blob: 7f9a3ea47418cf598b3084be4efd96b9884a10bf [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)) {
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
637static int interruptible_sleep_on_locked(wait_queue_head_t *fl_wait, int timeout)
638{
639 int result = 0;
640 DECLARE_WAITQUEUE(wait, current);
641
642 __set_current_state(TASK_INTERRUPTIBLE);
643 add_wait_queue(fl_wait, &wait);
644 if (timeout == 0)
645 schedule();
646 else
647 result = schedule_timeout(timeout);
648 if (signal_pending(current))
649 result = -ERESTARTSYS;
650 remove_wait_queue(fl_wait, &wait);
651 __set_current_state(TASK_RUNNING);
652 return result;
653}
654
655static int locks_block_on_timeout(struct file_lock *blocker, struct file_lock *waiter, int time)
656{
657 int result;
658 locks_insert_block(blocker, waiter);
659 result = interruptible_sleep_on_locked(&waiter->fl_wait, time);
660 __locks_delete_block(waiter);
661 return result;
662}
663
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400664void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500665posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666{
667 struct file_lock *cfl;
668
669 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800670 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 if (!IS_POSIX(cfl))
672 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400673 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 break;
675 }
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400676 if (cfl)
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500677 __locks_copy_lock(fl, cfl);
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400678 else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400679 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400681 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682}
683
684EXPORT_SYMBOL(posix_test_lock);
685
686/* This function tests for deadlock condition before putting a process to
687 * sleep. The detection scheme is no longer recursive. Recursive was neat,
688 * but dangerous - we risked stack corruption if the lock data was bad, or
689 * if the recursion was too deep for any other reason.
690 *
691 * We rely on the fact that a task can only be on one lock's wait queue
692 * at a time. When we find blocked_task on a wait queue we can re-search
693 * with blocked_task equal to that queue's owner, until either blocked_task
694 * isn't found, or blocked_task is found on a queue owned by my_task.
695 *
696 * Note: the above assumption may not be true when handling lock requests
697 * from a broken NFS client. But broken NFS clients have a lot more to
698 * worry about than proper deadlock detection anyway... --okir
699 */
Adrian Bunkb0904e12006-06-23 02:05:13 -0700700static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 struct file_lock *block_fl)
702{
Matthias Kaehlcke094f2822007-10-02 11:21:34 -0700703 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704
705next_task:
706 if (posix_same_owner(caller_fl, block_fl))
707 return 1;
Matthias Kaehlcke094f2822007-10-02 11:21:34 -0700708 list_for_each_entry(fl, &blocked_list, fl_link) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 if (posix_same_owner(fl, block_fl)) {
710 fl = fl->fl_next;
711 block_fl = fl;
712 goto next_task;
713 }
714 }
715 return 0;
716}
717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400719 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400720 *
721 * Note that if called with an FL_EXISTS argument, the caller may determine
722 * whether or not a lock was successfully freed by testing the return
723 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800725static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800727 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800729 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 int error = 0;
731 int found = 0;
732
733 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400734 if (request->fl_flags & FL_ACCESS)
735 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400736
737 if (request->fl_type != F_UNLCK) {
738 error = -ENOMEM;
739 new_fl = locks_alloc_lock();
740 if (new_fl == NULL)
741 goto out;
742 error = 0;
743 }
744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 for_each_lock(inode, before) {
746 struct file_lock *fl = *before;
747 if (IS_POSIX(fl))
748 break;
749 if (IS_LEASE(fl))
750 continue;
751 if (filp != fl->fl_file)
752 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800753 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 goto out;
755 found = 1;
756 locks_delete_lock(before);
757 break;
758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
Trond Myklebustf475ae92006-06-29 16:38:32 -0400760 if (request->fl_type == F_UNLCK) {
761 if ((request->fl_flags & FL_EXISTS) && !found)
762 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800763 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
766 /*
767 * If a higher-priority process was blocked on the old file lock,
768 * give it the opportunity to lock the file.
769 */
770 if (found)
771 cond_resched();
772
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400773find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 for_each_lock(inode, before) {
775 struct file_lock *fl = *before;
776 if (IS_POSIX(fl))
777 break;
778 if (IS_LEASE(fl))
779 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800780 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 continue;
782 error = -EAGAIN;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800783 if (request->fl_flags & FL_SLEEP)
784 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 goto out;
786 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400787 if (request->fl_flags & FL_ACCESS)
788 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800789 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700790 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800791 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400792 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794out:
795 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800796 if (new_fl)
797 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 return error;
799}
800
Marc Eshel150b3932007-01-18 16:15:35 -0500801static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802{
803 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700804 struct file_lock *new_fl = NULL;
805 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 struct file_lock *left = NULL;
807 struct file_lock *right = NULL;
808 struct file_lock **before;
809 int error, added = 0;
810
811 /*
812 * We may need two file_lock structures for this operation,
813 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700814 *
815 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700817 if (!(request->fl_flags & FL_ACCESS) &&
818 (request->fl_type != F_UNLCK ||
819 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
820 new_fl = locks_alloc_lock();
821 new_fl2 = locks_alloc_lock();
822 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
824 lock_kernel();
825 if (request->fl_type != F_UNLCK) {
826 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500827 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 if (!IS_POSIX(fl))
829 continue;
830 if (!posix_locks_conflict(request, fl))
831 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800832 if (conflock)
833 locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 error = -EAGAIN;
835 if (!(request->fl_flags & FL_SLEEP))
836 goto out;
837 error = -EDEADLK;
838 if (posix_locks_deadlock(request, fl))
839 goto out;
840 error = -EAGAIN;
841 locks_insert_block(fl, request);
842 goto out;
843 }
844 }
845
846 /* If we're just looking for a conflict, we're done. */
847 error = 0;
848 if (request->fl_flags & FL_ACCESS)
849 goto out;
850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 * Find the first old lock with the same owner as the new lock.
