blob: 592faadbcec12f19b2f6ea158e956e2a3b01c838 [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>
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000128#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130#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;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000188 fl->fl_nspid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 fl->fl_file = NULL;
190 fl->fl_flags = 0;
191 fl->fl_type = 0;
192 fl->fl_start = fl->fl_end = 0;
193 fl->fl_ops = NULL;
194 fl->fl_lmops = NULL;
195}
196
197EXPORT_SYMBOL(locks_init_lock);
198
199/*
200 * Initialises the fields of the file lock which are invariant for
201 * free file_locks.
202 */
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700203static void init_once(struct kmem_cache *cache, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204{
205 struct file_lock *lock = (struct file_lock *) foo;
206
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 locks_init_lock(lock);
208}
209
Trond Myklebust47831f32006-03-20 13:44:05 -0500210static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
211{
212 if (fl->fl_ops) {
213 if (fl->fl_ops->fl_copy_lock)
214 fl->fl_ops->fl_copy_lock(new, fl);
215 new->fl_ops = fl->fl_ops;
216 }
217 if (fl->fl_lmops) {
218 if (fl->fl_lmops->fl_copy_lock)
219 fl->fl_lmops->fl_copy_lock(new, fl);
220 new->fl_lmops = fl->fl_lmops;
221 }
222}
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224/*
225 * Initialize a new lock from an existing file_lock structure.
226 */
Trond Myklebust09969052006-03-20 13:44:38 -0500227static void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228{
229 new->fl_owner = fl->fl_owner;
230 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500231 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 new->fl_flags = fl->fl_flags;
233 new->fl_type = fl->fl_type;
234 new->fl_start = fl->fl_start;
235 new->fl_end = fl->fl_end;
Trond Myklebust09969052006-03-20 13:44:38 -0500236 new->fl_ops = NULL;
237 new->fl_lmops = NULL;
238}
239
240void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
241{
242 locks_release_private(new);
243
244 __locks_copy_lock(new, fl);
245 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 new->fl_ops = fl->fl_ops;
247 new->fl_lmops = fl->fl_lmops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500248
249 locks_copy_private(new, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250}
251
252EXPORT_SYMBOL(locks_copy_lock);
253
254static inline int flock_translate_cmd(int cmd) {
255 if (cmd & LOCK_MAND)
256 return cmd & (LOCK_MAND | LOCK_RW);
257 switch (cmd) {
258 case LOCK_SH:
259 return F_RDLCK;
260 case LOCK_EX:
261 return F_WRLCK;
262 case LOCK_UN:
263 return F_UNLCK;
264 }
265 return -EINVAL;
266}
267
268/* Fill in a file_lock structure with an appropriate FLOCK lock. */
269static int flock_make_lock(struct file *filp, struct file_lock **lock,
270 unsigned int cmd)
271{
272 struct file_lock *fl;
273 int type = flock_translate_cmd(cmd);
274 if (type < 0)
275 return type;
276
277 fl = locks_alloc_lock();
278 if (fl == NULL)
279 return -ENOMEM;
280
281 fl->fl_file = filp;
282 fl->fl_pid = current->tgid;
283 fl->fl_flags = FL_FLOCK;
284 fl->fl_type = type;
285 fl->fl_end = OFFSET_MAX;
286
287 *lock = fl;
288 return 0;
289}
290
291static int assign_type(struct file_lock *fl, int type)
292{
293 switch (type) {
294 case F_RDLCK:
295 case F_WRLCK:
296 case F_UNLCK:
297 fl->fl_type = type;
298 break;
299 default:
300 return -EINVAL;
301 }
302 return 0;
303}
304
305/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
306 * style lock.
307 */
308static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
309 struct flock *l)
310{
311 off_t start, end;
312
313 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700314 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 start = 0;
316 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700317 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 start = filp->f_pos;
319 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700320 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800321 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 break;
323 default:
324 return -EINVAL;
325 }
326
327 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
328 POSIX-2001 defines it. */
329 start += l->l_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 if (start < 0)
331 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700332 fl->fl_end = OFFSET_MAX;
333 if (l->l_len > 0) {
334 end = start + l->l_len - 1;
335 fl->fl_end = end;
336 } else if (l->l_len < 0) {
337 end = start - 1;
338 fl->fl_end = end;
339 start += l->l_len;
340 if (start < 0)
341 return -EINVAL;
342 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700344 if (fl->fl_end < fl->fl_start)
345 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346
347 fl->fl_owner = current->files;
348 fl->fl_pid = current->tgid;
349 fl->fl_file = filp;
350 fl->fl_flags = FL_POSIX;
351 fl->fl_ops = NULL;
352 fl->fl_lmops = NULL;
353
354 return assign_type(fl, l->l_type);
355}
356
357#if BITS_PER_LONG == 32
358static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
359 struct flock64 *l)
360{
361 loff_t start;
362
363 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700364 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 start = 0;
366 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700367 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 start = filp->f_pos;
369 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700370 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800371 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 break;
373 default:
374 return -EINVAL;
375 }
376
Trond Myklebust4c780a42005-10-18 14:20:21 -0700377 start += l->l_start;
378 if (start < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700380 fl->fl_end = OFFSET_MAX;
381 if (l->l_len > 0) {
382 fl->fl_end = start + l->l_len - 1;
383 } else if (l->l_len < 0) {
384 fl->fl_end = start - 1;
385 start += l->l_len;
386 if (start < 0)
387 return -EINVAL;
388 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700390 if (fl->fl_end < fl->fl_start)
391 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 fl->fl_owner = current->files;
394 fl->fl_pid = current->tgid;
395 fl->fl_file = filp;
396 fl->fl_flags = FL_POSIX;
397 fl->fl_ops = NULL;
398 fl->fl_lmops = NULL;
399
400 switch (l->l_type) {
401 case F_RDLCK:
402 case F_WRLCK:
403 case F_UNLCK:
404 fl->fl_type = l->l_type;
405 break;
406 default:
407 return -EINVAL;
408 }
409
410 return (0);
411}
412#endif
413
414/* default lease lock manager operations */
415static void lease_break_callback(struct file_lock *fl)
416{
417 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
418}
419
420static void lease_release_private_callback(struct file_lock *fl)
421{
422 if (!fl->fl_file)
423 return;
424
425 f_delown(fl->fl_file);
426 fl->fl_file->f_owner.signum = 0;
427}
428
Adrian Bunk75c96f82005-05-05 16:16:09 -0700429static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430{
431 return fl->fl_file == try->fl_file;
432}
433
Adrian Bunk75c96f82005-05-05 16:16:09 -0700434static struct lock_manager_operations lease_manager_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 .fl_break = lease_break_callback,
436 .fl_release_private = lease_release_private_callback,
437 .fl_mylease = lease_mylease_callback,
438 .fl_change = lease_modify,
439};
440
441/*
442 * Initialize a lease, use the default lock manager operations
443 */
444static int lease_init(struct file *filp, int type, struct file_lock *fl)
445 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400446 if (assign_type(fl, type) != 0)
447 return -EINVAL;
448
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 fl->fl_owner = current->files;
450 fl->fl_pid = current->tgid;
451
452 fl->fl_file = filp;
453 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 fl->fl_start = 0;
455 fl->fl_end = OFFSET_MAX;
456 fl->fl_ops = NULL;
457 fl->fl_lmops = &lease_manager_ops;
458 return 0;
459}
460
461/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500462static struct file_lock *lease_alloc(struct file *filp, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463{
464 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400465 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500468 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400471 if (error) {
472 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500473 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400474 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500475 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476}
477
478/* Check if two locks overlap each other.
