locks: add new fcntl cmd values for handling file private locks
[GitHub/LineageOS/android_kernel_samsung_universal7580.git] / fs / locks.c
1 /*
2 * linux/fs/locks.c
3 *
4 * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls.
5 * Doug Evans (dje@spiff.uucp), August 07, 1992
6 *
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.
11 *
12 * Miscellaneous edits, and a total rewrite of posix_lock_file() code.
13 * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994
14 *
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
18 *
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
26 *
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
30 *
31 * Implemented two lock personalities - FL_FLOCK and FL_POSIX.
32 *
33 * FL_POSIX locks are created with calls to fcntl() and lockf() through the
34 * fcntl() system call. They have the semantics described above.
35 *
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.
39 *
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
44 * unlocked).
45 *
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
51 *
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.
55 *
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.
60 *
61 * Initial implementation of mandatory locks. SunOS turned out to be
62 * a rotten model, so I implemented the "obvious" semantics.
63 * See 'Documentation/filesystems/mandatory-locking.txt' for details.
64 * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996.
65 *
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
69 * Manual, Section 2.
70 * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996.
71 *
72 * Tidied up block list handling. Added '/proc/locks' interface.
73 * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996.
74 *
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.
78 *
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.
84 *
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.
88 *
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.
93 *
94 * Made the block list a circular list to minimise searching in the list.
95 * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996.
96 *
97 * Made mandatory locking a mount option. Default is not to allow mandatory
98 * locking.
99 * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996.
100 *
101 * Some adaptations for NFS support.
102 * Olaf Kirch (okir@monad.swb.de), Dec 1996,
103 *
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.
106 *
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.
111 *
112 * Leases and LOCK_MAND
113 * Matthew Wilcox <willy@debian.org>, June, 2000.
114 * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000.
115 */
116
117 #include <linux/capability.h>
118 #include <linux/file.h>
119 #include <linux/fdtable.h>
120 #include <linux/fs.h>
121 #include <linux/init.h>
122 #include <linux/module.h>
123 #include <linux/security.h>
124 #include <linux/slab.h>
125 #include <linux/syscalls.h>
126 #include <linux/time.h>
127 #include <linux/rcupdate.h>
128 #include <linux/pid_namespace.h>
129
130 #include <asm/uaccess.h>
131
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)
135 #define IS_FILE_PVT(fl) (fl->fl_flags & FL_FILE_PVT)
136
137 static bool lease_breaking(struct file_lock *fl)
138 {
139 return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING);
140 }
141
142 static int target_leasetype(struct file_lock *fl)
143 {
144 if (fl->fl_flags & FL_UNLOCK_PENDING)
145 return F_UNLCK;
146 if (fl->fl_flags & FL_DOWNGRADE_PENDING)
147 return F_RDLCK;
148 return fl->fl_type;
149 }
150
151 int leases_enable = 1;
152 int lease_break_time = 45;
153
154 #define for_each_lock(inode, lockp) \
155 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
156
157 static LIST_HEAD(file_lock_list);
158 static LIST_HEAD(blocked_list);
159 static DEFINE_SPINLOCK(file_lock_lock);
160
161 /*
162 * Protects the two list heads above, plus the inode->i_flock list
163 */
164 void lock_flocks(void)
165 {
166 spin_lock(&file_lock_lock);
167 }
168 EXPORT_SYMBOL_GPL(lock_flocks);
169
170 void unlock_flocks(void)
171 {
172 spin_unlock(&file_lock_lock);
173 }
174 EXPORT_SYMBOL_GPL(unlock_flocks);
175
176 static struct kmem_cache *filelock_cache __read_mostly;
177
178 static void locks_init_lock_heads(struct file_lock *fl)
179 {
180 INIT_LIST_HEAD(&fl->fl_link);
181 INIT_LIST_HEAD(&fl->fl_block);
182 init_waitqueue_head(&fl->fl_wait);
183 }
184
185 /* Allocate an empty lock structure. */
186 struct file_lock *locks_alloc_lock(void)
187 {
188 struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL);
189
190 if (fl)
191 locks_init_lock_heads(fl);
192
193 return fl;
194 }
195 EXPORT_SYMBOL_GPL(locks_alloc_lock);
196
197 void locks_release_private(struct file_lock *fl)
198 {
199 if (fl->fl_ops) {
200 if (fl->fl_ops->fl_release_private)
201 fl->fl_ops->fl_release_private(fl);
202 fl->fl_ops = NULL;
203 }
204 fl->fl_lmops = NULL;
205
206 }
207 EXPORT_SYMBOL_GPL(locks_release_private);
208
209 /* Free a lock which is not in use. */
210 void locks_free_lock(struct file_lock *fl)
211 {
212 BUG_ON(waitqueue_active(&fl->fl_wait));
213 BUG_ON(!list_empty(&fl->fl_block));
214 BUG_ON(!list_empty(&fl->fl_link));
215
216 locks_release_private(fl);
217 kmem_cache_free(filelock_cache, fl);
218 }
219 EXPORT_SYMBOL(locks_free_lock);
220
221 void locks_init_lock(struct file_lock *fl)
222 {
223 memset(fl, 0, sizeof(struct file_lock));
224 locks_init_lock_heads(fl);
225 }
226
227 EXPORT_SYMBOL(locks_init_lock);
228
229 static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
230 {
231 if (fl->fl_ops) {
232 if (fl->fl_ops->fl_copy_lock)
233 fl->fl_ops->fl_copy_lock(new, fl);
234 new->fl_ops = fl->fl_ops;
235 }
236 if (fl->fl_lmops)
237 new->fl_lmops = fl->fl_lmops;
238 }
239
240 /*
241 * Initialize a new lock from an existing file_lock structure.
242 */
243 void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
244 {
245 new->fl_owner = fl->fl_owner;
246 new->fl_pid = fl->fl_pid;
247 new->fl_file = NULL;
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;
252 new->fl_ops = NULL;
253 new->fl_lmops = NULL;
254 }
255 EXPORT_SYMBOL(__locks_copy_lock);
256
257 void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
258 {
259 locks_release_private(new);
260
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;
265
266 locks_copy_private(new, fl);
267 }
268
269 EXPORT_SYMBOL(locks_copy_lock);
270
271 static inline int flock_translate_cmd(int cmd) {
272 if (cmd & LOCK_MAND)
273 return cmd & (LOCK_MAND | LOCK_RW);
274 switch (cmd) {
275 case LOCK_SH:
276 return F_RDLCK;
277 case LOCK_EX:
278 return F_WRLCK;
279 case LOCK_UN:
280 return F_UNLCK;
281 }
282 return -EINVAL;
283 }
284
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,
287 unsigned int cmd)
288 {
289 struct file_lock *fl;
290 int type = flock_translate_cmd(cmd);
291 if (type < 0)
292 return type;
293
294 fl = locks_alloc_lock();
295 if (fl == NULL)
296 return -ENOMEM;
297
298 fl->fl_file = filp;
299 fl->fl_pid = current->tgid;
300 fl->fl_flags = FL_FLOCK;
301 fl->fl_type = type;
302 fl->fl_end = OFFSET_MAX;
303
304 *lock = fl;
305 return 0;
306 }
307
308 static int assign_type(struct file_lock *fl, long type)
309 {
310 switch (type) {
311 case F_RDLCK:
312 case F_WRLCK:
313 case F_UNLCK:
314 fl->fl_type = type;
315 break;
316 default:
317 return -EINVAL;
318 }
319 return 0;
320 }
321
322 /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
323 * style lock.
