blob: ae9ded026b7cb18a783af01bffac31dd87d6a26e [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
Alexey Dobriyan7b021962009-09-21 17:01:12 -0700437static const 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;
Al Viro8737c932009-12-24 06:47:55 -05001185 int want_write = (mode & O_ACCMODE) != O_RDONLY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
Al Viro8737c932009-12-24 06:47:55 -05001187 new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188
1189 lock_kernel();
1190
1191 time_out_leases(inode);
1192
1193 flock = inode->i_flock;
1194 if ((flock == NULL) || !IS_LEASE(flock))
1195 goto out;
1196
1197 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1198 if (fl->fl_owner == current->files)
1199 i_have_this_lease = 1;
1200
Al Viro8737c932009-12-24 06:47:55 -05001201 if (want_write) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 /* If we want write access, we have to revoke any lease. */
1203 future = F_UNLCK | F_INPROGRESS;
1204 } else if (flock->fl_type & F_INPROGRESS) {
1205 /* If the lease is already being broken, we just leave it */
1206 future = flock->fl_type;
1207 } else if (flock->fl_type & F_WRLCK) {
1208 /* Downgrade the exclusive lease to a read-only lease. */
1209 future = F_RDLCK | F_INPROGRESS;
1210 } else {
1211 /* the existing lease was read-only, so we can read too. */
1212 goto out;
1213 }
1214
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001215 if (IS_ERR(new_fl) && !i_have_this_lease
1216 && ((mode & O_NONBLOCK) == 0)) {
1217 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 goto out;
1219 }
1220
1221 break_time = 0;
1222 if (lease_break_time > 0) {
1223 break_time = jiffies + lease_break_time * HZ;
1224 if (break_time == 0)
1225 break_time++; /* so that 0 means no break time */
1226 }
1227
1228 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1229 if (fl->fl_type != future) {
1230 fl->fl_type = future;
1231 fl->fl_break_time = break_time;
1232 /* lease must have lmops break callback */
1233 fl->fl_lmops->fl_break(fl);
1234 }
1235 }
1236
1237 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1238 error = -EWOULDBLOCK;
1239 goto out;
1240 }
1241
1242restart:
1243 break_time = flock->fl_break_time;
1244 if (break_time != 0) {
1245 break_time -= jiffies;
1246 if (break_time == 0)
1247 break_time++;
1248 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001249 locks_insert_block(flock, new_fl);
1250 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1251 !new_fl->fl_next, break_time);
1252 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 if (error >= 0) {
1254 if (error == 0)
1255 time_out_leases(inode);
1256 /* Wait for the next lease that has not been broken yet */
1257 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1258 flock = flock->fl_next) {
1259 if (flock->fl_type & F_INPROGRESS)
1260 goto restart;
1261 }
1262 error = 0;
1263 }
1264
1265out:
1266 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001267 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 locks_free_lock(new_fl);
1269 return error;
1270}
1271
1272EXPORT_SYMBOL(__break_lease);
1273
1274/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001275 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 * @inode: the inode
1277 * @time: pointer to a timespec which will contain the last modified time
1278 *
1279 * This is to force NFS clients to flush their caches for files with
1280 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001281 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 */
1283void lease_get_mtime(struct inode *inode, struct timespec *time)
1284{
1285 struct file_lock *flock = inode->i_flock;
1286 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1287 *time = current_fs_time(inode->i_sb);
1288 else
1289 *time = inode->i_mtime;
1290}
1291
1292EXPORT_SYMBOL(lease_get_mtime);
1293
1294/**
1295 * fcntl_getlease - Enquire what lease is currently active
1296 * @filp: the file
1297 *
1298 * The value returned by this function will be one of
1299 * (if no lease break is pending):
1300 *
1301 * %F_RDLCK to indicate a shared lease is held.
1302 *
1303 * %F_WRLCK to indicate an exclusive lease is held.
1304 *
1305 * %F_UNLCK to indicate no lease is held.
1306 *
1307 * (if a lease break is pending):
1308 *
1309 * %F_RDLCK to indicate an exclusive lease needs to be
1310 * changed to a shared lease (or removed).
1311 *
1312 * %F_UNLCK to indicate the lease needs to be removed.
