blob: c7c69d29a576506ea93ce82899df8f2791abe7d0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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/mandatory.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/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>
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700127#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
129#include <asm/semaphore.h>
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
136int leases_enable = 1;
137int lease_break_time = 45;
138
139#define for_each_lock(inode, lockp) \
140 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
141
Christoph Hellwig26bcbf92006-03-20 13:44:40 -0500142static LIST_HEAD(file_lock_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143static LIST_HEAD(blocked_list);
144
Christoph Lametere18b8902006-12-06 20:33:20 -0800145static struct kmem_cache *filelock_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146
147/* Allocate an empty lock structure. */
148static struct file_lock *locks_alloc_lock(void)
149{
Christoph Lametere94b1762006-12-06 20:33:17 -0800150 return kmem_cache_alloc(filelock_cache, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151}
152
Trond Myklebust47831f32006-03-20 13:44:05 -0500153static void locks_release_private(struct file_lock *fl)
154{
155 if (fl->fl_ops) {
156 if (fl->fl_ops->fl_release_private)
157 fl->fl_ops->fl_release_private(fl);
158 fl->fl_ops = NULL;
159 }
160 if (fl->fl_lmops) {
161 if (fl->fl_lmops->fl_release_private)
162 fl->fl_lmops->fl_release_private(fl);
163 fl->fl_lmops = NULL;
164 }
165
166}
167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168/* Free a lock which is not in use. */
Matt Mackall33443c42006-01-08 01:05:22 -0800169static void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800171 BUG_ON(waitqueue_active(&fl->fl_wait));
172 BUG_ON(!list_empty(&fl->fl_block));
173 BUG_ON(!list_empty(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
Trond Myklebust47831f32006-03-20 13:44:05 -0500175 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 kmem_cache_free(filelock_cache, fl);
177}
178
179void locks_init_lock(struct file_lock *fl)
180{
181 INIT_LIST_HEAD(&fl->fl_link);
182 INIT_LIST_HEAD(&fl->fl_block);
183 init_waitqueue_head(&fl->fl_wait);
184 fl->fl_next = NULL;
185 fl->fl_fasync = NULL;
186 fl->fl_owner = NULL;
187 fl->fl_pid = 0;
188 fl->fl_file = NULL;
189 fl->fl_flags = 0;
190 fl->fl_type = 0;
191 fl->fl_start = fl->fl_end = 0;
192 fl->fl_ops = NULL;
193 fl->fl_lmops = NULL;
194}
195
196EXPORT_SYMBOL(locks_init_lock);
197
198/*
199 * Initialises the fields of the file lock which are invariant for
200 * free file_locks.
201 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800202static void init_once(void *foo, struct kmem_cache *cache, unsigned long flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
204 struct file_lock *lock = (struct file_lock *) foo;
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 locks_init_lock(lock);
207}
208
Trond Myklebust47831f32006-03-20 13:44:05 -0500209static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
210{
211 if (fl->fl_ops) {
212 if (fl->fl_ops->fl_copy_lock)
213 fl->fl_ops->fl_copy_lock(new, fl);
214 new->fl_ops = fl->fl_ops;
215 }
216 if (fl->fl_lmops) {
217 if (fl->fl_lmops->fl_copy_lock)
218 fl->fl_lmops->fl_copy_lock(new, fl);
219 new->fl_lmops = fl->fl_lmops;
220 }
221}
222
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223/*
224 * Initialize a new lock from an existing file_lock structure.
225 */
Trond Myklebust09969052006-03-20 13:44:38 -0500226static void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
228 new->fl_owner = fl->fl_owner;
229 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500230 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 new->fl_flags = fl->fl_flags;
232 new->fl_type = fl->fl_type;
233 new->fl_start = fl->fl_start;
234 new->fl_end = fl->fl_end;
Trond Myklebust09969052006-03-20 13:44:38 -0500235 new->fl_ops = NULL;
236 new->fl_lmops = NULL;
237}
238
239void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
240{
241 locks_release_private(new);
242
243 __locks_copy_lock(new, fl);
244 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 new->fl_ops = fl->fl_ops;
246 new->fl_lmops = fl->fl_lmops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500247
248 locks_copy_private(new, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249}
250
251EXPORT_SYMBOL(locks_copy_lock);
252
253static inline int flock_translate_cmd(int cmd) {
254 if (cmd & LOCK_MAND)
255 return cmd & (LOCK_MAND | LOCK_RW);
256 switch (cmd) {
257 case LOCK_SH:
258 return F_RDLCK;
259 case LOCK_EX:
260 return F_WRLCK;
261 case LOCK_UN:
262 return F_UNLCK;
263 }
264 return -EINVAL;
265}
266
267/* Fill in a file_lock structure with an appropriate FLOCK lock. */
268static int flock_make_lock(struct file *filp, struct file_lock **lock,
269 unsigned int cmd)
270{
271 struct file_lock *fl;
272 int type = flock_translate_cmd(cmd);
273 if (type < 0)
274 return type;
275
276 fl = locks_alloc_lock();
277 if (fl == NULL)
278 return -ENOMEM;
279
280 fl->fl_file = filp;
281 fl->fl_pid = current->tgid;
282 fl->fl_flags = FL_FLOCK;
283 fl->fl_type = type;
284 fl->fl_end = OFFSET_MAX;
285
286 *lock = fl;
287 return 0;
288}
289
290static int assign_type(struct file_lock *fl, int type)
291{
292 switch (type) {
293 case F_RDLCK:
294 case F_WRLCK:
295 case F_UNLCK:
296 fl->fl_type = type;
297 break;
298 default:
299 return -EINVAL;
300 }
301 return 0;
302}
303
304/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
305 * style lock.
306 */
307static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
308 struct flock *l)
309{
310 off_t start, end;
311
312 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700313 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 start = 0;
315 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700316 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 start = filp->f_pos;
318 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700319 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800320 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 break;
322 default:
323 return -EINVAL;
324 }
325
326 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
327 POSIX-2001 defines it. */
328 start += l->l_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 if (start < 0)
330 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700331 fl->fl_end = OFFSET_MAX;
332 if (l->l_len > 0) {
333 end = start + l->l_len - 1;
334 fl->fl_end = end;
335 } else if (l->l_len < 0) {
336 end = start - 1;
337 fl->fl_end = end;
338 start += l->l_len;
339 if (start < 0)
340 return -EINVAL;
341 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700343 if (fl->fl_end < fl->fl_start)
344 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345
346 fl->fl_owner = current->files;
347 fl->fl_pid = current->tgid;
348 fl->fl_file = filp;
349 fl->fl_flags = FL_POSIX;
350 fl->fl_ops = NULL;
351 fl->fl_lmops = NULL;
352
353 return assign_type(fl, l->l_type);
354}
355
356#if BITS_PER_LONG == 32
357static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
358 struct flock64 *l)
359{
360 loff_t start;
361
362 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700363 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 start = 0;
365 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700366 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 start = filp->f_pos;
368 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700369 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800370 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 break;
372 default:
373 return -EINVAL;
374 }
375
Trond Myklebust4c780a42005-10-18 14:20:21 -0700376 start += l->l_start;
377 if (start < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700379 fl->fl_end = OFFSET_MAX;
380 if (l->l_len > 0) {
381 fl->fl_end = start + l->l_len - 1;
382 } else if (l->l_len < 0) {
383 fl->fl_end = start - 1;
384 start += l->l_len;
385 if (start < 0)
386 return -EINVAL;
387 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700389 if (fl->fl_end < fl->fl_start)
390 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 fl->fl_owner = current->files;
393 fl->fl_pid = current->tgid;
394 fl->fl_file = filp;
395 fl->fl_flags = FL_POSIX;
396 fl->fl_ops = NULL;
397 fl->fl_lmops = NULL;
398
399 switch (l->l_type) {
400 case F_RDLCK:
401 case F_WRLCK:
402 case F_UNLCK:
403 fl->fl_type = l->l_type;
404 break;
405 default:
406 return -EINVAL;
407 }
408
409 return (0);
410}
411#endif
412
413/* default lease lock manager operations */
414static void lease_break_callback(struct file_lock *fl)
415{
416 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
417}
418
419static void lease_release_private_callback(struct file_lock *fl)
420{
421 if (!fl->fl_file)
422 return;
423
424 f_delown(fl->fl_file);
425 fl->fl_file->f_owner.signum = 0;
426}
427
Adrian Bunk75c96f82005-05-05 16:16:09 -0700428static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429{
430 return fl->fl_file == try->fl_file;
431}
432
Adrian Bunk75c96f82005-05-05 16:16:09 -0700433static struct lock_manager_operations lease_manager_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 .fl_break = lease_break_callback,
435 .fl_release_private = lease_release_private_callback,
436 .fl_mylease = lease_mylease_callback,
437 .fl_change = lease_modify,
438};
439
440/*
441 * Initialize a lease, use the default lock manager operations
442 */
443static int lease_init(struct file *filp, int type, struct file_lock *fl)
444 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400445 if (assign_type(fl, type) != 0)
446 return -EINVAL;
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 fl->fl_owner = current->files;
449 fl->fl_pid = current->tgid;
450
451 fl->fl_file = filp;
452 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 fl->fl_start = 0;
454 fl->fl_end = OFFSET_MAX;
455 fl->fl_ops = NULL;
456 fl->fl_lmops = &lease_manager_ops;
457 return 0;
458}
459
460/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500461static struct file_lock *lease_alloc(struct file *filp, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462{
463 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400464 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500467 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
469 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400470 if (error) {
471 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500472 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400473 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500474 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
477/* Check if two locks overlap each other.
