blob: faddccb6336a8ecc6c6ad88edb18ad28108f92d9 [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 Lameter4ba9b9d2007-10-16 23:25:51 -0700202static void init_once(struct kmem_cache *cache, void *foo)
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)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400537 struct file_lock *waiter;
538
539 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 struct file_lock, fl_block);
541 __locks_delete_block(waiter);
542 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
543 waiter->fl_lmops->fl_notify(waiter);
544 else
545 wake_up(&waiter->fl_wait);
546 }
547}
548
549/* Insert file lock fl into an inode's lock list at the position indicated
550 * by pos. At the same time add the lock to the global file lock list.
551 */
552static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
553{
554 list_add(&fl->fl_link, &file_lock_list);
555
556 /* insert into file's list */
557 fl->fl_next = *pos;
558 *pos = fl;
559
560 if (fl->fl_ops && fl->fl_ops->fl_insert)
561 fl->fl_ops->fl_insert(fl);
562}
563
564/*
565 * Delete a lock and then free it.
566 * Wake up processes that are blocked waiting for this lock,
567 * notify the FS that the lock has been cleared and
568 * finally free the lock.
569 */
570static void locks_delete_lock(struct file_lock **thisfl_p)
571{
572 struct file_lock *fl = *thisfl_p;
573
574 *thisfl_p = fl->fl_next;
575 fl->fl_next = NULL;
576 list_del_init(&fl->fl_link);
577
578 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
579 if (fl->fl_fasync != NULL) {
580 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
581 fl->fl_fasync = NULL;
582 }
583
584 if (fl->fl_ops && fl->fl_ops->fl_remove)
585 fl->fl_ops->fl_remove(fl);
586
587 locks_wake_up_blocks(fl);
588 locks_free_lock(fl);
589}
590
591/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
592 * checks for shared/exclusive status of overlapping locks.
593 */
594static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
595{
596 if (sys_fl->fl_type == F_WRLCK)
597 return 1;
598 if (caller_fl->fl_type == F_WRLCK)
599 return 1;
600 return 0;
601}
602
603/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
604 * checking before calling the locks_conflict().
605 */
606static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
607{
608 /* POSIX locks owned by the same process do not conflict with
609 * each other.
610 */
611 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
612 return (0);
613
614 /* Check whether they overlap */
615 if (!locks_overlap(caller_fl, sys_fl))
616 return 0;
617
618 return (locks_conflict(caller_fl, sys_fl));
619}
620
621/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
622 * checking before calling the locks_conflict().
623 */
624static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
625{
626 /* FLOCK locks referring to the same filp do not conflict with
627 * each other.
628 */
629 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
630 return (0);
631 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
632 return 0;
633
634 return (locks_conflict(caller_fl, sys_fl));
635}
636
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400637void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500638posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639{
640 struct file_lock *cfl;
641
642 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800643 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 if (!IS_POSIX(cfl))
645 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400646 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 break;
648 }
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400649 if (cfl)
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500650 __locks_copy_lock(fl, cfl);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400651 else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400652 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400654 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655}
656
657EXPORT_SYMBOL(posix_test_lock);
658
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400659/*
660 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400662 * We attempt to detect deadlocks that are due purely to posix file
663 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400665 * We assume that a task can be waiting for at most one lock at a time.
666 * So for any acquired lock, the process holding that lock may be
667 * waiting on at most one other lock. That lock in turns may be held by
668 * someone waiting for at most one other lock. Given a requested lock
669 * caller_fl which is about to wait for a conflicting lock block_fl, we
670 * follow this chain of waiters to ensure we are not about to create a
671 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400672 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400673 * Since we do this before we ever put a process to sleep on a lock, we
674 * are ensured that there is never a cycle; that is what guarantees that
675 * the while() loop in posix_locks_deadlock() eventually completes.
676 *
677 * Note: the above assumption may not be true when handling lock
678 * requests from a broken NFS client. It may also fail in the presence
679 * of tasks (such as posix threads) sharing the same open file table.
680 *
681 * To handle those cases, we just bail out after a few iterations.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400683
684#define MAX_DEADLK_ITERATIONS 10
685
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400686/* Find a lock that the owner of the given block_fl is blocking on. */
687static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
688{
689 struct file_lock *fl;
690
691 list_for_each_entry(fl, &blocked_list, fl_link) {
692 if (posix_same_owner(fl, block_fl))
693 return fl->fl_next;
694 }
695 return NULL;
696}
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{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400701 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400703 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
704 if (i++ > MAX_DEADLK_ITERATIONS)
705 return 0;
706 if (posix_same_owner(caller_fl, block_fl))
707 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709 return 0;
710}
711
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400713 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400714 *
715 * Note that if called with an FL_EXISTS argument, the caller may determine
716 * whether or not a lock was successfully freed by testing the return
717 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800719static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800721 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800723 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 int error = 0;
725 int found = 0;
726
727 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400728 if (request->fl_flags & FL_ACCESS)
729 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400730
731 if (request->fl_type != F_UNLCK) {
732 error = -ENOMEM;
733 new_fl = locks_alloc_lock();
734 if (new_fl == NULL)
735 goto out;
736 error = 0;
737 }
738
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 for_each_lock(inode, before) {
740 struct file_lock *fl = *before;
741 if (IS_POSIX(fl))
742 break;
743 if (IS_LEASE(fl))
744 continue;
745 if (filp != fl->fl_file)
746 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800747 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto out;
749 found = 1;
750 locks_delete_lock(before);
751 break;
752 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
Trond Myklebustf475ae92006-06-29 16:38:32 -0400754 if (request->fl_type == F_UNLCK) {
755 if ((request->fl_flags & FL_EXISTS) && !found)
756 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800757 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760 /*
761 * If a higher-priority process was blocked on the old file lock,
762 * give it the opportunity to lock the file.
