blob: 19ee18a6829b9ae443778d63d57a8d8924131748 [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>
Al Viro9f3acc32008-04-24 07:44:08 -0400119#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#include <linux/fs.h>
121#include <linux/init.h>
122#include <linux/module.h>
123#include <linux/security.h>
124#include <linux/slab.h>
125#include <linux/smp_lock.h>
126#include <linux/syscalls.h>
127#include <linux/time.h>
Dipankar Sarma4fb3a532005-09-16 19:28:13 -0700128#include <linux/rcupdate.h>
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000129#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131#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
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500154void locks_release_private(struct file_lock *fl)
Trond Myklebust47831f32006-03-20 13:44:05 -0500155{
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}
Felix Blyakhera9e61e22009-03-31 15:12:56 -0500168EXPORT_SYMBOL_GPL(locks_release_private);
Trond Myklebust47831f32006-03-20 13:44:05 -0500169
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170/* Free a lock which is not in use. */
Matt Mackall33443c42006-01-08 01:05:22 -0800171static void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800173 BUG_ON(waitqueue_active(&fl->fl_wait));
174 BUG_ON(!list_empty(&fl->fl_block));
175 BUG_ON(!list_empty(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176
Trond Myklebust47831f32006-03-20 13:44:05 -0500177 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 kmem_cache_free(filelock_cache, fl);
179}
180
181void locks_init_lock(struct file_lock *fl)
182{
183 INIT_LIST_HEAD(&fl->fl_link);
184 INIT_LIST_HEAD(&fl->fl_block);
185 init_waitqueue_head(&fl->fl_wait);
186 fl->fl_next = NULL;
187 fl->fl_fasync = NULL;
188 fl->fl_owner = NULL;
189 fl->fl_pid = 0;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000190 fl->fl_nspid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 fl->fl_file = NULL;
192 fl->fl_flags = 0;
193 fl->fl_type = 0;
194 fl->fl_start = fl->fl_end = 0;
195 fl->fl_ops = NULL;
196 fl->fl_lmops = NULL;
197}
198
199EXPORT_SYMBOL(locks_init_lock);
200
201/*
202 * Initialises the fields of the file lock which are invariant for
203 * free file_locks.
204 */
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700205static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206{
207 struct file_lock *lock = (struct file_lock *) foo;
208
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 locks_init_lock(lock);
210}
211
Trond Myklebust47831f32006-03-20 13:44:05 -0500212static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
213{
214 if (fl->fl_ops) {
215 if (fl->fl_ops->fl_copy_lock)
216 fl->fl_ops->fl_copy_lock(new, fl);
217 new->fl_ops = fl->fl_ops;
218 }
219 if (fl->fl_lmops) {
220 if (fl->fl_lmops->fl_copy_lock)
221 fl->fl_lmops->fl_copy_lock(new, fl);
222 new->fl_lmops = fl->fl_lmops;
223 }
224}
225
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226/*
227 * Initialize a new lock from an existing file_lock structure.
228 */
J. Bruce Fields1a747ee2008-04-24 10:08:22 -0400229void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
231 new->fl_owner = fl->fl_owner;
232 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500233 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 new->fl_flags = fl->fl_flags;
235 new->fl_type = fl->fl_type;
236 new->fl_start = fl->fl_start;
237 new->fl_end = fl->fl_end;
Trond Myklebust09969052006-03-20 13:44:38 -0500238 new->fl_ops = NULL;
239 new->fl_lmops = NULL;
240}
Roland Dreier3dd7b712008-04-25 15:32:51 -0700241EXPORT_SYMBOL(__locks_copy_lock);
Trond Myklebust09969052006-03-20 13:44:38 -0500242
243void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
244{
245 locks_release_private(new);
246
247 __locks_copy_lock(new, fl);
248 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 new->fl_ops = fl->fl_ops;
250 new->fl_lmops = fl->fl_lmops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500251
252 locks_copy_private(new, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253}
254
255EXPORT_SYMBOL(locks_copy_lock);
256
257static inline int flock_translate_cmd(int cmd) {
258 if (cmd & LOCK_MAND)
259 return cmd & (LOCK_MAND | LOCK_RW);
260 switch (cmd) {
261 case LOCK_SH:
262 return F_RDLCK;
263 case LOCK_EX:
264 return F_WRLCK;
265 case LOCK_UN:
266 return F_UNLCK;
267 }
268 return -EINVAL;
269}
270
271/* Fill in a file_lock structure with an appropriate FLOCK lock. */
272static int flock_make_lock(struct file *filp, struct file_lock **lock,
273 unsigned int cmd)
274{
275 struct file_lock *fl;
276 int type = flock_translate_cmd(cmd);
277 if (type < 0)
278 return type;
279
280 fl = locks_alloc_lock();
281 if (fl == NULL)
282 return -ENOMEM;
283
284 fl->fl_file = filp;
285 fl->fl_pid = current->tgid;
286 fl->fl_flags = FL_FLOCK;
287 fl->fl_type = type;
288 fl->fl_end = OFFSET_MAX;
289
290 *lock = fl;
291 return 0;
292}
293
294static int assign_type(struct file_lock *fl, int type)
295{
296 switch (type) {
297 case F_RDLCK:
298 case F_WRLCK:
299 case F_UNLCK:
300 fl->fl_type = type;
301 break;
302 default:
303 return -EINVAL;
304 }
305 return 0;
306}
307
308/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
309 * style lock.
