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/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/vs_limit.h>
129 #include <asm/semaphore.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 LIST_HEAD(file_lock_list);
144 EXPORT_SYMBOL(file_lock_list);
146 static LIST_HEAD(blocked_list);
148 static kmem_cache_t *filelock_cache;
150 /* Allocate an empty lock structure. */
151 static struct file_lock *locks_alloc_lock(void)
153 if (!vx_locks_avail(1))
155 return kmem_cache_alloc(filelock_cache, SLAB_KERNEL);
158 /* Free a lock which is not in use. */
159 static inline void locks_free_lock(struct file_lock *fl)
167 if (waitqueue_active(&fl->fl_wait))
168 panic("Attempting to free lock with active wait queue");
170 if (!list_empty(&fl->fl_block))
171 panic("Attempting to free lock with active block list");
173 if (!list_empty(&fl->fl_link))
174 panic("Attempting to free lock on active lock list");
177 if (fl->fl_ops->fl_release_private)
178 fl->fl_ops->fl_release_private(fl);
183 if (fl->fl_lmops->fl_release_private)
184 fl->fl_lmops->fl_release_private(fl);
188 kmem_cache_free(filelock_cache, fl);
191 void locks_init_lock(struct file_lock *fl)
193 INIT_LIST_HEAD(&fl->fl_link);
194 INIT_LIST_HEAD(&fl->fl_block);
195 init_waitqueue_head(&fl->fl_wait);
197 fl->fl_fasync = NULL;
203 fl->fl_start = fl->fl_end = 0;
209 EXPORT_SYMBOL(locks_init_lock);
212 * Initialises the fields of the file lock which are invariant for
215 static void init_once(void *foo, kmem_cache_t *cache, unsigned long flags)
217 struct file_lock *lock = (struct file_lock *) foo;
219 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) !=
220 SLAB_CTOR_CONSTRUCTOR)
223 locks_init_lock(lock);
227 * Initialize a new lock from an existing file_lock structure.
229 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
231 new->fl_owner = fl->fl_owner;
232 new->fl_pid = fl->fl_pid;
233 new->fl_file = fl->fl_file;
234 new->fl_flags = fl->fl_flags;
235 new->fl_type = fl->fl_type;
236 new->fl_start = fl->fl_start;
237 new->fl_end = fl->fl_end;
238 new->fl_ops = fl->fl_ops;
239 new->fl_lmops = fl->fl_lmops;
240 if (fl->fl_ops && fl->fl_ops->fl_copy_lock)
241 fl->fl_ops->fl_copy_lock(new, fl);
242 if (fl->fl_lmops && fl->fl_lmops->fl_copy_lock)
243 fl->fl_lmops->fl_copy_lock(new, fl);
245 new->fl_xid = fl->fl_xid;
248 EXPORT_SYMBOL(locks_copy_lock);
250 static inline int flock_translate_cmd(int cmd) {
252 return cmd & (LOCK_MAND | LOCK_RW);
264 /* Fill in a file_lock structure with an appropriate FLOCK lock. */
265 static int flock_make_lock(struct file *filp, struct file_lock **lock,
268 struct file_lock *fl;
269 int type = flock_translate_cmd(cmd);
273 fl = locks_alloc_lock();
278 fl->fl_pid = current->tgid;
279 fl->fl_flags = FL_FLOCK;
281 fl->fl_end = OFFSET_MAX;
283 /* check against file xid maybe ? */
284 fl->fl_xid = vx_current_xid();
291 static int assign_type(struct file_lock *fl, int type)
305 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
308 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
313 switch (l->l_whence) {
321 start = i_size_read(filp->f_dentry->d_inode);
327 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
328 POSIX-2001 defines it. */
330 end = start + l->l_len - 1;
338 if (l->l_len > 0 && end < 0)
341 fl->fl_start = start; /* we record the absolute position */
344 fl->fl_end = OFFSET_MAX;
346 fl->fl_owner = current->files;
347 fl->fl_pid = current->tgid;
349 fl->fl_flags = FL_POSIX;
353 return assign_type(fl, l->l_type);
356 #if BITS_PER_LONG == 32
357 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
362 switch (l->l_whence) {
370 start = i_size_read(filp->f_dentry->d_inode);
376 if (((start += l->l_start) < 0) || (l->l_len < 0))
378 fl->fl_end = start + l->l_len - 1;
379 if (l->l_len > 0 && fl->fl_end < 0)
381 fl->fl_start = start; /* we record the absolute position */
383 fl->fl_end = OFFSET_MAX;
385 fl->fl_owner = current->files;
386 fl->fl_pid = current->tgid;
388 fl->fl_flags = FL_POSIX;
396 fl->fl_type = l->l_type;
406 /* default lease lock manager operations */
407 static void lease_break_callback(struct file_lock *fl)
409 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
412 static void lease_release_private_callback(struct file_lock *fl)
417 f_delown(fl->fl_file);
418 fl->fl_file->f_owner.signum = 0;
421 static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
423 return fl->fl_file == try->fl_file;
426 static struct lock_manager_operations lease_manager_ops = {
427 .fl_break = lease_break_callback,
428 .fl_release_private = lease_release_private_callback,
429 .fl_mylease = lease_mylease_callback,
430 .fl_change = lease_modify,
434 * Initialize a lease, use the default lock manager operations
436 static int lease_init(struct file *filp, int type, struct file_lock *fl)
438 fl->fl_owner = current->files;
439 fl->fl_pid = current->tgid;
442 fl->fl_flags = FL_LEASE;
443 if (assign_type(fl, type) != 0) {
448 fl->fl_end = OFFSET_MAX;
450 fl->fl_lmops = &lease_manager_ops;
454 /* Allocate a file_lock initialised to this type of lease */
455 static int lease_alloc(struct file *filp, int type, struct file_lock **flp)
457 struct file_lock *fl = locks_alloc_lock();
463 fl->fl_xid = vx_current_xid();
465 error = lease_init(filp, type, fl);
472 /* Check if two locks overlap each other.