763 */
764 if (found)
765 cond_resched();
766
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400767find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 for_each_lock(inode, before) {
769 struct file_lock *fl = *before;
770 if (IS_POSIX(fl))
771 break;
772 if (IS_LEASE(fl))
773 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800774 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 continue;
776 error = -EAGAIN;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800777 if (request->fl_flags & FL_SLEEP)
778 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 goto out;
780 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400781 if (request->fl_flags & FL_ACCESS)
782 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800783 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700784 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800785 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400786 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788out:
789 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800790 if (new_fl)
791 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 return error;
793}
794
Marc Eshel150b3932007-01-18 16:15:35 -0500795static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796{
797 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700798 struct file_lock *new_fl = NULL;
799 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 struct file_lock *left = NULL;
801 struct file_lock *right = NULL;
802 struct file_lock **before;
803 int error, added = 0;
804
805 /*
806 * We may need two file_lock structures for this operation,
807 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700808 *
809 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700811 if (!(request->fl_flags & FL_ACCESS) &&
812 (request->fl_type != F_UNLCK ||
813 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
814 new_fl = locks_alloc_lock();
815 new_fl2 = locks_alloc_lock();
816 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
818 lock_kernel();
819 if (request->fl_type != F_UNLCK) {
820 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500821 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 if (!IS_POSIX(fl))
823 continue;
824 if (!posix_locks_conflict(request, fl))
825 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800826 if (conflock)
827 locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 error = -EAGAIN;
829 if (!(request->fl_flags & FL_SLEEP))
830 goto out;
831 error = -EDEADLK;
832 if (posix_locks_deadlock(request, fl))
833 goto out;
834 error = -EAGAIN;
835 locks_insert_block(fl, request);
836 goto out;
837 }
838 }
839
840 /* If we're just looking for a conflict, we're done. */
841 error = 0;
842 if (request->fl_flags & FL_ACCESS)
843 goto out;
844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 * Find the first old lock with the same owner as the new lock.
847 */
848
849 before = &inode->i_flock;
850
851 /* First skip locks owned by other processes. */
852 while ((fl = *before) && (!IS_POSIX(fl) ||
853 !posix_same_owner(request, fl))) {
854 before = &fl->fl_next;
855 }
856
857 /* Process locks with this owner. */
858 while ((fl = *before) && posix_same_owner(request, fl)) {
859 /* Detect adjacent or overlapping regions (if same lock type)
860 */
861 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700862 /* In all comparisons of start vs end, use
863 * "start - 1" rather than "end + 1". If end
864 * is OFFSET_MAX, end + 1 will become negative.
865 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 if (fl->fl_end < request->fl_start - 1)
867 goto next_lock;
868 /* If the next lock in the list has entirely bigger
869 * addresses than the new one, insert the lock here.
870 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700871 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 break;
873
874 /* If we come here, the new and old lock are of the
875 * same type and adjacent or overlapping. Make one
876 * lock yielding from the lower start address of both
877 * locks to the higher end address.
878 */
879 if (fl->fl_start > request->fl_start)
880 fl->fl_start = request->fl_start;
881 else
882 request->fl_start = fl->fl_start;
883 if (fl->fl_end < request->fl_end)
884 fl->fl_end = request->fl_end;
885 else
886 request->fl_end = fl->fl_end;
887 if (added) {
888 locks_delete_lock(before);
889 continue;
890 }
891 request = fl;
892 added = 1;
893 }
894 else {
895 /* Processing for different lock types is a bit
896 * more complex.
897 */
898 if (fl->fl_end < request->fl_start)
899 goto next_lock;
900 if (fl->fl_start > request->fl_end)
901 break;
902 if (request->fl_type == F_UNLCK)
903 added = 1;
904 if (fl->fl_start < request->fl_start)
905 left = fl;
906 /* If the next lock in the list has a higher end
907 * address than the new one, insert the new one here.
908 */
909 if (fl->fl_end > request->fl_end) {
910 right = fl;
911 break;
912 }
913 if (fl->fl_start >= request->fl_start) {
914 /* The new lock completely replaces an old
915 * one (This may happen several times).
916 */
917 if (added) {
918 locks_delete_lock(before);
919 continue;
920 }
921 /* Replace the old lock with the new one.
922 * Wake up anybody waiting for the old one,
923 * as the change in lock type might satisfy
924 * their needs.
925 */
926 locks_wake_up_blocks(fl);
927 fl->fl_start = request->fl_start;
928 fl->fl_end = request->fl_end;
929 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500930 locks_release_private(fl);
931 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 request = fl;
933 added = 1;
934 }
935 }
936 /* Go on to next lock.
937 */
938 next_lock:
939 before = &fl->fl_next;
940 }
941
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700942 /*
943 * The above code only modifies existing locks in case of
944 * merging or replacing. If new lock(s) need to be inserted
945 * all modifications are done bellow this, so it's safe yet to
946 * bail out.
947 */
948 error = -ENOLCK; /* "no luck" */
949 if (right && left == right && !new_fl2)
950 goto out;
951
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 error = 0;
953 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400954 if (request->fl_type == F_UNLCK) {
955 if (request->fl_flags & FL_EXISTS)
956 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400958 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700959
960 if (!new_fl) {
961 error = -ENOLCK;
962 goto out;
963 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 locks_copy_lock(new_fl, request);
965 locks_insert_lock(before, new_fl);
966 new_fl = NULL;
967 }
968 if (right) {
969 if (left == right) {
970 /* The new lock breaks the old one in two pieces,
971 * so we have to use the second new lock.
972 */
973 left = new_fl2;
974 new_fl2 = NULL;
975 locks_copy_lock(left, right);
976 locks_insert_lock(before, left);
977 }
978 right->fl_start = request->fl_end + 1;
979 locks_wake_up_blocks(right);
980 }
981 if (left) {
982 left->fl_end = request->fl_start - 1;
983 locks_wake_up_blocks(left);
984 }
985 out:
986 unlock_kernel();
987 /*
988 * Free any unused locks.
