blob: d83fab1b77b56a4e065d6858440903ed6d70770e [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
130#include <asm/semaphore.h>
131#include <asm/uaccess.h>
132
133#define IS_POSIX(fl) (fl->fl_flags & FL_POSIX)
134#define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK)
135#define IS_LEASE(fl) (fl->fl_flags & FL_LEASE)
136
137int leases_enable = 1;
138int lease_break_time = 45;
139
140#define for_each_lock(inode, lockp) \
141 for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next)
142
Christoph Hellwig26bcbf92006-03-20 13:44:40 -0500143static LIST_HEAD(file_lock_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144static LIST_HEAD(blocked_list);
145
Christoph Lametere18b8902006-12-06 20:33:20 -0800146static struct kmem_cache *filelock_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
148/* Allocate an empty lock structure. */
149static struct file_lock *locks_alloc_lock(void)
150{
Christoph Lametere94b1762006-12-06 20:33:17 -0800151 return kmem_cache_alloc(filelock_cache, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152}
153
Trond Myklebust47831f32006-03-20 13:44:05 -0500154static void locks_release_private(struct file_lock *fl)
155{
156 if (fl->fl_ops) {
157 if (fl->fl_ops->fl_release_private)
158 fl->fl_ops->fl_release_private(fl);
159 fl->fl_ops = NULL;
160 }
161 if (fl->fl_lmops) {
162 if (fl->fl_lmops->fl_release_private)
163 fl->fl_lmops->fl_release_private(fl);
164 fl->fl_lmops = NULL;
165 }
166
167}
168
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169/* Free a lock which is not in use. */
Matt Mackall33443c42006-01-08 01:05:22 -0800170static void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800172 BUG_ON(waitqueue_active(&fl->fl_wait));
173 BUG_ON(!list_empty(&fl->fl_block));
174 BUG_ON(!list_empty(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175
Trond Myklebust47831f32006-03-20 13:44:05 -0500176 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 kmem_cache_free(filelock_cache, fl);
178}
179
180void locks_init_lock(struct file_lock *fl)
181{
182 INIT_LIST_HEAD(&fl->fl_link);
183 INIT_LIST_HEAD(&fl->fl_block);
184 init_waitqueue_head(&fl->fl_wait);
185 fl->fl_next = NULL;
186 fl->fl_fasync = NULL;
187 fl->fl_owner = NULL;
188 fl->fl_pid = 0;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000189 fl->fl_nspid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 fl->fl_file = NULL;
191 fl->fl_flags = 0;
192 fl->fl_type = 0;
193 fl->fl_start = fl->fl_end = 0;
194 fl->fl_ops = NULL;
195 fl->fl_lmops = NULL;
196}
197
198EXPORT_SYMBOL(locks_init_lock);
199
200/*
201 * Initialises the fields of the file lock which are invariant for
202 * free file_locks.
203 */
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700204static void init_once(struct kmem_cache *cache, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205{
206 struct file_lock *lock = (struct file_lock *) foo;
207
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 locks_init_lock(lock);
209}
210
Trond Myklebust47831f32006-03-20 13:44:05 -0500211static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
212{
213 if (fl->fl_ops) {
214 if (fl->fl_ops->fl_copy_lock)
215 fl->fl_ops->fl_copy_lock(new, fl);
216 new->fl_ops = fl->fl_ops;
217 }
218 if (fl->fl_lmops) {
219 if (fl->fl_lmops->fl_copy_lock)
220 fl->fl_lmops->fl_copy_lock(new, fl);
221 new->fl_lmops = fl->fl_lmops;
222 }
223}
224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225/*
226 * Initialize a new lock from an existing file_lock structure.
227 */
Trond Myklebust09969052006-03-20 13:44:38 -0500228static void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229{
230 new->fl_owner = fl->fl_owner;
231 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500232 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 new->fl_flags = fl->fl_flags;
234 new->fl_type = fl->fl_type;
235 new->fl_start = fl->fl_start;
236 new->fl_end = fl->fl_end;
Trond Myklebust09969052006-03-20 13:44:38 -0500237 new->fl_ops = NULL;
238 new->fl_lmops = NULL;
239}
240
241void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
242{
243 locks_release_private(new);
244
245 __locks_copy_lock(new, fl);
246 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 new->fl_ops = fl->fl_ops;
248 new->fl_lmops = fl->fl_lmops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500249
250 locks_copy_private(new, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251}
252
253EXPORT_SYMBOL(locks_copy_lock);
254
255static inline int flock_translate_cmd(int cmd) {
256 if (cmd & LOCK_MAND)
257 return cmd & (LOCK_MAND | LOCK_RW);
258 switch (cmd) {
259 case LOCK_SH:
260 return F_RDLCK;
261 case LOCK_EX:
262 return F_WRLCK;
263 case LOCK_UN:
264 return F_UNLCK;
265 }
266 return -EINVAL;
267}
268
269/* Fill in a file_lock structure with an appropriate FLOCK lock. */
270static int flock_make_lock(struct file *filp, struct file_lock **lock,
271 unsigned int cmd)
272{
273 struct file_lock *fl;
274 int type = flock_translate_cmd(cmd);
275 if (type < 0)
276 return type;
277
278 fl = locks_alloc_lock();
279 if (fl == NULL)
280 return -ENOMEM;
281
282 fl->fl_file = filp;
283 fl->fl_pid = current->tgid;
284 fl->fl_flags = FL_FLOCK;
285 fl->fl_type = type;
286 fl->fl_end = OFFSET_MAX;
287
288 *lock = fl;
289 return 0;
290}
291
292static int assign_type(struct file_lock *fl, int type)
293{
294 switch (type) {
295 case F_RDLCK:
296 case F_WRLCK:
297 case F_UNLCK:
298 fl->fl_type = type;
299 break;
300 default:
301 return -EINVAL;
302 }
303 return 0;
304}
305
306/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
307 * style lock.