324 */
325 static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
326 struct flock *l)
327 {
328 off_t start, end;
329
330 switch (l->l_whence) {
331 case SEEK_SET:
332 start = 0;
333 break;
334 case SEEK_CUR:
335 start = filp->f_pos;
336 break;
337 case SEEK_END:
338 start = i_size_read(file_inode(filp));
339 break;
340 default:
341 return -EINVAL;
342 }
343
344 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
345 POSIX-2001 defines it. */
346 start += l->l_start;
347 if (start < 0)
348 return -EINVAL;
349 fl->fl_end = OFFSET_MAX;
350 if (l->l_len > 0) {
351 end = start + l->l_len - 1;
352 fl->fl_end = end;
353 } else if (l->l_len < 0) {
354 end = start - 1;
355 fl->fl_end = end;
356 start += l->l_len;
357 if (start < 0)
358 return -EINVAL;
359 }
360 fl->fl_start = start; /* we record the absolute position */
361 if (fl->fl_end < fl->fl_start)
362 return -EOVERFLOW;
363
364 fl->fl_owner = current->files;
365 fl->fl_pid = current->tgid;
366 fl->fl_file = filp;
367 fl->fl_flags = FL_POSIX;
368 fl->fl_ops = NULL;
369 fl->fl_lmops = NULL;
370
371 return assign_type(fl, l->l_type);
372 }
373
374 #if BITS_PER_LONG == 32
375 static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
376 struct flock64 *l)
377 {
378 loff_t start;
379
380 switch (l->l_whence) {
381 case SEEK_SET:
382 start = 0;
383 break;
384 case SEEK_CUR:
385 start = filp->f_pos;
386 break;
387 case SEEK_END:
388 start = i_size_read(file_inode(filp));
389 break;
390 default:
391 return -EINVAL;
392 }
393
394 start += l->l_start;
395 if (start < 0)
396 return -EINVAL;
397 fl->fl_end = OFFSET_MAX;
398 if (l->l_len > 0) {
399 fl->fl_end = start + l->l_len - 1;
400 } else if (l->l_len < 0) {
401 fl->fl_end = start - 1;
402 start += l->l_len;
403 if (start < 0)
404 return -EINVAL;
405 }
406 fl->fl_start = start; /* we record the absolute position */
407 if (fl->fl_end < fl->fl_start)
408 return -EOVERFLOW;
409
410 fl->fl_owner = current->files;
411 fl->fl_pid = current->tgid;
412 fl->fl_file = filp;
413 fl->fl_flags = FL_POSIX;
414 fl->fl_ops = NULL;
415 fl->fl_lmops = NULL;
416
417 return assign_type(fl, l->l_type);
418 }
419 #endif
420
421 /* default lease lock manager operations */
422 static void lease_break_callback(struct file_lock *fl)
423 {
424 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
425 }
426
427 static const struct lock_manager_operations lease_manager_ops = {
428 .lm_break = lease_break_callback,
429 .lm_change = lease_modify,
430 };
431
432 /*
433 * Initialize a lease, use the default lock manager operations
434 */
435 static int lease_init(struct file *filp, long type, struct file_lock *fl)
436 {
437 if (assign_type(fl, type) != 0)
438 return -EINVAL;
439
440 fl->fl_owner = current->files;
441 fl->fl_pid = current->tgid;
442
443 fl->fl_file = filp;
444 fl->fl_flags = FL_LEASE;
445 fl->fl_start = 0;
446 fl->fl_end = OFFSET_MAX;
447 fl->fl_ops = NULL;
448 fl->fl_lmops = &lease_manager_ops;
449 return 0;
450 }
451
452 /* Allocate a file_lock initialised to this type of lease */
453 static struct file_lock *lease_alloc(struct file *filp, long type)
454 {
455 struct file_lock *fl = locks_alloc_lock();
456 int error = -ENOMEM;
457
458 if (fl == NULL)
459 return ERR_PTR(error);
460
461 error = lease_init(filp, type, fl);
462 if (error) {
463 locks_free_lock(fl);
464 return ERR_PTR(error);
465 }
466 return fl;
467 }
468
469 /* Check if two locks overlap each other.
470 */
471 static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
472 {
473 return ((fl1->fl_end >= fl2->fl_start) &&
474 (fl2->fl_end >= fl1->fl_start));
475 }
476
477 /*
478 * Check whether two locks have the same owner.
479 */
480 static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
481 {
482 if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner)
483 return fl2->fl_lmops == fl1->fl_lmops &&
484 fl1->fl_lmops->lm_compare_owner(fl1, fl2);
485 return fl1->fl_owner == fl2->fl_owner;
486 }
487
488 /* Remove waiter from blocker's block list.
489 * When blocker ends up pointing to itself then the list is empty.
490 */
491 static void __locks_delete_block(struct file_lock *waiter)
492 {
493 list_del_init(&waiter->fl_block);
494 list_del_init(&waiter->fl_link);
495 waiter->fl_next = NULL;
496 }
497
498 /*
499 */
500 void locks_delete_block(struct file_lock *waiter)
501 {
502 lock_flocks();
503 __locks_delete_block(waiter);
504 unlock_flocks();
505 }
506 EXPORT_SYMBOL(locks_delete_block);
507
508 /* Insert waiter into blocker's block list.
509 * We use a circular list so that processes can be easily woken up in
510 * the order they blocked. The documentation doesn't require this but
511 * it seems like the reasonable thing to do.
512 */
513 static void locks_insert_block(struct file_lock *blocker,
514 struct file_lock *waiter)
515 {
516 BUG_ON(!list_empty(&waiter->fl_block));
517 list_add_tail(&waiter->fl_block, &blocker->fl_block);
518 waiter->fl_next = blocker;
519 if (IS_POSIX(blocker))
520 list_add(&waiter->fl_link, &blocked_list);
521 }
522
523 /* Wake up processes blocked waiting for blocker.
524 * If told to wait then schedule the processes until the block list
525 * is empty, otherwise empty the block list ourselves.
526 */
527 static void locks_wake_up_blocks(struct file_lock *blocker)
528 {
529 while (!list_empty(&blocker->fl_block)) {
530 struct file_lock *waiter;
531
532 waiter = list_first_entry(&blocker->fl_block,
533 struct file_lock, fl_block);
534 __locks_delete_block(waiter);
535 if (waiter->fl_lmops && waiter->fl_lmops->lm_notify)
536 waiter->fl_lmops->lm_notify(waiter);
537 else
538 wake_up(&waiter->fl_wait);
539 }
540 }
541
542 /* Insert file lock fl into an inode's lock list at the position indicated
543 * by pos. At the same time add the lock to the global file lock list.
544 */
545 static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
546 {
547 list_add(&fl->fl_link, &file_lock_list);
548
549 fl->fl_nspid = get_pid(task_tgid(current));
550
551 /* insert into file's list */
552 fl->fl_next = *pos;
553 *pos = fl;
554 }
555
556 /*
557 * Delete a lock and then free it.
558 * Wake up processes that are blocked waiting for this lock,
559 * notify the FS that the lock has been cleared and
560 * finally free the lock.
561 */
562 static void locks_delete_lock(struct file_lock **thisfl_p)
563 {
564 struct file_lock *fl = *thisfl_p;
565
566 *thisfl_p = fl->fl_next;
567 fl->fl_next = NULL;
568 list_del_init(&fl->fl_link);
569
570 if (fl->fl_nspid) {
571 put_pid(fl->fl_nspid);
572 fl->fl_nspid = NULL;
573 }
574
575 locks_wake_up_blocks(fl);
576 locks_free_lock(fl);
577 }
578
579 /* Determine if lock sys_fl blocks lock caller_fl. Common functionality
580 * checks for shared/exclusive status of overlapping locks.
581 */
582 static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
583 {
584 if (sys_fl->fl_type == F_WRLCK)
585 return 1;
586 if (caller_fl->fl_type == F_WRLCK)
587 return 1;
588 return 0;
589 }
590
591 /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
592 * checking before calling the locks_conflict().
593 */
594 static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
595 {
596 /* POSIX locks owned by the same process do not conflict with
597 * each other.
598 */
599 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
600 return (0);
601
602 /* Check whether they overlap */
603 if (!locks_overlap(caller_fl, sys_fl))
604 return 0;
605
606 return (locks_conflict(caller_fl, sys_fl));
607 }
608
609 /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
610 * checking before calling the locks_conflict().
611 */
612 static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
613 {
614 /* FLOCK locks referring to the same filp do not conflict with
615 * each other.
616 */
617 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
618 return (0);
619 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
620 return 0;
621
622 return (locks_conflict(caller_fl, sys_fl));
623 }
624
625 void
626 posix_test_lock(struct file *filp, struct file_lock *fl)
627 {
628 struct file_lock *cfl;
629
630 lock_flocks();
631 for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) {
632 if (!IS_POSIX(cfl))
633 continue;
634 if (posix_locks_conflict(fl, cfl))
635 break;
636 }
637 if (cfl) {
638 __locks_copy_lock(fl, cfl);
639 if (cfl->fl_nspid)
640 fl->fl_pid = pid_vnr(cfl->fl_nspid);
641 } else
642 fl->fl_type = F_UNLCK;
643 unlock_flocks();
644 return;
645 }
646 EXPORT_SYMBOL(posix_test_lock);
647
648 /*
649 * Deadlock detection:
650 *
651 * We attempt to detect deadlocks that are due purely to posix file
652 * locks.
653 *
654 * We assume that a task can be waiting for at most one lock at a time.
655 * So for any acquired lock, the process holding that lock may be
656 * waiting on at most one other lock. That lock in turns may be held by
657 * someone waiting for at most one other lock. Given a requested lock
658 * caller_fl which is about to wait for a conflicting lock block_fl, we
659 * follow this chain of waiters to ensure we are not about to create a
660 * cycle.
661 *
662 * Since we do this before we ever put a process to sleep on a lock, we
663 * are ensured that there is never a cycle; that is what guarantees that
664 * the while() loop in posix_locks_deadlock() eventually completes.
665 *
666 * Note: the above assumption may not be true when handling lock
667 * requests from a broken NFS client. It may also fail in the presence
668 * of tasks (such as posix threads) sharing the same open file table.
669 *
670 * To handle those cases, we just bail out after a few iterations.