310 */
311static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
312 struct flock *l)
313{
314 off_t start, end;
315
316 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700317 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 start = 0;
319 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700320 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 start = filp->f_pos;
322 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700323 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800324 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 break;
326 default:
327 return -EINVAL;
328 }
329
330 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
331 POSIX-2001 defines it. */
332 start += l->l_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 if (start < 0)
334 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700335 fl->fl_end = OFFSET_MAX;
336 if (l->l_len > 0) {
337 end = start + l->l_len - 1;
338 fl->fl_end = end;
339 } else if (l->l_len < 0) {
340 end = start - 1;
341 fl->fl_end = end;
342 start += l->l_len;
343 if (start < 0)
344 return -EINVAL;
345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700347 if (fl->fl_end < fl->fl_start)
348 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
350 fl->fl_owner = current->files;
351 fl->fl_pid = current->tgid;
352 fl->fl_file = filp;
353 fl->fl_flags = FL_POSIX;
354 fl->fl_ops = NULL;
355 fl->fl_lmops = NULL;
356
357 return assign_type(fl, l->l_type);
358}
359
360#if BITS_PER_LONG == 32
361static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
362 struct flock64 *l)
363{
364 loff_t start;
365
366 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700367 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 start = 0;
369 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700370 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 start = filp->f_pos;
372 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700373 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800374 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 break;
376 default:
377 return -EINVAL;
378 }
379
Trond Myklebust4c780a42005-10-18 14:20:21 -0700380 start += l->l_start;
381 if (start < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700383 fl->fl_end = OFFSET_MAX;
384 if (l->l_len > 0) {
385 fl->fl_end = start + l->l_len - 1;
386 } else if (l->l_len < 0) {
387 fl->fl_end = start - 1;
388 start += l->l_len;
389 if (start < 0)
390 return -EINVAL;
391 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700393 if (fl->fl_end < fl->fl_start)
394 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
396 fl->fl_owner = current->files;
397 fl->fl_pid = current->tgid;
398 fl->fl_file = filp;
399 fl->fl_flags = FL_POSIX;
400 fl->fl_ops = NULL;
401 fl->fl_lmops = NULL;
402
403 switch (l->l_type) {
404 case F_RDLCK:
405 case F_WRLCK:
406 case F_UNLCK:
407 fl->fl_type = l->l_type;
408 break;
409 default:
410 return -EINVAL;
411 }
412
413 return (0);
414}
415#endif
416
417/* default lease lock manager operations */
418static void lease_break_callback(struct file_lock *fl)
419{
420 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
421}
422
423static void lease_release_private_callback(struct file_lock *fl)
424{
425 if (!fl->fl_file)
426 return;
427
428 f_delown(fl->fl_file);
429 fl->fl_file->f_owner.signum = 0;
430}
431
Adrian Bunk75c96f82005-05-05 16:16:09 -0700432static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
434 return fl->fl_file == try->fl_file;
435}
436
Adrian Bunk75c96f82005-05-05 16:16:09 -0700437static struct lock_manager_operations lease_manager_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 .fl_break = lease_break_callback,
439 .fl_release_private = lease_release_private_callback,
440 .fl_mylease = lease_mylease_callback,
441 .fl_change = lease_modify,
442};
443
444/*
445 * Initialize a lease, use the default lock manager operations
446 */
447static int lease_init(struct file *filp, int type, struct file_lock *fl)
448 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400449 if (assign_type(fl, type) != 0)
450 return -EINVAL;
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 fl->fl_owner = current->files;
453 fl->fl_pid = current->tgid;
454
455 fl->fl_file = filp;
456 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 fl->fl_start = 0;
458 fl->fl_end = OFFSET_MAX;
459 fl->fl_ops = NULL;
460 fl->fl_lmops = &lease_manager_ops;
461 return 0;
462}
463
464/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500465static struct file_lock *lease_alloc(struct file *filp, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
467 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400468 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
470 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500471 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
473 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400474 if (error) {
475 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500476 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400477 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500478 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479}
480
481/* Check if two locks overlap each other.
482 */
483static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
484{
485 return ((fl1->fl_end >= fl2->fl_start) &&
486 (fl2->fl_end >= fl1->fl_start));
487}
488
489/*
490 * Check whether two locks have the same owner.
491 */
Matt Mackall33443c42006-01-08 01:05:22 -0800492static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
495 return fl2->fl_lmops == fl1->fl_lmops &&
496 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
497 return fl1->fl_owner == fl2->fl_owner;
498}
499
500/* Remove waiter from blocker's block list.
501 * When blocker ends up pointing to itself then the list is empty.
502 */
Matt Mackall33443c42006-01-08 01:05:22 -0800503static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
505 list_del_init(&waiter->fl_block);
506 list_del_init(&waiter->fl_link);
507 waiter->fl_next = NULL;
508}
509
510/*
511 */
512static void locks_delete_block(struct file_lock *waiter)
513{
514 lock_kernel();
515 __locks_delete_block(waiter);
516 unlock_kernel();
517}
518
519/* Insert waiter into blocker's block list.
520 * We use a circular list so that processes can be easily woken up in
521 * the order they blocked. The documentation doesn't require this but
522 * it seems like the reasonable thing to do.
523 */
524static void locks_insert_block(struct file_lock *blocker,
525 struct file_lock *waiter)
526{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800527 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 list_add_tail(&waiter->fl_block, &blocker->fl_block);
529 waiter->fl_next = blocker;
530 if (IS_POSIX(blocker))
531 list_add(&waiter->fl_link, &blocked_list);
532}
533
534/* Wake up processes blocked waiting for blocker.
535 * If told to wait then schedule the processes until the block list
536 * is empty, otherwise empty the block list ourselves.
537 */
538static void locks_wake_up_blocks(struct file_lock *blocker)
539{
540 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400541 struct file_lock *waiter;
542
543 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 struct file_lock, fl_block);
545 __locks_delete_block(waiter);
546 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
547 waiter->fl_lmops->fl_notify(waiter);
548 else
549 wake_up(&waiter->fl_wait);
550 }
551}
552
553/* Insert file lock fl into an inode's lock list at the position indicated
554 * by pos. At the same time add the lock to the global file lock list.
555 */
556static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
557{
558 list_add(&fl->fl_link, &file_lock_list);
559
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000560 fl->fl_nspid = get_pid(task_tgid(current));
561
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 /* insert into file's list */
563 fl->fl_next = *pos;
564 *pos = fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565}
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
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000587 if (fl->fl_nspid) {
588 put_pid(fl->fl_nspid);
589 fl->fl_nspid = NULL;
590 }
591
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 locks_wake_up_blocks(fl);
593 locks_free_lock(fl);
594}
595
596/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
597 * checks for shared/exclusive status of overlapping locks.
598 */
599static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
600{
601 if (sys_fl->fl_type == F_WRLCK)
602 return 1;
603 if (caller_fl->fl_type == F_WRLCK)
604 return 1;
605 return 0;
606}
607
608/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
609 * checking before calling the locks_conflict().
610 */
611static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
612{
613 /* POSIX locks owned by the same process do not conflict with
614 * each other.