853 */
854
855 before = &inode->i_flock;
856
857 /* First skip locks owned by other processes. */
858 while ((fl = *before) && (!IS_POSIX(fl) ||
859 !posix_same_owner(request, fl))) {
860 before = &fl->fl_next;
861 }
862
863 /* Process locks with this owner. */
864 while ((fl = *before) && posix_same_owner(request, fl)) {
865 /* Detect adjacent or overlapping regions (if same lock type)
866 */
867 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700868 /* In all comparisons of start vs end, use
869 * "start - 1" rather than "end + 1". If end
870 * is OFFSET_MAX, end + 1 will become negative.
871 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 if (fl->fl_end < request->fl_start - 1)
873 goto next_lock;
874 /* If the next lock in the list has entirely bigger
875 * addresses than the new one, insert the lock here.
876 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700877 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 break;
879
880 /* If we come here, the new and old lock are of the
881 * same type and adjacent or overlapping. Make one
882 * lock yielding from the lower start address of both
883 * locks to the higher end address.
884 */
885 if (fl->fl_start > request->fl_start)
886 fl->fl_start = request->fl_start;
887 else
888 request->fl_start = fl->fl_start;
889 if (fl->fl_end < request->fl_end)
890 fl->fl_end = request->fl_end;
891 else
892 request->fl_end = fl->fl_end;
893 if (added) {
894 locks_delete_lock(before);
895 continue;
896 }
897 request = fl;
898 added = 1;
899 }
900 else {
901 /* Processing for different lock types is a bit
902 * more complex.
903 */
904 if (fl->fl_end < request->fl_start)
905 goto next_lock;
906 if (fl->fl_start > request->fl_end)
907 break;
908 if (request->fl_type == F_UNLCK)
909 added = 1;
910 if (fl->fl_start < request->fl_start)
911 left = fl;
912 /* If the next lock in the list has a higher end
913 * address than the new one, insert the new one here.
914 */
915 if (fl->fl_end > request->fl_end) {
916 right = fl;
917 break;
918 }
919 if (fl->fl_start >= request->fl_start) {
920 /* The new lock completely replaces an old
921 * one (This may happen several times).
922 */
923 if (added) {
924 locks_delete_lock(before);
925 continue;
926 }
927 /* Replace the old lock with the new one.
928 * Wake up anybody waiting for the old one,
929 * as the change in lock type might satisfy
930 * their needs.
931 */
932 locks_wake_up_blocks(fl);
933 fl->fl_start = request->fl_start;
934 fl->fl_end = request->fl_end;
935 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500936 locks_release_private(fl);
937 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 request = fl;
939 added = 1;
940 }
941 }
942 /* Go on to next lock.
943 */
944 next_lock:
945 before = &fl->fl_next;
946 }
947
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700948 /*
949 * The above code only modifies existing locks in case of
950 * merging or replacing. If new lock(s) need to be inserted
951 * all modifications are done bellow this, so it's safe yet to
952 * bail out.
953 */
954 error = -ENOLCK; /* "no luck" */
955 if (right && left == right && !new_fl2)
956 goto out;
957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 error = 0;
959 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400960 if (request->fl_type == F_UNLCK) {
961 if (request->fl_flags & FL_EXISTS)
962 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400964 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700965
966 if (!new_fl) {
967 error = -ENOLCK;
968 goto out;
969 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 locks_copy_lock(new_fl, request);
971 locks_insert_lock(before, new_fl);
972 new_fl = NULL;
973 }
974 if (right) {
975 if (left == right) {
976 /* The new lock breaks the old one in two pieces,
977 * so we have to use the second new lock.
978 */
979 left = new_fl2;
980 new_fl2 = NULL;
981 locks_copy_lock(left, right);
982 locks_insert_lock(before, left);
983 }
984 right->fl_start = request->fl_end + 1;
985 locks_wake_up_blocks(right);
986 }
987 if (left) {
988 left->fl_end = request->fl_start - 1;
989 locks_wake_up_blocks(left);
990 }
991 out:
992 unlock_kernel();
993 /*
994 * Free any unused locks.