989 */
990 if (new_fl)
991 locks_free_lock(new_fl);
992 if (new_fl2)
993 locks_free_lock(new_fl2);
994 return error;
995}
996
997/**
998 * posix_lock_file - Apply a POSIX-style lock to a file
999 * @filp: The file to apply the lock to
1000 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001001 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 *
1003 * Add a POSIX style lock to a file.
1004 * We merge adjacent & overlapping locks whenever possible.
1005 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001006 *
1007 * Note that if called with an FL_EXISTS argument, the caller may determine
1008 * whether or not a lock was successfully freed by testing the return
1009 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 */
Marc Eshel150b3932007-01-18 16:15:35 -05001011int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001012 struct file_lock *conflock)
1013{
Marc Eshel150b3932007-01-18 16:15:35 -05001014 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001015}
Marc Eshel150b3932007-01-18 16:15:35 -05001016EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
1018/**
1019 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1020 * @filp: The file to apply the lock to
1021 * @fl: The lock to be applied
1022 *
1023 * Add a POSIX style lock to a file.
1024 * We merge adjacent & overlapping locks whenever possible.
1025 * POSIX locks are sorted by owner task, then by starting address
1026 */
1027int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1028{
1029 int error;
1030 might_sleep ();
1031 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001032 error = posix_lock_file(filp, fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1034 break;
1035 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1036 if (!error)
1037 continue;
1038
1039 locks_delete_block(fl);
1040 break;
1041 }
1042 return error;
1043}
1044EXPORT_SYMBOL(posix_lock_file_wait);
1045
1046/**
1047 * locks_mandatory_locked - Check for an active lock
1048 * @inode: the file to check
1049 *
1050 * Searches the inode's list of locks to find any POSIX locks which conflict.
1051 * This function is called from locks_verify_locked() only.
1052 */
1053int locks_mandatory_locked(struct inode *inode)
1054{
1055 fl_owner_t owner = current->files;
1056 struct file_lock *fl;
1057
1058 /*
1059 * Search the lock list for this inode for any POSIX locks.
1060 */
1061 lock_kernel();
1062 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1063 if (!IS_POSIX(fl))
1064 continue;
1065 if (fl->fl_owner != owner)
1066 break;
1067 }
1068 unlock_kernel();
1069 return fl ? -EAGAIN : 0;
1070}
1071
1072/**
1073 * locks_mandatory_area - Check for a conflicting lock
1074 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1075 * for shared
1076 * @inode: the file to check
1077 * @filp: how the file was opened (if it was)
1078 * @offset: start of area to check
1079 * @count: length of area to check
1080 *
1081 * Searches the inode's list of locks to find any POSIX locks which conflict.
1082 * This function is called from rw_verify_area() and
1083 * locks_verify_truncate().
1084 */
1085int locks_mandatory_area(int read_write, struct inode *inode,
1086 struct file *filp, loff_t offset,
1087 size_t count)
1088{
1089 struct file_lock fl;
1090 int error;
1091
1092 locks_init_lock(&fl);
1093 fl.fl_owner = current->files;
1094 fl.fl_pid = current->tgid;
1095 fl.fl_file = filp;
1096 fl.fl_flags = FL_POSIX | FL_ACCESS;
1097 if (filp && !(filp->f_flags & O_NONBLOCK))
1098 fl.fl_flags |= FL_SLEEP;
1099 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1100 fl.fl_start = offset;
1101 fl.fl_end = offset + count - 1;
1102
1103 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001104 error = __posix_lock_file(inode, &fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 if (error != -EAGAIN)
1106 break;
1107 if (!(fl.fl_flags & FL_SLEEP))
1108 break;
1109 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1110 if (!error) {
1111 /*
1112 * If we've been sleeping someone might have
1113 * changed the permissions behind our back.
1114 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001115 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 continue;
1117 }
1118
1119 locks_delete_block(&fl);
1120 break;
1121 }
1122
1123 return error;
1124}
1125
1126EXPORT_SYMBOL(locks_mandatory_area);
1127
1128/* We already had a lease on this file; just change its type */
1129int lease_modify(struct file_lock **before, int arg)
1130{
1131 struct file_lock *fl = *before;
1132 int error = assign_type(fl, arg);
1133
1134 if (error)
1135 return error;
1136 locks_wake_up_blocks(fl);
1137 if (arg == F_UNLCK)
1138 locks_delete_lock(before);
1139 return 0;
1140}
1141
1142EXPORT_SYMBOL(lease_modify);
1143
1144static void time_out_leases(struct inode *inode)
1145{
1146 struct file_lock **before;
1147 struct file_lock *fl;
1148
1149 before = &inode->i_flock;
1150 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1151 if ((fl->fl_break_time == 0)
1152 || time_before(jiffies, fl->fl_break_time)) {
1153 before = &fl->fl_next;
1154 continue;
1155 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1157 if (fl == *before) /* lease_modify may have freed fl */
1158 before = &fl->fl_next;
1159 }
1160}
1161
1162/**
1163 * __break_lease - revoke all outstanding leases on file
1164 * @inode: the inode of the file to return
1165 * @mode: the open mode (read or write)
1166 *
david m. richter87250dd2007-05-09 16:10:27 -04001167 * break_lease (inlined for speed) has checked there already is at least
1168 * some kind of lock (maybe a lease) on this file. Leases are broken on
1169 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 * specified %O_NONBLOCK to your open().