308 */
309static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
310 struct flock *l)
311{
312 off_t start, end;
313
314 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700315 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 start = 0;
317 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700318 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 start = filp->f_pos;
320 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700321 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800322 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 break;
324 default:
325 return -EINVAL;
326 }
327
328 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
329 POSIX-2001 defines it. */
330 start += l->l_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 if (start < 0)
332 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700333 fl->fl_end = OFFSET_MAX;
334 if (l->l_len > 0) {
335 end = start + l->l_len - 1;
336 fl->fl_end = end;
337 } else if (l->l_len < 0) {
338 end = start - 1;
339 fl->fl_end = end;
340 start += l->l_len;
341 if (start < 0)
342 return -EINVAL;
343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700345 if (fl->fl_end < fl->fl_start)
346 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
348 fl->fl_owner = current->files;
349 fl->fl_pid = current->tgid;
350 fl->fl_file = filp;
351 fl->fl_flags = FL_POSIX;
352 fl->fl_ops = NULL;
353 fl->fl_lmops = NULL;
354
355 return assign_type(fl, l->l_type);
356}
357
358#if BITS_PER_LONG == 32
359static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
360 struct flock64 *l)
361{
362 loff_t start;
363
364 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700365 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 start = 0;
367 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700368 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 start = filp->f_pos;
370 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700371 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800372 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 break;
374 default:
375 return -EINVAL;
376 }
377
Trond Myklebust4c780a42005-10-18 14:20:21 -0700378 start += l->l_start;
379 if (start < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700381 fl->fl_end = OFFSET_MAX;
382 if (l->l_len > 0) {
383 fl->fl_end = start + l->l_len - 1;
384 } else if (l->l_len < 0) {
385 fl->fl_end = start - 1;
386 start += l->l_len;
387 if (start < 0)
388 return -EINVAL;
389 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700391 if (fl->fl_end < fl->fl_start)
392 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
394 fl->fl_owner = current->files;
395 fl->fl_pid = current->tgid;
396 fl->fl_file = filp;
397 fl->fl_flags = FL_POSIX;
398 fl->fl_ops = NULL;
399 fl->fl_lmops = NULL;
400
401 switch (l->l_type) {
402 case F_RDLCK:
403 case F_WRLCK:
404 case F_UNLCK:
405 fl->fl_type = l->l_type;
406 break;
407 default:
408 return -EINVAL;
409 }
410
411 return (0);
412}
413#endif
414
415/* default lease lock manager operations */
416static void lease_break_callback(struct file_lock *fl)
417{
418 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
419}
420
421static void lease_release_private_callback(struct file_lock *fl)
422{
423 if (!fl->fl_file)
424 return;
425
426 f_delown(fl->fl_file);
427 fl->fl_file->f_owner.signum = 0;
428}
429
Adrian Bunk75c96f82005-05-05 16:16:09 -0700430static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431{
432 return fl->fl_file == try->fl_file;
433}
434
Adrian Bunk75c96f82005-05-05 16:16:09 -0700435static struct lock_manager_operations lease_manager_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 .fl_break = lease_break_callback,
437 .fl_release_private = lease_release_private_callback,
438 .fl_mylease = lease_mylease_callback,
439 .fl_change = lease_modify,
440};
441
442/*
443 * Initialize a lease, use the default lock manager operations
444 */
445static int lease_init(struct file *filp, int type, struct file_lock *fl)
446 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400447 if (assign_type(fl, type) != 0)
448 return -EINVAL;
449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 fl->fl_owner = current->files;
451 fl->fl_pid = current->tgid;
452
453 fl->fl_file = filp;
454 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 fl->fl_start = 0;
456 fl->fl_end = OFFSET_MAX;
457 fl->fl_ops = NULL;
458 fl->fl_lmops = &lease_manager_ops;
459 return 0;
460}
461
462/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500463static struct file_lock *lease_alloc(struct file *filp, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464{
465 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400466 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
468 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500469 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400472 if (error) {
473 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500474 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400475 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500476 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477}
478
479/* Check if two locks overlap each other.
480 */
481static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
482{
483 return ((fl1->fl_end >= fl2->fl_start) &&
484 (fl2->fl_end >= fl1->fl_start));
485}
486
487/*
488 * Check whether two locks have the same owner.
489 */
Matt Mackall33443c42006-01-08 01:05:22 -0800490static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491{
492 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
493 return fl2->fl_lmops == fl1->fl_lmops &&
494 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
495 return fl1->fl_owner == fl2->fl_owner;
496}
497
498/* Remove waiter from blocker's block list.
499 * When blocker ends up pointing to itself then the list is empty.
500 */
Matt Mackall33443c42006-01-08 01:05:22 -0800501static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 list_del_init(&waiter->fl_block);
504 list_del_init(&waiter->fl_link);
505 waiter->fl_next = NULL;
506}
507
508/*
509 */
510static void locks_delete_block(struct file_lock *waiter)
511{
512 lock_kernel();
513 __locks_delete_block(waiter);
514 unlock_kernel();
515}
516
517/* Insert waiter into blocker's block list.
518 * We use a circular list so that processes can be easily woken up in
519 * the order they blocked. The documentation doesn't require this but
520 * it seems like the reasonable thing to do.
521 */
522static void locks_insert_block(struct file_lock *blocker,
523 struct file_lock *waiter)
524{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800525 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 list_add_tail(&waiter->fl_block, &blocker->fl_block);
527 waiter->fl_next = blocker;
528 if (IS_POSIX(blocker))
529 list_add(&waiter->fl_link, &blocked_list);
530}
531
532/* Wake up processes blocked waiting for blocker.
533 * If told to wait then schedule the processes until the block list
534 * is empty, otherwise empty the block list ourselves.
535 */
536static void locks_wake_up_blocks(struct file_lock *blocker)
537{
538 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400539 struct file_lock *waiter;
540
541 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 struct file_lock, fl_block);
543 __locks_delete_block(waiter);
544 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
545 waiter->fl_lmops->fl_notify(waiter);
546 else
547 wake_up(&waiter->fl_wait);
548 }
549}
550
551/* Insert file lock fl into an inode's lock list at the position indicated
552 * by pos. At the same time add the lock to the global file lock list.
553 */
554static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
555{
556 list_add(&fl->fl_link, &file_lock_list);
557
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000558 fl->fl_nspid = get_pid(task_tgid(current));
559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 /* insert into file's list */
561 fl->fl_next = *pos;
562 *pos = fl;
563
564 if (fl->fl_ops && fl->fl_ops->fl_insert)
565 fl->fl_ops->fl_insert(fl);
566}
567
568/*
569 * Delete a lock and then free it.
570 * Wake up processes that are blocked waiting for this lock,
571 * notify the FS that the lock has been cleared and
572 * finally free the lock.
573 */
574static void locks_delete_lock(struct file_lock **thisfl_p)
575{
576 struct file_lock *fl = *thisfl_p;
577
578 *thisfl_p = fl->fl_next;
579 fl->fl_next = NULL;
580 list_del_init(&fl->fl_link);
581
582 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
583 if (fl->fl_fasync != NULL) {
584 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
585 fl->fl_fasync = NULL;
586 }
587
588 if (fl->fl_ops && fl->fl_ops->fl_remove)
589 fl->fl_ops->fl_remove(fl);
590
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000591 if (fl->fl_nspid) {
592 put_pid(fl->fl_nspid);
593 fl->fl_nspid = NULL;
594 }
595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 locks_wake_up_blocks(fl);
597 locks_free_lock(fl);
598}
599
600/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
601 * checks for shared/exclusive status of overlapping locks.
602 */
603static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
604{
605 if (sys_fl->fl_type == F_WRLCK)
606 return 1;
607 if (caller_fl->fl_type == F_WRLCK)
608 return 1;
609 return 0;
610}
611
612/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
613 * checking before calling the locks_conflict().
614 */
615static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
616{
617 /* POSIX locks owned by the same process do not conflict with
618 * each other.
619 */
620 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
621 return (0);
622
623 /* Check whether they overlap */
624 if (!locks_overlap(caller_fl, sys_fl))
625 return 0;
626
627 return (locks_conflict(caller_fl, sys_fl));
628}
629
630/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
631 * checking before calling the locks_conflict().
632 */
633static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
634{
635 /* FLOCK locks referring to the same filp do not conflict with
636 * each other.
637 */
638 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
639 return (0);
640 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
641 return 0;
642
643 return (locks_conflict(caller_fl, sys_fl));
644}
645
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400646void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500647posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648{
649 struct file_lock *cfl;
650
651 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800652 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 if (!IS_POSIX(cfl))
654 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400655 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 break;
657 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000658 if (cfl) {
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500659 __locks_copy_lock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000660 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800661 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000662 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400663 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400665 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667EXPORT_SYMBOL(posix_test_lock);
668
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400669/*
670 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400672 * We attempt to detect deadlocks that are due purely to posix file
673 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400675 * We assume that a task can be waiting for at most one lock at a time.