995 */
996 if (new_fl)
997 locks_free_lock(new_fl);
998 if (new_fl2)
999 locks_free_lock(new_fl2);
1000 return error;
1001}
1002
1003/**
1004 * posix_lock_file - Apply a POSIX-style lock to a file
1005 * @filp: The file to apply the lock to
1006 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001007 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 *
1009 * Add a POSIX style lock to a file.
1010 * We merge adjacent & overlapping locks whenever possible.
1011 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001012 *
1013 * Note that if called with an FL_EXISTS argument, the caller may determine
1014 * whether or not a lock was successfully freed by testing the return
1015 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 */
Marc Eshel150b3932007-01-18 16:15:35 -05001017int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001018 struct file_lock *conflock)
1019{
Marc Eshel150b3932007-01-18 16:15:35 -05001020 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001021}
Marc Eshel150b3932007-01-18 16:15:35 -05001022EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023
1024/**
1025 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1026 * @filp: The file to apply the lock to
1027 * @fl: The lock to be applied
1028 *
1029 * Add a POSIX style lock to a file.
1030 * We merge adjacent & overlapping locks whenever possible.
1031 * POSIX locks are sorted by owner task, then by starting address
1032 */
1033int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1034{
1035 int error;
1036 might_sleep ();
1037 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001038 error = posix_lock_file(filp, fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1040 break;
1041 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1042 if (!error)
1043 continue;
1044
1045 locks_delete_block(fl);
1046 break;
1047 }
1048 return error;
1049}
1050EXPORT_SYMBOL(posix_lock_file_wait);
1051
1052/**
1053 * locks_mandatory_locked - Check for an active lock
1054 * @inode: the file to check
1055 *
1056 * Searches the inode's list of locks to find any POSIX locks which conflict.
1057 * This function is called from locks_verify_locked() only.
1058 */
1059int locks_mandatory_locked(struct inode *inode)
1060{
1061 fl_owner_t owner = current->files;
1062 struct file_lock *fl;
1063
1064 /*
1065 * Search the lock list for this inode for any POSIX locks.
1066 */
1067 lock_kernel();
1068 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1069 if (!IS_POSIX(fl))
1070 continue;
1071 if (fl->fl_owner != owner)
1072 break;
1073 }
1074 unlock_kernel();
1075 return fl ? -EAGAIN : 0;
1076}
1077
1078/**
1079 * locks_mandatory_area - Check for a conflicting lock
1080 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1081 * for shared
1082 * @inode: the file to check
1083 * @filp: how the file was opened (if it was)
1084 * @offset: start of area to check
1085 * @count: length of area to check
1086 *
1087 * Searches the inode's list of locks to find any POSIX locks which conflict.
1088 * This function is called from rw_verify_area() and
1089 * locks_verify_truncate().
1090 */
1091int locks_mandatory_area(int read_write, struct inode *inode,
1092 struct file *filp, loff_t offset,
1093 size_t count)
1094{
1095 struct file_lock fl;
1096 int error;
1097
1098 locks_init_lock(&fl);
1099 fl.fl_owner = current->files;
1100 fl.fl_pid = current->tgid;
1101 fl.fl_file = filp;
1102 fl.fl_flags = FL_POSIX | FL_ACCESS;
1103 if (filp && !(filp->f_flags & O_NONBLOCK))
1104 fl.fl_flags |= FL_SLEEP;
1105 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1106 fl.fl_start = offset;
1107 fl.fl_end = offset + count - 1;
1108
1109 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001110 error = __posix_lock_file(inode, &fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 if (error != -EAGAIN)
1112 break;
1113 if (!(fl.fl_flags & FL_SLEEP))
1114 break;
1115 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1116 if (!error) {
1117 /*
1118 * If we've been sleeping someone might have
1119 * changed the permissions behind our back.
1120 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001121 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 continue;
1123 }
1124
1125 locks_delete_block(&fl);
1126 break;
1127 }
1128
1129 return error;
1130}
1131
1132EXPORT_SYMBOL(locks_mandatory_area);
1133
1134/* We already had a lease on this file; just change its type */
1135int lease_modify(struct file_lock **before, int arg)
1136{
1137 struct file_lock *fl = *before;
1138 int error = assign_type(fl, arg);
1139
1140 if (error)
1141 return error;
1142 locks_wake_up_blocks(fl);
1143 if (arg == F_UNLCK)
1144 locks_delete_lock(before);
1145 return 0;
1146}
1147
1148EXPORT_SYMBOL(lease_modify);
1149
1150static void time_out_leases(struct inode *inode)
1151{
1152 struct file_lock **before;
1153 struct file_lock *fl;
1154
1155 before = &inode->i_flock;
1156 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1157 if ((fl->fl_break_time == 0)
1158 || time_before(jiffies, fl->fl_break_time)) {
1159 before = &fl->fl_next;
1160 continue;
1161 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1163 if (fl == *before) /* lease_modify may have freed fl */
1164 before = &fl->fl_next;
1165 }
1166}
1167
1168/**
1169 * __break_lease - revoke all outstanding leases on file
1170 * @inode: the inode of the file to return
1171 * @mode: the open mode (read or write)
1172 *
david m. richter87250dd2007-05-09 16:10:27 -04001173 * break_lease (inlined for speed) has checked there already is at least
1174 * some kind of lock (maybe a lease) on this file. Leases are broken on
1175 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 * specified %O_NONBLOCK to your open().
