof(flock)))
1815 inode = filp->f_path.dentry->d_inode;
1817 /* Don't allow mandatory locks on files that may be memory mapped
1820 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
1826 error = flock_to_posix_lock(filp, file_lock, &flock);
1829 if (cmd == F_SETLKW) {
1830 file_lock->fl_flags |= FL_SLEEP;
1834 switch (flock.l_type) {
1836 if (!(filp->f_mode & FMODE_READ))
1840 if (!(filp->f_mode & FMODE_WRITE))
1850 error = do_lock_file_wait(filp, cmd, file_lock);
1853 * Attempt to detect a close/fcntl race and recover by
1854 * releasing the lock that was just acquired.
1857 * we need that spin_lock here - it prevents reordering between
1858 * update of inode->i_flock and check for it done in close().
1859 * rcu_read_lock() wouldn't do.
1861 spin_lock(¤t->files->file_lock);
1863 spin_unlock(¤t->files->file_lock);
1864 if (!error && f != filp && flock.l_type != F_UNLCK) {
1865 flock.l_type = F_UNLCK;
1870 locks_free_lock(file_lock);
1874 #if BITS_PER_LONG == 32
1875 /* Report the first existing lock that would conflict with l.
1876 * This implements the F_GETLK command of fcntl().
1878 int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1880 struct file_lock file_lock;
1881 struct flock64 flock;
1885 if (copy_from_user(&flock, l, sizeof(flock)))
1888 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1891 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1895 error = vfs_test_lock(filp, &file_lock);
1899 flock.l_type = file_lock.fl_type;
1900 if (file_lock.fl_type != F_UNLCK)
1901 posix_lock_to_flock64(&flock, &file_lock);
1904 if (!copy_to_user(l, &flock, sizeof(flock)))
1911 /* Apply the lock described by l to an open file descriptor.
1912 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1914 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1915 struct flock64 __user *l)
1917 struct file_lock *file_lock = locks_alloc_lock();
1918 struct flock64 flock;
1919 struct inode *inode;
1923 if (file_lock == NULL)
1927 * This might block, so we do it before checking the inode.
1930 if (copy_from_user(&flock, l, sizeof(flock)))
1933 inode = filp->f_path.dentry->d_inode;
1935 /* Don't allow mandatory locks on files that may be memory mapped
1938 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
1944 error = flock64_to_posix_lock(filp, file_lock, &flock);
1947 if (cmd == F_SETLKW64) {
1948 file_lock->fl_flags |= FL_SLEEP;
1952 switch (flock.l_type) {
1954 if (!(filp->f_mode & FMODE_READ))
1958 if (!(filp->f_mode & FMODE_WRITE))
1968 error = do_lock_file_wait(filp, cmd, file_lock);
1971 * Attempt to detect a close/fcntl race and recover by
1972 * releasing the lock that was just acquired.
1974 spin_lock(¤t->files->file_lock);
1976 spin_unlock(¤t->files->file_lock);
1977 if (!error && f != filp && flock.l_type != F_UNLCK) {
1978 flock.l_type = F_UNLCK;
1983 locks_free_lock(file_lock);
1986 #endif /* BITS_PER_LONG == 32 */
1989 * This function is called when the file is being removed
1990 * from the task's fd array. POSIX locks belonging to this task
1991 * are deleted at this time.
1993 void locks_remove_posix(struct file *filp, fl_owner_t owner)
1995 struct file_lock lock;
1998 * If there are no locks held on this file, we don't need to call
1999 * posix_lock_file(). Another process could be setting a lock on this
2000 * file at the same time, but we wouldn't remove that lock anyway.
2002 if (!filp->f_path.dentry->d_inode->i_flock)
2005 lock.fl_type = F_UNLCK;
2006 lock.fl_flags = FL_POSIX | FL_CLOSE;
2008 lock.fl_end = OFFSET_MAX;
2009 lock.fl_owner = owner;
2010 lock.fl_pid = current->tgid;
2011 lock.fl_file = filp;
2013 lock.fl_lmops = NULL;
2015 vfs_lock_file(filp, F_SETLK, &lock, NULL);
2017 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2018 lock.fl_ops->fl_release_private(&lock);
2021 EXPORT_SYMBOL(locks_remove_posix);
2024 * This function is called on the last close of an open file.
2026 void locks_remove_flock(struct file *filp)
2028 struct inode * inode = filp->f_path.dentry->d_inode;
2029 struct file_lock *fl;
2030 struct file_lock **before;
2032 if (!inode->i_flock)
2035 if (filp->f_op && filp->f_op->flock) {
2036 struct file_lock fl = {
2037 .fl_pid = current->tgid,
2039 .fl_flags = FL_FLOCK,
2041 .fl_end = OFFSET_MAX,
2043 filp->f_op->flock(filp, F_SETLKW, &fl);
2044 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2045 fl.fl_ops->fl_release_private(&fl);
2049 before = &inode->i_flock;
2051 while ((fl = *before) != NULL) {
2052 if (fl->fl_file == filp) {
2054 locks_delete_lock(before);
2058 lease_modify(before, F_UNLCK);
2064 before = &fl->fl_next;
2070 * posix_unblock_lock - stop waiting for a file lock
2071 * @filp: how the file was opened
2072 * @waiter: the lock which was waiting
2074 * lockd needs to block waiting for locks.