676 * So for any acquired lock, the process holding that lock may be
677 * waiting on at most one other lock. That lock in turns may be held by
678 * someone waiting for at most one other lock. Given a requested lock
679 * caller_fl which is about to wait for a conflicting lock block_fl, we
680 * follow this chain of waiters to ensure we are not about to create a
681 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400682 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400683 * Since we do this before we ever put a process to sleep on a lock, we
684 * are ensured that there is never a cycle; that is what guarantees that
685 * the while() loop in posix_locks_deadlock() eventually completes.
686 *
687 * Note: the above assumption may not be true when handling lock
688 * requests from a broken NFS client. It may also fail in the presence
689 * of tasks (such as posix threads) sharing the same open file table.
690 *
691 * To handle those cases, we just bail out after a few iterations.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400693
694#define MAX_DEADLK_ITERATIONS 10
695
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400696/* Find a lock that the owner of the given block_fl is blocking on. */
697static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
698{
699 struct file_lock *fl;
700
701 list_for_each_entry(fl, &blocked_list, fl_link) {
702 if (posix_same_owner(fl, block_fl))
703 return fl->fl_next;
704 }
705 return NULL;
706}
707
Adrian Bunkb0904e12006-06-23 02:05:13 -0700708static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 struct file_lock *block_fl)
710{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400711 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400713 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
714 if (i++ > MAX_DEADLK_ITERATIONS)
715 return 0;
716 if (posix_same_owner(caller_fl, block_fl))
717 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 }
719 return 0;
720}
721
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400723 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400724 *
725 * Note that if called with an FL_EXISTS argument, the caller may determine
726 * whether or not a lock was successfully freed by testing the return
727 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800729static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800731 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800733 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 int error = 0;
735 int found = 0;
736
737 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400738 if (request->fl_flags & FL_ACCESS)
739 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400740
741 if (request->fl_type != F_UNLCK) {
742 error = -ENOMEM;
743 new_fl = locks_alloc_lock();
744 if (new_fl == NULL)
745 goto out;
746 error = 0;
747 }
748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 for_each_lock(inode, before) {
750 struct file_lock *fl = *before;
751 if (IS_POSIX(fl))
752 break;
753 if (IS_LEASE(fl))
754 continue;
755 if (filp != fl->fl_file)
756 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800757 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 goto out;
759 found = 1;
760 locks_delete_lock(before);
761 break;
762 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
Trond Myklebustf475ae92006-06-29 16:38:32 -0400764 if (request->fl_type == F_UNLCK) {
765 if ((request->fl_flags & FL_EXISTS) && !found)
766 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800767 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400768 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
770 /*
771 * If a higher-priority process was blocked on the old file lock,
772 * give it the opportunity to lock the file.
773 */
774 if (found)
775 cond_resched();
776
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400777find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 for_each_lock(inode, before) {
779 struct file_lock *fl = *before;
780 if (IS_POSIX(fl))
781 break;
782 if (IS_LEASE(fl))
783 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800784 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 continue;
786 error = -EAGAIN;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800787 if (request->fl_flags & FL_SLEEP)
788 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 goto out;
790 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400791 if (request->fl_flags & FL_ACCESS)
792 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800793 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700794 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800795 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400796 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
798out:
799 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800800 if (new_fl)
801 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return error;
803}
804
Marc Eshel150b3932007-01-18 16:15:35 -0500805static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
807 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700808 struct file_lock *new_fl = NULL;
809 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 struct file_lock *left = NULL;
811 struct file_lock *right = NULL;
812 struct file_lock **before;
813 int error, added = 0;
814
815 /*
816 * We may need two file_lock structures for this operation,
817 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700818 *
819 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700821 if (!(request->fl_flags & FL_ACCESS) &&
822 (request->fl_type != F_UNLCK ||
823 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
824 new_fl = locks_alloc_lock();
825 new_fl2 = locks_alloc_lock();
826 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828 lock_kernel();
829 if (request->fl_type != F_UNLCK) {
830 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500831 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (!IS_POSIX(fl))
833 continue;
834 if (!posix_locks_conflict(request, fl))
835 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800836 if (conflock)
837 locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 error = -EAGAIN;
839 if (!(request->fl_flags & FL_SLEEP))
840 goto out;
841 error = -EDEADLK;
842 if (posix_locks_deadlock(request, fl))
843 goto out;
844 error = -EAGAIN;
845 locks_insert_block(fl, request);
846 goto out;
847 }
848 }
849
850 /* If we're just looking for a conflict, we're done. */
851 error = 0;
852 if (request->fl_flags & FL_ACCESS)
853 goto out;
854
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 * Find the first old lock with the same owner as the new lock.
857 */
858
859 before = &inode->i_flock;
860
861 /* First skip locks owned by other processes. */
862 while ((fl = *before) && (!IS_POSIX(fl) ||
863 !posix_same_owner(request, fl))) {
864 before = &fl->fl_next;
865 }
866
867 /* Process locks with this owner. */
868 while ((fl = *before) && posix_same_owner(request, fl)) {
869 /* Detect adjacent or overlapping regions (if same lock type)
870 */
871 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700872 /* In all comparisons of start vs end, use
873 * "start - 1" rather than "end + 1". If end
874 * is OFFSET_MAX, end + 1 will become negative.
875 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 if (fl->fl_end < request->fl_start - 1)
877 goto next_lock;
878 /* If the next lock in the list has entirely bigger
879 * addresses than the new one, insert the lock here.
880 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700881 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 break;
883
884 /* If we come here, the new and old lock are of the
885 * same type and adjacent or overlapping. Make one
886 * lock yielding from the lower start address of both
887 * locks to the higher end address.
888 */
889 if (fl->fl_start > request->fl_start)
890 fl->fl_start = request->fl_start;
891 else
892 request->fl_start = fl->fl_start;
893 if (fl->fl_end < request->fl_end)
894 fl->fl_end = request->fl_end;
895 else
896 request->fl_end = fl->fl_end;
897 if (added) {
898 locks_delete_lock(before);
899 continue;
900 }
901 request = fl;
902 added = 1;
903 }
904 else {
905 /* Processing for different lock types is a bit
906 * more complex.
907 */
908 if (fl->fl_end < request->fl_start)
909 goto next_lock;
910 if (fl->fl_start > request->fl_end)
911 break;
912 if (request->fl_type == F_UNLCK)
913 added = 1;
914 if (fl->fl_start < request->fl_start)
915 left = fl;
916 /* If the next lock in the list has a higher end
917 * address than the new one, insert the new one here.
918 */
919 if (fl->fl_end > request->fl_end) {
920 right = fl;
921 break;
922 }
923 if (fl->fl_start >= request->fl_start) {
924 /* The new lock completely replaces an old
925 * one (This may happen several times).
926 */
927 if (added) {
928 locks_delete_lock(before);
929 continue;
930 }
931 /* Replace the old lock with the new one.
932 * Wake up anybody waiting for the old one,
933 * as the change in lock type might satisfy
934 * their needs.