671 */
672
673 #define MAX_DEADLK_ITERATIONS 10
674
675 /* Find a lock that the owner of the given block_fl is blocking on. */
676 static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
677 {
678 struct file_lock *fl;
679
680 list_for_each_entry(fl, &blocked_list, fl_link) {
681 if (posix_same_owner(fl, block_fl))
682 return fl->fl_next;
683 }
684 return NULL;
685 }
686
687 static int posix_locks_deadlock(struct file_lock *caller_fl,
688 struct file_lock *block_fl)
689 {
690 int i = 0;
691
692 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
693 if (i++ > MAX_DEADLK_ITERATIONS)
694 return 0;
695 if (posix_same_owner(caller_fl, block_fl))
696 return 1;
697 }
698 return 0;
699 }
700
701 /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
702 * after any leases, but before any posix locks.
703 *
704 * Note that if called with an FL_EXISTS argument, the caller may determine
705 * whether or not a lock was successfully freed by testing the return
706 * value for -ENOENT.
707 */
708 static int flock_lock_file(struct file *filp, struct file_lock *request)
709 {
710 struct file_lock *new_fl = NULL;
711 struct file_lock **before;
712 struct inode * inode = file_inode(filp);
713 int error = 0;
714 int found = 0;
715
716 if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) {
717 new_fl = locks_alloc_lock();
718 if (!new_fl)
719 return -ENOMEM;
720 }
721
722 lock_flocks();
723 if (request->fl_flags & FL_ACCESS)
724 goto find_conflict;
725
726 for_each_lock(inode, before) {
727 struct file_lock *fl = *before;
728 if (IS_POSIX(fl))
729 break;
730 if (IS_LEASE(fl))
731 continue;
732 if (filp != fl->fl_file)
733 continue;
734 if (request->fl_type == fl->fl_type)
735 goto out;
736 found = 1;
737 locks_delete_lock(before);
738 break;
739 }
740
741 if (request->fl_type == F_UNLCK) {
742 if ((request->fl_flags & FL_EXISTS) && !found)
743 error = -ENOENT;
744 goto out;
745 }
746
747 /*
748 * If a higher-priority process was blocked on the old file lock,
749 * give it the opportunity to lock the file.
750 */
751 if (found) {
752 unlock_flocks();
753 cond_resched();
754 lock_flocks();
755 }
756
757 find_conflict:
758 for_each_lock(inode, before) {
759 struct file_lock *fl = *before;
760 if (IS_POSIX(fl))
761 break;
762 if (IS_LEASE(fl))
763 continue;
764 if (!flock_locks_conflict(request, fl))
765 continue;
766 error = -EAGAIN;
767 if (!(request->fl_flags & FL_SLEEP))
768 goto out;
769 error = FILE_LOCK_DEFERRED;
770 locks_insert_block(fl, request);
771 goto out;
772 }
773 if (request->fl_flags & FL_ACCESS)
774 goto out;
775 locks_copy_lock(new_fl, request);
776 locks_insert_lock(before, new_fl);
777 new_fl = NULL;
778 error = 0;
779
780 out:
781 unlock_flocks();
782 if (new_fl)
783 locks_free_lock(new_fl);
784 return error;
785 }
786
787 static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
788 {
789 struct file_lock *fl;
790 struct file_lock *new_fl = NULL;
791 struct file_lock *new_fl2 = NULL;
792 struct file_lock *left = NULL;
793 struct file_lock *right = NULL;
794 struct file_lock **before;
795 int error, added = 0;
796
797 /*
798 * We may need two file_lock structures for this operation,
799 * so we get them in advance to avoid races.
800 *
801 * In some cases we can be sure, that no new locks will be needed
802 */
803 if (!(request->fl_flags & FL_ACCESS) &&
804 (request->fl_type != F_UNLCK ||
805 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
806 new_fl = locks_alloc_lock();
807 new_fl2 = locks_alloc_lock();
808 }
809
810 lock_flocks();
811 if (request->fl_type != F_UNLCK) {
812 for_each_lock(inode, before) {
813 fl = *before;
814 if (!IS_POSIX(fl))
815 continue;
816 if (!posix_locks_conflict(request, fl))
817 continue;
818 if (conflock)
819 __locks_copy_lock(conflock, fl);
820 error = -EAGAIN;
821 if (!(request->fl_flags & FL_SLEEP))
822 goto out;
823 error = -EDEADLK;
824 if (posix_locks_deadlock(request, fl))
825 goto out;
826 error = FILE_LOCK_DEFERRED;
827 locks_insert_block(fl, request);
828 goto out;
829 }
830 }
831
832 /* If we're just looking for a conflict, we're done. */
833 error = 0;
834 if (request->fl_flags & FL_ACCESS)
835 goto out;
836
837 /*
838 * Find the first old lock with the same owner as the new lock.
839 */
840
841 before = &inode->i_flock;
842
843 /* First skip locks owned by other processes. */
844 while ((fl = *before) && (!IS_POSIX(fl) ||
845 !posix_same_owner(request, fl))) {
846 before = &fl->fl_next;
847 }
848
849 /* Process locks with this owner. */
850 while ((fl = *before) && posix_same_owner(request, fl)) {
851 /* Detect adjacent or overlapping regions (if same lock type)
852 */
853 if (request->fl_type == fl->fl_type) {
854 /* In all comparisons of start vs end, use
855 * "start - 1" rather than "end + 1". If end
856 * is OFFSET_MAX, end + 1 will become negative.
857 */
858 if (fl->fl_end < request->fl_start - 1)
859 goto next_lock;
860 /* If the next lock in the list has entirely bigger
861 * addresses than the new one, insert the lock here.
862 */
863 if (fl->fl_start - 1 > request->fl_end)
864 break;
865
866 /* If we come here, the new and old lock are of the
867 * same type and adjacent or overlapping. Make one
868 * lock yielding from the lower start address of both
869 * locks to the higher end address.
870 */
871 if (fl->fl_start > request->fl_start)
872 fl->fl_start = request->fl_start;
873 else
874 request->fl_start = fl->fl_start;
875 if (fl->fl_end < request->fl_end)
876 fl->fl_end = request->fl_end;
877 else
878 request->fl_end = fl->fl_end;
879 if (added) {
880 locks_delete_lock(before);
881 continue;
882 }
883 request = fl;
884 added = 1;
885 }
886 else {
887 /* Processing for different lock types is a bit
888 * more complex.
889 */
890 if (fl->fl_end < request->fl_start)
891 goto next_lock;
892 if (fl->fl_start > request->fl_end)
893 break;
894 if (request->fl_type == F_UNLCK)
895 added = 1;
896 if (fl->fl_start < request->fl_start)
897 left = fl;
898 /* If the next lock in the list has a higher end
899 * address than the new one, insert the new one here.
900 */
901 if (fl->fl_end > request->fl_end) {
902 right = fl;
903 break;
904 }
905 if (fl->fl_start >= request->fl_start) {
906 /* The new lock completely replaces an old
907 * one (This may happen several times).
908 */
909 if (added) {
910 locks_delete_lock(before);
911 continue;
912 }
913 /* Replace the old lock with the new one.
914 * Wake up anybody waiting for the old one,
915 * as the change in lock type might satisfy
916 * their needs.
917 */
918 locks_wake_up_blocks(fl);
919 fl->fl_start = request->fl_start;
920 fl->fl_end = request->fl_end;
921 fl->fl_type = request->fl_type;
922 locks_release_private(fl);
923 locks_copy_private(fl, request);
924 request = fl;
925 added = 1;
926 }
927 }
928 /* Go on to next lock.
929 */
930 next_lock:
931 before = &fl->fl_next;
932 }
933
934 /*
935 * The above code only modifies existing locks in case of
936 * merging or replacing. If new lock(s) need to be inserted
937 * all modifications are done bellow this, so it's safe yet to
938 * bail out.
939 */
940 error = -ENOLCK; /* "no luck" */
941 if (right && left == right && !new_fl2)
942 goto out;
943
944 error = 0;
945 if (!added) {
946 if (request->fl_type == F_UNLCK) {
947 if (request->fl_flags & FL_EXISTS)
948 error = -ENOENT;
949 goto out;
950 }
951
952 if (!new_fl) {
953 error = -ENOLCK;
954 goto out;
955 }
956 locks_copy_lock(new_fl, request);
957 locks_insert_lock(before, new_fl);
958 new_fl = NULL;
959 }
960 if (right) {
961 if (left == right) {
962 /* The new lock breaks the old one in two pieces,
963 * so we have to use the second new lock.
964 */
965 left = new_fl2;
966 new_fl2 = NULL;
967 locks_copy_lock(left, right);
968 locks_insert_lock(before, left);
969 }
970 right->fl_start = request->fl_end + 1;
971 locks_wake_up_blocks(right);
972 }
973 if (left) {
974 left->fl_end = request->fl_start - 1;
975 locks_wake_up_blocks(left);
976 }
977 out:
978 unlock_flocks();
979 /*
980 * Free any unused locks.
981 */
982 if (new_fl)
983 locks_free_lock(new_fl);
984 if (new_fl2)
985 locks_free_lock(new_fl2);
986 return error;
987 }
988
989 /**
990 * posix_lock_file - Apply a POSIX-style lock to a file
991 * @filp: The file to apply the lock to
992 * @fl: The lock to be applied
993 * @conflock: Place to return a copy of the conflicting lock, if found.