615 */
616 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
617 return (0);
618
619 /* Check whether they overlap */
620 if (!locks_overlap(caller_fl, sys_fl))
621 return 0;
622
623 return (locks_conflict(caller_fl, sys_fl));
624}
625
626/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
627 * checking before calling the locks_conflict().
628 */
629static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
630{
631 /* FLOCK locks referring to the same filp do not conflict with
632 * each other.
633 */
634 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
635 return (0);
636 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
637 return 0;
638
639 return (locks_conflict(caller_fl, sys_fl));
640}
641
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400642void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500643posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644{
645 struct file_lock *cfl;
646
647 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800648 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 if (!IS_POSIX(cfl))
650 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400651 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 break;
653 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000654 if (cfl) {
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500655 __locks_copy_lock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000656 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800657 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000658 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400659 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400661 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663EXPORT_SYMBOL(posix_test_lock);
664
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400665/*
666 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400668 * We attempt to detect deadlocks that are due purely to posix file
669 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400671 * We assume that a task can be waiting for at most one lock at a time.
672 * So for any acquired lock, the process holding that lock may be
673 * waiting on at most one other lock. That lock in turns may be held by
674 * someone waiting for at most one other lock. Given a requested lock
675 * caller_fl which is about to wait for a conflicting lock block_fl, we
676 * follow this chain of waiters to ensure we are not about to create a
677 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400678 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400679 * Since we do this before we ever put a process to sleep on a lock, we
680 * are ensured that there is never a cycle; that is what guarantees that
681 * the while() loop in posix_locks_deadlock() eventually completes.
682 *
683 * Note: the above assumption may not be true when handling lock
684 * requests from a broken NFS client. It may also fail in the presence
685 * of tasks (such as posix threads) sharing the same open file table.
686 *
687 * To handle those cases, we just bail out after a few iterations.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400689
690#define MAX_DEADLK_ITERATIONS 10
691
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400692/* Find a lock that the owner of the given block_fl is blocking on. */
693static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
694{
695 struct file_lock *fl;
696
697 list_for_each_entry(fl, &blocked_list, fl_link) {
698 if (posix_same_owner(fl, block_fl))
699 return fl->fl_next;
700 }
701 return NULL;
702}
703
Adrian Bunkb0904e12006-06-23 02:05:13 -0700704static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 struct file_lock *block_fl)
706{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400707 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400709 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
710 if (i++ > MAX_DEADLK_ITERATIONS)
711 return 0;
712 if (posix_same_owner(caller_fl, block_fl))
713 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
715 return 0;
716}
717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400719 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400720 *
721 * Note that if called with an FL_EXISTS argument, the caller may determine
722 * whether or not a lock was successfully freed by testing the return
723 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800725static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800727 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800729 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 int error = 0;
731 int found = 0;
732
733 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400734 if (request->fl_flags & FL_ACCESS)
735 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400736
737 if (request->fl_type != F_UNLCK) {
738 error = -ENOMEM;
739 new_fl = locks_alloc_lock();
740 if (new_fl == NULL)
741 goto out;
742 error = 0;
743 }
744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 for_each_lock(inode, before) {
746 struct file_lock *fl = *before;
747 if (IS_POSIX(fl))
748 break;
749 if (IS_LEASE(fl))
750 continue;
751 if (filp != fl->fl_file)
752 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800753 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 goto out;
755 found = 1;
756 locks_delete_lock(before);
757 break;
758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
Trond Myklebustf475ae92006-06-29 16:38:32 -0400760 if (request->fl_type == F_UNLCK) {
761 if ((request->fl_flags & FL_EXISTS) && !found)
762 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800763 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
766 /*
767 * If a higher-priority process was blocked on the old file lock,
768 * give it the opportunity to lock the file.
769 */
770 if (found)
Frederic Weisbeckerdef01bc2009-07-16 15:44:29 +0200771 cond_resched();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400773find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 for_each_lock(inode, before) {
775 struct file_lock *fl = *before;
776 if (IS_POSIX(fl))
777 break;
778 if (IS_LEASE(fl))
779 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800780 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 continue;
782 error = -EAGAIN;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700783 if (!(request->fl_flags & FL_SLEEP))
784 goto out;
785 error = FILE_LOCK_DEFERRED;
786 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 goto out;
788 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400789 if (request->fl_flags & FL_ACCESS)
790 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800791 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700792 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800793 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400794 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
796out:
797 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800798 if (new_fl)
799 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 return error;
801}
802
Marc Eshel150b3932007-01-18 16:15:35 -0500803static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804{
805 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700806 struct file_lock *new_fl = NULL;
807 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 struct file_lock *left = NULL;
809 struct file_lock *right = NULL;
810 struct file_lock **before;
811 int error, added = 0;
812
813 /*
814 * We may need two file_lock structures for this operation,
815 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700816 *
817 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700819 if (!(request->fl_flags & FL_ACCESS) &&
820 (request->fl_type != F_UNLCK ||
821 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
822 new_fl = locks_alloc_lock();
823 new_fl2 = locks_alloc_lock();
824 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
826 lock_kernel();
827 if (request->fl_type != F_UNLCK) {
828 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500829 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 if (!IS_POSIX(fl))
831 continue;
832 if (!posix_locks_conflict(request, fl))
833 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800834 if (conflock)
J. Bruce Fields1a747ee2008-04-24 10:08:22 -0400835 __locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 error = -EAGAIN;
837 if (!(request->fl_flags & FL_SLEEP))
838 goto out;
839 error = -EDEADLK;
840 if (posix_locks_deadlock(request, fl))
841 goto out;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700842 error = FILE_LOCK_DEFERRED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 locks_insert_block(fl, request);
844 goto out;
845 }
846 }
847
848 /* If we're just looking for a conflict, we're done. */
849 error = 0;
850 if (request->fl_flags & FL_ACCESS)
851 goto out;
852
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 * Find the first old lock with the same owner as the new lock.
855 */
856
857 before = &inode->i_flock;
858
859 /* First skip locks owned by other processes. */
860 while ((fl = *before) && (!IS_POSIX(fl) ||
861 !posix_same_owner(request, fl))) {
862 before = &fl->fl_next;
863 }
864
865 /* Process locks with this owner. */
866 while ((fl = *before) && posix_same_owner(request, fl)) {
867 /* Detect adjacent or overlapping regions (if same lock type)
868 */
869 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700870 /* In all comparisons of start vs end, use
871 * "start - 1" rather than "end + 1". If end
872 * is OFFSET_MAX, end + 1 will become negative.