1177 */
1178int __break_lease(struct inode *inode, unsigned int mode)
1179{
1180 int error = 0, future;
1181 struct file_lock *new_fl, *flock;
1182 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 unsigned long break_time;
1184 int i_have_this_lease = 0;
1185
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001186 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
1188 lock_kernel();
1189
1190 time_out_leases(inode);
1191
1192 flock = inode->i_flock;
1193 if ((flock == NULL) || !IS_LEASE(flock))
1194 goto out;
1195
1196 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1197 if (fl->fl_owner == current->files)
1198 i_have_this_lease = 1;
1199
1200 if (mode & FMODE_WRITE) {
1201 /* If we want write access, we have to revoke any lease. */
1202 future = F_UNLCK | F_INPROGRESS;
1203 } else if (flock->fl_type & F_INPROGRESS) {
1204 /* If the lease is already being broken, we just leave it */
1205 future = flock->fl_type;
1206 } else if (flock->fl_type & F_WRLCK) {
1207 /* Downgrade the exclusive lease to a read-only lease. */
1208 future = F_RDLCK | F_INPROGRESS;
1209 } else {
1210 /* the existing lease was read-only, so we can read too. */
1211 goto out;
1212 }
1213
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001214 if (IS_ERR(new_fl) && !i_have_this_lease
1215 && ((mode & O_NONBLOCK) == 0)) {
1216 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 goto out;
1218 }
1219
1220 break_time = 0;
1221 if (lease_break_time > 0) {
1222 break_time = jiffies + lease_break_time * HZ;
1223 if (break_time == 0)
1224 break_time++; /* so that 0 means no break time */
1225 }
1226
1227 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1228 if (fl->fl_type != future) {
1229 fl->fl_type = future;
1230 fl->fl_break_time = break_time;
1231 /* lease must have lmops break callback */
1232 fl->fl_lmops->fl_break(fl);
1233 }
1234 }
1235
1236 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1237 error = -EWOULDBLOCK;
1238 goto out;
1239 }
1240
1241restart:
1242 break_time = flock->fl_break_time;
1243 if (break_time != 0) {
1244 break_time -= jiffies;
1245 if (break_time == 0)
1246 break_time++;
1247 }
1248 error = locks_block_on_timeout(flock, new_fl, break_time);
1249 if (error >= 0) {
1250 if (error == 0)
1251 time_out_leases(inode);
1252 /* Wait for the next lease that has not been broken yet */
1253 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1254 flock = flock->fl_next) {
1255 if (flock->fl_type & F_INPROGRESS)
1256 goto restart;
1257 }
1258 error = 0;
1259 }
1260
1261out:
1262 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001263 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 locks_free_lock(new_fl);
1265 return error;
1266}
1267
1268EXPORT_SYMBOL(__break_lease);
1269
1270/**
1271 * lease_get_mtime
1272 * @inode: the inode
1273 * @time: pointer to a timespec which will contain the last modified time
1274 *
1275 * This is to force NFS clients to flush their caches for files with
1276 * exclusive leases. The justification is that if someone has an
1277 * exclusive lease, then they could be modifiying it.
1278 */
1279void lease_get_mtime(struct inode *inode, struct timespec *time)
1280{
1281 struct file_lock *flock = inode->i_flock;
1282 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1283 *time = current_fs_time(inode->i_sb);
1284 else
1285 *time = inode->i_mtime;
1286}
1287
1288EXPORT_SYMBOL(lease_get_mtime);
1289
1290/**
1291 * fcntl_getlease - Enquire what lease is currently active
1292 * @filp: the file
1293 *
1294 * The value returned by this function will be one of
1295 * (if no lease break is pending):
1296 *
1297 * %F_RDLCK to indicate a shared lease is held.
1298 *
1299 * %F_WRLCK to indicate an exclusive lease is held.
1300 *
1301 * %F_UNLCK to indicate no lease is held.
1302 *
1303 * (if a lease break is pending):
1304 *
1305 * %F_RDLCK to indicate an exclusive lease needs to be
1306 * changed to a shared lease (or removed).
1307 *
1308 * %F_UNLCK to indicate the lease needs to be removed.
1309 *
1310 * XXX: sfr & willy disagree over whether F_INPROGRESS
1311 * should be returned to userspace.
1312 */
1313int fcntl_getlease(struct file *filp)
1314{
1315 struct file_lock *fl;
1316 int type = F_UNLCK;
1317
1318 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001319 time_out_leases(filp->f_path.dentry->d_inode);
1320 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 fl = fl->fl_next) {
1322 if (fl->fl_file == filp) {
1323 type = fl->fl_type & ~F_INPROGRESS;
1324 break;
1325 }
1326 }
1327 unlock_kernel();
1328 return type;
1329}
1330
1331/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001332 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 * @filp: file pointer
1334 * @arg: type of lease to obtain
1335 * @flp: input - file_lock to use, output - file_lock inserted
1336 *
1337 * The (input) flp->fl_lmops->fl_break function is required
1338 * by break_lease().