994 *
995 * Add a POSIX style lock to a file.
996 * We merge adjacent & overlapping locks whenever possible.
997 * POSIX locks are sorted by owner task, then by starting address
998 *
999 * Note that if called with an FL_EXISTS argument, the caller may determine
1000 * whether or not a lock was successfully freed by testing the return
1001 * value for -ENOENT.
1002 */
1003 int posix_lock_file(struct file *filp, struct file_lock *fl,
1004 struct file_lock *conflock)
1005 {
1006 return __posix_lock_file(file_inode(filp), fl, conflock);
1007 }
1008 EXPORT_SYMBOL(posix_lock_file);
1009
1010 /**
1011 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1012 * @filp: The file to apply the lock to
1013 * @fl: The lock to be applied
1014 *
1015 * Add a POSIX style lock to a file.
1016 * We merge adjacent & overlapping locks whenever possible.
1017 * POSIX locks are sorted by owner task, then by starting address
1018 */
1019 int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1020 {
1021 int error;
1022 might_sleep ();
1023 for (;;) {
1024 error = posix_lock_file(filp, fl, NULL);
1025 if (error != FILE_LOCK_DEFERRED)
1026 break;
1027 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1028 if (!error)
1029 continue;
1030
1031 locks_delete_block(fl);
1032 break;
1033 }
1034 return error;
1035 }
1036 EXPORT_SYMBOL(posix_lock_file_wait);
1037
1038 /**
1039 * locks_mandatory_locked - Check for an active lock
1040 * @inode: the file to check
1041 *
1042 * Searches the inode's list of locks to find any POSIX locks which conflict.
1043 * This function is called from locks_verify_locked() only.
1044 */
1045 int locks_mandatory_locked(struct inode *inode)
1046 {
1047 fl_owner_t owner = current->files;
1048 struct file_lock *fl;
1049
1050 /*
1051 * Search the lock list for this inode for any POSIX locks.
1052 */
1053 lock_flocks();
1054 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1055 if (!IS_POSIX(fl))
1056 continue;
1057 if (fl->fl_owner != owner)
1058 break;
1059 }
1060 unlock_flocks();
1061 return fl ? -EAGAIN : 0;
1062 }
1063
1064 /**
1065 * locks_mandatory_area - Check for a conflicting lock
1066 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1067 * for shared
1068 * @inode: the file to check
1069 * @filp: how the file was opened (if it was)
1070 * @offset: start of area to check
1071 * @count: length of area to check
1072 *
1073 * Searches the inode's list of locks to find any POSIX locks which conflict.
1074 * This function is called from rw_verify_area() and
1075 * locks_verify_truncate().
1076 */
1077 int locks_mandatory_area(int read_write, struct inode *inode,
1078 struct file *filp, loff_t offset,
1079 size_t count)
1080 {
1081 struct file_lock fl;
1082 int error;
1083
1084 locks_init_lock(&fl);
1085 fl.fl_owner = current->files;
1086 fl.fl_pid = current->tgid;
1087 fl.fl_file = filp;
1088 fl.fl_flags = FL_POSIX | FL_ACCESS;
1089 if (filp && !(filp->f_flags & O_NONBLOCK))
1090 fl.fl_flags |= FL_SLEEP;
1091 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1092 fl.fl_start = offset;
1093 fl.fl_end = offset + count - 1;
1094
1095 for (;;) {
1096 error = __posix_lock_file(inode, &fl, NULL);
1097 if (error != FILE_LOCK_DEFERRED)
1098 break;
1099 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1100 if (!error) {
1101 /*
1102 * If we've been sleeping someone might have
1103 * changed the permissions behind our back.
1104 */
1105 if (__mandatory_lock(inode))
1106 continue;
1107 }
1108
1109 locks_delete_block(&fl);
1110 break;
1111 }
1112
1113 return error;
1114 }
1115
1116 EXPORT_SYMBOL(locks_mandatory_area);
1117
1118 static void lease_clear_pending(struct file_lock *fl, int arg)
1119 {
1120 switch (arg) {
1121 case F_UNLCK:
1122 fl->fl_flags &= ~FL_UNLOCK_PENDING;
1123 /* fall through: */
1124 case F_RDLCK:
1125 fl->fl_flags &= ~FL_DOWNGRADE_PENDING;
1126 }
1127 }
1128
1129 /* We already had a lease on this file; just change its type */
1130 int lease_modify(struct file_lock **before, int arg)
1131 {
1132 struct file_lock *fl = *before;
1133 int error = assign_type(fl, arg);
1134
1135 if (error)
1136 return error;
1137 lease_clear_pending(fl, arg);
1138 locks_wake_up_blocks(fl);
1139 if (arg == F_UNLCK) {
1140 struct file *filp = fl->fl_file;
1141
1142 f_delown(filp);
1143 filp->f_owner.signum = 0;
1144 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
1145 if (fl->fl_fasync != NULL) {
1146 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
1147 fl->fl_fasync = NULL;
1148 }
1149 locks_delete_lock(before);
1150 }
1151 return 0;
1152 }
1153
1154 EXPORT_SYMBOL(lease_modify);
1155
1156 static bool past_time(unsigned long then)
1157 {
1158 if (!then)
1159 /* 0 is a special value meaning "this never expires": */
1160 return false;
1161 return time_after(jiffies, then);
1162 }
1163
1164 static void time_out_leases(struct inode *inode)
1165 {
1166 struct file_lock **before;
1167 struct file_lock *fl;
1168
1169 before = &inode->i_flock;
1170 while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) {
1171 if (past_time(fl->fl_downgrade_time))
1172 lease_modify(before, F_RDLCK);
1173 if (past_time(fl->fl_break_time))
1174 lease_modify(before, F_UNLCK);
1175 if (fl == *before) /* lease_modify may have freed fl */
1176 before = &fl->fl_next;
1177 }
1178 }
1179
1180 /**
1181 * __break_lease - revoke all outstanding leases on file
1182 * @inode: the inode of the file to return
1183 * @mode: the open mode (read or write)
1184 *
1185 * break_lease (inlined for speed) has checked there already is at least
1186 * some kind of lock (maybe a lease) on this file. Leases are broken on
1187 * a call to open() or truncate(). This function can sleep unless you
1188 * specified %O_NONBLOCK to your open().
1189 */
1190 int __break_lease(struct inode *inode, unsigned int mode)
1191 {
1192 int error = 0;
1193 struct file_lock *new_fl, *flock;
1194 struct file_lock *fl;
1195 unsigned long break_time;
1196 int i_have_this_lease = 0;
1197 int want_write = (mode & O_ACCMODE) != O_RDONLY;
1198
1199 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
1200 if (IS_ERR(new_fl))
1201 return PTR_ERR(new_fl);
1202
1203 lock_flocks();
1204
1205 time_out_leases(inode);
1206
1207 flock = inode->i_flock;
1208 if ((flock == NULL) || !IS_LEASE(flock))
1209 goto out;
1210
1211 if (!locks_conflict(flock, new_fl))
1212 goto out;
1213
1214 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1215 if (fl->fl_owner == current->files)
1216 i_have_this_lease = 1;
1217
1218 break_time = 0;
1219 if (lease_break_time > 0) {
1220 break_time = jiffies + lease_break_time * HZ;
1221 if (break_time == 0)
1222 break_time++; /* so that 0 means no break time */
1223 }
1224
1225 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1226 if (want_write) {
1227 if (fl->fl_flags & FL_UNLOCK_PENDING)
1228 continue;
1229 fl->fl_flags |= FL_UNLOCK_PENDING;
1230 fl->fl_break_time = break_time;
1231 } else {
1232 if (lease_breaking(flock))
1233 continue;
1234 fl->fl_flags |= FL_DOWNGRADE_PENDING;
1235 fl->fl_downgrade_time = break_time;
1236 }
1237 fl->fl_lmops->lm_break(fl);
1238 }
1239
1240 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1241 error = -EWOULDBLOCK;
1242 goto out;
1243 }
1244
1245 restart:
1246 break_time = flock->fl_break_time;
1247 if (break_time != 0)
1248 break_time -= jiffies;
1249 if (break_time == 0)
1250 break_time++;
1251 locks_insert_block(flock, new_fl);
1252 unlock_flocks();
1253 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1254 !new_fl->fl_next, break_time);
1255 lock_flocks();
1256 __locks_delete_block(new_fl);
1257 if (error >= 0) {
1258 if (error == 0)
1259 time_out_leases(inode);
1260 /*
1261 * Wait for the next conflicting lease that has not been
1262 * broken yet
1263 */
1264 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1265 flock = flock->fl_next) {
1266 if (locks_conflict(new_fl, flock))
1267 goto restart;
1268 }
1269 error = 0;
1270 }
1271
1272 out:
1273 unlock_flocks();
1274 locks_free_lock(new_fl);
1275 return error;
1276 }
1277
1278 EXPORT_SYMBOL(__break_lease);
1279
1280 /**
1281 * lease_get_mtime - get the last modified time of an inode
1282 * @inode: the inode
1283 * @time: pointer to a timespec which will contain the last modified time
1284 *
1285 * This is to force NFS clients to flush their caches for files with
1286 * exclusive leases. The justification is that if someone has an
1287 * exclusive lease, then they could be modifying it.