474 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
476 return ((fl1->fl_end >= fl2->fl_start) &&
477 (fl2->fl_end >= fl1->fl_start));
481 * Check whether two locks have the same owner.
484 posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
486 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
487 return fl2->fl_lmops == fl1->fl_lmops &&
488 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
489 return fl1->fl_owner == fl2->fl_owner;
492 /* Remove waiter from blocker's block list.
493 * When blocker ends up pointing to itself then the list is empty.
495 static inline void __locks_delete_block(struct file_lock *waiter)
497 list_del_init(&waiter->fl_block);
498 list_del_init(&waiter->fl_link);
499 waiter->fl_next = NULL;
504 static void locks_delete_block(struct file_lock *waiter)
507 __locks_delete_block(waiter);
511 /* Insert waiter into blocker's block list.
512 * We use a circular list so that processes can be easily woken up in
513 * the order they blocked. The documentation doesn't require this but
514 * it seems like the reasonable thing to do.
516 static void locks_insert_block(struct file_lock *blocker,
517 struct file_lock *waiter)
519 if (!list_empty(&waiter->fl_block)) {
520 printk(KERN_ERR "locks_insert_block: removing duplicated lock "
521 "(pid=%d %Ld-%Ld type=%d)\n", waiter->fl_pid,
522 waiter->fl_start, waiter->fl_end, waiter->fl_type);
523 __locks_delete_block(waiter);
525 list_add_tail(&waiter->fl_block, &blocker->fl_block);
526 waiter->fl_next = blocker;
527 if (IS_POSIX(blocker))
528 list_add(&waiter->fl_link, &blocked_list);
531 /* Wake up processes blocked waiting for blocker.
532 * If told to wait then schedule the processes until the block list
533 * is empty, otherwise empty the block list ourselves.
535 static void locks_wake_up_blocks(struct file_lock *blocker)
537 while (!list_empty(&blocker->fl_block)) {
538 struct file_lock *waiter = list_entry(blocker->fl_block.next,
539 struct file_lock, fl_block);
540 __locks_delete_block(waiter);
541 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
542 waiter->fl_lmops->fl_notify(waiter);
544 wake_up(&waiter->fl_wait);
548 /* Insert file lock fl into an inode's lock list at the position indicated
549 * by pos. At the same time add the lock to the global file lock list.
551 static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
553 list_add(&fl->fl_link, &file_lock_list);
555 /* insert into file's list */
559 if (fl->fl_ops && fl->fl_ops->fl_insert)
560 fl->fl_ops->fl_insert(fl);
564 * Delete a lock and then free it.
565 * Wake up processes that are blocked waiting for this lock,
566 * notify the FS that the lock has been cleared and
567 * finally free the lock.
569 static void locks_delete_lock(struct file_lock **thisfl_p)
571 struct file_lock *fl = *thisfl_p;
573 *thisfl_p = fl->fl_next;
575 list_del_init(&fl->fl_link);
577 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
578 if (fl->fl_fasync != NULL) {
579 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
580 fl->fl_fasync = NULL;
583 if (fl->fl_ops && fl->fl_ops->fl_remove)
584 fl->fl_ops->fl_remove(fl);
586 locks_wake_up_blocks(fl);
590 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
591 * checks for shared/exclusive status of overlapping locks.
593 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
595 if (sys_fl->fl_type == F_WRLCK)
597 if (caller_fl->fl_type == F_WRLCK)
602 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
603 * checking before calling the locks_conflict().
605 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
607 /* POSIX locks owned by the same process do not conflict with
610 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
613 /* Check whether they overlap */
614 if (!locks_overlap(caller_fl, sys_fl))
617 return (locks_conflict(caller_fl, sys_fl));
620 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
621 * checking before calling the locks_conflict().
623 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
625 /* FLOCK locks referring to the same filp do not conflict with
628 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
630 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
633 return (locks_conflict(caller_fl, sys_fl));
636 static int interruptible_sleep_on_locked(wait_queue_head_t *fl_wait, int timeout)
639 DECLARE_WAITQUEUE(wait, current);
641 __set_current_state(TASK_INTERRUPTIBLE);
642 add_wait_queue(fl_wait, &wait);
646 result = schedule_timeout(timeout);
647 if (signal_pending(current))
648 result = -ERESTARTSYS;
649 remove_wait_queue(fl_wait, &wait);
650 __set_current_state(TASK_RUNNING);
654 static int locks_block_on_timeout(struct file_lock *blocker, struct file_lock *waiter, int time)
657 locks_insert_block(blocker, waiter);
658 result = interruptible_sleep_on_locked(&waiter->fl_wait, time);
659 __locks_delete_block(waiter);
664 posix_test_lock(struct file *filp, struct file_lock *fl)
666 struct file_lock *cfl;
669 for (cfl = filp->f_dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
672 if (posix_locks_conflict(cfl, fl))
680 EXPORT_SYMBOL(posix_test_lock);
682 /* This function tests for deadlock condition before putting a process to
683 * sleep. The detection scheme is no longer recursive. Recursive was neat,
684 * but dangerous - we risked stack corruption if the lock data was bad, or
685 * if the recursion was too deep for any other reason.
687 * We rely on the fact that a task can only be on one lock's wait queue
688 * at a time. When we find blocked_task on a wait queue we can re-search
689 * with blocked_task equal to that queue's owner, until either blocked_task
690 * isn't found, or blocked_task is found on a queue owned by my_task.