2077 posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2082 if (waiter->fl_next)
2083 __locks_delete_block(waiter);
2090 EXPORT_SYMBOL(posix_unblock_lock);
2093 * vfs_cancel_lock - file byte range unblock lock
2094 * @filp: The file to apply the unblock to
2095 * @fl: The lock to be unblocked
2097 * Used by lock managers to cancel blocked requests
2099 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2101 if (filp->f_op && filp->f_op->lock)
2102 return filp->f_op->lock(filp, F_CANCELLK, fl);
2106 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2108 #ifdef CONFIG_PROC_FS
2109 #include <linux/proc_fs.h>
2110 #include <linux/seq_file.h>
2112 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2113 loff_t id, char *pfx)
2115 struct inode *inode = NULL;
2116 unsigned int fl_pid;
2119 fl_pid = pid_vnr(fl->fl_nspid);
2121 fl_pid = fl->fl_pid;
2123 if (fl->fl_file != NULL)
2124 inode = fl->fl_file->f_path.dentry->d_inode;
2126 seq_printf(f, "%lld:%s ", id, pfx);
2128 seq_printf(f, "%6s %s ",
2129 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2130 (inode == NULL) ? "*NOINODE*" :
2131 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2132 } else if (IS_FLOCK(fl)) {
2133 if (fl->fl_type & LOCK_MAND) {
2134 seq_printf(f, "FLOCK MSNFS ");
2136 seq_printf(f, "FLOCK ADVISORY ");
2138 } else if (IS_LEASE(fl)) {
2139 seq_printf(f, "LEASE ");
2140 if (fl->fl_type & F_INPROGRESS)
2141 seq_printf(f, "BREAKING ");
2142 else if (fl->fl_file)
2143 seq_printf(f, "ACTIVE ");
2145 seq_printf(f, "BREAKER ");
2147 seq_printf(f, "UNKNOWN UNKNOWN ");
2149 if (fl->fl_type & LOCK_MAND) {
2150 seq_printf(f, "%s ",
2151 (fl->fl_type & LOCK_READ)
2152 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2153 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2155 seq_printf(f, "%s ",
2156 (fl->fl_type & F_INPROGRESS)
2157 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2158 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2161 #ifdef WE_CAN_BREAK_LSLK_NOW
2162 seq_printf(f, "%d %s:%ld ", fl_pid,
2163 inode->i_sb->s_id, inode->i_ino);
2165 /* userspace relies on this representation of dev_t ;-( */
2166 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2167 MAJOR(inode->i_sb->s_dev),
2168 MINOR(inode->i_sb->s_dev), inode->i_ino);
2171 seq_printf(f, "%d <none>:0 ", fl_pid);
2174 if (fl->fl_end == OFFSET_MAX)
2175 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2177 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2179 seq_printf(f, "0 EOF\n");
2183 static int locks_show(struct seq_file *f, void *v)
2185 struct file_lock *fl, *bfl;
2187 fl = list_entry(v, struct file_lock, fl_link);
2189 lock_get_status(f, fl, *((loff_t *)f->private), "");
2191 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2192 lock_get_status(f, bfl, *((loff_t *)f->private), " ->");
2197 static void *locks_start(struct seq_file *f, loff_t *pos)
2199 loff_t *p = f->private;
2203 return seq_list_start(&file_lock_list, *pos);
2206 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2208 loff_t *p = f->private;
2210 return seq_list_next(v, &file_lock_list, pos);
2213 static void locks_stop(struct seq_file *f, void *v)
2218 static const struct seq_operations locks_seq_operations = {
2219 .start = locks_start,
2225 static int locks_open(struct inode *inode, struct file *filp)
2227 return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t));
2230 static const struct file_operations proc_locks_operations = {
2233 .llseek = seq_lseek,
2234 .release = seq_release_private,
2237 static int __init proc_locks_init(void)
2239 proc_create("locks", 0, NULL, &proc_locks_operations);
2242 module_init(proc_locks_init);
2246 * lock_may_read - checks that the region is free of locks
2247 * @inode: the inode that is being read
2248 * @start: the first byte to read
2249 * @len: the number of bytes to read
2251 * Emulates Windows locking requirements. Whole-file
2252 * mandatory locks (share modes) can prohibit a read and
2253 * byte-range POSIX locks can prohibit a read if they overlap.
2255 * N.B. this function is only ever called
2256 * from knfsd and ownership of locks is never checked.
2258 int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2260 struct file_lock *fl;
2263 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2265 if (fl->fl_type == F_RDLCK)
2267 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2269 } else if (IS_FLOCK(fl)) {
2270 if (!(fl->fl_type & LOCK_MAND))
2272 if (fl->fl_type & LOCK_READ)
2283 EXPORT_SYMBOL(lock_may_read);
2286 * lock_may_write - checks that the region is free of locks
2287 * @inode: the inode that is being written
2288 * @start: the first byte to write
2289 * @len: the number of bytes to write
2291 * Emulates Windows locking requirements. Whole-file
2292 * mandatory locks (share modes) can prohibit a write and
2293 * byte-range POSIX locks can prohibit a write if they overlap.
2295 * N.B. this function is only ever called
2296 * from knfsd and ownership of locks is never checked.
2298 int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2300 struct file_lock *fl;
2303 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2305 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2307 } else if (IS_FLOCK(fl)) {
2308 if (!(fl->fl_type & LOCK_MAND))
2310 if (fl->fl_type & LOCK_WRITE)
2321 EXPORT_SYMBOL(lock_may_write);
2323 static int __init filelock_init(void)
2325 filelock_cache = kmem_cache_create("file_lock_cache",
2326 sizeof(struct file_lock), 0, SLAB_PANIC,
2331 core_initcall(filelock_init);