873 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 if (fl->fl_end < request->fl_start - 1)
875 goto next_lock;
876 /* If the next lock in the list has entirely bigger
877 * addresses than the new one, insert the lock here.
878 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700879 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 break;
881
882 /* If we come here, the new and old lock are of the
883 * same type and adjacent or overlapping. Make one
884 * lock yielding from the lower start address of both
885 * locks to the higher end address.
886 */
887 if (fl->fl_start > request->fl_start)
888 fl->fl_start = request->fl_start;
889 else
890 request->fl_start = fl->fl_start;
891 if (fl->fl_end < request->fl_end)
892 fl->fl_end = request->fl_end;
893 else
894 request->fl_end = fl->fl_end;
895 if (added) {
896 locks_delete_lock(before);
897 continue;
898 }
899 request = fl;
900 added = 1;
901 }
902 else {
903 /* Processing for different lock types is a bit
904 * more complex.
905 */
906 if (fl->fl_end < request->fl_start)
907 goto next_lock;
908 if (fl->fl_start > request->fl_end)
909 break;
910 if (request->fl_type == F_UNLCK)
911 added = 1;
912 if (fl->fl_start < request->fl_start)
913 left = fl;
914 /* If the next lock in the list has a higher end
915 * address than the new one, insert the new one here.
916 */
917 if (fl->fl_end > request->fl_end) {
918 right = fl;
919 break;
920 }
921 if (fl->fl_start >= request->fl_start) {
922 /* The new lock completely replaces an old
923 * one (This may happen several times).
924 */
925 if (added) {
926 locks_delete_lock(before);
927 continue;
928 }
929 /* Replace the old lock with the new one.
930 * Wake up anybody waiting for the old one,
931 * as the change in lock type might satisfy
932 * their needs.
933 */
934 locks_wake_up_blocks(fl);
935 fl->fl_start = request->fl_start;
936 fl->fl_end = request->fl_end;
937 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500938 locks_release_private(fl);
939 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 request = fl;
941 added = 1;
942 }
943 }
944 /* Go on to next lock.
945 */
946 next_lock:
947 before = &fl->fl_next;
948 }
949
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700950 /*
951 * The above code only modifies existing locks in case of
952 * merging or replacing. If new lock(s) need to be inserted
953 * all modifications are done bellow this, so it's safe yet to
954 * bail out.
955 */
956 error = -ENOLCK; /* "no luck" */
957 if (right && left == right && !new_fl2)
958 goto out;
959
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 error = 0;
961 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400962 if (request->fl_type == F_UNLCK) {
963 if (request->fl_flags & FL_EXISTS)
964 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400966 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700967
968 if (!new_fl) {
969 error = -ENOLCK;
970 goto out;
971 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 locks_copy_lock(new_fl, request);
973 locks_insert_lock(before, new_fl);
974 new_fl = NULL;
975 }
976 if (right) {
977 if (left == right) {
978 /* The new lock breaks the old one in two pieces,
979 * so we have to use the second new lock.
980 */
981 left = new_fl2;
982 new_fl2 = NULL;
983 locks_copy_lock(left, right);
984 locks_insert_lock(before, left);
985 }
986 right->fl_start = request->fl_end + 1;
987 locks_wake_up_blocks(right);
988 }
989 if (left) {
990 left->fl_end = request->fl_start - 1;
991 locks_wake_up_blocks(left);
992 }
993 out:
994 unlock_kernel();
995 /*
996 * Free any unused locks.
997 */
998 if (new_fl)
999 locks_free_lock(new_fl);
1000 if (new_fl2)
1001 locks_free_lock(new_fl2);
1002 return error;
1003}
1004
1005/**
1006 * posix_lock_file - Apply a POSIX-style lock to a file
1007 * @filp: The file to apply the lock to
1008 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001009 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 *
1011 * Add a POSIX style lock to a file.
1012 * We merge adjacent & overlapping locks whenever possible.
1013 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001014 *
1015 * Note that if called with an FL_EXISTS argument, the caller may determine
1016 * whether or not a lock was successfully freed by testing the return
1017 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 */
Marc Eshel150b3932007-01-18 16:15:35 -05001019int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001020 struct file_lock *conflock)
1021{
Marc Eshel150b3932007-01-18 16:15:35 -05001022 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001023}
Marc Eshel150b3932007-01-18 16:15:35 -05001024EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
1026/**
1027 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1028 * @filp: The file to apply the lock to
1029 * @fl: The lock to be applied
1030 *
1031 * Add a POSIX style lock to a file.
1032 * We merge adjacent & overlapping locks whenever possible.
1033 * POSIX locks are sorted by owner task, then by starting address
1034 */
1035int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1036{
1037 int error;
1038 might_sleep ();
1039 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001040 error = posix_lock_file(filp, fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001041 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 break;
1043 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1044 if (!error)
1045 continue;
1046
1047 locks_delete_block(fl);
1048 break;
1049 }
1050 return error;
1051}
1052EXPORT_SYMBOL(posix_lock_file_wait);
1053
1054/**
1055 * locks_mandatory_locked - Check for an active lock
1056 * @inode: the file to check
1057 *
1058 * Searches the inode's list of locks to find any POSIX locks which conflict.
1059 * This function is called from locks_verify_locked() only.
1060 */
1061int locks_mandatory_locked(struct inode *inode)
1062{
1063 fl_owner_t owner = current->files;
1064 struct file_lock *fl;
1065
1066 /*
1067 * Search the lock list for this inode for any POSIX locks.
1068 */
1069 lock_kernel();
1070 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1071 if (!IS_POSIX(fl))
1072 continue;
1073 if (fl->fl_owner != owner)
1074 break;
1075 }
1076 unlock_kernel();
1077 return fl ? -EAGAIN : 0;
1078}
1079
1080/**
1081 * locks_mandatory_area - Check for a conflicting lock
1082 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1083 * for shared
1084 * @inode: the file to check
1085 * @filp: how the file was opened (if it was)
1086 * @offset: start of area to check
1087 * @count: length of area to check
1088 *
1089 * Searches the inode's list of locks to find any POSIX locks which conflict.