1313 *
1314 * XXX: sfr & willy disagree over whether F_INPROGRESS
1315 * should be returned to userspace.
1316 */
1317int fcntl_getlease(struct file *filp)
1318{
1319 struct file_lock *fl;
1320 int type = F_UNLCK;
1321
1322 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001323 time_out_leases(filp->f_path.dentry->d_inode);
1324 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 fl = fl->fl_next) {
1326 if (fl->fl_file == filp) {
1327 type = fl->fl_type & ~F_INPROGRESS;
1328 break;
1329 }
1330 }
1331 unlock_kernel();
1332 return type;
1333}
1334
1335/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001336 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 * @filp: file pointer
1338 * @arg: type of lease to obtain
1339 * @flp: input - file_lock to use, output - file_lock inserted
1340 *
1341 * The (input) flp->fl_lmops->fl_break function is required
1342 * by break_lease().
1343 *
1344 * Called with kernel lock held.
1345 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001346int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001348 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001349 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001350 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 struct inode *inode = dentry->d_inode;
1352 int error, rdlease_count = 0, wrlease_count = 0;
1353
David Howellsda9592e2008-11-14 10:39:05 +11001354 if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE))
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001355 return -EACCES;
1356 if (!S_ISREG(inode->i_mode))
1357 return -EINVAL;
1358 error = security_file_lock(filp, arg);
1359 if (error)
1360 return error;
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 time_out_leases(inode);
1363
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001364 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001366 lease = *flp;
1367
David M. Richter288b2fd2008-04-23 16:29:00 -04001368 if (arg != F_UNLCK) {
David M. Richter19085552008-04-23 16:29:01 -04001369 error = -ENOMEM;
1370 new_fl = locks_alloc_lock();
1371 if (new_fl == NULL)
1372 goto out;
1373
David M. Richter288b2fd2008-04-23 16:29:00 -04001374 error = -EAGAIN;
1375 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1376 goto out;
1377 if ((arg == F_WRLCK)
1378 && ((atomic_read(&dentry->d_count) > 1)
1379 || (atomic_read(&inode->i_count) > 1)))
1380 goto out;
David M. Richter288b2fd2008-04-23 16:29:00 -04001381 }
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 /*
1384 * At this point, we know that if there is an exclusive
1385 * lease on this file, then we hold it on this filp
1386 * (otherwise our open of this file would have blocked).
1387 * And if we are trying to acquire an exclusive lease,
1388 * then the file is not open by anyone (including us)
1389 * except for this filp.
1390 */
1391 for (before = &inode->i_flock;
1392 ((fl = *before) != NULL) && IS_LEASE(fl);
1393 before = &fl->fl_next) {
1394 if (lease->fl_lmops->fl_mylease(fl, lease))
1395 my_before = before;
1396 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1397 /*
1398 * Someone is in the process of opening this
1399 * file for writing so we may not take an
1400 * exclusive lease on it.
1401 */
1402 wrlease_count++;
1403 else
1404 rdlease_count++;
1405 }
1406
David M. Richter5fcc60c2008-04-23 16:28:59 -04001407 error = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 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;
David M. Richter9d91cdc2008-04-23 16:29:02 -04001494 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 int error;
1496
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 locks_init_lock(&fl);
1498 error = lease_init(filp, arg, &fl);
1499 if (error)
1500 return error;
1501
1502 lock_kernel();
1503
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001504 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001505 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 goto out_unlock;
1507
1508 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1509 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001510 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001512 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 time_out_leases(inode);
1514 goto out_unlock;
1515 }
1516
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001517 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518out_unlock:
1519 unlock_kernel();
1520 return error;
1521}
1522
1523/**
1524 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1525 * @filp: The file to apply the lock to
1526 * @fl: The lock to be applied
1527 *
1528 * Add a FLOCK style lock to a file.
1529 */
1530int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1531{
1532 int error;
1533 might_sleep();
1534 for (;;) {
1535 error = flock_lock_file(filp, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001536 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 break;
1538 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1539 if (!error)
1540 continue;
1541
1542 locks_delete_block(fl);
1543 break;
1544 }
1545 return error;
1546}
1547
1548EXPORT_SYMBOL(flock_lock_file_wait);
1549
1550/**
1551 * sys_flock: - flock() system call.