1288 */
1289 void lease_get_mtime(struct inode *inode, struct timespec *time)
1290 {
1291 struct file_lock *flock = inode->i_flock;
1292 if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK))
1293 *time = current_fs_time(inode->i_sb);
1294 else
1295 *time = inode->i_mtime;
1296 }
1297
1298 EXPORT_SYMBOL(lease_get_mtime);
1299
1300 /**
1301 * fcntl_getlease - Enquire what lease is currently active
1302 * @filp: the file
1303 *
1304 * The value returned by this function will be one of
1305 * (if no lease break is pending):
1306 *
1307 * %F_RDLCK to indicate a shared lease is held.
1308 *
1309 * %F_WRLCK to indicate an exclusive lease is held.
1310 *
1311 * %F_UNLCK to indicate no lease is held.
1312 *
1313 * (if a lease break is pending):
1314 *
1315 * %F_RDLCK to indicate an exclusive lease needs to be
1316 * changed to a shared lease (or removed).
1317 *
1318 * %F_UNLCK to indicate the lease needs to be removed.
1319 *
1320 * XXX: sfr & willy disagree over whether F_INPROGRESS
1321 * should be returned to userspace.
1322 */
1323 int fcntl_getlease(struct file *filp)
1324 {
1325 struct file_lock *fl;
1326 int type = F_UNLCK;
1327
1328 lock_flocks();
1329 time_out_leases(file_inode(filp));
1330 for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl);
1331 fl = fl->fl_next) {
1332 if (fl->fl_file == filp) {
1333 type = target_leasetype(fl);
1334 break;
1335 }
1336 }
1337 unlock_flocks();
1338 return type;
1339 }
1340
1341 int generic_add_lease(struct file *filp, long arg, struct file_lock **flp)
1342 {
1343 struct file_lock *fl, **before, **my_before = NULL, *lease;
1344 struct dentry *dentry = filp->f_path.dentry;
1345 struct inode *inode = dentry->d_inode;
1346 int error;
1347
1348 lease = *flp;
1349
1350 error = -EAGAIN;
1351 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1352 goto out;
1353 if ((arg == F_WRLCK)
1354 && ((dentry->d_count > 1)
1355 || (atomic_read(&inode->i_count) > 1)))
1356 goto out;
1357
1358 /*
1359 * At this point, we know that if there is an exclusive
1360 * lease on this file, then we hold it on this filp
1361 * (otherwise our open of this file would have blocked).
1362 * And if we are trying to acquire an exclusive lease,
1363 * then the file is not open by anyone (including us)
1364 * except for this filp.
1365 */
1366 error = -EAGAIN;
1367 for (before = &inode->i_flock;
1368 ((fl = *before) != NULL) && IS_LEASE(fl);
1369 before = &fl->fl_next) {
1370 if (fl->fl_file == filp) {
1371 my_before = before;
1372 continue;
1373 }
1374 /*
1375 * No exclusive leases if someone else has a lease on
1376 * this file:
1377 */
1378 if (arg == F_WRLCK)
1379 goto out;
1380 /*
1381 * Modifying our existing lease is OK, but no getting a
1382 * new lease if someone else is opening for write:
1383 */
1384 if (fl->fl_flags & FL_UNLOCK_PENDING)
1385 goto out;
1386 }
1387
1388 if (my_before != NULL) {
1389 error = lease->fl_lmops->lm_change(my_before, arg);
1390 if (!error)
1391 *flp = *my_before;
1392 goto out;
1393 }
1394
1395 error = -EINVAL;
1396 if (!leases_enable)
1397 goto out;
1398
1399 locks_insert_lock(before, lease);
1400 return 0;
1401
1402 out:
1403 return error;
1404 }
1405
1406 int generic_delete_lease(struct file *filp, struct file_lock **flp)
1407 {
1408 struct file_lock *fl, **before;
1409 struct dentry *dentry = filp->f_path.dentry;
1410 struct inode *inode = dentry->d_inode;
1411
1412 for (before = &inode->i_flock;
1413 ((fl = *before) != NULL) && IS_LEASE(fl);
1414 before = &fl->fl_next) {
1415 if (fl->fl_file != filp)
1416 continue;
1417 return (*flp)->fl_lmops->lm_change(before, F_UNLCK);
1418 }
1419 return -EAGAIN;
1420 }
1421
1422 /**
1423 * generic_setlease - sets a lease on an open file
1424 * @filp: file pointer
1425 * @arg: type of lease to obtain
1426 * @flp: input - file_lock to use, output - file_lock inserted
1427 *
1428 * The (input) flp->fl_lmops->lm_break function is required
1429 * by break_lease().
1430 *
1431 * Called with file_lock_lock held.
1432 */
1433 int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
1434 {
1435 struct dentry *dentry = filp->f_path.dentry;
1436 struct inode *inode = dentry->d_inode;
1437 int error;
1438
1439 if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE))
1440 return -EACCES;
1441 if (!S_ISREG(inode->i_mode))
1442 return -EINVAL;
1443 error = security_file_lock(filp, arg);
1444 if (error)
1445 return error;
1446
1447 time_out_leases(inode);
1448
1449 BUG_ON(!(*flp)->fl_lmops->lm_break);
1450
1451 switch (arg) {
1452 case F_UNLCK:
1453 return generic_delete_lease(filp, flp);
1454 case F_RDLCK:
1455 case F_WRLCK:
1456 return generic_add_lease(filp, arg, flp);
1457 default:
1458 return -EINVAL;
1459 }
1460 }
1461 EXPORT_SYMBOL(generic_setlease);
1462
1463 static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
1464 {
1465 if (filp->f_op && filp->f_op->setlease)
1466 return filp->f_op->setlease(filp, arg, lease);
1467 else
1468 return generic_setlease(filp, arg, lease);
1469 }
1470
1471 /**
1472 * vfs_setlease - sets a lease on an open file
1473 * @filp: file pointer
1474 * @arg: type of lease to obtain
1475 * @lease: file_lock to use
1476 *
1477 * Call this to establish a lease on the file.
1478 * The (*lease)->fl_lmops->lm_break operation must be set; if not,
1479 * break_lease will oops!
1480 *
1481 * This will call the filesystem's setlease file method, if
1482 * defined. Note that there is no getlease method; instead, the
1483 * filesystem setlease method should call back to setlease() to
1484 * add a lease to the inode's lease list, where fcntl_getlease() can
1485 * find it. Since fcntl_getlease() only reports whether the current
1486 * task holds a lease, a cluster filesystem need only do this for
1487 * leases held by processes on this node.
1488 *
1489 * There is also no break_lease method; filesystems that
1490 * handle their own leases should break leases themselves from the
1491 * filesystem's open, create, and (on truncate) setattr methods.
1492 *
1493 * Warning: the only current setlease methods exist only to disable
1494 * leases in certain cases. More vfs changes may be required to
1495 * allow a full filesystem lease implementation.
1496 */
1497
1498 int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
1499 {
1500 int error;
1501
1502 lock_flocks();
1503 error = __vfs_setlease(filp, arg, lease);
1504 unlock_flocks();
1505
1506 return error;
1507 }
1508 EXPORT_SYMBOL_GPL(vfs_setlease);
1509
1510 static int do_fcntl_delete_lease(struct file *filp)
1511 {
1512 struct file_lock fl, *flp = &fl;
1513
1514 lease_init(filp, F_UNLCK, flp);
1515
1516 return vfs_setlease(filp, F_UNLCK, &flp);
1517 }
1518
1519 static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg)
1520 {
1521 struct file_lock *fl, *ret;
1522 struct fasync_struct *new;
1523 int error;
1524
1525 fl = lease_alloc(filp, arg);
1526 if (IS_ERR(fl))
1527 return PTR_ERR(fl);
1528
1529 new = fasync_alloc();
1530 if (!new) {
1531 locks_free_lock(fl);
1532 return -ENOMEM;
1533 }
1534 ret = fl;
1535 lock_flocks();
1536 error = __vfs_setlease(filp, arg, &ret);
1537 if (error) {
1538 unlock_flocks();
1539 locks_free_lock(fl);
1540 goto out_free_fasync;
1541 }
1542 if (ret != fl)
1543 locks_free_lock(fl);
1544
1545 /*
1546 * fasync_insert_entry() returns the old entry if any.
1547 * If there was no old entry, then it used 'new' and
1548 * inserted it into the fasync list. Clear new so that
1549 * we don't release it here.
1550 */
1551 if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new))
1552 new = NULL;
1553
1554 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
1555 unlock_flocks();
1556
1557 out_free_fasync:
1558 if (new)
1559 fasync_free(new);
1560 return error;
1561 }
1562
1563 /**
1564 * fcntl_setlease - sets a lease on an open file
1565 * @fd: open file descriptor
1566 * @filp: file pointer
1567 * @arg: type of lease to obtain
1568 *
1569 * Call this fcntl to establish a lease on the file.