1171 */
1172int __break_lease(struct inode *inode, unsigned int mode)
1173{
1174 int error = 0, future;
1175 struct file_lock *new_fl, *flock;
1176 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 unsigned long break_time;
1178 int i_have_this_lease = 0;
1179
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001180 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
1182 lock_kernel();
1183
1184 time_out_leases(inode);
1185
1186 flock = inode->i_flock;
1187 if ((flock == NULL) || !IS_LEASE(flock))
1188 goto out;
1189
1190 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1191 if (fl->fl_owner == current->files)
1192 i_have_this_lease = 1;
1193
1194 if (mode & FMODE_WRITE) {
1195 /* If we want write access, we have to revoke any lease. */
1196 future = F_UNLCK | F_INPROGRESS;
1197 } else if (flock->fl_type & F_INPROGRESS) {
1198 /* If the lease is already being broken, we just leave it */
1199 future = flock->fl_type;
1200 } else if (flock->fl_type & F_WRLCK) {
1201 /* Downgrade the exclusive lease to a read-only lease. */
1202 future = F_RDLCK | F_INPROGRESS;
1203 } else {
1204 /* the existing lease was read-only, so we can read too. */
1205 goto out;
1206 }
1207
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001208 if (IS_ERR(new_fl) && !i_have_this_lease
1209 && ((mode & O_NONBLOCK) == 0)) {
1210 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 goto out;
1212 }
1213
1214 break_time = 0;
1215 if (lease_break_time > 0) {
1216 break_time = jiffies + lease_break_time * HZ;
1217 if (break_time == 0)
1218 break_time++; /* so that 0 means no break time */
1219 }
1220
1221 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1222 if (fl->fl_type != future) {
1223 fl->fl_type = future;
1224 fl->fl_break_time = break_time;
1225 /* lease must have lmops break callback */
1226 fl->fl_lmops->fl_break(fl);
1227 }
1228 }
1229
1230 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1231 error = -EWOULDBLOCK;
1232 goto out;
1233 }
1234
1235restart:
1236 break_time = flock->fl_break_time;
1237 if (break_time != 0) {
1238 break_time -= jiffies;
1239 if (break_time == 0)
1240 break_time++;
1241 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001242 locks_insert_block(flock, new_fl);
1243 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1244 !new_fl->fl_next, break_time);
1245 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 if (error >= 0) {
1247 if (error == 0)
1248 time_out_leases(inode);
1249 /* Wait for the next lease that has not been broken yet */
1250 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1251 flock = flock->fl_next) {
1252 if (flock->fl_type & F_INPROGRESS)
1253 goto restart;
1254 }
1255 error = 0;
1256 }
1257
1258out:
1259 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001260 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 locks_free_lock(new_fl);
1262 return error;
1263}
1264
1265EXPORT_SYMBOL(__break_lease);
1266
1267/**
1268 * lease_get_mtime
1269 * @inode: the inode
1270 * @time: pointer to a timespec which will contain the last modified time
1271 *
1272 * This is to force NFS clients to flush their caches for files with
1273 * exclusive leases. The justification is that if someone has an
1274 * exclusive lease, then they could be modifiying it.
1275 */
1276void lease_get_mtime(struct inode *inode, struct timespec *time)
1277{
1278 struct file_lock *flock = inode->i_flock;
1279 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1280 *time = current_fs_time(inode->i_sb);
1281 else
1282 *time = inode->i_mtime;
1283}
1284
1285EXPORT_SYMBOL(lease_get_mtime);
1286
1287/**
1288 * fcntl_getlease - Enquire what lease is currently active
1289 * @filp: the file
1290 *
1291 * The value returned by this function will be one of
1292 * (if no lease break is pending):
1293 *
1294 * %F_RDLCK to indicate a shared lease is held.
1295 *
1296 * %F_WRLCK to indicate an exclusive lease is held.
1297 *
1298 * %F_UNLCK to indicate no lease is held.
1299 *
1300 * (if a lease break is pending):
1301 *
1302 * %F_RDLCK to indicate an exclusive lease needs to be
1303 * changed to a shared lease (or removed).
1304 *
1305 * %F_UNLCK to indicate the lease needs to be removed.
1306 *
1307 * XXX: sfr & willy disagree over whether F_INPROGRESS
1308 * should be returned to userspace.
1309 */
1310int fcntl_getlease(struct file *filp)
1311{
1312 struct file_lock *fl;
1313 int type = F_UNLCK;
1314
1315 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001316 time_out_leases(filp->f_path.dentry->d_inode);
1317 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 fl = fl->fl_next) {
1319 if (fl->fl_file == filp) {
1320 type = fl->fl_type & ~F_INPROGRESS;
1321 break;
1322 }
1323 }
1324 unlock_kernel();
1325 return type;
1326}
1327
1328/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001329 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 * @filp: file pointer
1331 * @arg: type of lease to obtain
1332 * @flp: input - file_lock to use, output - file_lock inserted
1333 *
1334 * The (input) flp->fl_lmops->fl_break function is required
1335 * by break_lease().
1336 *
1337 * Called with kernel lock held.
1338 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001339int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001341 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001342 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001343 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 struct inode *inode = dentry->d_inode;
1345 int error, rdlease_count = 0, wrlease_count = 0;
1346
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001347 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1348 return -EACCES;
1349 if (!S_ISREG(inode->i_mode))
1350 return -EINVAL;
1351 error = security_file_lock(filp, arg);
1352 if (error)
1353 return error;
1354
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 time_out_leases(inode);
1356
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001357 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001359 lease = *flp;
1360
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 error = -EAGAIN;
1362 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1363 goto out;
1364 if ((arg == F_WRLCK)
1365 && ((atomic_read(&dentry->d_count) > 1)
1366 || (atomic_read(&inode->i_count) > 1)))
1367 goto out;
1368
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001369 error = -ENOMEM;
1370 new_fl = locks_alloc_lock();
1371 if (new_fl == NULL)
1372 goto out;
1373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 /*
1375 * At this point, we know that if there is an exclusive
1376 * lease on this file, then we hold it on this filp
1377 * (otherwise our open of this file would have blocked).
1378 * And if we are trying to acquire an exclusive lease,
1379 * then the file is not open by anyone (including us)
1380 * except for this filp.