935 */
936 locks_wake_up_blocks(fl);
937 fl->fl_start = request->fl_start;
938 fl->fl_end = request->fl_end;
939 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500940 locks_release_private(fl);
941 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 request = fl;
943 added = 1;
944 }
945 }
946 /* Go on to next lock.
947 */
948 next_lock:
949 before = &fl->fl_next;
950 }
951
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700952 /*
953 * The above code only modifies existing locks in case of
954 * merging or replacing. If new lock(s) need to be inserted
955 * all modifications are done bellow this, so it's safe yet to
956 * bail out.
957 */
958 error = -ENOLCK; /* "no luck" */
959 if (right && left == right && !new_fl2)
960 goto out;
961
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 error = 0;
963 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400964 if (request->fl_type == F_UNLCK) {
965 if (request->fl_flags & FL_EXISTS)
966 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400968 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700969
970 if (!new_fl) {
971 error = -ENOLCK;
972 goto out;
973 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 locks_copy_lock(new_fl, request);
975 locks_insert_lock(before, new_fl);
976 new_fl = NULL;
977 }
978 if (right) {
979 if (left == right) {
980 /* The new lock breaks the old one in two pieces,
981 * so we have to use the second new lock.
982 */
983 left = new_fl2;
984 new_fl2 = NULL;
985 locks_copy_lock(left, right);
986 locks_insert_lock(before, left);
987 }
988 right->fl_start = request->fl_end + 1;
989 locks_wake_up_blocks(right);
990 }
991 if (left) {
992 left->fl_end = request->fl_start - 1;
993 locks_wake_up_blocks(left);
994 }
995 out:
996 unlock_kernel();
997 /*
998 * Free any unused locks.
999 */
1000 if (new_fl)
1001 locks_free_lock(new_fl);
1002 if (new_fl2)
1003 locks_free_lock(new_fl2);
1004 return error;
1005}
1006
1007/**
1008 * posix_lock_file - Apply a POSIX-style lock to a file
1009 * @filp: The file to apply the lock to
1010 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001011 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 *
1013 * Add a POSIX style lock to a file.
1014 * We merge adjacent & overlapping locks whenever possible.
1015 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001016 *
1017 * Note that if called with an FL_EXISTS argument, the caller may determine
1018 * whether or not a lock was successfully freed by testing the return
1019 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 */
Marc Eshel150b3932007-01-18 16:15:35 -05001021int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001022 struct file_lock *conflock)
1023{
Marc Eshel150b3932007-01-18 16:15:35 -05001024 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001025}
Marc Eshel150b3932007-01-18 16:15:35 -05001026EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
1028/**
1029 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1030 * @filp: The file to apply the lock to
1031 * @fl: The lock to be applied
1032 *
1033 * Add a POSIX style lock to a file.
1034 * We merge adjacent & overlapping locks whenever possible.
1035 * POSIX locks are sorted by owner task, then by starting address
1036 */
1037int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1038{
1039 int error;
1040 might_sleep ();
1041 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001042 error = posix_lock_file(filp, fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1044 break;
1045 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1046 if (!error)
1047 continue;
1048
1049 locks_delete_block(fl);
1050 break;
1051 }
1052 return error;
1053}
1054EXPORT_SYMBOL(posix_lock_file_wait);
1055
1056/**
1057 * locks_mandatory_locked - Check for an active lock
1058 * @inode: the file to check
1059 *
1060 * Searches the inode's list of locks to find any POSIX locks which conflict.
1061 * This function is called from locks_verify_locked() only.
1062 */
1063int locks_mandatory_locked(struct inode *inode)
1064{
1065 fl_owner_t owner = current->files;
1066 struct file_lock *fl;
1067
1068 /*
1069 * Search the lock list for this inode for any POSIX locks.
1070 */
1071 lock_kernel();
1072 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1073 if (!IS_POSIX(fl))
1074 continue;
1075 if (fl->fl_owner != owner)
1076 break;
1077 }
1078 unlock_kernel();
1079 return fl ? -EAGAIN : 0;
1080}
1081
1082/**
1083 * locks_mandatory_area - Check for a conflicting lock
1084 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1085 * for shared
1086 * @inode: the file to check
1087 * @filp: how the file was opened (if it was)
1088 * @offset: start of area to check
1089 * @count: length of area to check
1090 *
1091 * Searches the inode's list of locks to find any POSIX locks which conflict.
1092 * This function is called from rw_verify_area() and
1093 * locks_verify_truncate().
1094 */
1095int locks_mandatory_area(int read_write, struct inode *inode,
1096 struct file *filp, loff_t offset,
1097 size_t count)
1098{
1099 struct file_lock fl;
1100 int error;
1101
1102 locks_init_lock(&fl);
1103 fl.fl_owner = current->files;
1104 fl.fl_pid = current->tgid;
1105 fl.fl_file = filp;
1106 fl.fl_flags = FL_POSIX | FL_ACCESS;
1107 if (filp && !(filp->f_flags & O_NONBLOCK))
1108 fl.fl_flags |= FL_SLEEP;
1109 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1110 fl.fl_start = offset;
1111 fl.fl_end = offset + count - 1;
1112
1113 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001114 error = __posix_lock_file(inode, &fl, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 if (error != -EAGAIN)
1116 break;
1117 if (!(fl.fl_flags & FL_SLEEP))
1118 break;
1119 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1120 if (!error) {
1121 /*
1122 * If we've been sleeping someone might have
1123 * changed the permissions behind our back.
1124 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001125 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 continue;
1127 }
1128
1129 locks_delete_block(&fl);
1130 break;
1131 }
1132
1133 return error;
1134}
1135
1136EXPORT_SYMBOL(locks_mandatory_area);
1137
1138/* We already had a lease on this file; just change its type */
1139int lease_modify(struct file_lock **before, int arg)
1140{
1141 struct file_lock *fl = *before;
1142 int error = assign_type(fl, arg);
1143
1144 if (error)
1145 return error;
1146 locks_wake_up_blocks(fl);
1147 if (arg == F_UNLCK)
1148 locks_delete_lock(before);
1149 return 0;
1150}
1151
1152EXPORT_SYMBOL(lease_modify);
1153
1154static void time_out_leases(struct inode *inode)
1155{
1156 struct file_lock **before;
1157 struct file_lock *fl;
1158
1159 before = &inode->i_flock;
1160 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1161 if ((fl->fl_break_time == 0)
1162 || time_before(jiffies, fl->fl_break_time)) {
1163 before = &fl->fl_next;
1164 continue;
1165 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1167 if (fl == *before) /* lease_modify may have freed fl */
1168 before = &fl->fl_next;
1169 }
1170}
1171
1172/**
1173 * __break_lease - revoke all outstanding leases on file
1174 * @inode: the inode of the file to return
1175 * @mode: the open mode (read or write)
1176 *
david m. richter87250dd2007-05-09 16:10:27 -04001177 * break_lease (inlined for speed) has checked there already is at least
1178 * some kind of lock (maybe a lease) on this file. Leases are broken on
1179 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 * specified %O_NONBLOCK to your open().