1570 * Note that you also need to call %F_SETSIG to
1571 * receive a signal when the lease is broken.
1572 */
1573 int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1574 {
1575 if (arg == F_UNLCK)
1576 return do_fcntl_delete_lease(filp);
1577 return do_fcntl_add_lease(fd, filp, arg);
1578 }
1579
1580 /**
1581 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1582 * @filp: The file to apply the lock to
1583 * @fl: The lock to be applied
1584 *
1585 * Add a FLOCK style lock to a file.
1586 */
1587 int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1588 {
1589 int error;
1590 might_sleep();
1591 for (;;) {
1592 error = flock_lock_file(filp, fl);
1593 if (error != FILE_LOCK_DEFERRED)
1594 break;
1595 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1596 if (!error)
1597 continue;
1598
1599 locks_delete_block(fl);
1600 break;
1601 }
1602 return error;
1603 }
1604
1605 EXPORT_SYMBOL(flock_lock_file_wait);
1606
1607 /**
1608 * sys_flock: - flock() system call.
1609 * @fd: the file descriptor to lock.
1610 * @cmd: the type of lock to apply.
1611 *
1612 * Apply a %FL_FLOCK style lock to an open file descriptor.
1613 * The @cmd can be one of
1614 *
1615 * %LOCK_SH -- a shared lock.
1616 *
1617 * %LOCK_EX -- an exclusive lock.
1618 *
1619 * %LOCK_UN -- remove an existing lock.
1620 *
1621 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1622 *
1623 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1624 * processes read and write access respectively.
1625 */
1626 SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
1627 {
1628 struct fd f = fdget(fd);
1629 struct file_lock *lock;
1630 int can_sleep, unlock;
1631 int error;
1632
1633 error = -EBADF;
1634 if (!f.file)
1635 goto out;
1636
1637 can_sleep = !(cmd & LOCK_NB);
1638 cmd &= ~LOCK_NB;
1639 unlock = (cmd == LOCK_UN);
1640
1641 if (!unlock && !(cmd & LOCK_MAND) &&
1642 !(f.file->f_mode & (FMODE_READ|FMODE_WRITE)))
1643 goto out_putf;
1644
1645 error = flock_make_lock(f.file, &lock, cmd);
1646 if (error)
1647 goto out_putf;
1648 if (can_sleep)
1649 lock->fl_flags |= FL_SLEEP;
1650
1651 error = security_file_lock(f.file, lock->fl_type);
1652 if (error)
1653 goto out_free;
1654
1655 if (f.file->f_op && f.file->f_op->flock)
1656 error = f.file->f_op->flock(f.file,
1657 (can_sleep) ? F_SETLKW : F_SETLK,
1658 lock);
1659 else
1660 error = flock_lock_file_wait(f.file, lock);
1661
1662 out_free:
1663 locks_free_lock(lock);
1664
1665 out_putf:
1666 fdput(f);
1667 out:
1668 return error;
1669 }
1670
1671 /**
1672 * vfs_test_lock - test file byte range lock
1673 * @filp: The file to test lock for
1674 * @fl: The lock to test; also used to hold result
1675 *
1676 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1677 * setting conf->fl_type to something other than F_UNLCK.
1678 */
1679 int vfs_test_lock(struct file *filp, struct file_lock *fl)
1680 {
1681 if (filp->f_op && filp->f_op->lock)
1682 return filp->f_op->lock(filp, F_GETLK, fl);
1683 posix_test_lock(filp, fl);
1684 return 0;
1685 }
1686 EXPORT_SYMBOL_GPL(vfs_test_lock);
1687
1688 static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1689 {
1690 flock->l_pid = IS_FILE_PVT(fl) ? -1 : fl->fl_pid;
1691 #if BITS_PER_LONG == 32
1692 /*
1693 * Make sure we can represent the posix lock via
1694 * legacy 32bit flock.
1695 */
1696 if (fl->fl_start > OFFT_OFFSET_MAX)
1697 return -EOVERFLOW;
1698 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1699 return -EOVERFLOW;
1700 #endif
1701 flock->l_start = fl->fl_start;
1702 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1703 fl->fl_end - fl->fl_start + 1;
1704 flock->l_whence = 0;
1705 flock->l_type = fl->fl_type;
1706 return 0;
1707 }
1708
1709 #if BITS_PER_LONG == 32
1710 static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1711 {
1712 flock->l_pid = IS_FILE_PVT(fl) ? -1 : fl->fl_pid;
1713 flock->l_start = fl->fl_start;
1714 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1715 fl->fl_end - fl->fl_start + 1;
1716 flock->l_whence = 0;
1717 flock->l_type = fl->fl_type;
1718 }
1719 #endif
1720
1721 /* Report the first existing lock that would conflict with l.
1722 * This implements the F_GETLK command of fcntl().
1723 */
1724 int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l)
1725 {
1726 struct file_lock file_lock;
1727 struct flock flock;
1728 int error;
1729
1730 error = -EFAULT;
1731 if (copy_from_user(&flock, l, sizeof(flock)))
1732 goto out;
1733 error = -EINVAL;
1734 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1735 goto out;
1736
1737 error = flock_to_posix_lock(filp, &file_lock, &flock);
1738 if (error)
1739 goto out;
1740
1741 if (cmd == F_GETLKP) {
1742 cmd = F_GETLK;
1743 file_lock.fl_flags |= FL_FILE_PVT;
1744 file_lock.fl_owner = (fl_owner_t)filp;
1745 }
1746
1747 error = vfs_test_lock(filp, &file_lock);
1748 if (error)
1749 goto out;
1750
1751 flock.l_type = file_lock.fl_type;
1752 if (file_lock.fl_type != F_UNLCK) {
1753 error = posix_lock_to_flock(&flock, &file_lock);
1754 if (error)
1755 goto out;
1756 }
1757 error = -EFAULT;
1758 if (!copy_to_user(l, &flock, sizeof(flock)))
1759 error = 0;
1760 out:
1761 return error;
1762 }
1763
1764 /**
1765 * vfs_lock_file - file byte range lock
1766 * @filp: The file to apply the lock to
1767 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1768 * @fl: The lock to be applied
1769 * @conf: Place to return a copy of the conflicting lock, if found.
1770 *
1771 * A caller that doesn't care about the conflicting lock may pass NULL
1772 * as the final argument.
1773 *
1774 * If the filesystem defines a private ->lock() method, then @conf will
1775 * be left unchanged; so a caller that cares should initialize it to
1776 * some acceptable default.
1777 *
1778 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1779 * locks, the ->lock() interface may return asynchronously, before the lock has
1780 * been granted or denied by the underlying filesystem, if (and only if)
1781 * lm_grant is set. Callers expecting ->lock() to return asynchronously
1782 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1783 * the request is for a blocking lock. When ->lock() does return asynchronously,
1784 * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock
1785 * request completes.
1786 * If the request is for non-blocking lock the file system should return
1787 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1788 * with the result. If the request timed out the callback routine will return a
1789 * nonzero return code and the file system should release the lock. The file
1790 * system is also responsible to keep a corresponding posix lock when it
1791 * grants a lock so the VFS can find out which locks are locally held and do
1792 * the correct lock cleanup when required.
1793 * The underlying filesystem must not drop the kernel lock or call
1794 * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED
1795 * return code.
1796 */
1797 int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
1798 {
1799 if (filp->f_op && filp->f_op->lock)
1800 return filp->f_op->lock(filp, cmd, fl);
1801 else
1802 return posix_lock_file(filp, fl, conf);
1803 }
1804 EXPORT_SYMBOL_GPL(vfs_lock_file);
1805
1806 static int do_lock_file_wait(struct file *filp, unsigned int cmd,
1807 struct file_lock *fl)
1808 {
1809 int error;
1810
1811 error = security_file_lock(filp, fl->fl_type);
1812 if (error)
1813 return error;
1814
1815 for (;;) {
1816 error = vfs_lock_file(filp, cmd, fl, NULL);
1817 if (error != FILE_LOCK_DEFERRED)
1818 break;
1819 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1820 if (!error)
1821 continue;
1822
1823 locks_delete_block(fl);
1824 break;
1825 }
1826
1827 return error;
1828 }
1829
1830 /* Apply the lock described by l to an open file descriptor.
1831 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1832 */
1833 int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1834 struct flock __user *l)
1835 {
1836 struct file_lock *file_lock = locks_alloc_lock();
1837 struct flock flock;
1838 struct inode *inode;
1839 struct file *f;
1840 int error;
1841
1842 if (file_lock == NULL)
1843 return -ENOLCK;
1844
1845 /*
1846 * This might block, so we do it before checking the inode.
1847 */
1848 error = -EFAULT;
1849 if (copy_from_user(&flock, l, sizeof(flock)))
1850 goto out;
1851
1852 inode = file_inode(filp);
1853
1854 /* Don't allow mandatory locks on files that may be memory mapped
1855 * and shared.