1381 */
1382 for (before = &inode->i_flock;
1383 ((fl = *before) != NULL) && IS_LEASE(fl);
1384 before = &fl->fl_next) {
1385 if (lease->fl_lmops->fl_mylease(fl, lease))
1386 my_before = before;
1387 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1388 /*
1389 * Someone is in the process of opening this
1390 * file for writing so we may not take an
1391 * exclusive lease on it.
1392 */
1393 wrlease_count++;
1394 else
1395 rdlease_count++;
1396 }
1397
1398 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1399 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1400 goto out;
1401
1402 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001403 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 error = lease->fl_lmops->fl_change(my_before, arg);
1405 goto out;
1406 }
1407
1408 error = 0;
1409 if (arg == F_UNLCK)
1410 goto out;
1411
1412 error = -EINVAL;
1413 if (!leases_enable)
1414 goto out;
1415
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001416 locks_copy_lock(new_fl, lease);
1417 locks_insert_lock(before, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001419 *flp = new_fl;
1420 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422out:
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001423 if (new_fl != NULL)
1424 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 return error;
1426}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001427EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
1429 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001430 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 * @filp: file pointer
1432 * @arg: type of lease to obtain
1433 * @lease: file_lock to use
1434 *
1435 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001436 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1437 * break_lease will oops!
1438 *
1439 * This will call the filesystem's setlease file method, if
1440 * defined. Note that there is no getlease method; instead, the
1441 * filesystem setlease method should call back to setlease() to
1442 * add a lease to the inode's lease list, where fcntl_getlease() can
1443 * find it. Since fcntl_getlease() only reports whether the current
1444 * task holds a lease, a cluster filesystem need only do this for
1445 * leases held by processes on this node.
1446 *
1447 * There is also no break_lease method; filesystems that
1448 * handle their own leases shoud break leases themselves from the
1449 * filesystem's open, create, and (on truncate) setattr methods.
1450 *
1451 * Warning: the only current setlease methods exist only to disable
1452 * leases in certain cases. More vfs changes may be required to
1453 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 */
1455
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001456int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 int error;
1459
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001461 if (filp->f_op && filp->f_op->setlease)
1462 error = filp->f_op->setlease(filp, arg, lease);
1463 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001464 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 unlock_kernel();
1466
1467 return error;
1468}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001469EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470
1471/**
1472 * fcntl_setlease - sets a lease on an open file
1473 * @fd: open file descriptor
1474 * @filp: file pointer
1475 * @arg: type of lease to obtain
1476 *
1477 * Call this fcntl to establish a lease on the file.
1478 * Note that you also need to call %F_SETSIG to
1479 * receive a signal when the lease is broken.
1480 */
1481int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1482{
1483 struct file_lock fl, *flp = &fl;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001484 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 struct inode *inode = dentry->d_inode;
1486 int error;
1487
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 locks_init_lock(&fl);
1489 error = lease_init(filp, arg, &fl);
1490 if (error)
1491 return error;
1492
1493 lock_kernel();
1494
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001495 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001496 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 goto out_unlock;
1498
1499 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1500 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001501 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001503 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 time_out_leases(inode);
1505 goto out_unlock;
1506 }
1507
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001508 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509out_unlock:
1510 unlock_kernel();
1511 return error;
1512}
1513
1514/**
1515 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1516 * @filp: The file to apply the lock to
1517 * @fl: The lock to be applied
1518 *
1519 * Add a FLOCK style lock to a file.
1520 */
1521int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1522{
1523 int error;
1524 might_sleep();
1525 for (;;) {
1526 error = flock_lock_file(filp, fl);
1527 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1528 break;
1529 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1530 if (!error)
1531 continue;
1532
1533 locks_delete_block(fl);
1534 break;
1535 }
1536 return error;
1537}
1538
1539EXPORT_SYMBOL(flock_lock_file_wait);
1540
1541/**
1542 * sys_flock: - flock() system call.
1543 * @fd: the file descriptor to lock.
1544 * @cmd: the type of lock to apply.
1545 *
1546 * Apply a %FL_FLOCK style lock to an open file descriptor.
1547 * The @cmd can be one of
1548 *
1549 * %LOCK_SH -- a shared lock.
1550 *
1551 * %LOCK_EX -- an exclusive lock.
1552 *
1553 * %LOCK_UN -- remove an existing lock.
1554 *
1555 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1556 *
1557 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1558 * processes read and write access respectively.
1559 */
1560asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1561{
1562 struct file *filp;
1563 struct file_lock *lock;
1564 int can_sleep, unlock;
1565 int error;
1566
1567 error = -EBADF;
1568 filp = fget(fd);
1569 if (!filp)
1570 goto out;
1571
1572 can_sleep = !(cmd & LOCK_NB);
1573 cmd &= ~LOCK_NB;
1574 unlock = (cmd == LOCK_UN);
1575
1576 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1577 goto out_putf;
1578
1579 error = flock_make_lock(filp, &lock, cmd);
1580 if (error)
1581 goto out_putf;
1582 if (can_sleep)
1583 lock->fl_flags |= FL_SLEEP;
1584
1585 error = security_file_lock(filp, cmd);
1586 if (error)
1587 goto out_free;
1588
1589 if (filp->f_op && filp->f_op->flock)
1590 error = filp->f_op->flock(filp,
1591 (can_sleep) ? F_SETLKW : F_SETLK,
1592 lock);
1593 else
1594 error = flock_lock_file_wait(filp, lock);
1595
1596 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001597 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599 out_putf:
1600 fput(filp);
1601 out:
1602 return error;
1603}
1604
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001605/**
1606 * vfs_test_lock - test file byte range lock
1607 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001608 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001609 *
1610 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1611 * setting conf->fl_type to something other than F_UNLCK.