478 */
479static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
480{
481 return ((fl1->fl_end >= fl2->fl_start) &&
482 (fl2->fl_end >= fl1->fl_start));
483}
484
485/*
486 * Check whether two locks have the same owner.
487 */
Matt Mackall33443c42006-01-08 01:05:22 -0800488static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
490 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
491 return fl2->fl_lmops == fl1->fl_lmops &&
492 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
493 return fl1->fl_owner == fl2->fl_owner;
494}
495
496/* Remove waiter from blocker's block list.
497 * When blocker ends up pointing to itself then the list is empty.
498 */
Matt Mackall33443c42006-01-08 01:05:22 -0800499static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500{
501 list_del_init(&waiter->fl_block);
502 list_del_init(&waiter->fl_link);
503 waiter->fl_next = NULL;
504}
505
506/*
507 */
508static void locks_delete_block(struct file_lock *waiter)
509{
510 lock_kernel();
511 __locks_delete_block(waiter);
512 unlock_kernel();
513}
514
515/* Insert waiter into blocker's block list.
516 * We use a circular list so that processes can be easily woken up in
517 * the order they blocked. The documentation doesn't require this but
518 * it seems like the reasonable thing to do.
519 */
520static void locks_insert_block(struct file_lock *blocker,
521 struct file_lock *waiter)
522{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800523 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 list_add_tail(&waiter->fl_block, &blocker->fl_block);
525 waiter->fl_next = blocker;
526 if (IS_POSIX(blocker))
527 list_add(&waiter->fl_link, &blocked_list);
528}
529
530/* Wake up processes blocked waiting for blocker.
531 * If told to wait then schedule the processes until the block list
532 * is empty, otherwise empty the block list ourselves.
533 */
534static void locks_wake_up_blocks(struct file_lock *blocker)
535{
536 while (!list_empty(&blocker->fl_block)) {
537 struct file_lock *waiter = list_entry(blocker->fl_block.next,
538 struct file_lock, fl_block);
539 __locks_delete_block(waiter);
540 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
541 waiter->fl_lmops->fl_notify(waiter);
542 else
543 wake_up(&waiter->fl_wait);
544 }
545}
546
547/* Insert file lock fl into an inode's lock list at the position indicated
548 * by pos. At the same time add the lock to the global file lock list.
549 */
550static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
551{
552 list_add(&fl->fl_link, &file_lock_list);
553
554 /* insert into file's list */
555 fl->fl_next = *pos;
556 *pos = fl;
557
558 if (fl->fl_ops && fl->fl_ops->fl_insert)
559 fl->fl_ops->fl_insert(fl);
560}
561
562/*
563 * Delete a lock and then free it.
564 * Wake up processes that are blocked waiting for this lock,
565 * notify the FS that the lock has been cleared and
566 * finally free the lock.
567 */
568static void locks_delete_lock(struct file_lock **thisfl_p)
569{
570 struct file_lock *fl = *thisfl_p;
571
572 *thisfl_p = fl->fl_next;
573 fl->fl_next = NULL;
574 list_del_init(&fl->fl_link);
575
576 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
577 if (fl->fl_fasync != NULL) {
578 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
579 fl->fl_fasync = NULL;
580 }
581
582 if (fl->fl_ops && fl->fl_ops->fl_remove)
583 fl->fl_ops->fl_remove(fl);
584
585 locks_wake_up_blocks(fl);
586 locks_free_lock(fl);
587}
588
589/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
590 * checks for shared/exclusive status of overlapping locks.
591 */
592static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
593{
594 if (sys_fl->fl_type == F_WRLCK)
595 return 1;
596 if (caller_fl->fl_type == F_WRLCK)
597 return 1;
598 return 0;
599}
600
601/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
602 * checking before calling the locks_conflict().
603 */
604static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
605{
606 /* POSIX locks owned by the same process do not conflict with
607 * each other.
608 */
609 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
610 return (0);
611
612 /* Check whether they overlap */
613 if (!locks_overlap(caller_fl, sys_fl))
614 return 0;
615
616 return (locks_conflict(caller_fl, sys_fl));
617}
618
619/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
620 * checking before calling the locks_conflict().
621 */
622static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
623{
624 /* FLOCK locks referring to the same filp do not conflict with
625 * each other.
626 */
627 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
628 return (0);
629 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
630 return 0;
631
632 return (locks_conflict(caller_fl, sys_fl));
633}
634
635static int interruptible_sleep_on_locked(wait_queue_head_t *fl_wait, int timeout)
636{
637 int result = 0;
638 DECLARE_WAITQUEUE(wait, current);
639
640 __set_current_state(TASK_INTERRUPTIBLE);
641 add_wait_queue(fl_wait, &wait);
642 if (timeout == 0)
643 schedule();
644 else
645 result = schedule_timeout(timeout);
646 if (signal_pending(current))
647 result = -ERESTARTSYS;
648 remove_wait_queue(fl_wait, &wait);
649 __set_current_state(TASK_RUNNING);
650 return result;
651}
652
653static int locks_block_on_timeout(struct file_lock *blocker, struct file_lock *waiter, int time)
654{
655 int result;
656 locks_insert_block(blocker, waiter);
657 result = interruptible_sleep_on_locked(&waiter->fl_wait, time);
658 __locks_delete_block(waiter);
659 return result;
660}
661
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400662void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500663posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664{
665 struct file_lock *cfl;
666
667 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800668 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 if (!IS_POSIX(cfl))
670 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400671 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 break;
673 }
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400674 if (cfl)
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500675 __locks_copy_lock(fl, cfl);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400676 else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400677 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400679 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680}
681
682EXPORT_SYMBOL(posix_test_lock);
683
684/* This function tests for deadlock condition before putting a process to
685 * sleep. The detection scheme is no longer recursive. Recursive was neat,
686 * but dangerous - we risked stack corruption if the lock data was bad, or
687 * if the recursion was too deep for any other reason.