692 * Note: the above assumption may not be true when handling lock requests
693 * from a broken NFS client. But broken NFS clients have a lot more to
694 * worry about than proper deadlock detection anyway... --okir
696 int posix_locks_deadlock(struct file_lock *caller_fl,
697 struct file_lock *block_fl)
699 struct list_head *tmp;
702 if (posix_same_owner(caller_fl, block_fl))
704 list_for_each(tmp, &blocked_list) {
705 struct file_lock *fl = list_entry(tmp, struct file_lock, fl_link);
706 if (posix_same_owner(fl, block_fl)) {
715 EXPORT_SYMBOL(posix_locks_deadlock);
717 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
718 * at the head of the list, but that's secret knowledge known only to
719 * flock_lock_file and posix_lock_file.
721 static int flock_lock_file(struct file *filp, struct file_lock *new_fl)
723 struct file_lock **before;
724 struct inode * inode = filp->f_dentry->d_inode;
729 for_each_lock(inode, before) {
730 struct file_lock *fl = *before;
735 if (filp != fl->fl_file)
737 if (new_fl->fl_type == fl->fl_type)
740 locks_delete_lock(before);
745 if (new_fl->fl_type == F_UNLCK)
749 * If a higher-priority process was blocked on the old file lock,
750 * give it the opportunity to lock the file.
756 for_each_lock(inode, before) {
757 struct file_lock *fl = *before;
762 if (!flock_locks_conflict(new_fl, fl))
765 if (new_fl->fl_flags & FL_SLEEP) {
766 locks_insert_block(fl, new_fl);
770 locks_insert_lock(&inode->i_flock, new_fl);
778 EXPORT_SYMBOL(posix_lock_file);
780 static int __posix_lock_file(struct inode *inode, struct file_lock *request)
782 struct file_lock *fl;
783 struct file_lock *new_fl, *new_fl2;
784 struct file_lock *left = NULL;
785 struct file_lock *right = NULL;
786 struct file_lock **before;
787 int error, added = 0;
790 * We may need two file_lock structures for this operation,
791 * so we get them in advance to avoid races.
793 new_fl = locks_alloc_lock();
794 new_fl->fl_xid = vx_current_xid();
795 vx_locks_inc(new_fl);
796 new_fl2 = locks_alloc_lock();
797 new_fl2->fl_xid = vx_current_xid();
798 vx_locks_inc(new_fl2);
801 if (request->fl_type != F_UNLCK) {
802 for_each_lock(inode, before) {
803 struct file_lock *fl = *before;
806 if (!posix_locks_conflict(request, fl))
809 if (!(request->fl_flags & FL_SLEEP))
812 if (posix_locks_deadlock(request, fl))
815 locks_insert_block(fl, request);
820 /* If we're just looking for a conflict, we're done. */
822 if (request->fl_flags & FL_ACCESS)
825 error = -ENOLCK; /* "no luck" */
826 if (!(new_fl && new_fl2))
830 * We've allocated the new locks in advance, so there are no
831 * errors possible (and no blocking operations) from here on.
833 * Find the first old lock with the same owner as the new lock.
836 before = &inode->i_flock;
838 /* First skip locks owned by other processes. */
839 while ((fl = *before) && (!IS_POSIX(fl) ||
840 !posix_same_owner(request, fl))) {
841 before = &fl->fl_next;
844 /* Process locks with this owner. */
845 while ((fl = *before) && posix_same_owner(request, fl)) {
846 /* Detect adjacent or overlapping regions (if same lock type)
848 if (request->fl_type == fl->fl_type) {
849 if (fl->fl_end < request->fl_start - 1)
851 /* If the next lock in the list has entirely bigger
852 * addresses than the new one, insert the lock here.
854 if (fl->fl_start > request->fl_end + 1)
857 /* If we come here, the new and old lock are of the
858 * same type and adjacent or overlapping. Make one
859 * lock yielding from the lower start address of both
860 * locks to the higher end address.
862 if (fl->fl_start > request->fl_start)
863 fl->fl_start = request->fl_start;
865 request->fl_start = fl->fl_start;
866 if (fl->fl_end < request->fl_end)
867 fl->fl_end = request->fl_end;
869 request->fl_end = fl->fl_end;
871 locks_delete_lock(before);
878 /* Processing for different lock types is a bit
881 if (fl->fl_end < request->fl_start)
883 if (fl->fl_start > request->fl_end)
885 if (request->fl_type == F_UNLCK)
887 if (fl->fl_start < request->fl_start)
889 /* If the next lock in the list has a higher end
890 * address than the new one, insert the new one here.
892 if (fl->fl_end > request->fl_end) {
896 if (fl->fl_start >= request->fl_start) {
897 /* The new lock completely replaces an old
898 * one (This may happen several times).
901 locks_delete_lock(before);
904 /* Replace the old lock with the new one.
905 * Wake up anybody waiting for the old one,
906 * as the change in lock type might satisfy
909 locks_wake_up_blocks(fl);
910 fl->fl_start = request->fl_start;
911 fl->fl_end = request->fl_end;
912 fl->fl_type = request->fl_type;
913 fl->fl_u = request->fl_u;
918 /* Go on to next lock.
921 before = &fl->fl_next;
926 if (request->fl_type == F_UNLCK)
928 locks_copy_lock(new_fl, request);
929 locks_insert_lock(before, new_fl);
934 /* The new lock breaks the old one in two pieces,
935 * so we have to use the second new lock.
939 locks_copy_lock(left, right);
940 locks_insert_lock(before, left);
942 right->fl_start = request->fl_end + 1;
943 locks_wake_up_blocks(right);
946 left->fl_end = request->fl_start - 1;
947 locks_wake_up_blocks(left);
952 * Free any unused locks.
955 locks_free_lock(new_fl);
957 locks_free_lock(new_fl2);
962 * posix_lock_file - Apply a POSIX-style lock to a file
963 * @filp: The file to apply the lock to
964 * @fl: The lock to be applied
966 * Add a POSIX style lock to a file.