1181 */
1182int __break_lease(struct inode *inode, unsigned int mode)
1183{
1184 int error = 0, future;
1185 struct file_lock *new_fl, *flock;
1186 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 unsigned long break_time;
1188 int i_have_this_lease = 0;
1189
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001190 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
1192 lock_kernel();
1193
1194 time_out_leases(inode);
1195
1196 flock = inode->i_flock;
1197 if ((flock == NULL) || !IS_LEASE(flock))
1198 goto out;
1199
1200 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1201 if (fl->fl_owner == current->files)
1202 i_have_this_lease = 1;
1203
1204 if (mode & FMODE_WRITE) {
1205 /* If we want write access, we have to revoke any lease. */
1206 future = F_UNLCK | F_INPROGRESS;
1207 } else if (flock->fl_type & F_INPROGRESS) {
1208 /* If the lease is already being broken, we just leave it */
1209 future = flock->fl_type;
1210 } else if (flock->fl_type & F_WRLCK) {
1211 /* Downgrade the exclusive lease to a read-only lease. */
1212 future = F_RDLCK | F_INPROGRESS;
1213 } else {
1214 /* the existing lease was read-only, so we can read too. */
1215 goto out;
1216 }
1217
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001218 if (IS_ERR(new_fl) && !i_have_this_lease
1219 && ((mode & O_NONBLOCK) == 0)) {
1220 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 goto out;
1222 }
1223
1224 break_time = 0;
1225 if (lease_break_time > 0) {
1226 break_time = jiffies + lease_break_time * HZ;
1227 if (break_time == 0)
1228 break_time++; /* so that 0 means no break time */
1229 }
1230
1231 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1232 if (fl->fl_type != future) {
1233 fl->fl_type = future;
1234 fl->fl_break_time = break_time;
1235 /* lease must have lmops break callback */
1236 fl->fl_lmops->fl_break(fl);
1237 }
1238 }
1239
1240 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1241 error = -EWOULDBLOCK;
1242 goto out;
1243 }
1244
1245restart:
1246 break_time = flock->fl_break_time;
1247 if (break_time != 0) {
1248 break_time -= jiffies;
1249 if (break_time == 0)
1250 break_time++;
1251 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001252 locks_insert_block(flock, new_fl);
1253 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1254 !new_fl->fl_next, break_time);
1255 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 if (error >= 0) {
1257 if (error == 0)
1258 time_out_leases(inode);
1259 /* Wait for the next lease that has not been broken yet */
1260 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1261 flock = flock->fl_next) {
1262 if (flock->fl_type & F_INPROGRESS)
1263 goto restart;
1264 }
1265 error = 0;
1266 }
1267
1268out:
1269 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001270 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 locks_free_lock(new_fl);
1272 return error;
1273}
1274
1275EXPORT_SYMBOL(__break_lease);
1276
1277/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001278 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 * @inode: the inode
1280 * @time: pointer to a timespec which will contain the last modified time
1281 *
1282 * This is to force NFS clients to flush their caches for files with
1283 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001284 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 */
1286void lease_get_mtime(struct inode *inode, struct timespec *time)
1287{
1288 struct file_lock *flock = inode->i_flock;
1289 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1290 *time = current_fs_time(inode->i_sb);
1291 else
1292 *time = inode->i_mtime;
1293}
1294
1295EXPORT_SYMBOL(lease_get_mtime);
1296
1297/**
1298 * fcntl_getlease - Enquire what lease is currently active
1299 * @filp: the file
1300 *
1301 * The value returned by this function will be one of
1302 * (if no lease break is pending):
1303 *
1304 * %F_RDLCK to indicate a shared lease is held.
1305 *
1306 * %F_WRLCK to indicate an exclusive lease is held.
1307 *
1308 * %F_UNLCK to indicate no lease is held.
1309 *
1310 * (if a lease break is pending):
1311 *
1312 * %F_RDLCK to indicate an exclusive lease needs to be
1313 * changed to a shared lease (or removed).
1314 *
1315 * %F_UNLCK to indicate the lease needs to be removed.
1316 *
1317 * XXX: sfr & willy disagree over whether F_INPROGRESS
1318 * should be returned to userspace.
1319 */
1320int fcntl_getlease(struct file *filp)
1321{
1322 struct file_lock *fl;
1323 int type = F_UNLCK;
1324
1325 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001326 time_out_leases(filp->f_path.dentry->d_inode);
1327 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 fl = fl->fl_next) {
1329 if (fl->fl_file == filp) {
1330 type = fl->fl_type & ~F_INPROGRESS;
1331 break;
1332 }
1333 }
1334 unlock_kernel();
1335 return type;
1336}
1337
1338/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001339 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 * @filp: file pointer
1341 * @arg: type of lease to obtain
1342 * @flp: input - file_lock to use, output - file_lock inserted
1343 *
1344 * The (input) flp->fl_lmops->fl_break function is required
1345 * by break_lease().
1346 *
1347 * Called with kernel lock held.
1348 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001349int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001351 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001352 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001353 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 struct inode *inode = dentry->d_inode;
1355 int error, rdlease_count = 0, wrlease_count = 0;
1356
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001357 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1358 return -EACCES;
1359 if (!S_ISREG(inode->i_mode))
1360 return -EINVAL;
1361 error = security_file_lock(filp, arg);
1362 if (error)
1363 return error;
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 time_out_leases(inode);
1366
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001367 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001369 lease = *flp;
1370
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 error = -EAGAIN;
1372 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1373 goto out;
1374 if ((arg == F_WRLCK)
1375 && ((atomic_read(&dentry->d_count) > 1)
1376 || (atomic_read(&inode->i_count) > 1)))
1377 goto out;
1378
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001379 error = -ENOMEM;
1380 new_fl = locks_alloc_lock();
1381 if (new_fl == NULL)
1382 goto out;
1383
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 /*
1385 * At this point, we know that if there is an exclusive
1386 * lease on this file, then we hold it on this filp
1387 * (otherwise our open of this file would have blocked).
1388 * And if we are trying to acquire an exclusive lease,
1389 * then the file is not open by anyone (including us)
1390 * except for this filp.
1391 */
1392 for (before = &inode->i_flock;
1393 ((fl = *before) != NULL) && IS_LEASE(fl);
1394 before = &fl->fl_next) {
1395 if (lease->fl_lmops->fl_mylease(fl, lease))
1396 my_before = before;
1397 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1398 /*
1399 * Someone is in the process of opening this
1400 * file for writing so we may not take an
1401 * exclusive lease on it.
1402 */
1403 wrlease_count++;
1404 else
1405 rdlease_count++;
1406 }
1407
1408 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1409 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1410 goto out;
1411
1412 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001413 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 error = lease->fl_lmops->fl_change(my_before, arg);
1415 goto out;
1416 }
1417
1418 error = 0;
1419 if (arg == F_UNLCK)
1420 goto out;
1421
1422 error = -EINVAL;
1423 if (!leases_enable)
1424 goto out;
1425
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001426 locks_copy_lock(new_fl, lease);
1427 locks_insert_lock(before, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001429 *flp = new_fl;
1430 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432out:
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001433 if (new_fl != NULL)
1434 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 return error;
1436}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001437EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
1439 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001440 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 * @filp: file pointer
1442 * @arg: type of lease to obtain
1443 * @lease: file_lock to use
1444 *
1445 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001446 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1447 * break_lease will oops!
1448 *
1449 * This will call the filesystem's setlease file method, if
1450 * defined. Note that there is no getlease method; instead, the
1451 * filesystem setlease method should call back to setlease() to
1452 * add a lease to the inode's lease list, where fcntl_getlease() can
1453 * find it. Since fcntl_getlease() only reports whether the current
1454 * task holds a lease, a cluster filesystem need only do this for
1455 * leases held by processes on this node.