688 *
689 * We rely on the fact that a task can only be on one lock's wait queue
690 * at a time. When we find blocked_task on a wait queue we can re-search
691 * with blocked_task equal to that queue's owner, until either blocked_task
692 * isn't found, or blocked_task is found on a queue owned by my_task.
693 *
694 * Note: the above assumption may not be true when handling lock requests
695 * from a broken NFS client. But broken NFS clients have a lot more to
696 * worry about than proper deadlock detection anyway... --okir
697 */
Adrian Bunkb0904e12006-06-23 02:05:13 -0700698static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 struct file_lock *block_fl)
700{
701 struct list_head *tmp;
702
703next_task:
704 if (posix_same_owner(caller_fl, block_fl))
705 return 1;
706 list_for_each(tmp, &blocked_list) {
707 struct file_lock *fl = list_entry(tmp, struct file_lock, fl_link);
708 if (posix_same_owner(fl, block_fl)) {
709 fl = fl->fl_next;
710 block_fl = fl;
711 goto next_task;
712 }
713 }
714 return 0;
715}
716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400718 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400719 *
720 * Note that if called with an FL_EXISTS argument, the caller may determine
721 * whether or not a lock was successfully freed by testing the return
722 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800724static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800726 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800728 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 int error = 0;
730 int found = 0;
731
732 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400733 if (request->fl_flags & FL_ACCESS)
734 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400735
736 if (request->fl_type != F_UNLCK) {
737 error = -ENOMEM;
738 new_fl = locks_alloc_lock();
739 if (new_fl == NULL)
740 goto out;
741 error = 0;
742 }
743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 for_each_lock(inode, before) {
745 struct file_lock *fl = *before;
746 if (IS_POSIX(fl))
747 break;
748 if (IS_LEASE(fl))
749 continue;
750 if (filp != fl->fl_file)
751 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800752 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 goto out;
754 found = 1;
755 locks_delete_lock(before);
756 break;
757 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Trond Myklebustf475ae92006-06-29 16:38:32 -0400759 if (request->fl_type == F_UNLCK) {
760 if ((request->fl_flags & FL_EXISTS) && !found)
761 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800762 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400763 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
765 /*
766 * If a higher-priority process was blocked on the old file lock,
767 * give it the opportunity to lock the file.
768 */
769 if (found)
770 cond_resched();
771
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400772find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 for_each_lock(inode, before) {
774 struct file_lock *fl = *before;
775 if (IS_POSIX(fl))
776 break;
777 if (IS_LEASE(fl))
778 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800779 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 continue;
781 error = -EAGAIN;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800782 if (request->fl_flags & FL_SLEEP)
783 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 goto out;
785 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400786 if (request->fl_flags & FL_ACCESS)
787 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800788 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700789 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800790 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400791 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793out:
794 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800795 if (new_fl)
796 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 return error;
798}
799
Marc Eshel150b3932007-01-18 16:15:35 -0500800static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801{
802 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700803 struct file_lock *new_fl = NULL;
804 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 struct file_lock *left = NULL;
806 struct file_lock *right = NULL;
807 struct file_lock **before;
808 int error, added = 0;
809
810 /*
811 * We may need two file_lock structures for this operation,
812 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700813 *
814 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700816 if (!(request->fl_flags & FL_ACCESS) &&
817 (request->fl_type != F_UNLCK ||
818 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
819 new_fl = locks_alloc_lock();
820 new_fl2 = locks_alloc_lock();
821 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822
823 lock_kernel();
824 if (request->fl_type != F_UNLCK) {
825 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500826 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 if (!IS_POSIX(fl))
828 continue;
829 if (!posix_locks_conflict(request, fl))
830 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800831 if (conflock)
832 locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 error = -EAGAIN;
834 if (!(request->fl_flags & FL_SLEEP))
835 goto out;
836 error = -EDEADLK;
837 if (posix_locks_deadlock(request, fl))
838 goto out;
839 error = -EAGAIN;
840 locks_insert_block(fl, request);
841 goto out;
842 }
843 }
844
845 /* If we're just looking for a conflict, we're done. */
846 error = 0;
847 if (request->fl_flags & FL_ACCESS)
848 goto out;
849
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 * Find the first old lock with the same owner as the new lock.
852 */
853
854 before = &inode->i_flock;
855
856 /* First skip locks owned by other processes. */
857 while ((fl = *before) && (!IS_POSIX(fl) ||
858 !posix_same_owner(request, fl))) {
859 before = &fl->fl_next;
860 }
861
862 /* Process locks with this owner. */
863 while ((fl = *before) && posix_same_owner(request, fl)) {
864 /* Detect adjacent or overlapping regions (if same lock type)
865 */
866 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700867 /* In all comparisons of start vs end, use
868 * "start - 1" rather than "end + 1". If end
869 * is OFFSET_MAX, end + 1 will become negative.
870 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 if (fl->fl_end < request->fl_start - 1)
872 goto next_lock;
873 /* If the next lock in the list has entirely bigger
874 * addresses than the new one, insert the lock here.
875 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700876 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 break;
878
879 /* If we come here, the new and old lock are of the
880 * same type and adjacent or overlapping. Make one
881 * lock yielding from the lower start address of both
882 * locks to the higher end address.
883 */
884 if (fl->fl_start > request->fl_start)
885 fl->fl_start = request->fl_start;
886 else
887 request->fl_start = fl->fl_start;
888 if (fl->fl_end < request->fl_end)
889 fl->fl_end = request->fl_end;
890 else
891 request->fl_end = fl->fl_end;
892 if (added) {
893 locks_delete_lock(before);
894 continue;
895 }
896 request = fl;
897 added = 1;
898 }
899 else {
900 /* Processing for different lock types is a bit
901 * more complex.
902 */
903 if (fl->fl_end < request->fl_start)
904 goto next_lock;
905 if (fl->fl_start > request->fl_end)
906 break;
907 if (request->fl_type == F_UNLCK)
908 added = 1;
909 if (fl->fl_start < request->fl_start)
910 left = fl;
911 /* If the next lock in the list has a higher end
912 * address than the new one, insert the new one here.
913 */
914 if (fl->fl_end > request->fl_end) {
915 right = fl;
916 break;
917 }
918 if (fl->fl_start >= request->fl_start) {
919 /* The new lock completely replaces an old
920 * one (This may happen several times).
921 */
922 if (added) {
923 locks_delete_lock(before);
924 continue;
925 }
926 /* Replace the old lock with the new one.
927 * Wake up anybody waiting for the old one,
928 * as the change in lock type might satisfy
929 * their needs.
930 */
931 locks_wake_up_blocks(fl);
932 fl->fl_start = request->fl_start;
933 fl->fl_end = request->fl_end;
934 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500935 locks_release_private(fl);
936 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 request = fl;
938 added = 1;
939 }
940 }
941 /* Go on to next lock.
942 */
943 next_lock:
944 before = &fl->fl_next;
945 }
946
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700947 /*
948 * The above code only modifies existing locks in case of
949 * merging or replacing. If new lock(s) need to be inserted
950 * all modifications are done bellow this, so it's safe yet to
951 * bail out.
952 */
953 error = -ENOLCK; /* "no luck" */
954 if (right && left == right && !new_fl2)
955 goto out;
956
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 error = 0;
958 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400959 if (request->fl_type == F_UNLCK) {
960 if (request->fl_flags & FL_EXISTS)
961 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400963 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700964
965 if (!new_fl) {
966 error = -ENOLCK;
967 goto out;
968 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 locks_copy_lock(new_fl, request);
970 locks_insert_lock(before, new_fl);
971 new_fl = NULL;
972 }
973 if (right) {
974 if (left == right) {
975 /* The new lock breaks the old one in two pieces,
976 * so we have to use the second new lock.