1339 *
1340 * Called with kernel lock held.
1341 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001342int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001344 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001345 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001346 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347 struct inode *inode = dentry->d_inode;
1348 int error, rdlease_count = 0, wrlease_count = 0;
1349
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001350 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1351 return -EACCES;
1352 if (!S_ISREG(inode->i_mode))
1353 return -EINVAL;
1354 error = security_file_lock(filp, arg);
1355 if (error)
1356 return error;
1357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 time_out_leases(inode);
1359
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001360 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001362 lease = *flp;
1363
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 error = -EAGAIN;
1365 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1366 goto out;
1367 if ((arg == F_WRLCK)
1368 && ((atomic_read(&dentry->d_count) > 1)
1369 || (atomic_read(&inode->i_count) > 1)))
1370 goto out;
1371
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001372 error = -ENOMEM;
1373 new_fl = locks_alloc_lock();
1374 if (new_fl == NULL)
1375 goto out;
1376
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 /*
1378 * At this point, we know that if there is an exclusive
1379 * lease on this file, then we hold it on this filp
1380 * (otherwise our open of this file would have blocked).
1381 * And if we are trying to acquire an exclusive lease,
1382 * then the file is not open by anyone (including us)
1383 * except for this filp.
1384 */
1385 for (before = &inode->i_flock;
1386 ((fl = *before) != NULL) && IS_LEASE(fl);
1387 before = &fl->fl_next) {
1388 if (lease->fl_lmops->fl_mylease(fl, lease))
1389 my_before = before;
1390 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1391 /*
1392 * Someone is in the process of opening this
1393 * file for writing so we may not take an
1394 * exclusive lease on it.
1395 */
1396 wrlease_count++;
1397 else
1398 rdlease_count++;
1399 }
1400
1401 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1402 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1403 goto out;
1404
1405 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001406 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 error = lease->fl_lmops->fl_change(my_before, arg);
1408 goto out;
1409 }
1410
1411 error = 0;
1412 if (arg == F_UNLCK)
1413 goto out;
1414
1415 error = -EINVAL;
1416 if (!leases_enable)
1417 goto out;
1418
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001419 locks_copy_lock(new_fl, lease);
1420 locks_insert_lock(before, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001422 *flp = new_fl;
1423 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425out:
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001426 if (new_fl != NULL)
1427 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 return error;
1429}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001430EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
1432 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001433 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 * @filp: file pointer
1435 * @arg: type of lease to obtain
1436 * @lease: file_lock to use
1437 *
1438 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001439 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1440 * break_lease will oops!
1441 *
1442 * This will call the filesystem's setlease file method, if
1443 * defined. Note that there is no getlease method; instead, the
1444 * filesystem setlease method should call back to setlease() to
1445 * add a lease to the inode's lease list, where fcntl_getlease() can
1446 * find it. Since fcntl_getlease() only reports whether the current
1447 * task holds a lease, a cluster filesystem need only do this for
1448 * leases held by processes on this node.
1449 *
1450 * There is also no break_lease method; filesystems that
1451 * handle their own leases shoud break leases themselves from the
1452 * filesystem's open, create, and (on truncate) setattr methods.
1453 *
1454 * Warning: the only current setlease methods exist only to disable
1455 * leases in certain cases. More vfs changes may be required to
1456 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 */
1458
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001459int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 int error;
1462
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001464 if (filp->f_op && filp->f_op->setlease)
1465 error = filp->f_op->setlease(filp, arg, lease);
1466 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001467 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 unlock_kernel();
1469
1470 return error;
1471}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001472EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473
1474/**
1475 * fcntl_setlease - sets a lease on an open file
1476 * @fd: open file descriptor
1477 * @filp: file pointer
1478 * @arg: type of lease to obtain
1479 *
1480 * Call this fcntl to establish a lease on the file.
1481 * Note that you also need to call %F_SETSIG to
1482 * receive a signal when the lease is broken.
1483 */
1484int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1485{
1486 struct file_lock fl, *flp = &fl;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001487 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 struct inode *inode = dentry->d_inode;
1489 int error;
1490
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 locks_init_lock(&fl);
1492 error = lease_init(filp, arg, &fl);
1493 if (error)
1494 return error;
1495
1496 lock_kernel();
1497
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001498 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001499 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 goto out_unlock;
1501
1502 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1503 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001504 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001506 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 time_out_leases(inode);
1508 goto out_unlock;
1509 }
1510
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001511 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512out_unlock:
1513 unlock_kernel();
1514 return error;
1515}
1516
1517/**
1518 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1519 * @filp: The file to apply the lock to
1520 * @fl: The lock to be applied
1521 *
1522 * Add a FLOCK style lock to a file.