967 * We merge adjacent & overlapping locks whenever possible.
968 * POSIX locks are sorted by owner task, then by starting address
970 int posix_lock_file(struct file *filp, struct file_lock *fl)
972 return __posix_lock_file(filp->f_dentry->d_inode, fl);
976 * posix_lock_file_wait - Apply a POSIX-style lock to a file
977 * @filp: The file to apply the lock to
978 * @fl: The lock to be applied
980 * Add a POSIX style lock to a file.
981 * We merge adjacent & overlapping locks whenever possible.
982 * POSIX locks are sorted by owner task, then by starting address
984 int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
989 error = __posix_lock_file(filp->f_dentry->d_inode, fl);
990 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
992 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
996 locks_delete_block(fl);
1001 EXPORT_SYMBOL(posix_lock_file_wait);
1004 * locks_mandatory_locked - Check for an active lock
1005 * @inode: the file to check
1007 * Searches the inode's list of locks to find any POSIX locks which conflict.
1008 * This function is called from locks_verify_locked() only.
1010 int locks_mandatory_locked(struct inode *inode)
1012 fl_owner_t owner = current->files;
1013 struct file_lock *fl;
1016 * Search the lock list for this inode for any POSIX locks.
1019 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1022 if (fl->fl_owner != owner)
1026 return fl ? -EAGAIN : 0;
1030 * locks_mandatory_area - Check for a conflicting lock
1031 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1033 * @inode: the file to check
1034 * @filp: how the file was opened (if it was)
1035 * @offset: start of area to check
1036 * @count: length of area to check
1038 * Searches the inode's list of locks to find any POSIX locks which conflict.
1039 * This function is called from rw_verify_area() and
1040 * locks_verify_truncate().
1042 int locks_mandatory_area(int read_write, struct inode *inode,
1043 struct file *filp, loff_t offset,
1046 struct file_lock fl;
1049 locks_init_lock(&fl);
1050 fl.fl_owner = current->files;
1051 fl.fl_pid = current->tgid;
1053 fl.fl_flags = FL_POSIX | FL_ACCESS;
1054 if (filp && !(filp->f_flags & O_NONBLOCK))
1055 fl.fl_flags |= FL_SLEEP;
1056 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1057 fl.fl_start = offset;
1058 fl.fl_end = offset + count - 1;
1061 error = __posix_lock_file(inode, &fl);
1062 if (error != -EAGAIN)
1064 if (!(fl.fl_flags & FL_SLEEP))
1066 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1069 * If we've been sleeping someone might have
1070 * changed the permissions behind our back.
1072 if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
1076 locks_delete_block(&fl);
1083 EXPORT_SYMBOL(locks_mandatory_area);
1085 /* We already had a lease on this file; just change its type */
1086 int lease_modify(struct file_lock **before, int arg)
1088 struct file_lock *fl = *before;
1089 int error = assign_type(fl, arg);
1093 locks_wake_up_blocks(fl);
1095 locks_delete_lock(before);
1099 EXPORT_SYMBOL(lease_modify);
1101 static void time_out_leases(struct inode *inode)
1103 struct file_lock **before;
1104 struct file_lock *fl;
1106 before = &inode->i_flock;
1107 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1108 if ((fl->fl_break_time == 0)
1109 || time_before(jiffies, fl->fl_break_time)) {
1110 before = &fl->fl_next;
1113 printk(KERN_INFO "lease broken - owner pid = %d\n", fl->fl_pid);
1114 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1115 if (fl == *before) /* lease_modify may have freed fl */
1116 before = &fl->fl_next;
1121 * __break_lease - revoke all outstanding leases on file
1122 * @inode: the inode of the file to return
1123 * @mode: the open mode (read or write)
1125 * break_lease (inlined for speed) has checked there already
1126 * is a lease on this file. Leases are broken on a call to open()
1127 * or truncate(). This function can sleep unless you
1128 * specified %O_NONBLOCK to your open().
1130 int __break_lease(struct inode *inode, unsigned int mode)
1132 int error = 0, future;
1133 struct file_lock *new_fl, *flock;
1134 struct file_lock *fl;
1136 unsigned long break_time;
1137 int i_have_this_lease = 0;
1139 alloc_err = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK,
1144 time_out_leases(inode);
1146 flock = inode->i_flock;
1147 if ((flock == NULL) || !IS_LEASE(flock))
1150 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1151 if (fl->fl_owner == current->files)
1152 i_have_this_lease = 1;
1154 if (mode & FMODE_WRITE) {
1155 /* If we want write access, we have to revoke any lease. */
1156 future = F_UNLCK | F_INPROGRESS;
1157 } else if (flock->fl_type & F_INPROGRESS) {
1158 /* If the lease is already being broken, we just leave it */
1159 future = flock->fl_type;
1160 } else if (flock->fl_type & F_WRLCK) {
1161 /* Downgrade the exclusive lease to a read-only lease. */
1162 future = F_RDLCK | F_INPROGRESS;
1164 /* the existing lease was read-only, so we can read too. */
1168 if (alloc_err && !i_have_this_lease && ((mode & O_NONBLOCK) == 0)) {
1174 if (lease_break_time > 0) {
1175 break_time = jiffies + lease_break_time * HZ;
1176 if (break_time == 0)
1177 break_time++; /* so that 0 means no break time */
1180 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1181 if (fl->fl_type != future) {
1182 fl->fl_type = future;
1183 fl->fl_break_time = break_time;
1184 /* lease must have lmops break callback */
1185 fl->fl_lmops->fl_break(fl);
1189 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1190 error = -EWOULDBLOCK;
1195 break_time = flock->fl_break_time;
1196 if (break_time != 0) {
1197 break_time -= jiffies;
1198 if (break_time == 0)
1201 error = locks_block_on_timeout(flock, new_fl, break_time);
1204 time_out_leases(inode);
1205 /* Wait for the next lease that has not been broken yet */
1206 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1207 flock = flock->fl_next) {
1208 if (flock->fl_type & F_INPROGRESS)
1217 locks_free_lock(new_fl);
1221 EXPORT_SYMBOL(__break_lease);
1226 * @time: pointer to a timespec which will contain the last modified time
1228 * This is to force NFS clients to flush their caches for files with
1229 * exclusive leases. The justification is that if someone has an
1230 * exclusive lease, then they could be modifiying it.