977 */
978 left = new_fl2;
979 new_fl2 = NULL;
980 locks_copy_lock(left, right);
981 locks_insert_lock(before, left);
982 }
983 right->fl_start = request->fl_end + 1;
984 locks_wake_up_blocks(right);
985 }
986 if (left) {
987 left->fl_end = request->fl_start - 1;
988 locks_wake_up_blocks(left);
989 }
990 out:
991 unlock_kernel();
992 /*
993 * Free any unused locks.
994 */
995 if (new_fl)
996 locks_free_lock(new_fl);
997 if (new_fl2)
998 locks_free_lock(new_fl2);
999 return error;
1000}
1001
1002/**
1003 * posix_lock_file - Apply a POSIX-style lock to a file
1004 * @filp: The file to apply the lock to
1005 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001006 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 *
1008 * Add a POSIX style lock to a file.
1009 * We merge adjacent & overlapping locks whenever possible.
1010 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001011 *
1012 * Note that if called with an FL_EXISTS argument, the caller may determine
1013 * whether or not a lock was successfully freed by testing the return
1014 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 */
Marc Eshel150b3932007-01-18 16:15:35 -05001016int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001017 struct file_lock *conflock)
1018{
Marc Eshel150b3932007-01-18 16:15:35 -05001019 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001020}
Marc Eshel150b3932007-01-18 16:15:35 -05001021EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
1023/**
1024 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1025 * @filp: The file to apply the lock to
1026 * @fl: The lock to be applied
1027 *
1028 * Add a POSIX style lock to a file.
1029 * We merge adjacent & overlapping locks whenever possible.
1030 * POSIX locks are sorted by owner task, then by starting address
1031 */
1032int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1033{
1034 int error;
1035 might_sleep ();
1036 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001037 error = posix_lock_file(filp, fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1039 break;
1040 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1041 if (!error)
1042 continue;
1043
1044 locks_delete_block(fl);
1045 break;
1046 }
1047 return error;
1048}
1049EXPORT_SYMBOL(posix_lock_file_wait);
1050
1051/**
1052 * locks_mandatory_locked - Check for an active lock
1053 * @inode: the file to check
1054 *
1055 * Searches the inode's list of locks to find any POSIX locks which conflict.
1056 * This function is called from locks_verify_locked() only.
1057 */
1058int locks_mandatory_locked(struct inode *inode)
1059{
1060 fl_owner_t owner = current->files;
1061 struct file_lock *fl;
1062
1063 /*
1064 * Search the lock list for this inode for any POSIX locks.
1065 */
1066 lock_kernel();
1067 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1068 if (!IS_POSIX(fl))
1069 continue;
1070 if (fl->fl_owner != owner)
1071 break;
1072 }
1073 unlock_kernel();
1074 return fl ? -EAGAIN : 0;
1075}
1076
1077/**
1078 * locks_mandatory_area - Check for a conflicting lock
1079 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1080 * for shared
1081 * @inode: the file to check
1082 * @filp: how the file was opened (if it was)
1083 * @offset: start of area to check
1084 * @count: length of area to check
1085 *
1086 * Searches the inode's list of locks to find any POSIX locks which conflict.
1087 * This function is called from rw_verify_area() and
1088 * locks_verify_truncate().
1089 */
1090int locks_mandatory_area(int read_write, struct inode *inode,
1091 struct file *filp, loff_t offset,
1092 size_t count)
1093{
1094 struct file_lock fl;
1095 int error;
1096
1097 locks_init_lock(&fl);
1098 fl.fl_owner = current->files;
1099 fl.fl_pid = current->tgid;
1100 fl.fl_file = filp;
1101 fl.fl_flags = FL_POSIX | FL_ACCESS;
1102 if (filp && !(filp->f_flags & O_NONBLOCK))
1103 fl.fl_flags |= FL_SLEEP;
1104 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1105 fl.fl_start = offset;
1106 fl.fl_end = offset + count - 1;
1107
1108 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001109 error = __posix_lock_file(inode, &fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 if (error != -EAGAIN)
1111 break;
1112 if (!(fl.fl_flags & FL_SLEEP))
1113 break;
1114 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1115 if (!error) {
1116 /*
1117 * If we've been sleeping someone might have
1118 * changed the permissions behind our back.
1119 */
1120 if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
1121 continue;
1122 }
1123
1124 locks_delete_block(&fl);
1125 break;
1126 }
1127
1128 return error;
1129}
1130
1131EXPORT_SYMBOL(locks_mandatory_area);
1132
1133/* We already had a lease on this file; just change its type */
1134int lease_modify(struct file_lock **before, int arg)
1135{
1136 struct file_lock *fl = *before;
1137 int error = assign_type(fl, arg);
1138
1139 if (error)
1140 return error;
1141 locks_wake_up_blocks(fl);
1142 if (arg == F_UNLCK)
1143 locks_delete_lock(before);
1144 return 0;
1145}
1146
1147EXPORT_SYMBOL(lease_modify);
1148
1149static void time_out_leases(struct inode *inode)
1150{
1151 struct file_lock **before;
1152 struct file_lock *fl;
1153
1154 before = &inode->i_flock;
1155 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1156 if ((fl->fl_break_time == 0)
1157 || time_before(jiffies, fl->fl_break_time)) {
1158 before = &fl->fl_next;
1159 continue;
1160 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1162 if (fl == *before) /* lease_modify may have freed fl */
1163 before = &fl->fl_next;
1164 }
1165}
1166
1167/**
1168 * __break_lease - revoke all outstanding leases on file
1169 * @inode: the inode of the file to return
1170 * @mode: the open mode (read or write)
1171 *
david m. richter87250dd2007-05-09 16:10:27 -04001172 * break_lease (inlined for speed) has checked there already is at least
1173 * some kind of lock (maybe a lease) on this file. Leases are broken on
1174 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 * specified %O_NONBLOCK to your open().
1176 */
1177int __break_lease(struct inode *inode, unsigned int mode)
1178{
1179 int error = 0, future;
1180 struct file_lock *new_fl, *flock;
1181 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 unsigned long break_time;
1183 int i_have_this_lease = 0;
1184
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001185 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 lock_kernel();
1188
1189 time_out_leases(inode);
1190
1191 flock = inode->i_flock;
1192 if ((flock == NULL) || !IS_LEASE(flock))
1193 goto out;
1194
1195 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1196 if (fl->fl_owner == current->files)
1197 i_have_this_lease = 1;
1198
1199 if (mode & FMODE_WRITE) {
1200 /* If we want write access, we have to revoke any lease. */
1201 future = F_UNLCK | F_INPROGRESS;
1202 } else if (flock->fl_type & F_INPROGRESS) {
1203 /* If the lease is already being broken, we just leave it */
1204 future = flock->fl_type;
1205 } else if (flock->fl_type & F_WRLCK) {
1206 /* Downgrade the exclusive lease to a read-only lease. */
1207 future = F_RDLCK | F_INPROGRESS;
1208 } else {
1209 /* the existing lease was read-only, so we can read too. */
1210 goto out;
1211 }
1212
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001213 if (IS_ERR(new_fl) && !i_have_this_lease
1214 && ((mode & O_NONBLOCK) == 0)) {
1215 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 goto out;
1217 }
1218
1219 break_time = 0;
1220 if (lease_break_time > 0) {
1221 break_time = jiffies + lease_break_time * HZ;
1222 if (break_time == 0)
1223 break_time++; /* so that 0 means no break time */
1224 }
1225
1226 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1227 if (fl->fl_type != future) {
1228 fl->fl_type = future;
1229 fl->fl_break_time = break_time;
1230 /* lease must have lmops break callback */
1231 fl->fl_lmops->fl_break(fl);
1232 }
1233 }
1234
1235 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1236 error = -EWOULDBLOCK;
1237 goto out;
1238 }
1239
1240restart:
1241 break_time = flock->fl_break_time;
1242 if (break_time != 0) {
1243 break_time -= jiffies;
1244 if (break_time == 0)
1245 break_time++;
1246 }
1247 error = locks_block_on_timeout(flock, new_fl, break_time);
1248 if (error >= 0) {
1249 if (error == 0)
1250 time_out_leases(inode);
1251 /* Wait for the next lease that has not been broken yet */
1252 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1253 flock = flock->fl_next) {
1254 if (flock->fl_type & F_INPROGRESS)
1255 goto restart;
1256 }
1257 error = 0;
1258 }
1259
1260out:
1261 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001262 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 locks_free_lock(new_fl);
1264 return error;
1265}
1266
1267EXPORT_SYMBOL(__break_lease);
1268
1269/**
1270 * lease_get_mtime
1271 * @inode: the inode
1272 * @time: pointer to a timespec which will contain the last modified time
1273 *
1274 * This is to force NFS clients to flush their caches for files with
1275 * exclusive leases. The justification is that if someone has an
1276 * exclusive lease, then they could be modifiying it.