479 */
480static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
481{
482 return ((fl1->fl_end >= fl2->fl_start) &&
483 (fl2->fl_end >= fl1->fl_start));
484}
485
486/*
487 * Check whether two locks have the same owner.
488 */
Matt Mackall33443c42006-01-08 01:05:22 -0800489static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490{
491 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
492 return fl2->fl_lmops == fl1->fl_lmops &&
493 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
494 return fl1->fl_owner == fl2->fl_owner;
495}
496
497/* Remove waiter from blocker's block list.
498 * When blocker ends up pointing to itself then the list is empty.
499 */
Matt Mackall33443c42006-01-08 01:05:22 -0800500static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501{
502 list_del_init(&waiter->fl_block);
503 list_del_init(&waiter->fl_link);
504 waiter->fl_next = NULL;
505}
506
507/*
508 */
509static void locks_delete_block(struct file_lock *waiter)
510{
511 lock_kernel();
512 __locks_delete_block(waiter);
513 unlock_kernel();
514}
515
516/* Insert waiter into blocker's block list.
517 * We use a circular list so that processes can be easily woken up in
518 * the order they blocked. The documentation doesn't require this but
519 * it seems like the reasonable thing to do.
520 */
521static void locks_insert_block(struct file_lock *blocker,
522 struct file_lock *waiter)
523{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800524 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 list_add_tail(&waiter->fl_block, &blocker->fl_block);
526 waiter->fl_next = blocker;
527 if (IS_POSIX(blocker))
528 list_add(&waiter->fl_link, &blocked_list);
529}
530
531/* Wake up processes blocked waiting for blocker.
532 * If told to wait then schedule the processes until the block list
533 * is empty, otherwise empty the block list ourselves.
534 */
535static void locks_wake_up_blocks(struct file_lock *blocker)
536{
537 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400538 struct file_lock *waiter;
539
540 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 struct file_lock, fl_block);
542 __locks_delete_block(waiter);
543 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
544 waiter->fl_lmops->fl_notify(waiter);
545 else
546 wake_up(&waiter->fl_wait);
547 }
548}
549
550/* Insert file lock fl into an inode's lock list at the position indicated
551 * by pos. At the same time add the lock to the global file lock list.
552 */
553static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
554{
555 list_add(&fl->fl_link, &file_lock_list);
556
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000557 fl->fl_nspid = get_pid(task_tgid(current));
558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 /* insert into file's list */
560 fl->fl_next = *pos;
561 *pos = fl;
562
563 if (fl->fl_ops && fl->fl_ops->fl_insert)
564 fl->fl_ops->fl_insert(fl);
565}
566
567/*
568 * Delete a lock and then free it.
569 * Wake up processes that are blocked waiting for this lock,
570 * notify the FS that the lock has been cleared and
571 * finally free the lock.
572 */
573static void locks_delete_lock(struct file_lock **thisfl_p)
574{
575 struct file_lock *fl = *thisfl_p;
576
577 *thisfl_p = fl->fl_next;
578 fl->fl_next = NULL;
579 list_del_init(&fl->fl_link);
580
581 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
582 if (fl->fl_fasync != NULL) {
583 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
584 fl->fl_fasync = NULL;
585 }
586
587 if (fl->fl_ops && fl->fl_ops->fl_remove)
588 fl->fl_ops->fl_remove(fl);
589
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000590 if (fl->fl_nspid) {
591 put_pid(fl->fl_nspid);
592 fl->fl_nspid = NULL;
593 }
594
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 locks_wake_up_blocks(fl);
596 locks_free_lock(fl);
597}
598
599/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
600 * checks for shared/exclusive status of overlapping locks.
601 */
602static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
603{
604 if (sys_fl->fl_type == F_WRLCK)
605 return 1;
606 if (caller_fl->fl_type == F_WRLCK)
607 return 1;
608 return 0;
609}
610
611/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
612 * checking before calling the locks_conflict().
613 */
614static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
615{
616 /* POSIX locks owned by the same process do not conflict with
617 * each other.
618 */
619 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
620 return (0);
621
622 /* Check whether they overlap */
623 if (!locks_overlap(caller_fl, sys_fl))
624 return 0;
625
626 return (locks_conflict(caller_fl, sys_fl));
627}
628
629/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
630 * checking before calling the locks_conflict().
631 */
632static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
633{
634 /* FLOCK locks referring to the same filp do not conflict with
635 * each other.
636 */
637 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
638 return (0);
639 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
640 return 0;
641
642 return (locks_conflict(caller_fl, sys_fl));
643}
644
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400645void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500646posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647{
648 struct file_lock *cfl;
649
650 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800651 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 if (!IS_POSIX(cfl))
653 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400654 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 break;
656 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000657 if (cfl) {
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500658 __locks_copy_lock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000659 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800660 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000661 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400662 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400664 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666EXPORT_SYMBOL(posix_test_lock);
667
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400668/*
669 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400671 * We attempt to detect deadlocks that are due purely to posix file
672 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400674 * We assume that a task can be waiting for at most one lock at a time.
675 * So for any acquired lock, the process holding that lock may be
676 * waiting on at most one other lock. That lock in turns may be held by
677 * someone waiting for at most one other lock. Given a requested lock
678 * caller_fl which is about to wait for a conflicting lock block_fl, we
679 * follow this chain of waiters to ensure we are not about to create a
680 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400681 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400682 * Since we do this before we ever put a process to sleep on a lock, we
683 * are ensured that there is never a cycle; that is what guarantees that
684 * the while() loop in posix_locks_deadlock() eventually completes.
685 *
686 * Note: the above assumption may not be true when handling lock
687 * requests from a broken NFS client. It may also fail in the presence
688 * of tasks (such as posix threads) sharing the same open file table.