1232 void lease_get_mtime(struct inode *inode, struct timespec *time)
1234 struct file_lock *flock = inode->i_flock;
1235 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1236 *time = current_fs_time(inode->i_sb);
1238 *time = inode->i_mtime;
1241 EXPORT_SYMBOL(lease_get_mtime);
1244 * fcntl_getlease - Enquire what lease is currently active
1247 * The value returned by this function will be one of
1248 * (if no lease break is pending):
1250 * %F_RDLCK to indicate a shared lease is held.
1252 * %F_WRLCK to indicate an exclusive lease is held.
1254 * %F_UNLCK to indicate no lease is held.
1256 * (if a lease break is pending):
1258 * %F_RDLCK to indicate an exclusive lease needs to be
1259 * changed to a shared lease (or removed).
1261 * %F_UNLCK to indicate the lease needs to be removed.
1263 * XXX: sfr & willy disagree over whether F_INPROGRESS
1264 * should be returned to userspace.
1266 int fcntl_getlease(struct file *filp)
1268 struct file_lock *fl;
1272 time_out_leases(filp->f_dentry->d_inode);
1273 for (fl = filp->f_dentry->d_inode->i_flock; fl && IS_LEASE(fl);
1275 if (fl->fl_file == filp) {
1276 type = fl->fl_type & ~F_INPROGRESS;
1285 * __setlease - sets a lease on an open file
1286 * @filp: file pointer
1287 * @arg: type of lease to obtain
1288 * @flp: input - file_lock to use, output - file_lock inserted
1290 * The (input) flp->fl_lmops->fl_break function is required
1293 * Called with kernel lock held.
1295 static int __setlease(struct file *filp, long arg, struct file_lock **flp)
1297 struct file_lock *fl, **before, **my_before = NULL, *lease = *flp;
1298 struct dentry *dentry = filp->f_dentry;
1299 struct inode *inode = dentry->d_inode;
1300 int error, rdlease_count = 0, wrlease_count = 0;
1302 time_out_leases(inode);
1305 if (!flp || !(*flp) || !(*flp)->fl_lmops || !(*flp)->fl_lmops->fl_break)
1309 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1311 if ((arg == F_WRLCK)
1312 && ((atomic_read(&dentry->d_count) > 1)
1313 || (atomic_read(&inode->i_count) > 1)))
1317 * At this point, we know that if there is an exclusive
1318 * lease on this file, then we hold it on this filp
1319 * (otherwise our open of this file would have blocked).
1320 * And if we are trying to acquire an exclusive lease,
1321 * then the file is not open by anyone (including us)
1322 * except for this filp.
1324 for (before = &inode->i_flock;
1325 ((fl = *before) != NULL) && IS_LEASE(fl);
1326 before = &fl->fl_next) {
1327 if (lease->fl_lmops->fl_mylease(fl, lease))
1329 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1331 * Someone is in the process of opening this
1332 * file for writing so we may not take an
1333 * exclusive lease on it.
1340 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1341 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1344 if (my_before != NULL) {
1345 error = lease->fl_lmops->fl_change(my_before, arg);
1357 error = lease_alloc(filp, arg, &fl);
1361 locks_copy_lock(fl, lease);
1363 locks_insert_lock(before, fl);
1371 * setlease - sets a lease on an open file
1372 * @filp: file pointer
1373 * @arg: type of lease to obtain
1374 * @lease: file_lock to use
1376 * Call this to establish a lease on the file.
1377 * The fl_lmops fl_break function is required by break_lease
1380 int setlease(struct file *filp, long arg, struct file_lock **lease)
1382 struct dentry *dentry = filp->f_dentry;
1383 struct inode *inode = dentry->d_inode;
1386 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1388 if (!S_ISREG(inode->i_mode))
1390 error = security_file_lock(filp, arg);
1395 error = __setlease(filp, arg, lease);
1401 EXPORT_SYMBOL(setlease);
1404 * fcntl_setlease - sets a lease on an open file
1405 * @fd: open file descriptor
1406 * @filp: file pointer
1407 * @arg: type of lease to obtain
1409 * Call this fcntl to establish a lease on the file.
1410 * Note that you also need to call %F_SETSIG to
1411 * receive a signal when the lease is broken.
1413 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1415 struct file_lock fl, *flp = &fl;
1416 struct dentry *dentry = filp->f_dentry;
1417 struct inode *inode = dentry->d_inode;
1420 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1422 if (!S_ISREG(inode->i_mode))
1424 error = security_file_lock(filp, arg);
1428 locks_init_lock(&fl);
1429 error = lease_init(filp, arg, &fl);
1435 error = __setlease(filp, arg, &flp);
1439 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1441 /* remove lease just inserted by __setlease */
1442 flp->fl_type = F_UNLCK | F_INPROGRESS;
1443 flp->fl_break_time = jiffies- 10;
1444 time_out_leases(inode);
1448 error = f_setown(filp, current->pid, 0);
1455 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1456 * @filp: The file to apply the lock to
1457 * @fl: The lock to be applied
1459 * Add a FLOCK style lock to a file.