1612 */
1613int vfs_test_lock(struct file *filp, struct file_lock *fl)
1614{
1615 if (filp->f_op && filp->f_op->lock)
1616 return filp->f_op->lock(filp, F_GETLK, fl);
1617 posix_test_lock(filp, fl);
1618 return 0;
1619}
1620EXPORT_SYMBOL_GPL(vfs_test_lock);
1621
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001622static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1623{
1624 flock->l_pid = fl->fl_pid;
1625#if BITS_PER_LONG == 32
1626 /*
1627 * Make sure we can represent the posix lock via
1628 * legacy 32bit flock.
1629 */
1630 if (fl->fl_start > OFFT_OFFSET_MAX)
1631 return -EOVERFLOW;
1632 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1633 return -EOVERFLOW;
1634#endif
1635 flock->l_start = fl->fl_start;
1636 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1637 fl->fl_end - fl->fl_start + 1;
1638 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001639 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001640 return 0;
1641}
1642
1643#if BITS_PER_LONG == 32
1644static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1645{
1646 flock->l_pid = fl->fl_pid;
1647 flock->l_start = fl->fl_start;
1648 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1649 fl->fl_end - fl->fl_start + 1;
1650 flock->l_whence = 0;
1651 flock->l_type = fl->fl_type;
1652}
1653#endif
1654
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655/* Report the first existing lock that would conflict with l.
1656 * This implements the F_GETLK command of fcntl().
1657 */
1658int fcntl_getlk(struct file *filp, struct flock __user *l)
1659{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001660 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 struct flock flock;
1662 int error;
1663
1664 error = -EFAULT;
1665 if (copy_from_user(&flock, l, sizeof(flock)))
1666 goto out;
1667 error = -EINVAL;
1668 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1669 goto out;
1670
1671 error = flock_to_posix_lock(filp, &file_lock, &flock);
1672 if (error)
1673 goto out;
1674
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001675 error = vfs_test_lock(filp, &file_lock);
1676 if (error)
1677 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001679 flock.l_type = file_lock.fl_type;
1680 if (file_lock.fl_type != F_UNLCK) {
1681 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001682 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 }
1685 error = -EFAULT;
1686 if (!copy_to_user(l, &flock, sizeof(flock)))
1687 error = 0;
1688out:
1689 return error;
1690}
1691
Marc Eshel7723ec92007-01-18 15:08:55 -05001692/**
1693 * vfs_lock_file - file byte range lock
1694 * @filp: The file to apply the lock to
1695 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1696 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001697 * @conf: Place to return a copy of the conflicting lock, if found.
1698 *
1699 * A caller that doesn't care about the conflicting lock may pass NULL
1700 * as the final argument.
1701 *
1702 * If the filesystem defines a private ->lock() method, then @conf will
1703 * be left unchanged; so a caller that cares should initialize it to
1704 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001705 *
1706 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1707 * locks, the ->lock() interface may return asynchronously, before the lock has
1708 * been granted or denied by the underlying filesystem, if (and only if)
1709 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1710 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1711 * the request is for a blocking lock. When ->lock() does return asynchronously,
1712 * it must return -EINPROGRESS, and call ->fl_grant() when the lock
1713 * request completes.
1714 * If the request is for non-blocking lock the file system should return
1715 * -EINPROGRESS then try to get the lock and call the callback routine with
1716 * the result. If the request timed out the callback routine will return a
1717 * nonzero return code and the file system should release the lock. The file
1718 * system is also responsible to keep a corresponding posix lock when it
1719 * grants a lock so the VFS can find out which locks are locally held and do
1720 * the correct lock cleanup when required.
1721 * The underlying filesystem must not drop the kernel lock or call
1722 * ->fl_grant() before returning to the caller with a -EINPROGRESS
1723 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001724 */
Marc Eshel150b3932007-01-18 16:15:35 -05001725int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001726{
1727 if (filp->f_op && filp->f_op->lock)
1728 return filp->f_op->lock(filp, cmd, fl);
1729 else
Marc Eshel150b3932007-01-18 16:15:35 -05001730 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001731}
1732EXPORT_SYMBOL_GPL(vfs_lock_file);
1733
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734/* Apply the lock described by l to an open file descriptor.
1735 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1736 */
Peter Staubachc2936212005-07-27 11:45:09 -07001737int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1738 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739{
1740 struct file_lock *file_lock = locks_alloc_lock();
1741 struct flock flock;
1742 struct inode *inode;
1743 int error;
1744
1745 if (file_lock == NULL)
1746 return -ENOLCK;
1747
1748 /*
1749 * This might block, so we do it before checking the inode.
1750 */
1751 error = -EFAULT;
1752 if (copy_from_user(&flock, l, sizeof(flock)))
1753 goto out;
1754
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001755 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756
1757 /* Don't allow mandatory locks on files that may be memory mapped
1758 * and shared.
1759 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001760 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 error = -EAGAIN;
1762 goto out;
1763 }
1764
Peter Staubachc2936212005-07-27 11:45:09 -07001765again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 error = flock_to_posix_lock(filp, file_lock, &flock);
1767 if (error)
1768 goto out;
1769 if (cmd == F_SETLKW) {
1770 file_lock->fl_flags |= FL_SLEEP;
1771 }
1772
1773 error = -EBADF;
1774 switch (flock.l_type) {
1775 case F_RDLCK:
1776 if (!(filp->f_mode & FMODE_READ))
1777 goto out;
1778 break;
1779 case F_WRLCK:
1780 if (!(filp->f_mode & FMODE_WRITE))
1781 goto out;
1782 break;
1783 case F_UNLCK:
1784 break;
1785 default:
1786 error = -EINVAL;
1787 goto out;
1788 }
1789
1790 error = security_file_lock(filp, file_lock->fl_type);
1791 if (error)
1792 goto out;
1793
Marc Eshel7723ec92007-01-18 15:08:55 -05001794 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001795 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001796 if (error != -EAGAIN || cmd == F_SETLK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001798 error = wait_event_interruptible(file_lock->fl_wait,
1799 !file_lock->fl_next);
1800 if (!error)
1801 continue;
1802
1803 locks_delete_block(file_lock);
1804 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 }
1806
Peter Staubachc2936212005-07-27 11:45:09 -07001807 /*
1808 * Attempt to detect a close/fcntl race and recover by
1809 * releasing the lock that was just acquired.