1856 */
1857 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
1858 error = -EAGAIN;
1859 goto out;
1860 }
1861
1862 error = flock_to_posix_lock(filp, file_lock, &flock);
1863 if (error)
1864 goto out;
1865
1866 /*
1867 * If the cmd is requesting file-private locks, then set the
1868 * FL_FILE_PVT flag and override the owner.
1869 */
1870 switch (cmd) {
1871 case F_SETLKP:
1872 cmd = F_SETLK;
1873 file_lock->fl_flags |= FL_FILE_PVT;
1874 file_lock->fl_owner = (fl_owner_t)filp;
1875 break;
1876 case F_SETLKPW:
1877 cmd = F_SETLKW;
1878 file_lock->fl_flags |= FL_FILE_PVT;
1879 file_lock->fl_owner = (fl_owner_t)filp;
1880 /* Fallthrough */
1881 case F_SETLKW:
1882 file_lock->fl_flags |= FL_SLEEP;
1883 }
1884
1885 error = -EBADF;
1886 switch (flock.l_type) {
1887 case F_RDLCK:
1888 if (!(filp->f_mode & FMODE_READ))
1889 goto out;
1890 break;
1891 case F_WRLCK:
1892 if (!(filp->f_mode & FMODE_WRITE))
1893 goto out;
1894 break;
1895 case F_UNLCK:
1896 break;
1897 default:
1898 error = -EINVAL;
1899 goto out;
1900 }
1901
1902 error = do_lock_file_wait(filp, cmd, file_lock);
1903
1904 /*
1905 * Attempt to detect a close/fcntl race and recover by
1906 * releasing the lock that was just acquired.
1907 */
1908 if (!error && file_lock->fl_type != F_UNLCK) {
1909 /*
1910 * We need that spin_lock here - it prevents reordering between
1911 * update of inode->i_flock and check for it done in
1912 * close(). rcu_read_lock() wouldn't do.
1913 */
1914 spin_lock(&current->files->file_lock);
1915 f = fcheck(fd);
1916 spin_unlock(&current->files->file_lock);
1917 if (f != filp) {
1918 file_lock->fl_type = F_UNLCK;
1919 error = do_lock_file_wait(filp, cmd, file_lock);
1920 WARN_ON_ONCE(error);
1921 error = -EBADF;
1922 }
1923 }
1924 out:
1925 locks_free_lock(file_lock);
1926 return error;
1927 }
1928
1929 #if BITS_PER_LONG == 32
1930 /* Report the first existing lock that would conflict with l.
1931 * This implements the F_GETLK command of fcntl().
1932 */
1933 int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l)
1934 {
1935 struct file_lock file_lock;
1936 struct flock64 flock;
1937 int error;
1938
1939 error = -EFAULT;
1940 if (copy_from_user(&flock, l, sizeof(flock)))
1941 goto out;
1942 error = -EINVAL;
1943 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1944 goto out;
1945
1946 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1947 if (error)
1948 goto out;
1949
1950 if (cmd == F_GETLKP) {
1951 cmd = F_GETLK64;
1952 file_lock.fl_flags |= FL_FILE_PVT;
1953 file_lock.fl_owner = (fl_owner_t)filp;
1954 }
1955
1956 error = vfs_test_lock(filp, &file_lock);
1957 if (error)
1958 goto out;
1959
1960 flock.l_type = file_lock.fl_type;
1961 if (file_lock.fl_type != F_UNLCK)
1962 posix_lock_to_flock64(&flock, &file_lock);
1963
1964 error = -EFAULT;
1965 if (!copy_to_user(l, &flock, sizeof(flock)))
1966 error = 0;
1967
1968 out:
1969 return error;
1970 }
1971
1972 /* Apply the lock described by l to an open file descriptor.
1973 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1974 */
1975 int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1976 struct flock64 __user *l)
1977 {
1978 struct file_lock *file_lock = locks_alloc_lock();
1979 struct flock64 flock;
1980 struct inode *inode;
1981 struct file *f;
1982 int error;
1983
1984 if (file_lock == NULL)
1985 return -ENOLCK;
1986
1987 /*
1988 * This might block, so we do it before checking the inode.
1989 */
1990 error = -EFAULT;
1991 if (copy_from_user(&flock, l, sizeof(flock)))
1992 goto out;
1993
1994 inode = file_inode(filp);
1995
1996 /* Don't allow mandatory locks on files that may be memory mapped
1997 * and shared.
1998 */
1999 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
2000 error = -EAGAIN;
2001 goto out;
2002 }
2003
2004 error = flock64_to_posix_lock(filp, file_lock, &flock);
2005 if (error)
2006 goto out;
2007
2008 /*
2009 * If the cmd is requesting file-private locks, then set the
2010 * FL_FILE_PVT flag and override the owner.
2011 */
2012 switch (cmd) {
2013 case F_SETLKP:
2014 cmd = F_SETLK64;
2015 file_lock->fl_flags |= FL_FILE_PVT;
2016 file_lock->fl_owner = (fl_owner_t)filp;
2017 break;
2018 case F_SETLKPW:
2019 cmd = F_SETLKW64;
2020 file_lock->fl_flags |= FL_FILE_PVT;
2021 file_lock->fl_owner = (fl_owner_t)filp;
2022 /* Fallthrough */
2023 case F_SETLKW64:
2024 file_lock->fl_flags |= FL_SLEEP;
2025 }
2026
2027 error = -EBADF;
2028 switch (flock.l_type) {
2029 case F_RDLCK:
2030 if (!(filp->f_mode & FMODE_READ))
2031 goto out;
2032 break;
2033 case F_WRLCK:
2034 if (!(filp->f_mode & FMODE_WRITE))
2035 goto out;
2036 break;
2037 case F_UNLCK:
2038 break;
2039 default:
2040 error = -EINVAL;
2041 goto out;
2042 }
2043
2044 error = do_lock_file_wait(filp, cmd, file_lock);
2045
2046 /*
2047 * Attempt to detect a close/fcntl race and recover by
2048 * releasing the lock that was just acquired.
2049 */
2050 if (!error && file_lock->fl_type != F_UNLCK) {
2051 /*
2052 * We need that spin_lock here - it prevents reordering between
2053 * update of inode->i_flock and check for it done in
2054 * close(). rcu_read_lock() wouldn't do.
2055 */
2056 spin_lock(&current->files->file_lock);
2057 f = fcheck(fd);
2058 spin_unlock(&current->files->file_lock);
2059 if (f != filp) {
2060 file_lock->fl_type = F_UNLCK;
2061 error = do_lock_file_wait(filp, cmd, file_lock);
2062 WARN_ON_ONCE(error);
2063 error = -EBADF;
2064 }
2065 }
2066 out:
2067 locks_free_lock(file_lock);
2068 return error;
2069 }
2070 #endif /* BITS_PER_LONG == 32 */
2071
2072 /*
2073 * This function is called when the file is being removed
2074 * from the task's fd array. POSIX locks belonging to this task
2075 * are deleted at this time.
2076 */
2077 void locks_remove_posix(struct file *filp, fl_owner_t owner)
2078 {
2079 struct file_lock lock;
2080
2081 /*
2082 * If there are no locks held on this file, we don't need to call
2083 * posix_lock_file(). Another process could be setting a lock on this
2084 * file at the same time, but we wouldn't remove that lock anyway.
2085 */
2086 if (!file_inode(filp)->i_flock)
2087 return;
2088
2089 lock.fl_type = F_UNLCK;
2090 lock.fl_flags = FL_POSIX | FL_CLOSE;
2091 lock.fl_start = 0;
2092 lock.fl_end = OFFSET_MAX;
2093 lock.fl_owner = owner;
2094 lock.fl_pid = current->tgid;
2095 lock.fl_file = filp;
2096 lock.fl_ops = NULL;
2097 lock.fl_lmops = NULL;
2098
2099 vfs_lock_file(filp, F_SETLK, &lock, NULL);
2100
2101 if (lock.fl_ops && lock.fl_ops->fl_release_private)
2102 lock.fl_ops->fl_release_private(&lock);
2103 }
2104
2105 EXPORT_SYMBOL(locks_remove_posix);
2106
2107 /*
2108 * This function is called on the last close of an open file.