1277 */
1278void lease_get_mtime(struct inode *inode, struct timespec *time)
1279{
1280 struct file_lock *flock = inode->i_flock;
1281 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1282 *time = current_fs_time(inode->i_sb);
1283 else
1284 *time = inode->i_mtime;
1285}
1286
1287EXPORT_SYMBOL(lease_get_mtime);
1288
1289/**
1290 * fcntl_getlease - Enquire what lease is currently active
1291 * @filp: the file
1292 *
1293 * The value returned by this function will be one of
1294 * (if no lease break is pending):
1295 *
1296 * %F_RDLCK to indicate a shared lease is held.
1297 *
1298 * %F_WRLCK to indicate an exclusive lease is held.
1299 *
1300 * %F_UNLCK to indicate no lease is held.
1301 *
1302 * (if a lease break is pending):
1303 *
1304 * %F_RDLCK to indicate an exclusive lease needs to be
1305 * changed to a shared lease (or removed).
1306 *
1307 * %F_UNLCK to indicate the lease needs to be removed.
1308 *
1309 * XXX: sfr & willy disagree over whether F_INPROGRESS
1310 * should be returned to userspace.
1311 */
1312int fcntl_getlease(struct file *filp)
1313{
1314 struct file_lock *fl;
1315 int type = F_UNLCK;
1316
1317 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001318 time_out_leases(filp->f_path.dentry->d_inode);
1319 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 fl = fl->fl_next) {
1321 if (fl->fl_file == filp) {
1322 type = fl->fl_type & ~F_INPROGRESS;
1323 break;
1324 }
1325 }
1326 unlock_kernel();
1327 return type;
1328}
1329
1330/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001331 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 * @filp: file pointer
1333 * @arg: type of lease to obtain
1334 * @flp: input - file_lock to use, output - file_lock inserted
1335 *
1336 * The (input) flp->fl_lmops->fl_break function is required
1337 * by break_lease().
1338 *
1339 * Called with kernel lock held.
1340 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001341int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001343 struct file_lock *fl, **before, **my_before = NULL, *lease;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001344 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 struct inode *inode = dentry->d_inode;
1346 int error, rdlease_count = 0, wrlease_count = 0;
1347
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001348 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1349 return -EACCES;
1350 if (!S_ISREG(inode->i_mode))
1351 return -EINVAL;
1352 error = security_file_lock(filp, arg);
1353 if (error)
1354 return error;
1355
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 time_out_leases(inode);
1357
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001358 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001360 lease = *flp;
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 error = -EAGAIN;
1363 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1364 goto out;
1365 if ((arg == F_WRLCK)
1366 && ((atomic_read(&dentry->d_count) > 1)
1367 || (atomic_read(&inode->i_count) > 1)))
1368 goto out;
1369
1370 /*
1371 * At this point, we know that if there is an exclusive
1372 * lease on this file, then we hold it on this filp
1373 * (otherwise our open of this file would have blocked).
1374 * And if we are trying to acquire an exclusive lease,
1375 * then the file is not open by anyone (including us)
1376 * except for this filp.
1377 */
1378 for (before = &inode->i_flock;
1379 ((fl = *before) != NULL) && IS_LEASE(fl);
1380 before = &fl->fl_next) {
1381 if (lease->fl_lmops->fl_mylease(fl, lease))
1382 my_before = before;
1383 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1384 /*
1385 * Someone is in the process of opening this
1386 * file for writing so we may not take an
1387 * exclusive lease on it.
1388 */
1389 wrlease_count++;
1390 else
1391 rdlease_count++;
1392 }
1393
1394 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1395 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1396 goto out;
1397
1398 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001399 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 error = lease->fl_lmops->fl_change(my_before, arg);
1401 goto out;
1402 }
1403
1404 error = 0;
1405 if (arg == F_UNLCK)
1406 goto out;
1407
1408 error = -EINVAL;
1409 if (!leases_enable)
1410 goto out;
1411
Trond Myklebust74361cb2006-08-14 08:54:48 -07001412 error = -ENOMEM;
1413 fl = locks_alloc_lock();
1414 if (fl == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 goto out;
1416
1417 locks_copy_lock(fl, lease);
1418
1419 locks_insert_lock(before, fl);
1420
1421 *flp = fl;
Trond Myklebust74361cb2006-08-14 08:54:48 -07001422 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423out:
1424 return error;
1425}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001426EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
1428 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001429 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 * @filp: file pointer
1431 * @arg: type of lease to obtain
1432 * @lease: file_lock to use
1433 *
1434 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001435 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1436 * break_lease will oops!
1437 *
1438 * This will call the filesystem's setlease file method, if
1439 * defined. Note that there is no getlease method; instead, the
1440 * filesystem setlease method should call back to setlease() to
1441 * add a lease to the inode's lease list, where fcntl_getlease() can
1442 * find it. Since fcntl_getlease() only reports whether the current
1443 * task holds a lease, a cluster filesystem need only do this for
1444 * leases held by processes on this node.
1445 *
1446 * There is also no break_lease method; filesystems that
1447 * handle their own leases shoud break leases themselves from the
1448 * filesystem's open, create, and (on truncate) setattr methods.
1449 *
1450 * Warning: the only current setlease methods exist only to disable
1451 * leases in certain cases. More vfs changes may be required to
1452 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 */
1454
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001455int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 int error;
1458
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001460 if (filp->f_op && filp->f_op->setlease)
1461 error = filp->f_op->setlease(filp, arg, lease);
1462 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001463 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 unlock_kernel();
1465
1466 return error;
1467}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001468EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469
1470/**
1471 * fcntl_setlease - sets a lease on an open file
1472 * @fd: open file descriptor
1473 * @filp: file pointer
1474 * @arg: type of lease to obtain
1475 *
1476 * Call this fcntl to establish a lease on the file.
1477 * Note that you also need to call %F_SETSIG to
1478 * receive a signal when the lease is broken.
1479 */
1480int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1481{
1482 struct file_lock fl, *flp = &fl;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001483 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 struct inode *inode = dentry->d_inode;
1485 int error;
1486
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 locks_init_lock(&fl);
1488 error = lease_init(filp, arg, &fl);
1489 if (error)
1490 return error;
1491
1492 lock_kernel();
1493
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001494 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001495 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 goto out_unlock;
1497
1498 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1499 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001500 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001502 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 time_out_leases(inode);
1504 goto out_unlock;
1505 }
1506
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001507 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508out_unlock:
1509 unlock_kernel();
1510 return error;
1511}
1512
1513/**
1514 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1515 * @filp: The file to apply the lock to
1516 * @fl: The lock to be applied
1517 *
1518 * Add a FLOCK style lock to a file.