1810 */
1811 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1812 flock.l_type = F_UNLCK;
1813 goto again;
1814 }
1815
1816out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 locks_free_lock(file_lock);
1818 return error;
1819}
1820
1821#if BITS_PER_LONG == 32
1822/* Report the first existing lock that would conflict with l.
1823 * This implements the F_GETLK command of fcntl().
1824 */
1825int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1826{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001827 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 struct flock64 flock;
1829 int error;
1830
1831 error = -EFAULT;
1832 if (copy_from_user(&flock, l, sizeof(flock)))
1833 goto out;
1834 error = -EINVAL;
1835 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1836 goto out;
1837
1838 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1839 if (error)
1840 goto out;
1841
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001842 error = vfs_test_lock(filp, &file_lock);
1843 if (error)
1844 goto out;
1845
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001846 flock.l_type = file_lock.fl_type;
1847 if (file_lock.fl_type != F_UNLCK)
1848 posix_lock_to_flock64(&flock, &file_lock);
1849
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 error = -EFAULT;
1851 if (!copy_to_user(l, &flock, sizeof(flock)))
1852 error = 0;
1853
1854out:
1855 return error;
1856}
1857
1858/* Apply the lock described by l to an open file descriptor.
1859 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1860 */
Peter Staubachc2936212005-07-27 11:45:09 -07001861int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1862 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863{
1864 struct file_lock *file_lock = locks_alloc_lock();
1865 struct flock64 flock;
1866 struct inode *inode;
1867 int error;
1868
1869 if (file_lock == NULL)
1870 return -ENOLCK;
1871
1872 /*
1873 * This might block, so we do it before checking the inode.
1874 */
1875 error = -EFAULT;
1876 if (copy_from_user(&flock, l, sizeof(flock)))
1877 goto out;
1878
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001879 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
1881 /* Don't allow mandatory locks on files that may be memory mapped
1882 * and shared.
1883 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001884 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 error = -EAGAIN;
1886 goto out;
1887 }
1888
Peter Staubachc2936212005-07-27 11:45:09 -07001889again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 error = flock64_to_posix_lock(filp, file_lock, &flock);
1891 if (error)
1892 goto out;
1893 if (cmd == F_SETLKW64) {
1894 file_lock->fl_flags |= FL_SLEEP;
1895 }
1896
1897 error = -EBADF;
1898 switch (flock.l_type) {
1899 case F_RDLCK:
1900 if (!(filp->f_mode & FMODE_READ))
1901 goto out;
1902 break;
1903 case F_WRLCK:
1904 if (!(filp->f_mode & FMODE_WRITE))
1905 goto out;
1906 break;
1907 case F_UNLCK:
1908 break;
1909 default:
1910 error = -EINVAL;
1911 goto out;
1912 }
1913
1914 error = security_file_lock(filp, file_lock->fl_type);
1915 if (error)
1916 goto out;
1917
Marc Eshel7723ec92007-01-18 15:08:55 -05001918 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001919 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001920 if (error != -EAGAIN || cmd == F_SETLK64)
Peter Staubachc2936212005-07-27 11:45:09 -07001921 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001922 error = wait_event_interruptible(file_lock->fl_wait,
1923 !file_lock->fl_next);
1924 if (!error)
1925 continue;
1926
1927 locks_delete_block(file_lock);
1928 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 }
1930
Peter Staubachc2936212005-07-27 11:45:09 -07001931 /*
1932 * Attempt to detect a close/fcntl race and recover by
1933 * releasing the lock that was just acquired.
1934 */
1935 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1936 flock.l_type = F_UNLCK;
1937 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 }
1939
1940out:
1941 locks_free_lock(file_lock);
1942 return error;
1943}
1944#endif /* BITS_PER_LONG == 32 */
1945
1946/*
1947 * This function is called when the file is being removed
1948 * from the task's fd array. POSIX locks belonging to this task
1949 * are deleted at this time.
1950 */
1951void locks_remove_posix(struct file *filp, fl_owner_t owner)
1952{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001953 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954
1955 /*
1956 * If there are no locks held on this file, we don't need to call
1957 * posix_lock_file(). Another process could be setting a lock on this
1958 * file at the same time, but we wouldn't remove that lock anyway.
1959 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001960 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 return;
1962
1963 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001964 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 lock.fl_start = 0;
1966 lock.fl_end = OFFSET_MAX;
1967 lock.fl_owner = owner;
1968 lock.fl_pid = current->tgid;
1969 lock.fl_file = filp;
1970 lock.fl_ops = NULL;
1971 lock.fl_lmops = NULL;
1972
Marc Eshel150b3932007-01-18 16:15:35 -05001973 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1976 lock.fl_ops->fl_release_private(&lock);
1977}
1978
1979EXPORT_SYMBOL(locks_remove_posix);
1980
1981/*
1982 * This function is called on the last close of an open file.
1983 */
1984void locks_remove_flock(struct file *filp)
1985{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001986 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 struct file_lock *fl;
1988 struct file_lock **before;
1989
1990 if (!inode->i_flock)
1991 return;
1992
1993 if (filp->f_op && filp->f_op->flock) {
1994 struct file_lock fl = {
1995 .fl_pid = current->tgid,
1996 .fl_file = filp,
1997 .fl_flags = FL_FLOCK,
1998 .fl_type = F_UNLCK,
1999 .fl_end = OFFSET_MAX,
2000 };
2001 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002002 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2003 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 }
2005
2006 lock_kernel();
2007 before = &inode->i_flock;
2008
2009 while ((fl = *before) != NULL) {
2010 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002011 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 locks_delete_lock(before);
2013 continue;
2014 }
2015 if (IS_LEASE(fl)) {
2016 lease_modify(before, F_UNLCK);
2017 continue;
2018 }
2019 /* What? */
2020 BUG();
2021 }
2022 before = &fl->fl_next;
2023 }
2024 unlock_kernel();
2025}
2026
2027/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 * posix_unblock_lock - stop waiting for a file lock
2029 * @filp: how the file was opened
2030 * @waiter: the lock which was waiting
2031 *
2032 * lockd needs to block waiting for locks.