1523 */
1524int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1525{
1526 int error;
1527 might_sleep();
1528 for (;;) {
1529 error = flock_lock_file(filp, fl);
1530 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1531 break;
1532 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1533 if (!error)
1534 continue;
1535
1536 locks_delete_block(fl);
1537 break;
1538 }
1539 return error;
1540}
1541
1542EXPORT_SYMBOL(flock_lock_file_wait);
1543
1544/**
1545 * sys_flock: - flock() system call.
1546 * @fd: the file descriptor to lock.
1547 * @cmd: the type of lock to apply.
1548 *
1549 * Apply a %FL_FLOCK style lock to an open file descriptor.
1550 * The @cmd can be one of
1551 *
1552 * %LOCK_SH -- a shared lock.
1553 *
1554 * %LOCK_EX -- an exclusive lock.
1555 *
1556 * %LOCK_UN -- remove an existing lock.
1557 *
1558 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1559 *
1560 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1561 * processes read and write access respectively.
1562 */
1563asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1564{
1565 struct file *filp;
1566 struct file_lock *lock;
1567 int can_sleep, unlock;
1568 int error;
1569
1570 error = -EBADF;
1571 filp = fget(fd);
1572 if (!filp)
1573 goto out;
1574
1575 can_sleep = !(cmd & LOCK_NB);
1576 cmd &= ~LOCK_NB;
1577 unlock = (cmd == LOCK_UN);
1578
1579 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1580 goto out_putf;
1581
1582 error = flock_make_lock(filp, &lock, cmd);
1583 if (error)
1584 goto out_putf;
1585 if (can_sleep)
1586 lock->fl_flags |= FL_SLEEP;
1587
1588 error = security_file_lock(filp, cmd);
1589 if (error)
1590 goto out_free;
1591
1592 if (filp->f_op && filp->f_op->flock)
1593 error = filp->f_op->flock(filp,
1594 (can_sleep) ? F_SETLKW : F_SETLK,
1595 lock);
1596 else
1597 error = flock_lock_file_wait(filp, lock);
1598
1599 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001600 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
1602 out_putf:
1603 fput(filp);
1604 out:
1605 return error;
1606}
1607
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001608/**
1609 * vfs_test_lock - test file byte range lock
1610 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001611 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001612 *
1613 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1614 * setting conf->fl_type to something other than F_UNLCK.
1615 */
1616int vfs_test_lock(struct file *filp, struct file_lock *fl)
1617{
1618 if (filp->f_op && filp->f_op->lock)
1619 return filp->f_op->lock(filp, F_GETLK, fl);
1620 posix_test_lock(filp, fl);
1621 return 0;
1622}
1623EXPORT_SYMBOL_GPL(vfs_test_lock);
1624
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001625static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1626{
1627 flock->l_pid = fl->fl_pid;
1628#if BITS_PER_LONG == 32
1629 /*
1630 * Make sure we can represent the posix lock via
1631 * legacy 32bit flock.
1632 */
1633 if (fl->fl_start > OFFT_OFFSET_MAX)
1634 return -EOVERFLOW;
1635 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1636 return -EOVERFLOW;
1637#endif
1638 flock->l_start = fl->fl_start;
1639 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1640 fl->fl_end - fl->fl_start + 1;
1641 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001642 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001643 return 0;
1644}
1645
1646#if BITS_PER_LONG == 32
1647static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1648{
1649 flock->l_pid = fl->fl_pid;
1650 flock->l_start = fl->fl_start;
1651 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1652 fl->fl_end - fl->fl_start + 1;
1653 flock->l_whence = 0;
1654 flock->l_type = fl->fl_type;
1655}
1656#endif
1657
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658/* Report the first existing lock that would conflict with l.
1659 * This implements the F_GETLK command of fcntl().
1660 */
1661int fcntl_getlk(struct file *filp, struct flock __user *l)
1662{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001663 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 struct flock flock;
1665 int error;
1666
1667 error = -EFAULT;
1668 if (copy_from_user(&flock, l, sizeof(flock)))
1669 goto out;
1670 error = -EINVAL;
1671 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1672 goto out;
1673
1674 error = flock_to_posix_lock(filp, &file_lock, &flock);
1675 if (error)
1676 goto out;
1677
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001678 error = vfs_test_lock(filp, &file_lock);
1679 if (error)
1680 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001682 flock.l_type = file_lock.fl_type;
1683 if (file_lock.fl_type != F_UNLCK) {
1684 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001685 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 }
1688 error = -EFAULT;
1689 if (!copy_to_user(l, &flock, sizeof(flock)))
1690 error = 0;
1691out:
1692 return error;
1693}
1694
Marc Eshel7723ec92007-01-18 15:08:55 -05001695/**
1696 * vfs_lock_file - file byte range lock
1697 * @filp: The file to apply the lock to
1698 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1699 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001700 * @conf: Place to return a copy of the conflicting lock, if found.
1701 *
1702 * A caller that doesn't care about the conflicting lock may pass NULL
1703 * as the final argument.