1456 *
1457 * There is also no break_lease method; filesystems that
1458 * handle their own leases shoud break leases themselves from the
1459 * filesystem's open, create, and (on truncate) setattr methods.
1460 *
1461 * Warning: the only current setlease methods exist only to disable
1462 * leases in certain cases. More vfs changes may be required to
1463 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 */
1465
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001466int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 int error;
1469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001471 if (filp->f_op && filp->f_op->setlease)
1472 error = filp->f_op->setlease(filp, arg, lease);
1473 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001474 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 unlock_kernel();
1476
1477 return error;
1478}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001479EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
1481/**
1482 * fcntl_setlease - sets a lease on an open file
1483 * @fd: open file descriptor
1484 * @filp: file pointer
1485 * @arg: type of lease to obtain
1486 *
1487 * Call this fcntl to establish a lease on the file.
1488 * Note that you also need to call %F_SETSIG to
1489 * receive a signal when the lease is broken.
1490 */
1491int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1492{
1493 struct file_lock fl, *flp = &fl;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001494 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 struct inode *inode = dentry->d_inode;
1496 int error;
1497
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 locks_init_lock(&fl);
1499 error = lease_init(filp, arg, &fl);
1500 if (error)
1501 return error;
1502
1503 lock_kernel();
1504
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001505 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001506 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 goto out_unlock;
1508
1509 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1510 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001511 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001513 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 time_out_leases(inode);
1515 goto out_unlock;
1516 }
1517
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001518 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519out_unlock:
1520 unlock_kernel();
1521 return error;
1522}
1523
1524/**
1525 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1526 * @filp: The file to apply the lock to
1527 * @fl: The lock to be applied
1528 *
1529 * Add a FLOCK style lock to a file.
1530 */
1531int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1532{
1533 int error;
1534 might_sleep();
1535 for (;;) {
1536 error = flock_lock_file(filp, fl);
1537 if ((error != -EAGAIN) || !(fl->fl_flags & FL_SLEEP))
1538 break;
1539 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1540 if (!error)
1541 continue;
1542
1543 locks_delete_block(fl);
1544 break;
1545 }
1546 return error;
1547}
1548
1549EXPORT_SYMBOL(flock_lock_file_wait);
1550
1551/**
1552 * sys_flock: - flock() system call.
1553 * @fd: the file descriptor to lock.
1554 * @cmd: the type of lock to apply.
1555 *
1556 * Apply a %FL_FLOCK style lock to an open file descriptor.
1557 * The @cmd can be one of
1558 *
1559 * %LOCK_SH -- a shared lock.
1560 *
1561 * %LOCK_EX -- an exclusive lock.
1562 *
1563 * %LOCK_UN -- remove an existing lock.
1564 *
1565 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1566 *
1567 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1568 * processes read and write access respectively.
1569 */
1570asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1571{
1572 struct file *filp;
1573 struct file_lock *lock;
1574 int can_sleep, unlock;
1575 int error;
1576
1577 error = -EBADF;
1578 filp = fget(fd);
1579 if (!filp)
1580 goto out;
1581
1582 can_sleep = !(cmd & LOCK_NB);
1583 cmd &= ~LOCK_NB;
1584 unlock = (cmd == LOCK_UN);
1585
1586 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1587 goto out_putf;
1588
1589 error = flock_make_lock(filp, &lock, cmd);
1590 if (error)
1591 goto out_putf;
1592 if (can_sleep)
1593 lock->fl_flags |= FL_SLEEP;
1594
1595 error = security_file_lock(filp, cmd);
1596 if (error)
1597 goto out_free;
1598
1599 if (filp->f_op && filp->f_op->flock)
1600 error = filp->f_op->flock(filp,
1601 (can_sleep) ? F_SETLKW : F_SETLK,
1602 lock);
1603 else
1604 error = flock_lock_file_wait(filp, lock);
1605
1606 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001607 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608
1609 out_putf:
1610 fput(filp);
1611 out:
1612 return error;
1613}
1614
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001615/**
1616 * vfs_test_lock - test file byte range lock
1617 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001618 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001619 *
1620 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1621 * setting conf->fl_type to something other than F_UNLCK.
1622 */
1623int vfs_test_lock(struct file *filp, struct file_lock *fl)
1624{
1625 if (filp->f_op && filp->f_op->lock)
1626 return filp->f_op->lock(filp, F_GETLK, fl);
1627 posix_test_lock(filp, fl);
1628 return 0;
1629}
1630EXPORT_SYMBOL_GPL(vfs_test_lock);
1631
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001632static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1633{
1634 flock->l_pid = fl->fl_pid;
1635#if BITS_PER_LONG == 32
1636 /*
1637 * Make sure we can represent the posix lock via
1638 * legacy 32bit flock.
1639 */
1640 if (fl->fl_start > OFFT_OFFSET_MAX)
1641 return -EOVERFLOW;
1642 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1643 return -EOVERFLOW;
1644#endif
1645 flock->l_start = fl->fl_start;
1646 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1647 fl->fl_end - fl->fl_start + 1;
1648 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001649 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001650 return 0;
1651}
1652
1653#if BITS_PER_LONG == 32
1654static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1655{
1656 flock->l_pid = fl->fl_pid;
1657 flock->l_start = fl->fl_start;
1658 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1659 fl->fl_end - fl->fl_start + 1;
1660 flock->l_whence = 0;
1661 flock->l_type = fl->fl_type;
1662}
1663#endif
1664
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665/* Report the first existing lock that would conflict with l.
1666 * This implements the F_GETLK command of fcntl().
1667 */
1668int fcntl_getlk(struct file *filp, struct flock __user *l)
1669{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001670 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 struct flock flock;
1672 int error;
1673
1674 error = -EFAULT;
1675 if (copy_from_user(&flock, l, sizeof(flock)))
1676 goto out;
1677 error = -EINVAL;
1678 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1679 goto out;
1680
1681 error = flock_to_posix_lock(filp, &file_lock, &flock);
1682 if (error)
1683 goto out;
1684
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001685 error = vfs_test_lock(filp, &file_lock);
1686 if (error)
1687 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001689 flock.l_type = file_lock.fl_type;
1690 if (file_lock.fl_type != F_UNLCK) {
1691 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001692 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 }
1695 error = -EFAULT;
1696 if (!copy_to_user(l, &flock, sizeof(flock)))
1697 error = 0;
1698out:
1699 return error;
1700}
1701
Marc Eshel7723ec92007-01-18 15:08:55 -05001702/**
1703 * vfs_lock_file - file byte range lock
1704 * @filp: The file to apply the lock to
1705 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1706 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001707 * @conf: Place to return a copy of the conflicting lock, if found.
1708 *
1709 * A caller that doesn't care about the conflicting lock may pass NULL
1710 * as the final argument.
1711 *
1712 * If the filesystem defines a private ->lock() method, then @conf will
1713 * be left unchanged; so a caller that cares should initialize it to
1714 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001715 *
1716 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1717 * locks, the ->lock() interface may return asynchronously, before the lock has
1718 * been granted or denied by the underlying filesystem, if (and only if)
1719 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1720 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1721 * the request is for a blocking lock. When ->lock() does return asynchronously,
1722 * it must return -EINPROGRESS, and call ->fl_grant() when the lock
1723 * request completes.