1519 */
1520int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1521{
1522 int error;
1523 might_sleep();
1524 for (;;) {
1525 error = flock_lock_file(filp, fl);
1526 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1527 break;
1528 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1529 if (!error)
1530 continue;
1531
1532 locks_delete_block(fl);
1533 break;
1534 }
1535 return error;
1536}
1537
1538EXPORT_SYMBOL(flock_lock_file_wait);
1539
1540/**
1541 * sys_flock: - flock() system call.
1542 * @fd: the file descriptor to lock.
1543 * @cmd: the type of lock to apply.
1544 *
1545 * Apply a %FL_FLOCK style lock to an open file descriptor.
1546 * The @cmd can be one of
1547 *
1548 * %LOCK_SH -- a shared lock.
1549 *
1550 * %LOCK_EX -- an exclusive lock.
1551 *
1552 * %LOCK_UN -- remove an existing lock.
1553 *
1554 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1555 *
1556 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1557 * processes read and write access respectively.
1558 */
1559asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1560{
1561 struct file *filp;
1562 struct file_lock *lock;
1563 int can_sleep, unlock;
1564 int error;
1565
1566 error = -EBADF;
1567 filp = fget(fd);
1568 if (!filp)
1569 goto out;
1570
1571 can_sleep = !(cmd & LOCK_NB);
1572 cmd &= ~LOCK_NB;
1573 unlock = (cmd == LOCK_UN);
1574
1575 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1576 goto out_putf;
1577
1578 error = flock_make_lock(filp, &lock, cmd);
1579 if (error)
1580 goto out_putf;
1581 if (can_sleep)
1582 lock->fl_flags |= FL_SLEEP;
1583
1584 error = security_file_lock(filp, cmd);
1585 if (error)
1586 goto out_free;
1587
1588 if (filp->f_op && filp->f_op->flock)
1589 error = filp->f_op->flock(filp,
1590 (can_sleep) ? F_SETLKW : F_SETLK,
1591 lock);
1592 else
1593 error = flock_lock_file_wait(filp, lock);
1594
1595 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001596 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
1598 out_putf:
1599 fput(filp);
1600 out:
1601 return error;
1602}
1603
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001604/**
1605 * vfs_test_lock - test file byte range lock
1606 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001607 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001608 *
1609 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1610 * setting conf->fl_type to something other than F_UNLCK.
1611 */
1612int vfs_test_lock(struct file *filp, struct file_lock *fl)
1613{
1614 if (filp->f_op && filp->f_op->lock)
1615 return filp->f_op->lock(filp, F_GETLK, fl);
1616 posix_test_lock(filp, fl);
1617 return 0;
1618}
1619EXPORT_SYMBOL_GPL(vfs_test_lock);
1620
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001621static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1622{
1623 flock->l_pid = fl->fl_pid;
1624#if BITS_PER_LONG == 32
1625 /*
1626 * Make sure we can represent the posix lock via
1627 * legacy 32bit flock.
1628 */
1629 if (fl->fl_start > OFFT_OFFSET_MAX)
1630 return -EOVERFLOW;
1631 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1632 return -EOVERFLOW;
1633#endif
1634 flock->l_start = fl->fl_start;
1635 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1636 fl->fl_end - fl->fl_start + 1;
1637 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001638 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001639 return 0;
1640}
1641
1642#if BITS_PER_LONG == 32
1643static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1644{
1645 flock->l_pid = fl->fl_pid;
1646 flock->l_start = fl->fl_start;
1647 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1648 fl->fl_end - fl->fl_start + 1;
1649 flock->l_whence = 0;
1650 flock->l_type = fl->fl_type;
1651}
1652#endif
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654/* Report the first existing lock that would conflict with l.
1655 * This implements the F_GETLK command of fcntl().
1656 */
1657int fcntl_getlk(struct file *filp, struct flock __user *l)
1658{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001659 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 struct flock flock;
1661 int error;
1662
1663 error = -EFAULT;
1664 if (copy_from_user(&flock, l, sizeof(flock)))
1665 goto out;
1666 error = -EINVAL;
1667 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1668 goto out;
1669
1670 error = flock_to_posix_lock(filp, &file_lock, &flock);
1671 if (error)
1672 goto out;
1673
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001674 error = vfs_test_lock(filp, &file_lock);
1675 if (error)
1676 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001678 flock.l_type = file_lock.fl_type;
1679 if (file_lock.fl_type != F_UNLCK) {
1680 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001681 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 }
1684 error = -EFAULT;
1685 if (!copy_to_user(l, &flock, sizeof(flock)))
1686 error = 0;
1687out:
1688 return error;
1689}
1690
Marc Eshel7723ec92007-01-18 15:08:55 -05001691/**
1692 * vfs_lock_file - file byte range lock
1693 * @filp: The file to apply the lock to
1694 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1695 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001696 * @conf: Place to return a copy of the conflicting lock, if found.
1697 *
1698 * A caller that doesn't care about the conflicting lock may pass NULL
1699 * as the final argument.
1700 *
1701 * If the filesystem defines a private ->lock() method, then @conf will
1702 * be left unchanged; so a caller that cares should initialize it to
1703 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001704 *
1705 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1706 * locks, the ->lock() interface may return asynchronously, before the lock has
1707 * been granted or denied by the underlying filesystem, if (and only if)
1708 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1709 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1710 * the request is for a blocking lock. When ->lock() does return asynchronously,
1711 * it must return -EINPROGRESS, and call ->fl_grant() when the lock
1712 * request completes.
1713 * If the request is for non-blocking lock the file system should return
1714 * -EINPROGRESS then try to get the lock and call the callback routine with
1715 * the result. If the request timed out the callback routine will return a
1716 * nonzero return code and the file system should release the lock. The file
1717 * system is also responsible to keep a corresponding posix lock when it
1718 * grants a lock so the VFS can find out which locks are locally held and do
1719 * the correct lock cleanup when required.
1720 * The underlying filesystem must not drop the kernel lock or call
1721 * ->fl_grant() before returning to the caller with a -EINPROGRESS
1722 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001723 */
Marc Eshel150b3932007-01-18 16:15:35 -05001724int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001725{
1726 if (filp->f_op && filp->f_op->lock)
1727 return filp->f_op->lock(filp, cmd, fl);
1728 else
Marc Eshel150b3932007-01-18 16:15:35 -05001729 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001730}
1731EXPORT_SYMBOL_GPL(vfs_lock_file);
1732
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733/* Apply the lock described by l to an open file descriptor.
1734 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1735 */
Peter Staubachc2936212005-07-27 11:45:09 -07001736int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1737 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738{
1739 struct file_lock *file_lock = locks_alloc_lock();
1740 struct flock flock;
1741 struct inode *inode;
1742 int error;
1743
1744 if (file_lock == NULL)
1745 return -ENOLCK;
1746
1747 /*
1748 * This might block, so we do it before checking the inode.
1749 */
1750 error = -EFAULT;
1751 if (copy_from_user(&flock, l, sizeof(flock)))
1752 goto out;
1753
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001754 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755
1756 /* Don't allow mandatory locks on files that may be memory mapped
1757 * and shared.