1552 * @fd: the file descriptor to lock.
1553 * @cmd: the type of lock to apply.
1554 *
1555 * Apply a %FL_FLOCK style lock to an open file descriptor.
1556 * The @cmd can be one of
1557 *
1558 * %LOCK_SH -- a shared lock.
1559 *
1560 * %LOCK_EX -- an exclusive lock.
1561 *
1562 * %LOCK_UN -- remove an existing lock.
1563 *
1564 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1565 *
1566 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1567 * processes read and write access respectively.
1568 */
Heiko Carstens002c8972009-01-14 14:14:18 +01001569SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570{
1571 struct file *filp;
1572 struct file_lock *lock;
1573 int can_sleep, unlock;
1574 int error;
1575
1576 error = -EBADF;
1577 filp = fget(fd);
1578 if (!filp)
1579 goto out;
1580
1581 can_sleep = !(cmd & LOCK_NB);
1582 cmd &= ~LOCK_NB;
1583 unlock = (cmd == LOCK_UN);
1584
Al Viroaeb5d722008-09-02 15:28:45 -04001585 if (!unlock && !(cmd & LOCK_MAND) &&
1586 !(filp->f_mode & (FMODE_READ|FMODE_WRITE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 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
Sten Spans713c0ec2009-07-16 09:41:39 +02001595 error = security_file_lock(filp, lock->fl_type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 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,
Miklos Szeredibde74e42008-07-25 01:48:57 -07001722 * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05001723 * request completes.
1724 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07001725 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1726 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05001727 * 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
Miklos Szeredibde74e42008-07-25 01:48:57 -07001732 * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05001733 * 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
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001744static int do_lock_file_wait(struct file *filp, unsigned int cmd,
1745 struct file_lock *fl)
1746{
1747 int error;
1748
1749 error = security_file_lock(filp, fl->fl_type);
1750 if (error)
1751 return error;
1752
Miklos Szeredi764c76b2008-07-25 01:48:58 -07001753 for (;;) {
1754 error = vfs_lock_file(filp, cmd, fl, NULL);
1755 if (error != FILE_LOCK_DEFERRED)
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001756 break;
Miklos Szeredi764c76b2008-07-25 01:48:58 -07001757 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1758 if (!error)
1759 continue;
1760
1761 locks_delete_block(fl);
1762 break;
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001763 }
1764
1765 return error;
1766}
1767
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768/* Apply the lock described by l to an open file descriptor.
1769 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1770 */
Peter Staubachc2936212005-07-27 11:45:09 -07001771int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1772 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773{
1774 struct file_lock *file_lock = locks_alloc_lock();
1775 struct flock flock;
1776 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001777 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 int error;
1779
1780 if (file_lock == NULL)
1781 return -ENOLCK;
1782
1783 /*
1784 * This might block, so we do it before checking the inode.
1785 */
1786 error = -EFAULT;
1787 if (copy_from_user(&flock, l, sizeof(flock)))
1788 goto out;
1789
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001790 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 /* Don't allow mandatory locks on files that may be memory mapped
1793 * and shared.
1794 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001795 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 error = -EAGAIN;
1797 goto out;
1798 }
1799
Peter Staubachc2936212005-07-27 11:45:09 -07001800again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 error = flock_to_posix_lock(filp, file_lock, &flock);
1802 if (error)
1803 goto out;
1804 if (cmd == F_SETLKW) {
1805 file_lock->fl_flags |= FL_SLEEP;
1806 }
1807
1808 error = -EBADF;
1809 switch (flock.l_type) {
1810 case F_RDLCK:
1811 if (!(filp->f_mode & FMODE_READ))
1812 goto out;
1813 break;
1814 case F_WRLCK:
1815 if (!(filp->f_mode & FMODE_WRITE))
1816 goto out;
1817 break;
1818 case F_UNLCK:
1819 break;
1820 default:
1821 error = -EINVAL;
1822 goto out;
1823 }
1824
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001825 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826
Peter Staubachc2936212005-07-27 11:45:09 -07001827 /*
1828 * Attempt to detect a close/fcntl race and recover by
1829 * releasing the lock that was just acquired.