1090 * This function is called from rw_verify_area() and
1091 * locks_verify_truncate().
1092 */
1093int locks_mandatory_area(int read_write, struct inode *inode,
1094 struct file *filp, loff_t offset,
1095 size_t count)
1096{
1097 struct file_lock fl;
1098 int error;
1099
1100 locks_init_lock(&fl);
1101 fl.fl_owner = current->files;
1102 fl.fl_pid = current->tgid;
1103 fl.fl_file = filp;
1104 fl.fl_flags = FL_POSIX | FL_ACCESS;
1105 if (filp && !(filp->f_flags & O_NONBLOCK))
1106 fl.fl_flags |= FL_SLEEP;
1107 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1108 fl.fl_start = offset;
1109 fl.fl_end = offset + count - 1;
1110
1111 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001112 error = __posix_lock_file(inode, &fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001113 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 break;
1115 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1116 if (!error) {
1117 /*
1118 * If we've been sleeping someone might have
1119 * changed the permissions behind our back.
1120 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001121 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 continue;
1123 }
1124
1125 locks_delete_block(&fl);
1126 break;
1127 }
1128
1129 return error;
1130}
1131
1132EXPORT_SYMBOL(locks_mandatory_area);
1133
1134/* We already had a lease on this file; just change its type */
1135int lease_modify(struct file_lock **before, int arg)
1136{
1137 struct file_lock *fl = *before;
1138 int error = assign_type(fl, arg);
1139
1140 if (error)
1141 return error;
1142 locks_wake_up_blocks(fl);
1143 if (arg == F_UNLCK)
1144 locks_delete_lock(before);
1145 return 0;
1146}
1147
1148EXPORT_SYMBOL(lease_modify);
1149
1150static void time_out_leases(struct inode *inode)
1151{
1152 struct file_lock **before;
1153 struct file_lock *fl;
1154
1155 before = &inode->i_flock;
1156 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1157 if ((fl->fl_break_time == 0)
1158 || time_before(jiffies, fl->fl_break_time)) {
1159 before = &fl->fl_next;
1160 continue;
1161 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1163 if (fl == *before) /* lease_modify may have freed fl */
1164 before = &fl->fl_next;
1165 }
1166}
1167
1168/**
1169 * __break_lease - revoke all outstanding leases on file
1170 * @inode: the inode of the file to return
1171 * @mode: the open mode (read or write)
1172 *
david m. richter87250dd2007-05-09 16:10:27 -04001173 * break_lease (inlined for speed) has checked there already is at least
1174 * some kind of lock (maybe a lease) on this file. Leases are broken on
1175 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 * specified %O_NONBLOCK to your open().
1177 */
1178int __break_lease(struct inode *inode, unsigned int mode)
1179{
1180 int error = 0, future;
1181 struct file_lock *new_fl, *flock;
1182 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 unsigned long break_time;
1184 int i_have_this_lease = 0;
1185
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001186 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
1188 lock_kernel();
1189
1190 time_out_leases(inode);
1191
1192 flock = inode->i_flock;
1193 if ((flock == NULL) || !IS_LEASE(flock))
1194 goto out;
1195
1196 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1197 if (fl->fl_owner == current->files)
1198 i_have_this_lease = 1;
1199
1200 if (mode & FMODE_WRITE) {
1201 /* If we want write access, we have to revoke any lease. */
1202 future = F_UNLCK | F_INPROGRESS;
1203 } else if (flock->fl_type & F_INPROGRESS) {
1204 /* If the lease is already being broken, we just leave it */
1205 future = flock->fl_type;
1206 } else if (flock->fl_type & F_WRLCK) {
1207 /* Downgrade the exclusive lease to a read-only lease. */
1208 future = F_RDLCK | F_INPROGRESS;
1209 } else {
1210 /* the existing lease was read-only, so we can read too. */
1211 goto out;
1212 }
1213
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001214 if (IS_ERR(new_fl) && !i_have_this_lease
1215 && ((mode & O_NONBLOCK) == 0)) {
1216 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 goto out;
1218 }
1219
1220 break_time = 0;
1221 if (lease_break_time > 0) {
1222 break_time = jiffies + lease_break_time * HZ;
1223 if (break_time == 0)
1224 break_time++; /* so that 0 means no break time */
1225 }
1226
1227 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1228 if (fl->fl_type != future) {
1229 fl->fl_type = future;
1230 fl->fl_break_time = break_time;
1231 /* lease must have lmops break callback */
1232 fl->fl_lmops->fl_break(fl);
1233 }
1234 }
1235
1236 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1237 error = -EWOULDBLOCK;
1238 goto out;
1239 }
1240
1241restart:
1242 break_time = flock->fl_break_time;
1243 if (break_time != 0) {
1244 break_time -= jiffies;
1245 if (break_time == 0)
1246 break_time++;
1247 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001248 locks_insert_block(flock, new_fl);
1249 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1250 !new_fl->fl_next, break_time);
1251 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 if (error >= 0) {
1253 if (error == 0)
1254 time_out_leases(inode);
1255 /* Wait for the next lease that has not been broken yet */
1256 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1257 flock = flock->fl_next) {
1258 if (flock->fl_type & F_INPROGRESS)
1259 goto restart;
1260 }
1261 error = 0;
1262 }
1263
1264out:
1265 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001266 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 locks_free_lock(new_fl);
1268 return error;
1269}
1270
1271EXPORT_SYMBOL(__break_lease);
1272
1273/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001274 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 * @inode: the inode
1276 * @time: pointer to a timespec which will contain the last modified time
1277 *
1278 * This is to force NFS clients to flush their caches for files with
1279 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001280 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 */
1282void lease_get_mtime(struct inode *inode, struct timespec *time)
1283{
1284 struct file_lock *flock = inode->i_flock;
1285 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1286 *time = current_fs_time(inode->i_sb);
1287 else
1288 *time = inode->i_mtime;
1289}
1290
1291EXPORT_SYMBOL(lease_get_mtime);
1292
1293/**
1294 * fcntl_getlease - Enquire what lease is currently active
1295 * @filp: the file
1296 *
1297 * The value returned by this function will be one of
1298 * (if no lease break is pending):
1299 *
1300 * %F_RDLCK to indicate a shared lease is held.