689 *
690 * To handle those cases, we just bail out after a few iterations.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400692
693#define MAX_DEADLK_ITERATIONS 10
694
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400695/* Find a lock that the owner of the given block_fl is blocking on. */
696static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
697{
698 struct file_lock *fl;
699
700 list_for_each_entry(fl, &blocked_list, fl_link) {
701 if (posix_same_owner(fl, block_fl))
702 return fl->fl_next;
703 }
704 return NULL;
705}
706
Adrian Bunkb0904e12006-06-23 02:05:13 -0700707static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 struct file_lock *block_fl)
709{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400710 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400712 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
713 if (i++ > MAX_DEADLK_ITERATIONS)
714 return 0;
715 if (posix_same_owner(caller_fl, block_fl))
716 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 }
718 return 0;
719}
720
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400722 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400723 *
724 * Note that if called with an FL_EXISTS argument, the caller may determine
725 * whether or not a lock was successfully freed by testing the return
726 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800728static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800730 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800732 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 int error = 0;
734 int found = 0;
735
736 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400737 if (request->fl_flags & FL_ACCESS)
738 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400739
740 if (request->fl_type != F_UNLCK) {
741 error = -ENOMEM;
742 new_fl = locks_alloc_lock();
743 if (new_fl == NULL)
744 goto out;
745 error = 0;
746 }
747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 for_each_lock(inode, before) {
749 struct file_lock *fl = *before;
750 if (IS_POSIX(fl))
751 break;
752 if (IS_LEASE(fl))
753 continue;
754 if (filp != fl->fl_file)
755 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800756 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 goto out;
758 found = 1;
759 locks_delete_lock(before);
760 break;
761 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Trond Myklebustf475ae92006-06-29 16:38:32 -0400763 if (request->fl_type == F_UNLCK) {
764 if ((request->fl_flags & FL_EXISTS) && !found)
765 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800766 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400767 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769 /*
770 * If a higher-priority process was blocked on the old file lock,
771 * give it the opportunity to lock the file.
772 */
773 if (found)
774 cond_resched();
775
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400776find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 for_each_lock(inode, before) {
778 struct file_lock *fl = *before;
779 if (IS_POSIX(fl))
780 break;
781 if (IS_LEASE(fl))
782 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800783 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 continue;
785 error = -EAGAIN;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800786 if (request->fl_flags & FL_SLEEP)
787 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 goto out;
789 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400790 if (request->fl_flags & FL_ACCESS)
791 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800792 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700793 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800794 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400795 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
797out:
798 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800799 if (new_fl)
800 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 return error;
802}
803
Marc Eshel150b3932007-01-18 16:15:35 -0500804static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700807 struct file_lock *new_fl = NULL;
808 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 struct file_lock *left = NULL;
810 struct file_lock *right = NULL;
811 struct file_lock **before;
812 int error, added = 0;
813
814 /*
815 * We may need two file_lock structures for this operation,
816 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700817 *
818 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700820 if (!(request->fl_flags & FL_ACCESS) &&
821 (request->fl_type != F_UNLCK ||
822 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
823 new_fl = locks_alloc_lock();
824 new_fl2 = locks_alloc_lock();
825 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
827 lock_kernel();
828 if (request->fl_type != F_UNLCK) {
829 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500830 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 if (!IS_POSIX(fl))
832 continue;
833 if (!posix_locks_conflict(request, fl))
834 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800835 if (conflock)
836 locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 error = -EAGAIN;
838 if (!(request->fl_flags & FL_SLEEP))
839 goto out;
840 error = -EDEADLK;
841 if (posix_locks_deadlock(request, fl))
842 goto out;
843 error = -EAGAIN;
844 locks_insert_block(fl, request);
845 goto out;
846 }
847 }
848
849 /* If we're just looking for a conflict, we're done. */
850 error = 0;
851 if (request->fl_flags & FL_ACCESS)
852 goto out;
853
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 * Find the first old lock with the same owner as the new lock.
856 */
857
858 before = &inode->i_flock;
859
860 /* First skip locks owned by other processes. */
861 while ((fl = *before) && (!IS_POSIX(fl) ||
862 !posix_same_owner(request, fl))) {
863 before = &fl->fl_next;
864 }
865
866 /* Process locks with this owner. */
867 while ((fl = *before) && posix_same_owner(request, fl)) {
868 /* Detect adjacent or overlapping regions (if same lock type)
869 */
870 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700871 /* In all comparisons of start vs end, use
872 * "start - 1" rather than "end + 1". If end
873 * is OFFSET_MAX, end + 1 will become negative.
874 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 if (fl->fl_end < request->fl_start - 1)
876 goto next_lock;
877 /* If the next lock in the list has entirely bigger
878 * addresses than the new one, insert the lock here.
879 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700880 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 break;
882
883 /* If we come here, the new and old lock are of the
884 * same type and adjacent or overlapping. Make one
885 * lock yielding from the lower start address of both
886 * locks to the higher end address.
887 */
888 if (fl->fl_start > request->fl_start)
889 fl->fl_start = request->fl_start;
890 else
891 request->fl_start = fl->fl_start;
892 if (fl->fl_end < request->fl_end)
893 fl->fl_end = request->fl_end;
894 else
895 request->fl_end = fl->fl_end;
896 if (added) {
897 locks_delete_lock(before);
898 continue;
899 }
900 request = fl;
901 added = 1;
902 }
903 else {
904 /* Processing for different lock types is a bit
905 * more complex.
906 */
907 if (fl->fl_end < request->fl_start)
908 goto next_lock;
909 if (fl->fl_start > request->fl_end)
910 break;
911 if (request->fl_type == F_UNLCK)
912 added = 1;
913 if (fl->fl_start < request->fl_start)
914 left = fl;
915 /* If the next lock in the list has a higher end
916 * address than the new one, insert the new one here.
917 */
918 if (fl->fl_end > request->fl_end) {
919 right = fl;
920 break;
921 }
922 if (fl->fl_start >= request->fl_start) {
923 /* The new lock completely replaces an old
924 * one (This may happen several times).
925 */
926 if (added) {
927 locks_delete_lock(before);
928 continue;
929 }
930 /* Replace the old lock with the new one.
931 * Wake up anybody waiting for the old one,
932 * as the change in lock type might satisfy
933 * their needs.
934 */
935 locks_wake_up_blocks(fl);
936 fl->fl_start = request->fl_start;
937 fl->fl_end = request->fl_end;
938 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500939 locks_release_private(fl);
940 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 request = fl;
942 added = 1;
943 }
944 }
945 /* Go on to next lock.
946 */
947 next_lock:
948 before = &fl->fl_next;
949 }
950
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700951 /*
952 * The above code only modifies existing locks in case of
953 * merging or replacing. If new lock(s) need to be inserted
954 * all modifications are done bellow this, so it's safe yet to
955 * bail out.
956 */
957 error = -ENOLCK; /* "no luck" */
958 if (right && left == right && !new_fl2)
959 goto out;
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 error = 0;
962 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400963 if (request->fl_type == F_UNLCK) {
964 if (request->fl_flags & FL_EXISTS)
965 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400967 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700968
969 if (!new_fl) {
970 error = -ENOLCK;
971 goto out;
972 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 locks_copy_lock(new_fl, request);
974 locks_insert_lock(before, new_fl);
975 new_fl = NULL;
976 }
977 if (right) {
978 if (left == right) {
979 /* The new lock breaks the old one in two pieces,
980 * so we have to use the second new lock.
981 */
982 left = new_fl2;
983 new_fl2 = NULL;
984 locks_copy_lock(left, right);
985 locks_insert_lock(before, left);
986 }
987 right->fl_start = request->fl_end + 1;
988 locks_wake_up_blocks(right);
989 }
990 if (left) {
991 left->fl_end = request->fl_start - 1;
992 locks_wake_up_blocks(left);
993 }
994 out:
995 unlock_kernel();
996 /*
997 * Free any unused locks.