1830 */
Al Viro0b2bac22008-05-06 13:58:34 -04001831 /*
1832 * we need that spin_lock here - it prevents reordering between
1833 * update of inode->i_flock and check for it done in close().
1834 * rcu_read_lock() wouldn't do.
1835 */
1836 spin_lock(&current->files->file_lock);
1837 f = fcheck(fd);
1838 spin_unlock(&current->files->file_lock);
1839 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001840 flock.l_type = F_UNLCK;
1841 goto again;
1842 }
1843
1844out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 locks_free_lock(file_lock);
1846 return error;
1847}
1848
1849#if BITS_PER_LONG == 32
1850/* Report the first existing lock that would conflict with l.
1851 * This implements the F_GETLK command of fcntl().
1852 */
1853int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1854{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001855 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 struct flock64 flock;
1857 int error;
1858
1859 error = -EFAULT;
1860 if (copy_from_user(&flock, l, sizeof(flock)))
1861 goto out;
1862 error = -EINVAL;
1863 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1864 goto out;
1865
1866 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1867 if (error)
1868 goto out;
1869
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001870 error = vfs_test_lock(filp, &file_lock);
1871 if (error)
1872 goto out;
1873
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001874 flock.l_type = file_lock.fl_type;
1875 if (file_lock.fl_type != F_UNLCK)
1876 posix_lock_to_flock64(&flock, &file_lock);
1877
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 error = -EFAULT;
1879 if (!copy_to_user(l, &flock, sizeof(flock)))
1880 error = 0;
1881
1882out:
1883 return error;
1884}
1885
1886/* Apply the lock described by l to an open file descriptor.
1887 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1888 */
Peter Staubachc2936212005-07-27 11:45:09 -07001889int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1890 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891{
1892 struct file_lock *file_lock = locks_alloc_lock();
1893 struct flock64 flock;
1894 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001895 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 int error;
1897
1898 if (file_lock == NULL)
1899 return -ENOLCK;
1900
1901 /*
1902 * This might block, so we do it before checking the inode.
1903 */
1904 error = -EFAULT;
1905 if (copy_from_user(&flock, l, sizeof(flock)))
1906 goto out;
1907
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001908 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
1910 /* Don't allow mandatory locks on files that may be memory mapped
1911 * and shared.
1912 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001913 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 error = -EAGAIN;
1915 goto out;
1916 }
1917
Peter Staubachc2936212005-07-27 11:45:09 -07001918again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 error = flock64_to_posix_lock(filp, file_lock, &flock);
1920 if (error)
1921 goto out;
1922 if (cmd == F_SETLKW64) {
1923 file_lock->fl_flags |= FL_SLEEP;
1924 }
1925
1926 error = -EBADF;
1927 switch (flock.l_type) {
1928 case F_RDLCK:
1929 if (!(filp->f_mode & FMODE_READ))
1930 goto out;
1931 break;
1932 case F_WRLCK:
1933 if (!(filp->f_mode & FMODE_WRITE))
1934 goto out;
1935 break;
1936 case F_UNLCK:
1937 break;
1938 default:
1939 error = -EINVAL;
1940 goto out;
1941 }
1942
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001943 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944
Peter Staubachc2936212005-07-27 11:45:09 -07001945 /*
1946 * Attempt to detect a close/fcntl race and recover by
1947 * releasing the lock that was just acquired.
1948 */
Al Viro0b2bac22008-05-06 13:58:34 -04001949 spin_lock(&current->files->file_lock);
1950 f = fcheck(fd);
1951 spin_unlock(&current->files->file_lock);
1952 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001953 flock.l_type = F_UNLCK;
1954 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 }
1956
1957out:
1958 locks_free_lock(file_lock);
1959 return error;
1960}
1961#endif /* BITS_PER_LONG == 32 */
1962
1963/*
1964 * This function is called when the file is being removed
1965 * from the task's fd array. POSIX locks belonging to this task
1966 * are deleted at this time.
1967 */
1968void locks_remove_posix(struct file *filp, fl_owner_t owner)
1969{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001970 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
1972 /*
1973 * If there are no locks held on this file, we don't need to call
1974 * posix_lock_file(). Another process could be setting a lock on this
1975 * file at the same time, but we wouldn't remove that lock anyway.