1461 int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1466 error = flock_lock_file(filp, fl);
1467 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1469 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1473 locks_delete_block(fl);
1479 EXPORT_SYMBOL(flock_lock_file_wait);
1482 * sys_flock: - flock() system call.
1483 * @fd: the file descriptor to lock.
1484 * @cmd: the type of lock to apply.
1486 * Apply a %FL_FLOCK style lock to an open file descriptor.
1487 * The @cmd can be one of
1489 * %LOCK_SH -- a shared lock.
1491 * %LOCK_EX -- an exclusive lock.
1493 * %LOCK_UN -- remove an existing lock.
1495 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1497 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1498 * processes read and write access respectively.
1500 asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1503 struct file_lock *lock;
1504 int can_sleep, unlock;
1512 can_sleep = !(cmd & LOCK_NB);
1514 unlock = (cmd == LOCK_UN);
1516 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1519 error = flock_make_lock(filp, &lock, cmd);
1523 lock->fl_flags |= FL_SLEEP;
1525 error = security_file_lock(filp, cmd);
1529 if (filp->f_op && filp->f_op->flock)
1530 error = filp->f_op->flock(filp,
1531 (can_sleep) ? F_SETLKW : F_SETLK,
1534 error = flock_lock_file_wait(filp, lock);
1537 if (list_empty(&lock->fl_link)) {
1538 locks_free_lock(lock);
1547 /* Report the first existing lock that would conflict with l.
1548 * This implements the F_GETLK command of fcntl().
1550 int fcntl_getlk(struct file *filp, struct flock __user *l)
1552 struct file_lock *fl, file_lock;
1557 if (copy_from_user(&flock, l, sizeof(flock)))
1560 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1563 error = flock_to_posix_lock(filp, &file_lock, &flock);
1567 if (filp->f_op && filp->f_op->lock) {
1568 error = filp->f_op->lock(filp, F_GETLK, &file_lock);
1572 fl = (file_lock.fl_type == F_UNLCK ? NULL : &file_lock);
1574 fl = posix_test_lock(filp, &file_lock);
1577 flock.l_type = F_UNLCK;
1579 flock.l_pid = fl->fl_pid;
1580 #if BITS_PER_LONG == 32
1582 * Make sure we can represent the posix lock via
1583 * legacy 32bit flock.
1586 if (fl->fl_start > OFFT_OFFSET_MAX)
1588 if ((fl->fl_end != OFFSET_MAX)
1589 && (fl->fl_end > OFFT_OFFSET_MAX))
1592 flock.l_start = fl->fl_start;
1593 flock.l_len = fl->fl_end == OFFSET_MAX ? 0 :
1594 fl->fl_end - fl->fl_start + 1;
1596 flock.l_type = fl->fl_type;
1599 if (!copy_to_user(l, &flock, sizeof(flock)))
1605 /* Apply the lock described by l to an open file descriptor.
1606 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1608 int fcntl_setlk(struct file *filp, unsigned int cmd, struct flock __user *l)
1610 struct file_lock *file_lock = locks_alloc_lock();
1612 struct inode *inode;
1615 if (file_lock == NULL)
1618 file_lock->fl_xid = vx_current_xid();
1619 vx_locks_inc(file_lock);
1622 * This might block, so we do it before checking the inode.
1625 if (copy_from_user(&flock, l, sizeof(flock)))
1628 inode = filp->f_dentry->d_inode;
1630 /* Don't allow mandatory locks on files that may be memory mapped
1633 if (IS_MANDLOCK(inode) &&
1634 (inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
1635 mapping_writably_mapped(filp->f_mapping)) {
1640 error = flock_to_posix_lock(filp, file_lock, &flock);
1643 if (cmd == F_SETLKW) {
1644 file_lock->fl_flags |= FL_SLEEP;
1648 switch (flock.l_type) {
1650 if (!(filp->f_mode & FMODE_READ))
1654 if (!(filp->f_mode & FMODE_WRITE))
1664 error = security_file_lock(filp, file_lock->fl_type);
1668 if (filp->f_op && filp->f_op->lock != NULL) {
1669 error = filp->f_op->lock(filp, cmd, file_lock);
1674 error = __posix_lock_file(inode, file_lock);
1675 if ((error != -EAGAIN) || (cmd == F_SETLK))
1677 error = wait_event_interruptible(file_lock->fl_wait,
1678 !file_lock->fl_next);
1682 locks_delete_block(file_lock);
1687 locks_free_lock(file_lock);
1691 #if BITS_PER_LONG == 32
1692 /* Report the first existing lock that would conflict with l.
1693 * This implements the F_GETLK command of fcntl().
1695 int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1697 struct file_lock *fl, file_lock;
1698 struct flock64 flock;
1702 if (copy_from_user(&flock, l, sizeof(flock)))
1705 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1708 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1712 if (filp->f_op && filp->f_op->lock) {
1713 error = filp->f_op->lock(filp, F_GETLK, &file_lock);
1717 fl = (file_lock.fl_type == F_UNLCK ? NULL : &file_lock);
1719 fl = posix_test_lock(filp, &file_lock);
1722 flock.l_type = F_UNLCK;
1724 flock.l_pid = fl->fl_pid;
1725 flock.l_start = fl->fl_start;
1726 flock.l_len = fl->fl_end == OFFSET_MAX ? 0 :
1727 fl->fl_end - fl->fl_start + 1;
1729 flock.l_type = fl->fl_type;
1732 if (!copy_to_user(l, &flock, sizeof(flock)))
1739 /* Apply the lock described by l to an open file descriptor.
1740 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1742 int fcntl_setlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
1744 struct file_lock *file_lock = locks_alloc_lock();
1745 struct flock64 flock;
1746 struct inode *inode;
1749 if (file_lock == NULL)
1752 file_lock->fl_xid = vx_current_xid();
1753 vx_locks_inc(file_lock);
1756 * This might block, so we do it before checking the inode.