998 */
999 if (new_fl)
1000 locks_free_lock(new_fl);
1001 if (new_fl2)
1002 locks_free_lock(new_fl2);
1003 return error;
1004}
1005
1006/**
1007 * posix_lock_file - Apply a POSIX-style lock to a file
1008 * @filp: The file to apply the lock to
1009 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001010 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 *
1012 * Add a POSIX style lock to a file.
1013 * We merge adjacent & overlapping locks whenever possible.
1014 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001015 *
1016 * Note that if called with an FL_EXISTS argument, the caller may determine
1017 * whether or not a lock was successfully freed by testing the return
1018 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 */
Marc Eshel150b3932007-01-18 16:15:35 -05001020int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001021 struct file_lock *conflock)
1022{
Marc Eshel150b3932007-01-18 16:15:35 -05001023 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001024}
Marc Eshel150b3932007-01-18 16:15:35 -05001025EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026
1027/**
1028 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1029 * @filp: The file to apply the lock to
1030 * @fl: The lock to be applied
1031 *
1032 * Add a POSIX style lock to a file.
1033 * We merge adjacent & overlapping locks whenever possible.
1034 * POSIX locks are sorted by owner task, then by starting address
1035 */
1036int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1037{
1038 int error;
1039 might_sleep ();
1040 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001041 error = posix_lock_file(filp, fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1043 break;
1044 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1045 if (!error)
1046 continue;
1047
1048 locks_delete_block(fl);
1049 break;
1050 }
1051 return error;
1052}
1053EXPORT_SYMBOL(posix_lock_file_wait);
1054
1055/**
1056 * locks_mandatory_locked - Check for an active lock
1057 * @inode: the file to check
1058 *
1059 * Searches the inode's list of locks to find any POSIX locks which conflict.
1060 * This function is called from locks_verify_locked() only.
1061 */
1062int locks_mandatory_locked(struct inode *inode)
1063{
1064 fl_owner_t owner = current->files;
1065 struct file_lock *fl;
1066
1067 /*
1068 * Search the lock list for this inode for any POSIX locks.
1069 */
1070 lock_kernel();
1071 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1072 if (!IS_POSIX(fl))
1073 continue;
1074 if (fl->fl_owner != owner)
1075 break;
1076 }
1077 unlock_kernel();
1078 return fl ? -EAGAIN : 0;
1079}
1080
1081/**
1082 * locks_mandatory_area - Check for a conflicting lock
1083 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1084 * for shared
1085 * @inode: the file to check
1086 * @filp: how the file was opened (if it was)
1087 * @offset: start of area to check
1088 * @count: length of area to check
1089 *
1090 * Searches the inode's list of locks to find any POSIX locks which conflict.
1091 * This function is called from rw_verify_area() and
1092 * locks_verify_truncate().
1093 */
1094int locks_mandatory_area(int read_write, struct inode *inode,
1095 struct file *filp, loff_t offset,
1096 size_t count)
1097{
1098 struct file_lock fl;
1099 int error;
1100
1101 locks_init_lock(&fl);
1102 fl.fl_owner = current->files;
1103 fl.fl_pid = current->tgid;
1104 fl.fl_file = filp;
1105 fl.fl_flags = FL_POSIX | FL_ACCESS;
1106 if (filp && !(filp->f_flags & O_NONBLOCK))
1107 fl.fl_flags |= FL_SLEEP;
1108 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1109 fl.fl_start = offset;
1110 fl.fl_end = offset + count - 1;
1111
1112 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001113 error = __posix_lock_file(inode, &fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 if (error != -EAGAIN)
1115 break;
1116 if (!(fl.fl_flags & FL_SLEEP))
1117 break;
1118 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1119 if (!error) {
1120 /*
1121 * If we've been sleeping someone might have
1122 * changed the permissions behind our back.
1123 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001124 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 continue;
1126 }
1127
1128 locks_delete_block(&fl);
1129 break;
1130 }
1131
1132 return error;
1133}
1134
1135EXPORT_SYMBOL(locks_mandatory_area);
1136
1137/* We already had a lease on this file; just change its type */
1138int lease_modify(struct file_lock **before, int arg)
1139{
1140 struct file_lock *fl = *before;
1141 int error = assign_type(fl, arg);
1142
1143 if (error)
1144 return error;
1145 locks_wake_up_blocks(fl);
1146 if (arg == F_UNLCK)
1147 locks_delete_lock(before);
1148 return 0;
1149}
1150
1151EXPORT_SYMBOL(lease_modify);
1152
1153static void time_out_leases(struct inode *inode)
1154{
1155 struct file_lock **before;
1156 struct file_lock *fl;
1157
1158 before = &inode->i_flock;
1159 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1160 if ((fl->fl_break_time == 0)
1161 || time_before(jiffies, fl->fl_break_time)) {
1162 before = &fl->fl_next;
1163 continue;
1164 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1166 if (fl == *before) /* lease_modify may have freed fl */
1167 before = &fl->fl_next;
1168 }
1169}
1170
1171/**
1172 * __break_lease - revoke all outstanding leases on file
1173 * @inode: the inode of the file to return
1174 * @mode: the open mode (read or write)
1175 *
david m. richter87250dd2007-05-09 16:10:27 -04001176 * break_lease (inlined for speed) has checked there already is at least
1177 * some kind of lock (maybe a lease) on this file. Leases are broken on
1178 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 * specified %O_NONBLOCK to your open().