1301 *
1302 * %F_WRLCK to indicate an exclusive lease is held.
1303 *
1304 * %F_UNLCK to indicate no lease is held.
1305 *
1306 * (if a lease break is pending):
1307 *
1308 * %F_RDLCK to indicate an exclusive lease needs to be
1309 * changed to a shared lease (or removed).
1310 *
1311 * %F_UNLCK to indicate the lease needs to be removed.
1312 *
1313 * XXX: sfr & willy disagree over whether F_INPROGRESS
1314 * should be returned to userspace.
1315 */
1316int fcntl_getlease(struct file *filp)
1317{
1318 struct file_lock *fl;
1319 int type = F_UNLCK;
1320
1321 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001322 time_out_leases(filp->f_path.dentry->d_inode);
1323 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 fl = fl->fl_next) {
1325 if (fl->fl_file == filp) {
1326 type = fl->fl_type & ~F_INPROGRESS;
1327 break;
1328 }
1329 }
1330 unlock_kernel();
1331 return type;
1332}
1333
1334/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001335 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 * @filp: file pointer
1337 * @arg: type of lease to obtain
1338 * @flp: input - file_lock to use, output - file_lock inserted
1339 *
1340 * The (input) flp->fl_lmops->fl_break function is required
1341 * by break_lease().
1342 *
1343 * Called with kernel lock held.
1344 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001345int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001347 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001348 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001349 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 struct inode *inode = dentry->d_inode;
1351 int error, rdlease_count = 0, wrlease_count = 0;
1352
David Howellsda9592e2008-11-14 10:39:05 +11001353 if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE))
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001354 return -EACCES;
1355 if (!S_ISREG(inode->i_mode))
1356 return -EINVAL;
1357 error = security_file_lock(filp, arg);
1358 if (error)
1359 return error;
1360
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 time_out_leases(inode);
1362
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001363 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001365 lease = *flp;
1366
David M. Richter288b2fd2008-04-23 16:29:00 -04001367 if (arg != F_UNLCK) {
David M. Richter19085552008-04-23 16:29:01 -04001368 error = -ENOMEM;
1369 new_fl = locks_alloc_lock();
1370 if (new_fl == NULL)
1371 goto out;
1372
David M. Richter288b2fd2008-04-23 16:29:00 -04001373 error = -EAGAIN;
1374 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1375 goto out;
1376 if ((arg == F_WRLCK)
1377 && ((atomic_read(&dentry->d_count) > 1)
1378 || (atomic_read(&inode->i_count) > 1)))
1379 goto out;
David M. Richter288b2fd2008-04-23 16:29:00 -04001380 }
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001381
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 /*
1383 * At this point, we know that if there is an exclusive
1384 * lease on this file, then we hold it on this filp
1385 * (otherwise our open of this file would have blocked).
1386 * And if we are trying to acquire an exclusive lease,
1387 * then the file is not open by anyone (including us)
1388 * except for this filp.
1389 */
1390 for (before = &inode->i_flock;
1391 ((fl = *before) != NULL) && IS_LEASE(fl);
1392 before = &fl->fl_next) {
1393 if (lease->fl_lmops->fl_mylease(fl, lease))
1394 my_before = before;
1395 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1396 /*
1397 * Someone is in the process of opening this
1398 * file for writing so we may not take an
1399 * exclusive lease on it.
1400 */
1401 wrlease_count++;
1402 else
1403 rdlease_count++;
1404 }
1405
David M. Richter5fcc60c2008-04-23 16:28:59 -04001406 error = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 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;
David M. Richter9d91cdc2008-04-23 16:29:02 -04001493 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 int error;
1495
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 locks_init_lock(&fl);
1497 error = lease_init(filp, arg, &fl);
1498 if (error)
1499 return error;
1500
1501 lock_kernel();
1502
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001503 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001504 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 goto out_unlock;
1506
1507 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1508 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001509 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001511 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 time_out_leases(inode);
1513 goto out_unlock;
1514 }
1515
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001516 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517out_unlock:
1518 unlock_kernel();
1519 return error;
1520}
1521
1522/**
1523 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1524 * @filp: The file to apply the lock to
1525 * @fl: The lock to be applied
1526 *
1527 * Add a FLOCK style lock to a file.
1528 */
1529int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1530{
1531 int error;
1532 might_sleep();
1533 for (;;) {
1534 error = flock_lock_file(filp, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001535 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 break;
1537 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1538 if (!error)
1539 continue;
1540
1541 locks_delete_block(fl);
1542 break;
1543 }
1544 return error;
1545}
1546
1547EXPORT_SYMBOL(flock_lock_file_wait);
1548
1549/**
1550 * sys_flock: - flock() system call.
1551 * @fd: the file descriptor to lock.
1552 * @cmd: the type of lock to apply.
1553 *
1554 * Apply a %FL_FLOCK style lock to an open file descriptor.
1555 * The @cmd can be one of
1556 *
1557 * %LOCK_SH -- a shared lock.
1558 *
1559 * %LOCK_EX -- an exclusive lock.
1560 *
1561 * %LOCK_UN -- remove an existing lock.
1562 *
1563 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1564 *
1565 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1566 * processes read and write access respectively.
1567 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001568SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569{
1570 struct file *filp;
1571 struct file_lock *lock;
1572 int can_sleep, unlock;
1573 int error;
1574
1575 error = -EBADF;
1576 filp = fget(fd);
1577 if (!filp)
1578 goto out;
1579
1580 can_sleep = !(cmd & LOCK_NB);
1581 cmd &= ~LOCK_NB;
1582 unlock = (cmd == LOCK_UN);
1583
Al Viroaeb5d722008-09-02 15:28:45 -04001584 if (!unlock && !(cmd & LOCK_MAND) &&
1585 !(filp->f_mode & (FMODE_READ|FMODE_WRITE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 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
Sten Spans713c0ec2009-07-16 09:41:39 +02001594 error = security_file_lock(filp, lock->fl_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 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,
Miklos Szeredibde74e42008-07-25 01:48:57 -07001721 * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05001722 * request completes.