2033 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002034int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2036{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002037 int status = 0;
2038
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002040 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002042 else
2043 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002044 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002045 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046}
2047
2048EXPORT_SYMBOL(posix_unblock_lock);
2049
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002050/**
2051 * vfs_cancel_lock - file byte range unblock lock
2052 * @filp: The file to apply the unblock to
2053 * @fl: The lock to be unblocked
2054 *
2055 * Used by lock managers to cancel blocked requests
2056 */
2057int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2058{
2059 if (filp->f_op && filp->f_op->lock)
2060 return filp->f_op->lock(filp, F_CANCELLK, fl);
2061 return 0;
2062}
2063
2064EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2065
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002066#ifdef CONFIG_PROC_FS
2067#include <linux/seq_file.h>
2068
2069static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2070 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071{
2072 struct inode *inode = NULL;
2073
2074 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002075 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002077 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002079 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2081 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002082 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 } else if (IS_FLOCK(fl)) {
2084 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002085 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002087 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 }
2089 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002090 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002092 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002094 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002096 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002098 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 }
2100 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002101 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 (fl->fl_type & LOCK_READ)
2103 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2104 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2105 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002106 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 (fl->fl_type & F_INPROGRESS)
2108 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2109 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2110 }
2111 if (inode) {
2112#ifdef WE_CAN_BREAK_LSLK_NOW
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002113 seq_printf(f, "%d %s:%ld ", fl->fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 inode->i_sb->s_id, inode->i_ino);
2115#else
2116 /* userspace relies on this representation of dev_t ;-( */
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002117 seq_printf(f, "%d %02x:%02x:%ld ", fl->fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 MAJOR(inode->i_sb->s_dev),
2119 MINOR(inode->i_sb->s_dev), inode->i_ino);
2120#endif
2121 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002122 seq_printf(f, "%d <none>:0 ", fl->fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 }
2124 if (IS_POSIX(fl)) {
2125 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002126 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002128 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002130 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 }
2132}
2133
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002134static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002136 struct file_lock *fl, *bfl;
2137
2138 fl = list_entry(v, struct file_lock, fl_link);
2139
2140 lock_get_status(f, fl, (long)f->private, "");
2141
2142 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2143 lock_get_status(f, bfl, (long)f->private, " ->");
2144
2145 f->private++;
2146 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147}
2148
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002149static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002152 f->private = (void *)1;
2153 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154}
2155
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002156static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2157{
2158 return seq_list_next(v, &file_lock_list, pos);
2159}
2160
2161static void locks_stop(struct seq_file *f, void *v)
2162{
2163 unlock_kernel();
2164}
2165
2166struct seq_operations locks_seq_operations = {
2167 .start = locks_start,
2168 .next = locks_next,
2169 .stop = locks_stop,
2170 .show = locks_show,
2171};
2172#endif
2173
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174/**
2175 * lock_may_read - checks that the region is free of locks
2176 * @inode: the inode that is being read
2177 * @start: the first byte to read
2178 * @len: the number of bytes to read
2179 *
2180 * Emulates Windows locking requirements. Whole-file
2181 * mandatory locks (share modes) can prohibit a read and
2182 * byte-range POSIX locks can prohibit a read if they overlap.
2183 *
2184 * N.B. this function is only ever called
2185 * from knfsd and ownership of locks is never checked.
2186 */
2187int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2188{
2189 struct file_lock *fl;
2190 int result = 1;
2191 lock_kernel();
2192 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2193 if (IS_POSIX(fl)) {
2194 if (fl->fl_type == F_RDLCK)
2195 continue;
2196 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2197 continue;
2198 } else if (IS_FLOCK(fl)) {
2199 if (!(fl->fl_type & LOCK_MAND))
2200 continue;
2201 if (fl->fl_type & LOCK_READ)
2202 continue;
2203 } else
2204 continue;
2205 result = 0;
2206 break;
2207 }
2208 unlock_kernel();
2209 return result;
2210}
2211
2212EXPORT_SYMBOL(lock_may_read);
2213
2214/**
2215 * lock_may_write - checks that the region is free of locks
2216 * @inode: the inode that is being written
2217 * @start: the first byte to write
2218 * @len: the number of bytes to write
2219 *
2220 * Emulates Windows locking requirements. Whole-file
2221 * mandatory locks (share modes) can prohibit a write and
2222 * byte-range POSIX locks can prohibit a write if they overlap.
2223 *
2224 * N.B. this function is only ever called
2225 * from knfsd and ownership of locks is never checked.
2226 */
2227int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2228{
2229 struct file_lock *fl;
2230 int result = 1;
2231 lock_kernel();
2232 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2233 if (IS_POSIX(fl)) {
2234 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2235 continue;
2236 } else if (IS_FLOCK(fl)) {
2237 if (!(fl->fl_type & LOCK_MAND))
2238 continue;
2239 if (fl->fl_type & LOCK_WRITE)
2240 continue;
2241 } else
2242 continue;
2243 result = 0;
2244 break;
2245 }
2246 unlock_kernel();
2247 return result;
2248}
2249
2250EXPORT_SYMBOL(lock_may_write);
2251
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252static int __init filelock_init(void)
2253{
2254 filelock_cache = kmem_cache_create("file_lock_cache",
2255 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002256 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257 return 0;
2258}
2259
2260core_initcall(filelock_init);