2109 */
2110 void locks_remove_file(struct file *filp)
2111 {
2112 struct inode * inode = file_inode(filp);
2113 struct file_lock *fl;
2114 struct file_lock **before;
2115
2116 if (!inode->i_flock)
2117 return;
2118
2119 locks_remove_posix(filp, (fl_owner_t)filp);
2120
2121 if (filp->f_op && filp->f_op->flock) {
2122 struct file_lock fl = {
2123 .fl_pid = current->tgid,
2124 .fl_file = filp,
2125 .fl_flags = FL_FLOCK,
2126 .fl_type = F_UNLCK,
2127 .fl_end = OFFSET_MAX,
2128 };
2129 filp->f_op->flock(filp, F_SETLKW, &fl);
2130 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2131 fl.fl_ops->fl_release_private(&fl);
2132 }
2133
2134 lock_flocks();
2135 before = &inode->i_flock;
2136
2137 while ((fl = *before) != NULL) {
2138 if (fl->fl_file == filp) {
2139 if (IS_FLOCK(fl)) {
2140 locks_delete_lock(before);
2141 continue;
2142 }
2143 if (IS_LEASE(fl)) {
2144 lease_modify(before, F_UNLCK);
2145 continue;
2146 }
2147 /* What? */
2148 BUG();
2149 }
2150 before = &fl->fl_next;
2151 }
2152 unlock_flocks();
2153 }
2154
2155 /**
2156 * posix_unblock_lock - stop waiting for a file lock
2157 * @filp: how the file was opened
2158 * @waiter: the lock which was waiting
2159 *
2160 * lockd needs to block waiting for locks.
2161 */
2162 int
2163 posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2164 {
2165 int status = 0;
2166
2167 lock_flocks();
2168 if (waiter->fl_next)
2169 __locks_delete_block(waiter);
2170 else
2171 status = -ENOENT;
2172 unlock_flocks();
2173 return status;
2174 }
2175
2176 EXPORT_SYMBOL(posix_unblock_lock);
2177
2178 /**
2179 * vfs_cancel_lock - file byte range unblock lock
2180 * @filp: The file to apply the unblock to
2181 * @fl: The lock to be unblocked
2182 *
2183 * Used by lock managers to cancel blocked requests
2184 */
2185 int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2186 {
2187 if (filp->f_op && filp->f_op->lock)
2188 return filp->f_op->lock(filp, F_CANCELLK, fl);
2189 return 0;
2190 }
2191
2192 EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2193
2194 #ifdef CONFIG_PROC_FS
2195 #include <linux/proc_fs.h>
2196 #include <linux/seq_file.h>
2197
2198 static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2199 loff_t id, char *pfx)
2200 {
2201 struct inode *inode = NULL;
2202 unsigned int fl_pid;
2203
2204 if (fl->fl_nspid)
2205 fl_pid = pid_vnr(fl->fl_nspid);
2206 else
2207 fl_pid = fl->fl_pid;
2208
2209 if (fl->fl_file != NULL)
2210 inode = file_inode(fl->fl_file);
2211
2212 seq_printf(f, "%lld:%s ", id, pfx);
2213 if (IS_POSIX(fl)) {
2214 if (fl->fl_flags & FL_ACCESS)
2215 seq_printf(f, "ACCESS");
2216 else if (IS_FILE_PVT(fl))
2217 seq_printf(f, "FLPVT ");
2218 else
2219 seq_printf(f, "POSIX ");
2220
2221 seq_printf(f, " %s ",
2222 (inode == NULL) ? "*NOINODE*" :
2223 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
2224 } else if (IS_FLOCK(fl)) {
2225 if (fl->fl_type & LOCK_MAND) {
2226 seq_printf(f, "FLOCK MSNFS ");
2227 } else {
2228 seq_printf(f, "FLOCK ADVISORY ");
2229 }
2230 } else if (IS_LEASE(fl)) {
2231 seq_printf(f, "LEASE ");
2232 if (lease_breaking(fl))
2233 seq_printf(f, "BREAKING ");
2234 else if (fl->fl_file)
2235 seq_printf(f, "ACTIVE ");
2236 else
2237 seq_printf(f, "BREAKER ");
2238 } else {
2239 seq_printf(f, "UNKNOWN UNKNOWN ");
2240 }
2241 if (fl->fl_type & LOCK_MAND) {
2242 seq_printf(f, "%s ",
2243 (fl->fl_type & LOCK_READ)
2244 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2245 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2246 } else {
2247 seq_printf(f, "%s ",
2248 (lease_breaking(fl))
2249 ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ "
2250 : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ ");
2251 }
2252 if (inode) {
2253 #ifdef WE_CAN_BREAK_LSLK_NOW
2254 seq_printf(f, "%d %s:%ld ", fl_pid,
2255 inode->i_sb->s_id, inode->i_ino);
2256 #else
2257 /* userspace relies on this representation of dev_t ;-( */
2258 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
2259 MAJOR(inode->i_sb->s_dev),
2260 MINOR(inode->i_sb->s_dev), inode->i_ino);
2261 #endif
2262 } else {
2263 seq_printf(f, "%d <none>:0 ", fl_pid);
2264 }
2265 if (IS_POSIX(fl)) {
2266 if (fl->fl_end == OFFSET_MAX)
2267 seq_printf(f, "%Ld EOF\n", fl->fl_start);
2268 else
2269 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
2270 } else {
2271 seq_printf(f, "0 EOF\n");
2272 }
2273 }
2274
2275 static int locks_show(struct seq_file *f, void *v)
2276 {
2277 struct file_lock *fl, *bfl;
2278
2279 fl = list_entry(v, struct file_lock, fl_link);
2280
2281 lock_get_status(f, fl, *((loff_t *)f->private), "");
2282
2283 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2284 lock_get_status(f, bfl, *((loff_t *)f->private), " ->");
2285
2286 return 0;
2287 }
2288
2289 static void *locks_start(struct seq_file *f, loff_t *pos)
2290 {
2291 loff_t *p = f->private;
2292
2293 lock_flocks();
2294 *p = (*pos + 1);
2295 return seq_list_start(&file_lock_list, *pos);
2296 }
2297
2298 static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2299 {
2300 loff_t *p = f->private;
2301 ++*p;
2302 return seq_list_next(v, &file_lock_list, pos);
2303 }
2304
2305 static void locks_stop(struct seq_file *f, void *v)
2306 {
2307 unlock_flocks();
2308 }
2309
2310 static const struct seq_operations locks_seq_operations = {
2311 .start = locks_start,
2312 .next = locks_next,
2313 .stop = locks_stop,
2314 .show = locks_show,
2315 };
2316
2317 static int locks_open(struct inode *inode, struct file *filp)
2318 {
2319 return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t));
2320 }
2321
2322 static const struct file_operations proc_locks_operations = {
2323 .open = locks_open,
2324 .read = seq_read,
2325 .llseek = seq_lseek,
2326 .release = seq_release_private,
2327 };
2328
2329 static int __init proc_locks_init(void)
2330 {
2331 proc_create("locks", 0, NULL, &proc_locks_operations);
2332 return 0;
2333 }
2334 module_init(proc_locks_init);
2335 #endif
2336
2337 /**
2338 * lock_may_read - checks that the region is free of locks
2339 * @inode: the inode that is being read
2340 * @start: the first byte to read
2341 * @len: the number of bytes to read
2342 *
2343 * Emulates Windows locking requirements. Whole-file
2344 * mandatory locks (share modes) can prohibit a read and
2345 * byte-range POSIX locks can prohibit a read if they overlap.
2346 *
2347 * N.B. this function is only ever called
2348 * from knfsd and ownership of locks is never checked.
2349 */
2350 int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2351 {
2352 struct file_lock *fl;
2353 int result = 1;
2354 lock_flocks();
2355 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2356 if (IS_POSIX(fl)) {
2357 if (fl->fl_type == F_RDLCK)
2358 continue;
2359 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2360 continue;
2361 } else if (IS_FLOCK(fl)) {
2362 if (!(fl->fl_type & LOCK_MAND))
2363 continue;
2364 if (fl->fl_type & LOCK_READ)
2365 continue;
2366 } else
2367 continue;
2368 result = 0;
2369 break;
2370 }
2371 unlock_flocks();
2372 return result;
2373 }
2374
2375 EXPORT_SYMBOL(lock_may_read);
2376
2377 /**
2378 * lock_may_write - checks that the region is free of locks
2379 * @inode: the inode that is being written
2380 * @start: the first byte to write
2381 * @len: the number of bytes to write
2382 *
2383 * Emulates Windows locking requirements. Whole-file
2384 * mandatory locks (share modes) can prohibit a write and
2385 * byte-range POSIX locks can prohibit a write if they overlap.
2386 *
2387 * N.B. this function is only ever called
2388 * from knfsd and ownership of locks is never checked.
2389 */
2390 int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2391 {
2392 struct file_lock *fl;
2393 int result = 1;
2394 lock_flocks();
2395 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2396 if (IS_POSIX(fl)) {
2397 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2398 continue;
2399 } else if (IS_FLOCK(fl)) {
2400 if (!(fl->fl_type & LOCK_MAND))
2401 continue;
2402 if (fl->fl_type & LOCK_WRITE)
2403 continue;
2404 } else
2405 continue;
2406 result = 0;
2407 break;
2408 }
2409 unlock_flocks();
2410 return result;
2411 }
2412
2413 EXPORT_SYMBOL(lock_may_write);
2414
2415 static int __init filelock_init(void)
2416 {
2417 filelock_cache = kmem_cache_create("file_lock_cache",
2418 sizeof(struct file_lock), 0, SLAB_PANIC, NULL);
2419
2420 return 0;
2421 }
2422
2423 core_initcall(filelock_init);