1724 * If the request is for non-blocking lock the file system should return
1725 * -EINPROGRESS then try to get the lock and call the callback routine with
1726 * the result. If the request timed out the callback routine will return a
1727 * nonzero return code and the file system should release the lock. The file
1728 * system is also responsible to keep a corresponding posix lock when it
1729 * grants a lock so the VFS can find out which locks are locally held and do
1730 * the correct lock cleanup when required.
1731 * The underlying filesystem must not drop the kernel lock or call
1732 * ->fl_grant() before returning to the caller with a -EINPROGRESS
1733 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001734 */
Marc Eshel150b3932007-01-18 16:15:35 -05001735int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001736{
1737 if (filp->f_op && filp->f_op->lock)
1738 return filp->f_op->lock(filp, cmd, fl);
1739 else
Marc Eshel150b3932007-01-18 16:15:35 -05001740 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001741}
1742EXPORT_SYMBOL_GPL(vfs_lock_file);
1743
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744/* Apply the lock described by l to an open file descriptor.
1745 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1746 */
Peter Staubachc2936212005-07-27 11:45:09 -07001747int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1748 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749{
1750 struct file_lock *file_lock = locks_alloc_lock();
1751 struct flock flock;
1752 struct inode *inode;
1753 int error;
1754
1755 if (file_lock == NULL)
1756 return -ENOLCK;
1757
1758 /*
1759 * This might block, so we do it before checking the inode.
1760 */
1761 error = -EFAULT;
1762 if (copy_from_user(&flock, l, sizeof(flock)))
1763 goto out;
1764
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001765 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
1767 /* Don't allow mandatory locks on files that may be memory mapped
1768 * and shared.
1769 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001770 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 error = -EAGAIN;
1772 goto out;
1773 }
1774
Peter Staubachc2936212005-07-27 11:45:09 -07001775again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 error = flock_to_posix_lock(filp, file_lock, &flock);
1777 if (error)
1778 goto out;
1779 if (cmd == F_SETLKW) {
1780 file_lock->fl_flags |= FL_SLEEP;
1781 }
1782
1783 error = -EBADF;
1784 switch (flock.l_type) {
1785 case F_RDLCK:
1786 if (!(filp->f_mode & FMODE_READ))
1787 goto out;
1788 break;
1789 case F_WRLCK:
1790 if (!(filp->f_mode & FMODE_WRITE))
1791 goto out;
1792 break;
1793 case F_UNLCK:
1794 break;
1795 default:
1796 error = -EINVAL;
1797 goto out;
1798 }
1799
1800 error = security_file_lock(filp, file_lock->fl_type);
1801 if (error)
1802 goto out;
1803
Marc Eshel7723ec92007-01-18 15:08:55 -05001804 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001805 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001806 if (error != -EAGAIN || cmd == F_SETLK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001808 error = wait_event_interruptible(file_lock->fl_wait,
1809 !file_lock->fl_next);
1810 if (!error)
1811 continue;
1812
1813 locks_delete_block(file_lock);
1814 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 }
1816
Peter Staubachc2936212005-07-27 11:45:09 -07001817 /*
1818 * Attempt to detect a close/fcntl race and recover by
1819 * releasing the lock that was just acquired.
1820 */
1821 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1822 flock.l_type = F_UNLCK;
1823 goto again;
1824 }
1825
1826out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 locks_free_lock(file_lock);
1828 return error;
1829}
1830
1831#if BITS_PER_LONG == 32
1832/* Report the first existing lock that would conflict with l.
1833 * This implements the F_GETLK command of fcntl().
1834 */
1835int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1836{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001837 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 struct flock64 flock;
1839 int error;
1840
1841 error = -EFAULT;
1842 if (copy_from_user(&flock, l, sizeof(flock)))
1843 goto out;
1844 error = -EINVAL;
1845 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1846 goto out;
1847
1848 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1849 if (error)
1850 goto out;
1851
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001852 error = vfs_test_lock(filp, &file_lock);
1853 if (error)
1854 goto out;
1855
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001856 flock.l_type = file_lock.fl_type;
1857 if (file_lock.fl_type != F_UNLCK)
1858 posix_lock_to_flock64(&flock, &file_lock);
1859
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 error = -EFAULT;
1861 if (!copy_to_user(l, &flock, sizeof(flock)))
1862 error = 0;
1863
1864out:
1865 return error;
1866}
1867
1868/* Apply the lock described by l to an open file descriptor.
1869 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1870 */
Peter Staubachc2936212005-07-27 11:45:09 -07001871int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1872 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873{
1874 struct file_lock *file_lock = locks_alloc_lock();
1875 struct flock64 flock;
1876 struct inode *inode;
1877 int error;
1878
1879 if (file_lock == NULL)
1880 return -ENOLCK;
1881
1882 /*
1883 * This might block, so we do it before checking the inode.
1884 */
1885 error = -EFAULT;
1886 if (copy_from_user(&flock, l, sizeof(flock)))
1887 goto out;
1888
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001889 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890
1891 /* Don't allow mandatory locks on files that may be memory mapped
1892 * and shared.
1893 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001894 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 error = -EAGAIN;
1896 goto out;
1897 }
1898
Peter Staubachc2936212005-07-27 11:45:09 -07001899again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 error = flock64_to_posix_lock(filp, file_lock, &flock);
1901 if (error)
1902 goto out;
1903 if (cmd == F_SETLKW64) {
1904 file_lock->fl_flags |= FL_SLEEP;
1905 }
1906
1907 error = -EBADF;
1908 switch (flock.l_type) {
1909 case F_RDLCK:
1910 if (!(filp->f_mode & FMODE_READ))
1911 goto out;
1912 break;
1913 case F_WRLCK:
1914 if (!(filp->f_mode & FMODE_WRITE))
1915 goto out;
1916 break;
1917 case F_UNLCK:
1918 break;
1919 default:
1920 error = -EINVAL;
1921 goto out;
1922 }
1923
1924 error = security_file_lock(filp, file_lock->fl_type);
1925 if (error)
1926 goto out;
1927
Marc Eshel7723ec92007-01-18 15:08:55 -05001928 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001929 error = vfs_lock_file(filp, cmd, file_lock, NULL);
Marc Eshel7723ec92007-01-18 15:08:55 -05001930 if (error != -EAGAIN || cmd == F_SETLK64)
Peter Staubachc2936212005-07-27 11:45:09 -07001931 break;
Marc Eshel7723ec92007-01-18 15:08:55 -05001932 error = wait_event_interruptible(file_lock->fl_wait,
1933 !file_lock->fl_next);
1934 if (!error)
1935 continue;
1936
1937 locks_delete_block(file_lock);
1938 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 }
1940
Peter Staubachc2936212005-07-27 11:45:09 -07001941 /*
1942 * Attempt to detect a close/fcntl race and recover by
1943 * releasing the lock that was just acquired.
1944 */
1945 if (!error && fcheck(fd) != filp && flock.l_type != F_UNLCK) {
1946 flock.l_type = F_UNLCK;
1947 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 }
1949
1950out:
1951 locks_free_lock(file_lock);
1952 return error;
1953}
1954#endif /* BITS_PER_LONG == 32 */
1955
1956/*
1957 * This function is called when the file is being removed
1958 * from the task's fd array. POSIX locks belonging to this task
1959 * are deleted at this time.