1180 */
1181int __break_lease(struct inode *inode, unsigned int mode)
1182{
1183 int error = 0, future;
1184 struct file_lock *new_fl, *flock;
1185 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 unsigned long break_time;
1187 int i_have_this_lease = 0;
1188
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001189 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
1191 lock_kernel();
1192
1193 time_out_leases(inode);
1194
1195 flock = inode->i_flock;
1196 if ((flock == NULL) || !IS_LEASE(flock))
1197 goto out;
1198
1199 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1200 if (fl->fl_owner == current->files)
1201 i_have_this_lease = 1;
1202
1203 if (mode & FMODE_WRITE) {
1204 /* If we want write access, we have to revoke any lease. */
1205 future = F_UNLCK | F_INPROGRESS;
1206 } else if (flock->fl_type & F_INPROGRESS) {
1207 /* If the lease is already being broken, we just leave it */
1208 future = flock->fl_type;
1209 } else if (flock->fl_type & F_WRLCK) {
1210 /* Downgrade the exclusive lease to a read-only lease. */
1211 future = F_RDLCK | F_INPROGRESS;
1212 } else {
1213 /* the existing lease was read-only, so we can read too. */
1214 goto out;
1215 }
1216
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001217 if (IS_ERR(new_fl) && !i_have_this_lease
1218 && ((mode & O_NONBLOCK) == 0)) {
1219 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 goto out;
1221 }
1222
1223 break_time = 0;
1224 if (lease_break_time > 0) {
1225 break_time = jiffies + lease_break_time * HZ;
1226 if (break_time == 0)
1227 break_time++; /* so that 0 means no break time */
1228 }
1229
1230 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1231 if (fl->fl_type != future) {
1232 fl->fl_type = future;
1233 fl->fl_break_time = break_time;
1234 /* lease must have lmops break callback */
1235 fl->fl_lmops->fl_break(fl);
1236 }
1237 }
1238
1239 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1240 error = -EWOULDBLOCK;
1241 goto out;
1242 }
1243
1244restart:
1245 break_time = flock->fl_break_time;
1246 if (break_time != 0) {
1247 break_time -= jiffies;
1248 if (break_time == 0)
1249 break_time++;
1250 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001251 locks_insert_block(flock, new_fl);
1252 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1253 !new_fl->fl_next, break_time);
1254 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 if (error >= 0) {
1256 if (error == 0)
1257 time_out_leases(inode);
1258 /* Wait for the next lease that has not been broken yet */
1259 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1260 flock = flock->fl_next) {
1261 if (flock->fl_type & F_INPROGRESS)
1262 goto restart;
1263 }
1264 error = 0;
1265 }
1266
1267out:
1268 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001269 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 locks_free_lock(new_fl);
1271 return error;
1272}
1273
1274EXPORT_SYMBOL(__break_lease);
1275
1276/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001277 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 * @inode: the inode
1279 * @time: pointer to a timespec which will contain the last modified time
1280 *
1281 * This is to force NFS clients to flush their caches for files with
1282 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001283 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 */
1285void lease_get_mtime(struct inode *inode, struct timespec *time)
1286{
1287 struct file_lock *flock = inode->i_flock;
1288 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1289 *time = current_fs_time(inode->i_sb);
1290 else
1291 *time = inode->i_mtime;
1292}
1293
1294EXPORT_SYMBOL(lease_get_mtime);
1295
1296/**
1297 * fcntl_getlease - Enquire what lease is currently active
1298 * @filp: the file
1299 *
1300 * The value returned by this function will be one of
1301 * (if no lease break is pending):
1302 *
1303 * %F_RDLCK to indicate a shared lease is held.
1304 *
1305 * %F_WRLCK to indicate an exclusive lease is held.
1306 *
1307 * %F_UNLCK to indicate no lease is held.
1308 *
1309 * (if a lease break is pending):
1310 *
1311 * %F_RDLCK to indicate an exclusive lease needs to be
1312 * changed to a shared lease (or removed).
1313 *
1314 * %F_UNLCK to indicate the lease needs to be removed.
1315 *
1316 * XXX: sfr & willy disagree over whether F_INPROGRESS
1317 * should be returned to userspace.
1318 */
1319int fcntl_getlease(struct file *filp)
1320{
1321 struct file_lock *fl;
1322 int type = F_UNLCK;
1323
1324 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001325 time_out_leases(filp->f_path.dentry->d_inode);
1326 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 fl = fl->fl_next) {
1328 if (fl->fl_file == filp) {
1329 type = fl->fl_type & ~F_INPROGRESS;
1330 break;
1331 }
1332 }
1333 unlock_kernel();
1334 return type;
1335}
1336
1337/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001338 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 * @filp: file pointer
1340 * @arg: type of lease to obtain
1341 * @flp: input - file_lock to use, output - file_lock inserted
1342 *
1343 * The (input) flp->fl_lmops->fl_break function is required
1344 * by break_lease().
1345 *
1346 * Called with kernel lock held.
1347 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001348int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001350 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001351 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001352 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 struct inode *inode = dentry->d_inode;
1354 int error, rdlease_count = 0, wrlease_count = 0;
1355
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001356 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1357 return -EACCES;
1358 if (!S_ISREG(inode->i_mode))
1359 return -EINVAL;
1360 error = security_file_lock(filp, arg);
1361 if (error)
1362 return error;
1363
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 time_out_leases(inode);
1365
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001366 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001368 lease = *flp;
1369
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 error = -EAGAIN;
1371 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1372 goto out;
1373 if ((arg == F_WRLCK)
1374 && ((atomic_read(&dentry->d_count) > 1)
1375 || (atomic_read(&inode->i_count) > 1)))
1376 goto out;
1377
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001378 error = -ENOMEM;
1379 new_fl = locks_alloc_lock();
1380 if (new_fl == NULL)
1381 goto out;
1382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 /*
1384 * At this point, we know that if there is an exclusive
1385 * lease on this file, then we hold it on this filp
1386 * (otherwise our open of this file would have blocked).
1387 * And if we are trying to acquire an exclusive lease,
1388 * then the file is not open by anyone (including us)
1389 * except for this filp.
1390 */
1391 for (before = &inode->i_flock;
1392 ((fl = *before) != NULL) && IS_LEASE(fl);
1393 before = &fl->fl_next) {
1394 if (lease->fl_lmops->fl_mylease(fl, lease))
1395 my_before = before;
1396 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1397 /*
1398 * Someone is in the process of opening this
1399 * file for writing so we may not take an
1400 * exclusive lease on it.
1401 */
1402 wrlease_count++;
1403 else
1404 rdlease_count++;
1405 }
1406
1407 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1408 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1409 goto out;
1410
1411 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001412 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 error = lease->fl_lmops->fl_change(my_before, arg);
1414 goto out;
1415 }
1416
1417 error = 0;
1418 if (arg == F_UNLCK)
1419 goto out;
1420
1421 error = -EINVAL;
1422 if (!leases_enable)
1423 goto out;
1424
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001425 locks_copy_lock(new_fl, lease);
1426 locks_insert_lock(before, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001428 *flp = new_fl;
1429 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431out:
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001432 if (new_fl != NULL)
1433 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 return error;
1435}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001436EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
1438 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001439 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 * @filp: file pointer
1441 * @arg: type of lease to obtain
1442 * @lease: file_lock to use
1443 *
1444 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001445 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1446 * break_lease will oops!
1447 *
1448 * This will call the filesystem's setlease file method, if
1449 * defined. Note that there is no getlease method; instead, the
1450 * filesystem setlease method should call back to setlease() to
1451 * add a lease to the inode's lease list, where fcntl_getlease() can
1452 * find it. Since fcntl_getlease() only reports whether the current
1453 * task holds a lease, a cluster filesystem need only do this for
1454 * leases held by processes on this node.
1455 *
1456 * There is also no break_lease method; filesystems that
1457 * handle their own leases shoud break leases themselves from the
1458 * filesystem's open, create, and (on truncate) setattr methods.
1459 *
1460 * Warning: the only current setlease methods exist only to disable
1461 * leases in certain cases. More vfs changes may be required to
1462 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 */
1464
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001465int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 int error;
1468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001470 if (filp->f_op && filp->f_op->setlease)
1471 error = filp->f_op->setlease(filp, arg, lease);
1472 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001473 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 unlock_kernel();
1475
1476 return error;
1477}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001478EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
1480/**
1481 * fcntl_setlease - sets a lease on an open file
1482 * @fd: open file descriptor
1483 * @filp: file pointer
1484 * @arg: type of lease to obtain
1485 *
1486 * Call this fcntl to establish a lease on the file.
1487 * Note that you also need to call %F_SETSIG to
1488 * receive a signal when the lease is broken.