1704 *
1705 * If the filesystem defines a private ->lock() method, then @conf will
1706 * be left unchanged; so a caller that cares should initialize it to
1707 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001708 *
1709 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1710 * locks, the ->lock() interface may return asynchronously, before the lock has
1711 * been granted or denied by the underlying filesystem, if (and only if)
1712 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1713 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1714 * the request is for a blocking lock. When ->lock() does return asynchronously,
1715 * it must return -EINPROGRESS, and call ->fl_grant() when the lock
1716 * request completes.
1717 * If the request is for non-blocking lock the file system should return
1718 * -EINPROGRESS then try to get the lock and call the callback routine with
1719 * the result. If the request timed out the callback routine will return a
1720 * nonzero return code and the file system should release the lock. The file
1721 * system is also responsible to keep a corresponding posix lock when it
1722 * grants a lock so the VFS can find out which locks are locally held and do
1723 * the correct lock cleanup when required.
1724 * The underlying filesystem must not drop the kernel lock or call
1725 * ->fl_grant() before returning to the caller with a -EINPROGRESS
1726 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001727 */
Marc Eshel150b3932007-01-18 16:15:35 -05001728int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001729{
1730 if (filp->f_op && filp->f_op->lock)
1731 return filp->f_op->lock(filp, cmd, fl);
1732 else
Marc Eshel150b3932007-01-18 16:15:35 -05001733 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001734}
1735EXPORT_SYMBOL_GPL(vfs_lock_file);
1736
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737/* Apply the lock described by l to an open file descriptor.
1738 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1739 */
Peter Staubachc2936212005-07-27 11:45:09 -07001740int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1741 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742{
1743 struct file_lock *file_lock = locks_alloc_lock();
1744 struct flock flock;
1745 struct inode *inode;
1746 int error;
1747
1748 if (file_lock == NULL)
1749 return -ENOLCK;
1750
1751 /*
1752 * This might block, so we do it before checking the inode.
1753 */
1754 error = -EFAULT;
1755 if (copy_from_user(&flock, l, sizeof(flock)))
1756 goto out;
1757
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001758 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
1760 /* Don't allow mandatory locks on files that may be memory mapped
1761 * and shared.
1762 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001763 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 error = -EAGAIN;
1765 goto out;
1766 }
1767
Peter Staubachc2936212005-07-27 11:45:09 -07001768again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 error = flock_to_posix_lock(filp, file_lock, &flock);
1770 if (error)
1771 goto out;
1772 if (cmd == F_SETLKW) {
1773 file_lock->fl_flags |= FL_SLEEP;
1774 }
1775
1776 error = -EBADF;
1777 switch (flock.l_type) {
1778 case F_RDLCK:
1779 if (!(filp->f_mode & FMODE_READ))
1780 goto out;
1781 break;
1782 case F_WRLCK:
1783 if (!(filp->f_mode & FMODE_WRITE))
1784 goto out;
1785 break;
1786 case F_UNLCK:
1787 break;
1788 default:
1789 error = -EINVAL;
1790 goto out;
1791 }
1792
1793 error = security_file_lock(filp, file_lock->fl_type);
1794 if (error)
1795 goto out;
1796
Marc Eshel7723ec92007-01-18 15:08:55 -05001797 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001798 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001799 if (error != -EAGAIN || cmd == F_SETLK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001801 error = wait_event_interruptible(file_lock->fl_wait,
1802 !file_lock->fl_next);
1803 if (!error)
1804 continue;
1805
1806 locks_delete_block(file_lock);
1807 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 }
1809
Peter Staubachc2936212005-07-27 11:45:09 -07001810 /*
1811 * Attempt to detect a close/fcntl race and recover by
1812 * releasing the lock that was just acquired.
1813 */
1814 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1815 flock.l_type = F_UNLCK;
1816 goto again;
1817 }
1818
1819out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 locks_free_lock(file_lock);
1821 return error;
1822}
1823
1824#if BITS_PER_LONG == 32
1825/* Report the first existing lock that would conflict with l.
1826 * This implements the F_GETLK command of fcntl().
1827 */
1828int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1829{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001830 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 struct flock64 flock;
1832 int error;
1833
1834 error = -EFAULT;
1835 if (copy_from_user(&flock, l, sizeof(flock)))
1836 goto out;
1837 error = -EINVAL;
1838 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1839 goto out;
1840
1841 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1842 if (error)
1843 goto out;
1844
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001845 error = vfs_test_lock(filp, &file_lock);
1846 if (error)
1847 goto out;
1848
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001849 flock.l_type = file_lock.fl_type;
1850 if (file_lock.fl_type != F_UNLCK)
1851 posix_lock_to_flock64(&flock, &file_lock);
1852
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 error = -EFAULT;
1854 if (!copy_to_user(l, &flock, sizeof(flock)))
1855 error = 0;
1856
1857out:
1858 return error;
1859}
1860
1861/* Apply the lock described by l to an open file descriptor.
1862 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1863 */
Peter Staubachc2936212005-07-27 11:45:09 -07001864int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1865 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866{
1867 struct file_lock *file_lock = locks_alloc_lock();
1868 struct flock64 flock;
1869 struct inode *inode;
1870 int error;
1871
1872 if (file_lock == NULL)
1873 return -ENOLCK;
1874
1875 /*
1876 * This might block, so we do it before checking the inode.