1960 */
1961void locks_remove_posix(struct file *filp, fl_owner_t owner)
1962{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001963 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964
1965 /*
1966 * If there are no locks held on this file, we don't need to call
1967 * posix_lock_file(). Another process could be setting a lock on this
1968 * file at the same time, but we wouldn't remove that lock anyway.
1969 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001970 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 return;
1972
1973 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001974 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975 lock.fl_start = 0;
1976 lock.fl_end = OFFSET_MAX;
1977 lock.fl_owner = owner;
1978 lock.fl_pid = current->tgid;
1979 lock.fl_file = filp;
1980 lock.fl_ops = NULL;
1981 lock.fl_lmops = NULL;
1982
Marc Eshel150b3932007-01-18 16:15:35 -05001983 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1986 lock.fl_ops->fl_release_private(&lock);
1987}
1988
1989EXPORT_SYMBOL(locks_remove_posix);
1990
1991/*
1992 * This function is called on the last close of an open file.
1993 */
1994void locks_remove_flock(struct file *filp)
1995{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001996 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 struct file_lock *fl;
1998 struct file_lock **before;
1999
2000 if (!inode->i_flock)
2001 return;
2002
2003 if (filp->f_op && filp->f_op->flock) {
2004 struct file_lock fl = {
2005 .fl_pid = current->tgid,
2006 .fl_file = filp,
2007 .fl_flags = FL_FLOCK,
2008 .fl_type = F_UNLCK,
2009 .fl_end = OFFSET_MAX,
2010 };
2011 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002012 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2013 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 }
2015
2016 lock_kernel();
2017 before = &inode->i_flock;
2018
2019 while ((fl = *before) != NULL) {
2020 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002021 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 locks_delete_lock(before);
2023 continue;
2024 }
2025 if (IS_LEASE(fl)) {
2026 lease_modify(before, F_UNLCK);
2027 continue;
2028 }
2029 /* What? */
2030 BUG();
2031 }
2032 before = &fl->fl_next;
2033 }
2034 unlock_kernel();
2035}
2036
2037/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 * posix_unblock_lock - stop waiting for a file lock
2039 * @filp: how the file was opened
2040 * @waiter: the lock which was waiting
2041 *
2042 * lockd needs to block waiting for locks.
2043 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002044int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2046{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002047 int status = 0;
2048
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002050 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002052 else
2053 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002054 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002055 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056}
2057
2058EXPORT_SYMBOL(posix_unblock_lock);
2059
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002060/**
2061 * vfs_cancel_lock - file byte range unblock lock
2062 * @filp: The file to apply the unblock to
2063 * @fl: The lock to be unblocked
2064 *
2065 * Used by lock managers to cancel blocked requests
2066 */
2067int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2068{
2069 if (filp->f_op && filp->f_op->lock)
2070 return filp->f_op->lock(filp, F_CANCELLK, fl);
2071 return 0;
2072}
2073
2074EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2075
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002076#ifdef CONFIG_PROC_FS
2077#include <linux/seq_file.h>
2078
2079static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2080 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081{
2082 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002083 unsigned int fl_pid;
2084
2085 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002086 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002087 else
2088 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089
2090 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002091 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002093 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002095 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2097 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002098 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 } else if (IS_FLOCK(fl)) {
2100 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002101 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002103 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 }
2105 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002106 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002108 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002110 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002112 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002114 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 }
2116 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002117 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 (fl->fl_type & LOCK_READ)
2119 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2120 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2121 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002122 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 (fl->fl_type & F_INPROGRESS)
2124 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2125 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2126 }
2127 if (inode) {
2128#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002129 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 inode->i_sb->s_id, inode->i_ino);
2131#else
2132 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002133 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134 MAJOR(inode->i_sb->s_dev),
2135 MINOR(inode->i_sb->s_dev), inode->i_ino);
2136#endif
2137 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002138 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 }
2140 if (IS_POSIX(fl)) {
2141 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002142 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002144 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002146 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 }
2148}
2149
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002150static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002152 struct file_lock *fl, *bfl;
2153
2154 fl = list_entry(v, struct file_lock, fl_link);
2155
2156 lock_get_status(f, fl, (long)f->private, "");
2157
2158 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2159 lock_get_status(f, bfl, (long)f->private, " ->");
2160
2161 f->private++;
2162 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163}
2164
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002165static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002166{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002168 f->private = (void *)1;
2169 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170}
2171
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002172static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2173{
2174 return seq_list_next(v, &file_lock_list, pos);
2175}
2176
2177static void locks_stop(struct seq_file *f, void *v)
2178{
2179 unlock_kernel();
2180}
2181
2182struct seq_operations locks_seq_operations = {
2183 .start = locks_start,
2184 .next = locks_next,
2185 .stop = locks_stop,
2186 .show = locks_show,
2187};
2188#endif
2189
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190/**
2191 * lock_may_read - checks that the region is free of locks
2192 * @inode: the inode that is being read
2193 * @start: the first byte to read
2194 * @len: the number of bytes to read
2195 *
2196 * Emulates Windows locking requirements. Whole-file
2197 * mandatory locks (share modes) can prohibit a read and
2198 * byte-range POSIX locks can prohibit a read if they overlap.
2199 *
2200 * N.B. this function is only ever called
2201 * from knfsd and ownership of locks is never checked.
2202 */
2203int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2204{
2205 struct file_lock *fl;
2206 int result = 1;
2207 lock_kernel();
2208 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2209 if (IS_POSIX(fl)) {
2210 if (fl->fl_type == F_RDLCK)
2211 continue;
2212 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2213 continue;
2214 } else if (IS_FLOCK(fl)) {
2215 if (!(fl->fl_type & LOCK_MAND))
2216 continue;
2217 if (fl->fl_type & LOCK_READ)
2218 continue;
2219 } else
2220 continue;
2221 result = 0;
2222 break;
2223 }
2224 unlock_kernel();
2225 return result;
2226}
2227
2228EXPORT_SYMBOL(lock_may_read);
2229
2230/**
2231 * lock_may_write - checks that the region is free of locks
2232 * @inode: the inode that is being written
2233 * @start: the first byte to write
2234 * @len: the number of bytes to write
2235 *
2236 * Emulates Windows locking requirements. Whole-file
2237 * mandatory locks (share modes) can prohibit a write and
2238 * byte-range POSIX locks can prohibit a write if they overlap.
2239 *
2240 * N.B. this function is only ever called
2241 * from knfsd and ownership of locks is never checked.
2242 */
2243int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2244{
2245 struct file_lock *fl;
2246 int result = 1;
2247 lock_kernel();
2248 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2249 if (IS_POSIX(fl)) {
2250 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2251 continue;
2252 } else if (IS_FLOCK(fl)) {
2253 if (!(fl->fl_type & LOCK_MAND))
2254 continue;
2255 if (fl->fl_type & LOCK_WRITE)
2256 continue;
2257 } else
2258 continue;
2259 result = 0;
2260 break;
2261 }
2262 unlock_kernel();
2263 return result;
2264}
2265
2266EXPORT_SYMBOL(lock_may_write);
2267
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268static int __init filelock_init(void)
2269{
2270 filelock_cache = kmem_cache_create("file_lock_cache",
2271 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002272 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002273 return 0;
2274}
2275
2276core_initcall(filelock_init);