1489 */
1490int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1491{
1492 struct file_lock fl, *flp = &fl;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001493 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 struct inode *inode = dentry->d_inode;
1495 int error;
1496
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 locks_init_lock(&fl);
1498 error = lease_init(filp, arg, &fl);
1499 if (error)
1500 return error;
1501
1502 lock_kernel();
1503
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001504 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001505 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 goto out_unlock;
1507
1508 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1509 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001510 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001512 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 time_out_leases(inode);
1514 goto out_unlock;
1515 }
1516
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001517 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518out_unlock:
1519 unlock_kernel();
1520 return error;
1521}
1522
1523/**
1524 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1525 * @filp: The file to apply the lock to
1526 * @fl: The lock to be applied
1527 *
1528 * Add a FLOCK style lock to a file.
1529 */
1530int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1531{
1532 int error;
1533 might_sleep();
1534 for (;;) {
1535 error = flock_lock_file(filp, fl);
1536 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1537 break;
1538 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1539 if (!error)
1540 continue;
1541
1542 locks_delete_block(fl);
1543 break;
1544 }
1545 return error;
1546}
1547
1548EXPORT_SYMBOL(flock_lock_file_wait);
1549
1550/**
1551 * sys_flock: - flock() system call.
1552 * @fd: the file descriptor to lock.
1553 * @cmd: the type of lock to apply.
1554 *
1555 * Apply a %FL_FLOCK style lock to an open file descriptor.
1556 * The @cmd can be one of
1557 *
1558 * %LOCK_SH -- a shared lock.
1559 *
1560 * %LOCK_EX -- an exclusive lock.
1561 *
1562 * %LOCK_UN -- remove an existing lock.
1563 *
1564 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1565 *
1566 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1567 * processes read and write access respectively.
1568 */
1569asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1570{
1571 struct file *filp;
1572 struct file_lock *lock;
1573 int can_sleep, unlock;
1574 int error;
1575
1576 error = -EBADF;
1577 filp = fget(fd);
1578 if (!filp)
1579 goto out;
1580
1581 can_sleep = !(cmd & LOCK_NB);
1582 cmd &= ~LOCK_NB;
1583 unlock = (cmd == LOCK_UN);
1584
1585 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1586 goto out_putf;
1587
1588 error = flock_make_lock(filp, &lock, cmd);
1589 if (error)
1590 goto out_putf;
1591 if (can_sleep)
1592 lock->fl_flags |= FL_SLEEP;
1593
1594 error = security_file_lock(filp, cmd);
1595 if (error)
1596 goto out_free;
1597
1598 if (filp->f_op && filp->f_op->flock)
1599 error = filp->f_op->flock(filp,
1600 (can_sleep) ? F_SETLKW : F_SETLK,
1601 lock);
1602 else
1603 error = flock_lock_file_wait(filp, lock);
1604
1605 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001606 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607
1608 out_putf:
1609 fput(filp);
1610 out:
1611 return error;
1612}
1613
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001614/**
1615 * vfs_test_lock - test file byte range lock
1616 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001617 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001618 *
1619 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1620 * setting conf->fl_type to something other than F_UNLCK.
1621 */
1622int vfs_test_lock(struct file *filp, struct file_lock *fl)
1623{
1624 if (filp->f_op && filp->f_op->lock)
1625 return filp->f_op->lock(filp, F_GETLK, fl);
1626 posix_test_lock(filp, fl);
1627 return 0;
1628}
1629EXPORT_SYMBOL_GPL(vfs_test_lock);
1630
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001631static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1632{
1633 flock->l_pid = fl->fl_pid;
1634#if BITS_PER_LONG == 32
1635 /*
1636 * Make sure we can represent the posix lock via
1637 * legacy 32bit flock.
1638 */
1639 if (fl->fl_start > OFFT_OFFSET_MAX)
1640 return -EOVERFLOW;
1641 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1642 return -EOVERFLOW;
1643#endif
1644 flock->l_start = fl->fl_start;
1645 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1646 fl->fl_end - fl->fl_start + 1;
1647 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001648 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001649 return 0;
1650}
1651
1652#if BITS_PER_LONG == 32
1653static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1654{
1655 flock->l_pid = fl->fl_pid;
1656 flock->l_start = fl->fl_start;
1657 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1658 fl->fl_end - fl->fl_start + 1;
1659 flock->l_whence = 0;
1660 flock->l_type = fl->fl_type;
1661}
1662#endif
1663
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664/* Report the first existing lock that would conflict with l.
1665 * This implements the F_GETLK command of fcntl().
1666 */
1667int fcntl_getlk(struct file *filp, struct flock __user *l)
1668{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001669 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 struct flock flock;
1671 int error;
1672
1673 error = -EFAULT;
1674 if (copy_from_user(&flock, l, sizeof(flock)))
1675 goto out;
1676 error = -EINVAL;
1677 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1678 goto out;
1679
1680 error = flock_to_posix_lock(filp, &file_lock, &flock);
1681 if (error)
1682 goto out;
1683
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001684 error = vfs_test_lock(filp, &file_lock);
1685 if (error)
1686 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001688 flock.l_type = file_lock.fl_type;
1689 if (file_lock.fl_type != F_UNLCK) {
1690 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001691 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 }
1694 error = -EFAULT;
1695 if (!copy_to_user(l, &flock, sizeof(flock)))
1696 error = 0;
1697out:
1698 return error;
1699}
1700
Marc Eshel7723ec92007-01-18 15:08:55 -05001701/**
1702 * vfs_lock_file - file byte range lock
1703 * @filp: The file to apply the lock to
1704 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1705 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001706 * @conf: Place to return a copy of the conflicting lock, if found.
1707 *
1708 * A caller that doesn't care about the conflicting lock may pass NULL
1709 * as the final argument.
1710 *
1711 * If the filesystem defines a private ->lock() method, then @conf will
1712 * be left unchanged; so a caller that cares should initialize it to
1713 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001714 *
1715 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1716 * locks, the ->lock() interface may return asynchronously, before the lock has
1717 * been granted or denied by the underlying filesystem, if (and only if)
1718 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1719 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1720 * the request is for a blocking lock. When ->lock() does return asynchronously,
1721 * it must return -EINPROGRESS, and call ->fl_grant() when the lock
1722 * request completes.
1723 * If the request is for non-blocking lock the file system should return
1724 * -EINPROGRESS then try to get the lock and call the callback routine with
1725 * the result. If the request timed out the callback routine will return a
1726 * nonzero return code and the file system should release the lock. The file
1727 * system is also responsible to keep a corresponding posix lock when it
1728 * grants a lock so the VFS can find out which locks are locally held and do
1729 * the correct lock cleanup when required.
1730 * The underlying filesystem must not drop the kernel lock or call
1731 * ->fl_grant() before returning to the caller with a -EINPROGRESS
1732 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001733 */
Marc Eshel150b3932007-01-18 16:15:35 -05001734int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001735{
1736 if (filp->f_op && filp->f_op->lock)
1737 return filp->f_op->lock(filp, cmd, fl);
1738 else
Marc Eshel150b3932007-01-18 16:15:35 -05001739 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001740}
1741EXPORT_SYMBOL_GPL(vfs_lock_file);
1742
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743/* Apply the lock described by l to an open file descriptor.