1759 if (copy_from_user(&flock, l, sizeof(flock)))
1762 inode = filp->f_dentry->d_inode;
1764 /* Don't allow mandatory locks on files that may be memory mapped
1767 if (IS_MANDLOCK(inode) &&
1768 (inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
1769 mapping_writably_mapped(filp->f_mapping)) {
1774 error = flock64_to_posix_lock(filp, file_lock, &flock);
1777 if (cmd == F_SETLKW64) {
1778 file_lock->fl_flags |= FL_SLEEP;
1782 switch (flock.l_type) {
1784 if (!(filp->f_mode & FMODE_READ))
1788 if (!(filp->f_mode & FMODE_WRITE))
1798 error = security_file_lock(filp, file_lock->fl_type);
1802 if (filp->f_op && filp->f_op->lock != NULL) {
1803 error = filp->f_op->lock(filp, cmd, file_lock);
1808 error = __posix_lock_file(inode, file_lock);
1809 if ((error != -EAGAIN) || (cmd == F_SETLK64))
1811 error = wait_event_interruptible(file_lock->fl_wait,
1812 !file_lock->fl_next);
1816 locks_delete_block(file_lock);
1821 locks_free_lock(file_lock);
1824 #endif /* BITS_PER_LONG == 32 */
1827 * This function is called when the file is being removed
1828 * from the task's fd array. POSIX locks belonging to this task
1829 * are deleted at this time.
1831 void locks_remove_posix(struct file *filp, fl_owner_t owner)
1833 struct file_lock lock, **before;
1836 * If there are no locks held on this file, we don't need to call
1837 * posix_lock_file(). Another process could be setting a lock on this
1838 * file at the same time, but we wouldn't remove that lock anyway.
1840 before = &filp->f_dentry->d_inode->i_flock;
1841 if (*before == NULL)
1844 lock.fl_type = F_UNLCK;
1845 lock.fl_flags = FL_POSIX;
1847 lock.fl_end = OFFSET_MAX;
1848 lock.fl_owner = owner;
1849 lock.fl_pid = current->tgid;
1850 lock.fl_file = filp;
1852 lock.fl_lmops = NULL;
1854 if (filp->f_op && filp->f_op->lock != NULL) {
1855 filp->f_op->lock(filp, F_SETLK, &lock);
1859 /* Can't use posix_lock_file here; we need to remove it no matter
1860 * which pid we have.
1863 while (*before != NULL) {
1864 struct file_lock *fl = *before;
1865 if (IS_POSIX(fl) && posix_same_owner(fl, &lock)) {
1866 locks_delete_lock(before);
1869 before = &fl->fl_next;
1873 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1874 lock.fl_ops->fl_release_private(&lock);
1877 EXPORT_SYMBOL(locks_remove_posix);
1880 * This function is called on the last close of an open file.
1882 void locks_remove_flock(struct file *filp)
1884 struct inode * inode = filp->f_dentry->d_inode;
1885 struct file_lock *fl;
1886 struct file_lock **before;
1888 if (!inode->i_flock)
1891 if (filp->f_op && filp->f_op->flock) {
1892 struct file_lock fl = {
1893 .fl_pid = current->tgid,
1895 .fl_flags = FL_FLOCK,
1897 .fl_end = OFFSET_MAX,
1899 filp->f_op->flock(filp, F_SETLKW, &fl);
1903 before = &inode->i_flock;
1905 while ((fl = *before) != NULL) {
1906 if (fl->fl_file == filp) {
1908 * We might have a POSIX lock that was created at the same time
1909 * the filp was closed for the last time. Just remove that too,
1910 * regardless of ownership, since nobody can own it.
1912 if (IS_FLOCK(fl) || IS_POSIX(fl)) {
1913 locks_delete_lock(before);
1917 lease_modify(before, F_UNLCK);
1923 before = &fl->fl_next;
1929 * posix_block_lock - blocks waiting for a file lock
1930 * @blocker: the lock which is blocking
1931 * @waiter: the lock which conflicts and has to wait
1933 * lockd needs to block waiting for locks.
1936 posix_block_lock(struct file_lock *blocker, struct file_lock *waiter)
1938 locks_insert_block(blocker, waiter);
1941 EXPORT_SYMBOL(posix_block_lock);
1944 * posix_unblock_lock - stop waiting for a file lock
1945 * @filp: how the file was opened
1946 * @waiter: the lock which was waiting
1948 * lockd needs to block waiting for locks.
1951 posix_unblock_lock(struct file *filp, struct file_lock *waiter)
1954 * A remote machine may cancel the lock request after it's been
1955 * granted locally. If that happens, we need to delete the lock.
1958 if (waiter->fl_next) {
1959 __locks_delete_block(waiter);
1963 waiter->fl_type = F_UNLCK;
1964 posix_lock_file(filp, waiter);
1968 EXPORT_SYMBOL(posix_unblock_lock);
1970 static void lock_get_status(char* out, struct file_lock *fl, int id, char *pfx)
1972 struct inode *inode = NULL;
1974 if (fl->fl_file != NULL)
1975 inode = fl->fl_file->f_dentry->d_inode;
1977 out += sprintf(out, "%d:%s ", id, pfx);
1979 out += sprintf(out, "%6s %s ",
1980 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
1981 (inode == NULL) ? "*NOINODE*" :
1982 (IS_MANDLOCK(inode) &&
1983 (inode->i_mode & (S_IXGRP | S_ISGID)) == S_ISGID) ?