1723 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07001724 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1725 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05001726 * 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
Miklos Szeredibde74e42008-07-25 01:48:57 -07001731 * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05001732 * 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
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001743static int do_lock_file_wait(struct file *filp, unsigned int cmd,
1744 struct file_lock *fl)
1745{
1746 int error;
1747
1748 error = security_file_lock(filp, fl->fl_type);
1749 if (error)
1750 return error;
1751
Miklos Szeredi764c76b2008-07-25 01:48:58 -07001752 for (;;) {
1753 error = vfs_lock_file(filp, cmd, fl, NULL);
1754 if (error != FILE_LOCK_DEFERRED)
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001755 break;
Miklos Szeredi764c76b2008-07-25 01:48:58 -07001756 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1757 if (!error)
1758 continue;
1759
1760 locks_delete_block(fl);
1761 break;
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001762 }
1763
1764 return error;
1765}
1766
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767/* Apply the lock described by l to an open file descriptor.
1768 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1769 */
Peter Staubachc2936212005-07-27 11:45:09 -07001770int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1771 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772{
1773 struct file_lock *file_lock = locks_alloc_lock();
1774 struct flock flock;
1775 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001776 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 int error;
1778
1779 if (file_lock == NULL)
1780 return -ENOLCK;
1781
1782 /*
1783 * This might block, so we do it before checking the inode.
1784 */
1785 error = -EFAULT;
1786 if (copy_from_user(&flock, l, sizeof(flock)))
1787 goto out;
1788
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001789 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
1791 /* Don't allow mandatory locks on files that may be memory mapped
1792 * and shared.
1793 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001794 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 error = -EAGAIN;
1796 goto out;
1797 }
1798
Peter Staubachc2936212005-07-27 11:45:09 -07001799again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 error = flock_to_posix_lock(filp, file_lock, &flock);
1801 if (error)
1802 goto out;
1803 if (cmd == F_SETLKW) {
1804 file_lock->fl_flags |= FL_SLEEP;
1805 }
1806
1807 error = -EBADF;
1808 switch (flock.l_type) {
1809 case F_RDLCK:
1810 if (!(filp->f_mode & FMODE_READ))
1811 goto out;
1812 break;
1813 case F_WRLCK:
1814 if (!(filp->f_mode & FMODE_WRITE))
1815 goto out;
1816 break;
1817 case F_UNLCK:
1818 break;
1819 default:
1820 error = -EINVAL;
1821 goto out;
1822 }
1823
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001824 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825
Peter Staubachc2936212005-07-27 11:45:09 -07001826 /*
1827 * Attempt to detect a close/fcntl race and recover by
1828 * releasing the lock that was just acquired.
1829 */
Al Viro0b2bac22008-05-06 13:58:34 -04001830 /*
1831 * we need that spin_lock here - it prevents reordering between
1832 * update of inode->i_flock and check for it done in close().
1833 * rcu_read_lock() wouldn't do.
1834 */
1835 spin_lock(&current->files->file_lock);
1836 f = fcheck(fd);
1837 spin_unlock(&current->files->file_lock);
1838 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001839 flock.l_type = F_UNLCK;
1840 goto again;
1841 }
1842
1843out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 locks_free_lock(file_lock);
1845 return error;
1846}
1847
1848#if BITS_PER_LONG == 32
1849/* Report the first existing lock that would conflict with l.
1850 * This implements the F_GETLK command of fcntl().
1851 */
1852int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1853{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001854 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 struct flock64 flock;
1856 int error;
1857
1858 error = -EFAULT;
1859 if (copy_from_user(&flock, l, sizeof(flock)))
1860 goto out;
1861 error = -EINVAL;
1862 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1863 goto out;
1864
1865 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1866 if (error)
1867 goto out;
1868
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001869 error = vfs_test_lock(filp, &file_lock);
1870 if (error)
1871 goto out;
1872
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001873 flock.l_type = file_lock.fl_type;
1874 if (file_lock.fl_type != F_UNLCK)
1875 posix_lock_to_flock64(&flock, &file_lock);
1876
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877 error = -EFAULT;
1878 if (!copy_to_user(l, &flock, sizeof(flock)))
1879 error = 0;
1880
1881out:
1882 return error;
1883}
1884
1885/* Apply the lock described by l to an open file descriptor.
1886 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1887 */
Peter Staubachc2936212005-07-27 11:45:09 -07001888int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1889 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890{
1891 struct file_lock *file_lock = locks_alloc_lock();
1892 struct flock64 flock;
1893 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001894 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 int error;
1896
1897 if (file_lock == NULL)
1898 return -ENOLCK;
1899
1900 /*
1901 * This might block, so we do it before checking the inode.
1902 */
1903 error = -EFAULT;
1904 if (copy_from_user(&flock, l, sizeof(flock)))
1905 goto out;
1906
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001907 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908
1909 /* Don't allow mandatory locks on files that may be memory mapped
1910 * and shared.
1911 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001912 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 error = -EAGAIN;
1914 goto out;
1915 }
1916
Peter Staubachc2936212005-07-27 11:45:09 -07001917again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 error = flock64_to_posix_lock(filp, file_lock, &flock);
1919 if (error)
1920 goto out;
1921 if (cmd == F_SETLKW64) {
1922 file_lock->fl_flags |= FL_SLEEP;
1923 }
1924
1925 error = -EBADF;
1926 switch (flock.l_type) {
1927 case F_RDLCK:
1928 if (!(filp->f_mode & FMODE_READ))
1929 goto out;
1930 break;
1931 case F_WRLCK:
1932 if (!(filp->f_mode & FMODE_WRITE))
1933 goto out;
1934 break;
1935 case F_UNLCK:
1936 break;
1937 default:
1938 error = -EINVAL;
1939 goto out;
1940 }
1941
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001942 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943
Peter Staubachc2936212005-07-27 11:45:09 -07001944 /*
1945 * Attempt to detect a close/fcntl race and recover by
1946 * releasing the lock that was just acquired.
1947 */
Al Viro0b2bac22008-05-06 13:58:34 -04001948 spin_lock(&current->files->file_lock);
1949 f = fcheck(fd);
1950 spin_unlock(&current->files->file_lock);
1951 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001952 flock.l_type = F_UNLCK;
1953 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 }
1955
1956out:
1957 locks_free_lock(file_lock);
1958 return error;
1959}
1960#endif /* BITS_PER_LONG == 32 */
1961
1962/*
1963 * This function is called when the file is being removed
1964 * from the task's fd array. POSIX locks belonging to this task
1965 * are deleted at this time.