1744 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1745 */
Peter Staubachc2936212005-07-27 11:45:09 -07001746int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1747 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748{
1749 struct file_lock *file_lock = locks_alloc_lock();
1750 struct flock flock;
1751 struct inode *inode;
1752 int error;
1753
1754 if (file_lock == NULL)
1755 return -ENOLCK;
1756
1757 /*
1758 * This might block, so we do it before checking the inode.
1759 */
1760 error = -EFAULT;
1761 if (copy_from_user(&flock, l, sizeof(flock)))
1762 goto out;
1763
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001764 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765
1766 /* Don't allow mandatory locks on files that may be memory mapped
1767 * and shared.
1768 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001769 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 error = -EAGAIN;
1771 goto out;
1772 }
1773
Peter Staubachc2936212005-07-27 11:45:09 -07001774again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 error = flock_to_posix_lock(filp, file_lock, &flock);
1776 if (error)
1777 goto out;
1778 if (cmd == F_SETLKW) {
1779 file_lock->fl_flags |= FL_SLEEP;
1780 }
1781
1782 error = -EBADF;
1783 switch (flock.l_type) {
1784 case F_RDLCK:
1785 if (!(filp->f_mode & FMODE_READ))
1786 goto out;
1787 break;
1788 case F_WRLCK:
1789 if (!(filp->f_mode & FMODE_WRITE))
1790 goto out;
1791 break;
1792 case F_UNLCK:
1793 break;
1794 default:
1795 error = -EINVAL;
1796 goto out;
1797 }
1798
1799 error = security_file_lock(filp, file_lock->fl_type);
1800 if (error)
1801 goto out;
1802
J. Bruce Fields19e729a2008-04-14 15:03:02 -04001803 if (filp->f_op && filp->f_op->lock != NULL)
1804 error = filp->f_op->lock(filp, cmd, file_lock);
1805 else {
1806 for (;;) {
1807 error = posix_lock_file(filp, file_lock, NULL);
1808 if (error != -EAGAIN || cmd == F_SETLK)
1809 break;
1810 error = wait_event_interruptible(file_lock->fl_wait,
1811 !file_lock->fl_next);
1812 if (!error)
1813 continue;
Marc Eshel7723ec92007-01-18 15:08:55 -05001814
J. Bruce Fields19e729a2008-04-14 15:03:02 -04001815 locks_delete_block(file_lock);
1816 break;
1817 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 }
1819
Peter Staubachc2936212005-07-27 11:45:09 -07001820 /*
1821 * Attempt to detect a close/fcntl race and recover by
1822 * releasing the lock that was just acquired.
1823 */
1824 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1825 flock.l_type = F_UNLCK;
1826 goto again;
1827 }
1828
1829out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 locks_free_lock(file_lock);
1831 return error;
1832}
1833
1834#if BITS_PER_LONG == 32
1835/* Report the first existing lock that would conflict with l.
1836 * This implements the F_GETLK command of fcntl().
1837 */
1838int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1839{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001840 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841 struct flock64 flock;
1842 int error;
1843
1844 error = -EFAULT;
1845 if (copy_from_user(&flock, l, sizeof(flock)))
1846 goto out;
1847 error = -EINVAL;
1848 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1849 goto out;
1850
1851 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1852 if (error)
1853 goto out;
1854
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001855 error = vfs_test_lock(filp, &file_lock);
1856 if (error)
1857 goto out;
1858
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001859 flock.l_type = file_lock.fl_type;
1860 if (file_lock.fl_type != F_UNLCK)
1861 posix_lock_to_flock64(&flock, &file_lock);
1862
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 error = -EFAULT;
1864 if (!copy_to_user(l, &flock, sizeof(flock)))
1865 error = 0;
1866
1867out:
1868 return error;
1869}
1870
1871/* Apply the lock described by l to an open file descriptor.
1872 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1873 */
Peter Staubachc2936212005-07-27 11:45:09 -07001874int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1875 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
1877 struct file_lock *file_lock = locks_alloc_lock();
1878 struct flock64 flock;
1879 struct inode *inode;
1880 int error;
1881
1882 if (file_lock == NULL)
1883 return -ENOLCK;
1884
1885 /*
1886 * This might block, so we do it before checking the inode.
1887 */
1888 error = -EFAULT;
1889 if (copy_from_user(&flock, l, sizeof(flock)))
1890 goto out;
1891
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001892 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894 /* Don't allow mandatory locks on files that may be memory mapped
1895 * and shared.
1896 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001897 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 error = -EAGAIN;
1899 goto out;
1900 }
1901
Peter Staubachc2936212005-07-27 11:45:09 -07001902again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 error = flock64_to_posix_lock(filp, file_lock, &flock);
1904 if (error)
1905 goto out;
1906 if (cmd == F_SETLKW64) {
1907 file_lock->fl_flags |= FL_SLEEP;
1908 }
1909
1910 error = -EBADF;
1911 switch (flock.l_type) {
1912 case F_RDLCK:
1913 if (!(filp->f_mode & FMODE_READ))
1914 goto out;
1915 break;
1916 case F_WRLCK:
1917 if (!(filp->f_mode & FMODE_WRITE))
1918 goto out;
1919 break;
1920 case F_UNLCK:
1921 break;
1922 default:
1923 error = -EINVAL;
1924 goto out;
1925 }
1926
1927 error = security_file_lock(filp, file_lock->fl_type);
1928 if (error)
1929 goto out;
1930
J. Bruce Fields19e729a2008-04-14 15:03:02 -04001931 if (filp->f_op && filp->f_op->lock != NULL)
1932 error = filp->f_op->lock(filp, cmd, file_lock);
1933 else {
1934 for (;;) {
1935 error = posix_lock_file(filp, file_lock, NULL);
1936 if (error != -EAGAIN || cmd == F_SETLK64)
1937 break;
1938 error = wait_event_interruptible(file_lock->fl_wait,
1939 !file_lock->fl_next);
1940 if (!error)
1941 continue;
Marc Eshel7723ec92007-01-18 15:08:55 -05001942
J. Bruce Fields19e729a2008-04-14 15:03:02 -04001943 locks_delete_block(file_lock);
1944 break;
1945 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 }
1947
Peter Staubachc2936212005-07-27 11:45:09 -07001948 /*
1949 * Attempt to detect a close/fcntl race and recover by
1950 * releasing the lock that was just acquired.
1951 */
1952 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1953 flock.l_type = F_UNLCK;
1954 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 }
1956
1957out:
1958 locks_free_lock(file_lock);
1959 return error;
1960}
1961#endif /* BITS_PER_LONG == 32 */
1962
1963/*
1964 * This function is called when the file is being removed
1965 * from the task's fd array. POSIX locks belonging to this task
1966 * are deleted at this time.
1967 */
1968void locks_remove_posix(struct file *filp, fl_owner_t owner)
1969{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001970 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
1972 /*
1973 * If there are no locks held on this file, we don't need to call
1974 * posix_lock_file(). Another process could be setting a lock on this
1975 * file at the same time, but we wouldn't remove that lock anyway.