1758 */
1759 if (IS_MANDLOCK(inode) &&
1760 (inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
1761 mapping_writably_mapped(filp->f_mapping)) {
1762 error = -EAGAIN;
1763 goto out;
1764 }
1765
Peter Staubachc2936212005-07-27 11:45:09 -07001766again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 error = flock_to_posix_lock(filp, file_lock, &flock);
1768 if (error)
1769 goto out;
1770 if (cmd == F_SETLKW) {
1771 file_lock->fl_flags |= FL_SLEEP;
1772 }
1773
1774 error = -EBADF;
1775 switch (flock.l_type) {
1776 case F_RDLCK:
1777 if (!(filp->f_mode & FMODE_READ))
1778 goto out;
1779 break;
1780 case F_WRLCK:
1781 if (!(filp->f_mode & FMODE_WRITE))
1782 goto out;
1783 break;
1784 case F_UNLCK:
1785 break;
1786 default:
1787 error = -EINVAL;
1788 goto out;
1789 }
1790
1791 error = security_file_lock(filp, file_lock->fl_type);
1792 if (error)
1793 goto out;
1794
Marc Eshel7723ec92007-01-18 15:08:55 -05001795 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001796 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001797 if (error != -EAGAIN || cmd == F_SETLK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001799 error = wait_event_interruptible(file_lock->fl_wait,
1800 !file_lock->fl_next);
1801 if (!error)
1802 continue;
1803
1804 locks_delete_block(file_lock);
1805 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 }
1807
Peter Staubachc2936212005-07-27 11:45:09 -07001808 /*
1809 * Attempt to detect a close/fcntl race and recover by
1810 * releasing the lock that was just acquired.
1811 */
1812 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1813 flock.l_type = F_UNLCK;
1814 goto again;
1815 }
1816
1817out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 locks_free_lock(file_lock);
1819 return error;
1820}
1821
1822#if BITS_PER_LONG == 32
1823/* Report the first existing lock that would conflict with l.
1824 * This implements the F_GETLK command of fcntl().
1825 */
1826int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1827{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001828 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 struct flock64 flock;
1830 int error;
1831
1832 error = -EFAULT;
1833 if (copy_from_user(&flock, l, sizeof(flock)))
1834 goto out;
1835 error = -EINVAL;
1836 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1837 goto out;
1838
1839 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1840 if (error)
1841 goto out;
1842
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001843 error = vfs_test_lock(filp, &file_lock);
1844 if (error)
1845 goto out;
1846
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001847 flock.l_type = file_lock.fl_type;
1848 if (file_lock.fl_type != F_UNLCK)
1849 posix_lock_to_flock64(&flock, &file_lock);
1850
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 error = -EFAULT;
1852 if (!copy_to_user(l, &flock, sizeof(flock)))
1853 error = 0;
1854
1855out:
1856 return error;
1857}
1858
1859/* Apply the lock described by l to an open file descriptor.
1860 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1861 */
Peter Staubachc2936212005-07-27 11:45:09 -07001862int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1863 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864{
1865 struct file_lock *file_lock = locks_alloc_lock();
1866 struct flock64 flock;
1867 struct inode *inode;
1868 int error;
1869
1870 if (file_lock == NULL)
1871 return -ENOLCK;
1872
1873 /*
1874 * This might block, so we do it before checking the inode.
1875 */
1876 error = -EFAULT;
1877 if (copy_from_user(&flock, l, sizeof(flock)))
1878 goto out;
1879
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001880 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881
1882 /* Don't allow mandatory locks on files that may be memory mapped
1883 * and shared.
1884 */
1885 if (IS_MANDLOCK(inode) &&
1886 (inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
1887 mapping_writably_mapped(filp->f_mapping)) {
1888 error = -EAGAIN;
1889 goto out;
1890 }
1891
Peter Staubachc2936212005-07-27 11:45:09 -07001892again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 error = flock64_to_posix_lock(filp, file_lock, &flock);
1894 if (error)
1895 goto out;
1896 if (cmd == F_SETLKW64) {
1897 file_lock->fl_flags |= FL_SLEEP;
1898 }
1899
1900 error = -EBADF;
1901 switch (flock.l_type) {
1902 case F_RDLCK:
1903 if (!(filp->f_mode & FMODE_READ))
1904 goto out;
1905 break;
1906 case F_WRLCK:
1907 if (!(filp->f_mode & FMODE_WRITE))
1908 goto out;
1909 break;
1910 case F_UNLCK:
1911 break;
1912 default:
1913 error = -EINVAL;
1914 goto out;
1915 }
1916
1917 error = security_file_lock(filp, file_lock->fl_type);
1918 if (error)
1919 goto out;
1920
Marc Eshel7723ec92007-01-18 15:08:55 -05001921 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001922 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001923 if (error != -EAGAIN || cmd == F_SETLK64)
Peter Staubachc2936212005-07-27 11:45:09 -07001924 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001925 error = wait_event_interruptible(file_lock->fl_wait,
1926 !file_lock->fl_next);
1927 if (!error)
1928 continue;
1929
1930 locks_delete_block(file_lock);
1931 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 }
1933
Peter Staubachc2936212005-07-27 11:45:09 -07001934 /*
1935 * Attempt to detect a close/fcntl race and recover by
1936 * releasing the lock that was just acquired.
1937 */
1938 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1939 flock.l_type = F_UNLCK;
1940 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 }
1942
1943out:
1944 locks_free_lock(file_lock);
1945 return error;
1946}
1947#endif /* BITS_PER_LONG == 32 */
1948
1949/*
1950 * This function is called when the file is being removed
1951 * from the task's fd array. POSIX locks belonging to this task
1952 * are deleted at this time.
1953 */
1954void locks_remove_posix(struct file *filp, fl_owner_t owner)
1955{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001956 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958 /*
1959 * If there are no locks held on this file, we don't need to call
1960 * posix_lock_file(). Another process could be setting a lock on this
1961 * file at the same time, but we wouldn't remove that lock anyway.
1962 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001963 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 return;
1965
1966 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001967 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 lock.fl_start = 0;
1969 lock.fl_end = OFFSET_MAX;
1970 lock.fl_owner = owner;
1971 lock.fl_pid = current->tgid;
1972 lock.fl_file = filp;
1973 lock.fl_ops = NULL;
1974 lock.fl_lmops = NULL;
1975
Marc Eshel150b3932007-01-18 16:15:35 -05001976 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1979 lock.fl_ops->fl_release_private(&lock);
1980}
1981
1982EXPORT_SYMBOL(locks_remove_posix);
1983
1984/*
1985 * This function is called on the last close of an open file.
1986 */
1987void locks_remove_flock(struct file *filp)
1988{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001989 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 struct file_lock *fl;
1991 struct file_lock **before;
1992
1993 if (!inode->i_flock)
1994 return;
1995
1996 if (filp->f_op && filp->f_op->flock) {
1997 struct file_lock fl = {
1998 .fl_pid = current->tgid,
1999 .fl_file = filp,
2000 .fl_flags = FL_FLOCK,
2001 .fl_type = F_UNLCK,
2002 .fl_end = OFFSET_MAX,
2003 };
2004 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002005 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2006 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 }
2008
2009 lock_kernel();
2010 before = &inode->i_flock;
2011
2012 while ((fl = *before) != NULL) {
2013 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002014 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 locks_delete_lock(before);
2016 continue;
2017 }
2018 if (IS_LEASE(fl)) {
2019 lease_modify(before, F_UNLCK);
2020 continue;
2021 }
2022 /* What? */
2023 BUG();
2024 }
2025 before = &fl->fl_next;
2026 }
2027 unlock_kernel();
2028}
2029
2030/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 * posix_unblock_lock - stop waiting for a file lock
2032 * @filp: how the file was opened
2033 * @waiter: the lock which was waiting
2034 *
2035 * lockd needs to block waiting for locks.