1966 */
1967void locks_remove_posix(struct file *filp, fl_owner_t owner)
1968{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001969 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970
1971 /*
1972 * If there are no locks held on this file, we don't need to call
1973 * posix_lock_file(). Another process could be setting a lock on this
1974 * file at the same time, but we wouldn't remove that lock anyway.
1975 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001976 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 return;
1978
1979 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001980 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 lock.fl_start = 0;
1982 lock.fl_end = OFFSET_MAX;
1983 lock.fl_owner = owner;
1984 lock.fl_pid = current->tgid;
1985 lock.fl_file = filp;
1986 lock.fl_ops = NULL;
1987 lock.fl_lmops = NULL;
1988
Marc Eshel150b3932007-01-18 16:15:35 -05001989 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1992 lock.fl_ops->fl_release_private(&lock);
1993}
1994
1995EXPORT_SYMBOL(locks_remove_posix);
1996
1997/*
1998 * This function is called on the last close of an open file.
1999 */
2000void locks_remove_flock(struct file *filp)
2001{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002002 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 struct file_lock *fl;
2004 struct file_lock **before;
2005
2006 if (!inode->i_flock)
2007 return;
2008
2009 if (filp->f_op && filp->f_op->flock) {
2010 struct file_lock fl = {
2011 .fl_pid = current->tgid,
2012 .fl_file = filp,
2013 .fl_flags = FL_FLOCK,
2014 .fl_type = F_UNLCK,
2015 .fl_end = OFFSET_MAX,
2016 };
2017 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002018 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2019 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 }
2021
2022 lock_kernel();
2023 before = &inode->i_flock;
2024
2025 while ((fl = *before) != NULL) {
2026 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002027 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 locks_delete_lock(before);
2029 continue;
2030 }
2031 if (IS_LEASE(fl)) {
2032 lease_modify(before, F_UNLCK);
2033 continue;
2034 }
2035 /* What? */
2036 BUG();
2037 }
2038 before = &fl->fl_next;
2039 }
2040 unlock_kernel();
2041}
2042
2043/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 * posix_unblock_lock - stop waiting for a file lock
2045 * @filp: how the file was opened
2046 * @waiter: the lock which was waiting
2047 *
2048 * lockd needs to block waiting for locks.
2049 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002050int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2052{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002053 int status = 0;
2054
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002056 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002058 else
2059 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002060 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002061 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062}
2063
2064EXPORT_SYMBOL(posix_unblock_lock);
2065
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002066/**
2067 * vfs_cancel_lock - file byte range unblock lock
2068 * @filp: The file to apply the unblock to
2069 * @fl: The lock to be unblocked
2070 *
2071 * Used by lock managers to cancel blocked requests
2072 */
2073int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2074{
2075 if (filp->f_op && filp->f_op->lock)
2076 return filp->f_op->lock(filp, F_CANCELLK, fl);
2077 return 0;
2078}
2079
2080EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2081
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002082#ifdef CONFIG_PROC_FS
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002083#include <linux/proc_fs.h>
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002084#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
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002189static const struct seq_operations locks_seq_operations = {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002190 .start = locks_start,
2191 .next = locks_next,
2192 .stop = locks_stop,
2193 .show = locks_show,
2194};
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002195
2196static int locks_open(struct inode *inode, struct file *filp)
2197{
2198 return seq_open(filp, &locks_seq_operations);
2199}
2200
2201static const struct file_operations proc_locks_operations = {
2202 .open = locks_open,
2203 .read = seq_read,
2204 .llseek = seq_lseek,
2205 .release = seq_release,
2206};
2207
2208static int __init proc_locks_init(void)
2209{
2210 proc_create("locks", 0, NULL, &proc_locks_operations);
2211 return 0;
2212}
2213module_init(proc_locks_init);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002214#endif
2215
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216/**
2217 * lock_may_read - checks that the region is free of locks
2218 * @inode: the inode that is being read
2219 * @start: the first byte to read
2220 * @len: the number of bytes to read
2221 *
2222 * Emulates Windows locking requirements. Whole-file
2223 * mandatory locks (share modes) can prohibit a read and
2224 * byte-range POSIX locks can prohibit a read if they overlap.
2225 *
2226 * N.B. this function is only ever called
2227 * from knfsd and ownership of locks is never checked.
2228 */
2229int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2230{
2231 struct file_lock *fl;
2232 int result = 1;
2233 lock_kernel();
2234 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2235 if (IS_POSIX(fl)) {
2236 if (fl->fl_type == F_RDLCK)
2237 continue;
2238 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2239 continue;
2240 } else if (IS_FLOCK(fl)) {
2241 if (!(fl->fl_type & LOCK_MAND))
2242 continue;
2243 if (fl->fl_type & LOCK_READ)
2244 continue;
2245 } else
2246 continue;
2247 result = 0;
2248 break;
2249 }
2250 unlock_kernel();
2251 return result;
2252}
2253
2254EXPORT_SYMBOL(lock_may_read);
2255
2256/**
2257 * lock_may_write - checks that the region is free of locks
2258 * @inode: the inode that is being written
2259 * @start: the first byte to write
2260 * @len: the number of bytes to write
2261 *
2262 * Emulates Windows locking requirements. Whole-file
2263 * mandatory locks (share modes) can prohibit a write and
2264 * byte-range POSIX locks can prohibit a write if they overlap.
2265 *
2266 * N.B. this function is only ever called
2267 * from knfsd and ownership of locks is never checked.
2268 */
2269int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2270{
2271 struct file_lock *fl;
2272 int result = 1;
2273 lock_kernel();
2274 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2275 if (IS_POSIX(fl)) {
2276 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2277 continue;
2278 } else if (IS_FLOCK(fl)) {
2279 if (!(fl->fl_type & LOCK_MAND))
2280 continue;
2281 if (fl->fl_type & LOCK_WRITE)
2282 continue;
2283 } else
2284 continue;
2285 result = 0;
2286 break;
2287 }
2288 unlock_kernel();
2289 return result;
2290}
2291
2292EXPORT_SYMBOL(lock_may_write);
2293
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294static int __init filelock_init(void)
2295{
2296 filelock_cache = kmem_cache_create("file_lock_cache",
2297 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002298 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 return 0;
2300}
2301
2302core_initcall(filelock_init);