1976 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001977 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 return;
1979
1980 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001981 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 lock.fl_start = 0;
1983 lock.fl_end = OFFSET_MAX;
1984 lock.fl_owner = owner;
1985 lock.fl_pid = current->tgid;
1986 lock.fl_file = filp;
1987 lock.fl_ops = NULL;
1988 lock.fl_lmops = NULL;
1989
Marc Eshel150b3932007-01-18 16:15:35 -05001990 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1993 lock.fl_ops->fl_release_private(&lock);
1994}
1995
1996EXPORT_SYMBOL(locks_remove_posix);
1997
1998/*
1999 * This function is called on the last close of an open file.
2000 */
2001void locks_remove_flock(struct file *filp)
2002{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002003 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 struct file_lock *fl;
2005 struct file_lock **before;
2006
2007 if (!inode->i_flock)
2008 return;
2009
2010 if (filp->f_op && filp->f_op->flock) {
2011 struct file_lock fl = {
2012 .fl_pid = current->tgid,
2013 .fl_file = filp,
2014 .fl_flags = FL_FLOCK,
2015 .fl_type = F_UNLCK,
2016 .fl_end = OFFSET_MAX,
2017 };
2018 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002019 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2020 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 }
2022
2023 lock_kernel();
2024 before = &inode->i_flock;
2025
2026 while ((fl = *before) != NULL) {
2027 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002028 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 locks_delete_lock(before);
2030 continue;
2031 }
2032 if (IS_LEASE(fl)) {
2033 lease_modify(before, F_UNLCK);
2034 continue;
2035 }
2036 /* What? */
2037 BUG();
2038 }
2039 before = &fl->fl_next;
2040 }
2041 unlock_kernel();
2042}
2043
2044/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 * posix_unblock_lock - stop waiting for a file lock
2046 * @filp: how the file was opened
2047 * @waiter: the lock which was waiting
2048 *
2049 * lockd needs to block waiting for locks.
2050 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002051int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2053{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002054 int status = 0;
2055
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002057 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002059 else
2060 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002061 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002062 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063}
2064
2065EXPORT_SYMBOL(posix_unblock_lock);
2066
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002067/**
2068 * vfs_cancel_lock - file byte range unblock lock
2069 * @filp: The file to apply the unblock to
2070 * @fl: The lock to be unblocked
2071 *
2072 * Used by lock managers to cancel blocked requests
2073 */
2074int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2075{
2076 if (filp->f_op && filp->f_op->lock)
2077 return filp->f_op->lock(filp, F_CANCELLK, fl);
2078 return 0;
2079}
2080
2081EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2082
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002083#ifdef CONFIG_PROC_FS
2084#include <linux/seq_file.h>
2085
2086static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2087 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088{
2089 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002090 unsigned int fl_pid;
2091
2092 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002093 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002094 else
2095 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
2097 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002098 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002100 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002102 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2104 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002105 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 } else if (IS_FLOCK(fl)) {
2107 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002108 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002110 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 }
2112 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002113 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002115 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002117 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002119 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002121 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 }
2123 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002124 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 (fl->fl_type & LOCK_READ)
2126 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2127 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2128 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002129 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 (fl->fl_type & F_INPROGRESS)
2131 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2132 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2133 }
2134 if (inode) {
2135#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002136 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 inode->i_sb->s_id, inode->i_ino);
2138#else
2139 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002140 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 MAJOR(inode->i_sb->s_dev),
2142 MINOR(inode->i_sb->s_dev), inode->i_ino);
2143#endif
2144 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002145 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 }
2147 if (IS_POSIX(fl)) {
2148 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002149 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002151 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002153 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 }
2155}
2156
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002157static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002159 struct file_lock *fl, *bfl;
2160
2161 fl = list_entry(v, struct file_lock, fl_link);
2162
2163 lock_get_status(f, fl, (long)f->private, "");
2164
2165 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2166 lock_get_status(f, bfl, (long)f->private, " ->");
2167
2168 f->private++;
2169 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170}
2171
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002172static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002175 f->private = (void *)1;
2176 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177}
2178
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002179static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2180{
2181 return seq_list_next(v, &file_lock_list, pos);
2182}
2183
2184static void locks_stop(struct seq_file *f, void *v)
2185{
2186 unlock_kernel();
2187}
2188
2189struct seq_operations locks_seq_operations = {
2190 .start = locks_start,
2191 .next = locks_next,
2192 .stop = locks_stop,
2193 .show = locks_show,
2194};
2195#endif
2196
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197/**
2198 * lock_may_read - checks that the region is free of locks
2199 * @inode: the inode that is being read
2200 * @start: the first byte to read
2201 * @len: the number of bytes to read
2202 *
2203 * Emulates Windows locking requirements. Whole-file
2204 * mandatory locks (share modes) can prohibit a read and
2205 * byte-range POSIX locks can prohibit a read if they overlap.
2206 *
2207 * N.B. this function is only ever called
2208 * from knfsd and ownership of locks is never checked.
2209 */
2210int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2211{
2212 struct file_lock *fl;
2213 int result = 1;
2214 lock_kernel();
2215 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2216 if (IS_POSIX(fl)) {
2217 if (fl->fl_type == F_RDLCK)
2218 continue;
2219 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2220 continue;
2221 } else if (IS_FLOCK(fl)) {
2222 if (!(fl->fl_type & LOCK_MAND))
2223 continue;
2224 if (fl->fl_type & LOCK_READ)
2225 continue;
2226 } else
2227 continue;
2228 result = 0;
2229 break;
2230 }
2231 unlock_kernel();
2232 return result;
2233}
2234
2235EXPORT_SYMBOL(lock_may_read);
2236
2237/**
2238 * lock_may_write - checks that the region is free of locks
2239 * @inode: the inode that is being written
2240 * @start: the first byte to write
2241 * @len: the number of bytes to write
2242 *
2243 * Emulates Windows locking requirements. Whole-file
2244 * mandatory locks (share modes) can prohibit a write and
2245 * byte-range POSIX locks can prohibit a write if they overlap.
2246 *
2247 * N.B. this function is only ever called
2248 * from knfsd and ownership of locks is never checked.
2249 */
2250int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2251{
2252 struct file_lock *fl;
2253 int result = 1;
2254 lock_kernel();
2255 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2256 if (IS_POSIX(fl)) {
2257 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2258 continue;
2259 } else if (IS_FLOCK(fl)) {
2260 if (!(fl->fl_type & LOCK_MAND))
2261 continue;
2262 if (fl->fl_type & LOCK_WRITE)
2263 continue;
2264 } else
2265 continue;
2266 result = 0;
2267 break;
2268 }
2269 unlock_kernel();
2270 return result;
2271}
2272
2273EXPORT_SYMBOL(lock_may_write);
2274
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275static int __init filelock_init(void)
2276{
2277 filelock_cache = kmem_cache_create("file_lock_cache",
2278 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002279 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 return 0;
2281}
2282
2283core_initcall(filelock_init);