1984 "MANDATORY" : "ADVISORY ");
1985 } else if (IS_FLOCK(fl)) {
1986 if (fl->fl_type & LOCK_MAND) {
1987 out += sprintf(out, "FLOCK MSNFS ");
1989 out += sprintf(out, "FLOCK ADVISORY ");
1991 } else if (IS_LEASE(fl)) {
1992 out += sprintf(out, "LEASE ");
1993 if (fl->fl_type & F_INPROGRESS)
1994 out += sprintf(out, "BREAKING ");
1995 else if (fl->fl_file)
1996 out += sprintf(out, "ACTIVE ");
1998 out += sprintf(out, "BREAKER ");
2000 out += sprintf(out, "UNKNOWN UNKNOWN ");
2002 if (fl->fl_type & LOCK_MAND) {
2003 out += sprintf(out, "%s ",
2004 (fl->fl_type & LOCK_READ)
2005 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2006 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2008 out += sprintf(out, "%s ",
2009 (fl->fl_type & F_INPROGRESS)
2010 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2011 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2014 #ifdef WE_CAN_BREAK_LSLK_NOW
2015 out += sprintf(out, "%d %s:%ld ", fl->fl_pid,
2016 inode->i_sb->s_id, inode->i_ino);
2018 /* userspace relies on this representation of dev_t ;-( */
2019 out += sprintf(out, "%d %02x:%02x:%ld ", fl->fl_pid,
2020 MAJOR(inode->i_sb->s_dev),
2021 MINOR(inode->i_sb->s_dev), inode->i_ino);
2024 out += sprintf(out, "%d <none>:0 ", fl->fl_pid);
2027 if (fl->fl_end == OFFSET_MAX)
2028 out += sprintf(out, "%Ld EOF\n", fl->fl_start);
2030 out += sprintf(out, "%Ld %Ld\n", fl->fl_start,
2033 out += sprintf(out, "0 EOF\n");
2037 static void move_lock_status(char **p, off_t* pos, off_t offset)
2041 if(*pos >= offset) {
2042 /* the complete line is valid */
2047 if(*pos+len > offset) {
2048 /* use the second part of the line */
2049 int i = offset-*pos;
2050 memmove(*p,*p+i,len-i);
2055 /* discard the complete line */
2060 * get_locks_status - reports lock usage in /proc/locks
2061 * @buffer: address in userspace to write into
2063 * @offset: how far we are through the buffer
2064 * @length: how much to read
2067 int get_locks_status(char *buffer, char **start, off_t offset, int length)
2069 struct list_head *tmp;
2075 list_for_each(tmp, &file_lock_list) {
2076 struct list_head *btmp;
2077 struct file_lock *fl = list_entry(tmp, struct file_lock, fl_link);
2078 lock_get_status(q, fl, ++i, "");
2079 move_lock_status(&q, &pos, offset);
2081 if(pos >= offset+length)
2084 list_for_each(btmp, &fl->fl_block) {
2085 struct file_lock *bfl = list_entry(btmp,
2086 struct file_lock, fl_block);
2087 lock_get_status(q, bfl, i, " ->");
2088 move_lock_status(&q, &pos, offset);
2090 if(pos >= offset+length)
2097 if(q-buffer < length)
2103 * lock_may_read - checks that the region is free of locks
2104 * @inode: the inode that is being read
2105 * @start: the first byte to read
2106 * @len: the number of bytes to read
2108 * Emulates Windows locking requirements. Whole-file
2109 * mandatory locks (share modes) can prohibit a read and
2110 * byte-range POSIX locks can prohibit a read if they overlap.
2112 * N.B. this function is only ever called
2113 * from knfsd and ownership of locks is never checked.
2115 int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2117 struct file_lock *fl;
2120 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2122 if (fl->fl_type == F_RDLCK)
2124 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2126 } else if (IS_FLOCK(fl)) {
2127 if (!(fl->fl_type & LOCK_MAND))
2129 if (fl->fl_type & LOCK_READ)
2140 EXPORT_SYMBOL(lock_may_read);
2143 * lock_may_write - checks that the region is free of locks
2144 * @inode: the inode that is being written
2145 * @start: the first byte to write
2146 * @len: the number of bytes to write
2148 * Emulates Windows locking requirements. Whole-file
2149 * mandatory locks (share modes) can prohibit a write and
2150 * byte-range POSIX locks can prohibit a write if they overlap.
2152 * N.B. this function is only ever called
2153 * from knfsd and ownership of locks is never checked.
2155 int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2157 struct file_lock *fl;
2160 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2162 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2164 } else if (IS_FLOCK(fl)) {
2165 if (!(fl->fl_type & LOCK_MAND))
2167 if (fl->fl_type & LOCK_WRITE)
2178 EXPORT_SYMBOL(lock_may_write);
2180 static inline void __steal_locks(struct file *file, fl_owner_t from)
2182 struct inode *inode = file->f_dentry->d_inode;
2183 struct file_lock *fl = inode->i_flock;
2186 if (fl->fl_file == file && fl->fl_owner == from)
2187 fl->fl_owner = current->files;
2192 /* When getting ready for executing a binary, we make sure that current
2193 * has a files_struct on its own. Before dropping the old files_struct,
2194 * we take over ownership of all locks for all file descriptors we own.
2195 * Note that we may accidentally steal a lock for a file that a sibling
2196 * has created since the unshare_files() call.
2198 void steal_locks(fl_owner_t from)
2200 struct files_struct *files = current->files;
2211 if (i >= files->max_fdset || i >= files->max_fds)
2213 set = files->open_fds->fds_bits[j++];
2216 struct file *file = files->fd[i];
2218 __steal_locks(file, from);
2226 EXPORT_SYMBOL(steal_locks);
2228 static int __init filelock_init(void)
2230 filelock_cache = kmem_cache_create("file_lock_cache",
2231 sizeof(struct file_lock), 0, SLAB_PANIC,
2236 core_initcall(filelock_init);