1877 */
1878 error = -EFAULT;
1879 if (copy_from_user(&flock, l, sizeof(flock)))
1880 goto out;
1881
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001882 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883
1884 /* Don't allow mandatory locks on files that may be memory mapped
1885 * and shared.
1886 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001887 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 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
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002069#ifdef CONFIG_PROC_FS
2070#include <linux/seq_file.h>
2071
2072static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2073 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074{
2075 struct inode *inode = NULL;
2076
2077 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002078 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002080 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002082 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2084 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002085 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 } else if (IS_FLOCK(fl)) {
2087 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002088 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002090 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 }
2092 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002093 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002095 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002097 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002099 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002101 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 }
2103 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002104 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 (fl->fl_type & LOCK_READ)
2106 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2107 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2108 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002109 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 (fl->fl_type & F_INPROGRESS)
2111 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2112 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2113 }
2114 if (inode) {
2115#ifdef WE_CAN_BREAK_LSLK_NOW
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002116 seq_printf(f, "%d %s:%ld ", fl->fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 inode->i_sb->s_id, inode->i_ino);
2118#else
2119 /* userspace relies on this representation of dev_t ;-( */
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002120 seq_printf(f, "%d %02x:%02x:%ld ", fl->fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 MAJOR(inode->i_sb->s_dev),
2122 MINOR(inode->i_sb->s_dev), inode->i_ino);
2123#endif
2124 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002125 seq_printf(f, "%d <none>:0 ", fl->fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 }
2127 if (IS_POSIX(fl)) {
2128 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002129 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002131 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002133 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134 }
2135}
2136
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002137static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002139 struct file_lock *fl, *bfl;
2140
2141 fl = list_entry(v, struct file_lock, fl_link);
2142
2143 lock_get_status(f, fl, (long)f->private, "");
2144
2145 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2146 lock_get_status(f, bfl, (long)f->private, " ->");
2147
2148 f->private++;
2149 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150}
2151
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002152static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002155 f->private = (void *)1;
2156 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157}
2158
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002159static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2160{
2161 return seq_list_next(v, &file_lock_list, pos);
2162}
2163
2164static void locks_stop(struct seq_file *f, void *v)
2165{
2166 unlock_kernel();
2167}
2168
2169struct seq_operations locks_seq_operations = {
2170 .start = locks_start,
2171 .next = locks_next,
2172 .stop = locks_stop,
2173 .show = locks_show,
2174};
2175#endif
2176
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177/**
2178 * lock_may_read - checks that the region is free of locks
2179 * @inode: the inode that is being read
2180 * @start: the first byte to read
2181 * @len: the number of bytes to read
2182 *
2183 * Emulates Windows locking requirements. Whole-file
2184 * mandatory locks (share modes) can prohibit a read and
2185 * byte-range POSIX locks can prohibit a read if they overlap.
2186 *
2187 * N.B. this function is only ever called
2188 * from knfsd and ownership of locks is never checked.
2189 */
2190int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2191{
2192 struct file_lock *fl;
2193 int result = 1;
2194 lock_kernel();
2195 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2196 if (IS_POSIX(fl)) {
2197 if (fl->fl_type == F_RDLCK)
2198 continue;
2199 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2200 continue;
2201 } else if (IS_FLOCK(fl)) {
2202 if (!(fl->fl_type & LOCK_MAND))
2203 continue;
2204 if (fl->fl_type & LOCK_READ)
2205 continue;
2206 } else
2207 continue;
2208 result = 0;
2209 break;
2210 }
2211 unlock_kernel();
2212 return result;
2213}
2214
2215EXPORT_SYMBOL(lock_may_read);
2216
2217/**
2218 * lock_may_write - checks that the region is free of locks
2219 * @inode: the inode that is being written
2220 * @start: the first byte to write
2221 * @len: the number of bytes to write
2222 *
2223 * Emulates Windows locking requirements. Whole-file
2224 * mandatory locks (share modes) can prohibit a write and
2225 * byte-range POSIX locks can prohibit a write if they overlap.
2226 *
2227 * N.B. this function is only ever called
2228 * from knfsd and ownership of locks is never checked.
2229 */
2230int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2231{
2232 struct file_lock *fl;
2233 int result = 1;
2234 lock_kernel();
2235 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2236 if (IS_POSIX(fl)) {
2237 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2238 continue;
2239 } else if (IS_FLOCK(fl)) {
2240 if (!(fl->fl_type & LOCK_MAND))
2241 continue;
2242 if (fl->fl_type & LOCK_WRITE)
2243 continue;
2244 } else
2245 continue;
2246 result = 0;
2247 break;
2248 }
2249 unlock_kernel();
2250 return result;
2251}
2252
2253EXPORT_SYMBOL(lock_may_write);
2254
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255static int __init filelock_init(void)
2256{
2257 filelock_cache = kmem_cache_create("file_lock_cache",
2258 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002259 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260 return 0;
2261}
2262
2263core_initcall(filelock_init);