2036 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002037int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2039{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002040 int status = 0;
2041
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002043 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002045 else
2046 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002047 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002048 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049}
2050
2051EXPORT_SYMBOL(posix_unblock_lock);
2052
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002053/**
2054 * vfs_cancel_lock - file byte range unblock lock
2055 * @filp: The file to apply the unblock to
2056 * @fl: The lock to be unblocked
2057 *
2058 * Used by lock managers to cancel blocked requests
2059 */
2060int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2061{
2062 if (filp->f_op && filp->f_op->lock)
2063 return filp->f_op->lock(filp, F_CANCELLK, fl);
2064 return 0;
2065}
2066
2067EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2068
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069static void lock_get_status(char* out, struct file_lock *fl, int id, char *pfx)
2070{
2071 struct inode *inode = NULL;
2072
2073 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002074 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
2076 out += sprintf(out, "%d:%s ", id, pfx);
2077 if (IS_POSIX(fl)) {
2078 out += sprintf(out, "%6s %s ",
2079 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2080 (inode == NULL) ? "*NOINODE*" :
2081 (IS_MANDLOCK(inode) &&
2082 (inode->i_mode & (S_IXGRP | S_ISGID)) == S_ISGID) ?
2083 "MANDATORY" : "ADVISORY ");
2084 } else if (IS_FLOCK(fl)) {
2085 if (fl->fl_type & LOCK_MAND) {
2086 out += sprintf(out, "FLOCK MSNFS ");
2087 } else {
2088 out += sprintf(out, "FLOCK ADVISORY ");
2089 }
2090 } else if (IS_LEASE(fl)) {
2091 out += sprintf(out, "LEASE ");
2092 if (fl->fl_type & F_INPROGRESS)
2093 out += sprintf(out, "BREAKING ");
2094 else if (fl->fl_file)
2095 out += sprintf(out, "ACTIVE ");
2096 else
2097 out += sprintf(out, "BREAKER ");
2098 } else {
2099 out += sprintf(out, "UNKNOWN UNKNOWN ");
2100 }
2101 if (fl->fl_type & LOCK_MAND) {
2102 out += sprintf(out, "%s ",
2103 (fl->fl_type & LOCK_READ)
2104 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2105 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2106 } else {
2107 out += sprintf(out, "%s ",
2108 (fl->fl_type & F_INPROGRESS)
2109 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2110 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2111 }
2112 if (inode) {
2113#ifdef WE_CAN_BREAK_LSLK_NOW
2114 out += sprintf(out, "%d %s:%ld ", fl->fl_pid,
2115 inode->i_sb->s_id, inode->i_ino);
2116#else
2117 /* userspace relies on this representation of dev_t ;-( */
2118 out += sprintf(out, "%d %02x:%02x:%ld ", fl->fl_pid,
2119 MAJOR(inode->i_sb->s_dev),
2120 MINOR(inode->i_sb->s_dev), inode->i_ino);
2121#endif
2122 } else {
2123 out += sprintf(out, "%d <none>:0 ", fl->fl_pid);
2124 }
2125 if (IS_POSIX(fl)) {
2126 if (fl->fl_end == OFFSET_MAX)
2127 out += sprintf(out, "%Ld EOF\n", fl->fl_start);
2128 else
2129 out += sprintf(out, "%Ld %Ld\n", fl->fl_start,
2130 fl->fl_end);
2131 } else {
2132 out += sprintf(out, "0 EOF\n");
2133 }
2134}
2135
2136static void move_lock_status(char **p, off_t* pos, off_t offset)
2137{
2138 int len;
2139 len = strlen(*p);
2140 if(*pos >= offset) {
2141 /* the complete line is valid */
2142 *p += len;
2143 *pos += len;
2144 return;
2145 }
2146 if(*pos+len > offset) {
2147 /* use the second part of the line */
2148 int i = offset-*pos;
2149 memmove(*p,*p+i,len-i);
2150 *p += len-i;
2151 *pos += len;
2152 return;
2153 }
2154 /* discard the complete line */
2155 *pos += len;
2156}
2157
2158/**
2159 * get_locks_status - reports lock usage in /proc/locks
2160 * @buffer: address in userspace to write into
2161 * @start: ?
2162 * @offset: how far we are through the buffer
2163 * @length: how much to read
2164 */
2165
2166int get_locks_status(char *buffer, char **start, off_t offset, int length)
2167{
2168 struct list_head *tmp;
2169 char *q = buffer;
2170 off_t pos = 0;
2171 int i = 0;
2172
2173 lock_kernel();
2174 list_for_each(tmp, &file_lock_list) {
2175 struct list_head *btmp;
2176 struct file_lock *fl = list_entry(tmp, struct file_lock, fl_link);
2177 lock_get_status(q, fl, ++i, "");
2178 move_lock_status(&q, &pos, offset);
2179
2180 if(pos >= offset+length)
2181 goto done;
2182
2183 list_for_each(btmp, &fl->fl_block) {
2184 struct file_lock *bfl = list_entry(btmp,
2185 struct file_lock, fl_block);
2186 lock_get_status(q, bfl, i, " ->");
2187 move_lock_status(&q, &pos, offset);
2188
2189 if(pos >= offset+length)
2190 goto done;
2191 }
2192 }
2193done:
2194 unlock_kernel();
2195 *start = buffer;
2196 if(q-buffer < length)
2197 return (q-buffer);
2198 return length;
2199}
2200
2201/**
2202 * lock_may_read - checks that the region is free of locks
2203 * @inode: the inode that is being read
2204 * @start: the first byte to read
2205 * @len: the number of bytes to read
2206 *
2207 * Emulates Windows locking requirements. Whole-file
2208 * mandatory locks (share modes) can prohibit a read and
2209 * byte-range POSIX locks can prohibit a read if they overlap.
2210 *
2211 * N.B. this function is only ever called
2212 * from knfsd and ownership of locks is never checked.
2213 */
2214int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2215{
2216 struct file_lock *fl;
2217 int result = 1;
2218 lock_kernel();
2219 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2220 if (IS_POSIX(fl)) {
2221 if (fl->fl_type == F_RDLCK)
2222 continue;
2223 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2224 continue;
2225 } else if (IS_FLOCK(fl)) {
2226 if (!(fl->fl_type & LOCK_MAND))
2227 continue;
2228 if (fl->fl_type & LOCK_READ)
2229 continue;
2230 } else
2231 continue;
2232 result = 0;
2233 break;
2234 }
2235 unlock_kernel();
2236 return result;
2237}
2238
2239EXPORT_SYMBOL(lock_may_read);
2240
2241/**
2242 * lock_may_write - checks that the region is free of locks
2243 * @inode: the inode that is being written
2244 * @start: the first byte to write
2245 * @len: the number of bytes to write
2246 *
2247 * Emulates Windows locking requirements. Whole-file
2248 * mandatory locks (share modes) can prohibit a write and
2249 * byte-range POSIX locks can prohibit a write if they overlap.
2250 *
2251 * N.B. this function is only ever called
2252 * from knfsd and ownership of locks is never checked.
2253 */
2254int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2255{
2256 struct file_lock *fl;
2257 int result = 1;
2258 lock_kernel();
2259 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2260 if (IS_POSIX(fl)) {
2261 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2262 continue;
2263 } else if (IS_FLOCK(fl)) {
2264 if (!(fl->fl_type & LOCK_MAND))
2265 continue;
2266 if (fl->fl_type & LOCK_WRITE)
2267 continue;
2268 } else
2269 continue;
2270 result = 0;
2271 break;
2272 }
2273 unlock_kernel();
2274 return result;
2275}
2276
2277EXPORT_SYMBOL(lock_may_write);
2278
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279static int __init filelock_init(void)
2280{
2281 filelock_cache = kmem_cache_create("file_lock_cache",
2282 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002283 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 return 0;
2285}
2286
2287core_initcall(filelock_init);