1976 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001977 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 return;
1979
1980 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001981 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 lock.fl_start = 0;
1983 lock.fl_end = OFFSET_MAX;
1984 lock.fl_owner = owner;
1985 lock.fl_pid = current->tgid;
1986 lock.fl_file = filp;
1987 lock.fl_ops = NULL;
1988 lock.fl_lmops = NULL;
1989
Marc Eshel150b3932007-01-18 16:15:35 -05001990 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1993 lock.fl_ops->fl_release_private(&lock);
1994}
1995
1996EXPORT_SYMBOL(locks_remove_posix);
1997
1998/*
1999 * This function is called on the last close of an open file.
2000 */
2001void locks_remove_flock(struct file *filp)
2002{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002003 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 struct file_lock *fl;
2005 struct file_lock **before;
2006
2007 if (!inode->i_flock)
2008 return;
2009
2010 if (filp->f_op && filp->f_op->flock) {
2011 struct file_lock fl = {
2012 .fl_pid = current->tgid,
2013 .fl_file = filp,
2014 .fl_flags = FL_FLOCK,
2015 .fl_type = F_UNLCK,
2016 .fl_end = OFFSET_MAX,
2017 };
2018 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002019 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2020 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 }
2022
2023 lock_kernel();
2024 before = &inode->i_flock;
2025
2026 while ((fl = *before) != NULL) {
2027 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002028 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 locks_delete_lock(before);
2030 continue;
2031 }
2032 if (IS_LEASE(fl)) {
2033 lease_modify(before, F_UNLCK);
2034 continue;
2035 }
2036 /* What? */
2037 BUG();
2038 }
2039 before = &fl->fl_next;
2040 }
2041 unlock_kernel();
2042}
2043
2044/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 * posix_unblock_lock - stop waiting for a file lock
2046 * @filp: how the file was opened
2047 * @waiter: the lock which was waiting
2048 *
2049 * lockd needs to block waiting for locks.
2050 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002051int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2053{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002054 int status = 0;
2055
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002057 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002059 else
2060 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002061 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002062 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063}
2064
2065EXPORT_SYMBOL(posix_unblock_lock);
2066
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002067/**
2068 * vfs_cancel_lock - file byte range unblock lock
2069 * @filp: The file to apply the unblock to
2070 * @fl: The lock to be unblocked
2071 *
2072 * Used by lock managers to cancel blocked requests
2073 */
2074int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2075{
2076 if (filp->f_op && filp->f_op->lock)
2077 return filp->f_op->lock(filp, F_CANCELLK, fl);
2078 return 0;
2079}
2080
2081EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2082
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002083#ifdef CONFIG_PROC_FS
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002084#include <linux/proc_fs.h>
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002085#include <linux/seq_file.h>
2086
2087static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2088 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089{
2090 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002091 unsigned int fl_pid;
2092
2093 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002094 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002095 else
2096 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
2098 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002099 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002101 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002103 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2105 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002106 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 } else if (IS_FLOCK(fl)) {
2108 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002109 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002111 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 }
2113 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002114 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002116 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002118 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002120 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002122 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 }
2124 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002125 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 (fl->fl_type & LOCK_READ)
2127 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2128 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2129 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002130 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 (fl->fl_type & F_INPROGRESS)
2132 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2133 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2134 }
2135 if (inode) {
2136#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002137 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 inode->i_sb->s_id, inode->i_ino);
2139#else
2140 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002141 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 MAJOR(inode->i_sb->s_dev),
2143 MINOR(inode->i_sb->s_dev), inode->i_ino);
2144#endif
2145 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002146 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 }
2148 if (IS_POSIX(fl)) {
2149 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002150 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002152 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002154 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 }
2156}
2157
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002158static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002160 struct file_lock *fl, *bfl;
2161
2162 fl = list_entry(v, struct file_lock, fl_link);
2163
2164 lock_get_status(f, fl, (long)f->private, "");
2165
2166 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2167 lock_get_status(f, bfl, (long)f->private, " ->");
2168
2169 f->private++;
2170 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171}
2172
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002173static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002176 f->private = (void *)1;
2177 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178}
2179
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002180static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2181{
2182 return seq_list_next(v, &file_lock_list, pos);
2183}
2184
2185static void locks_stop(struct seq_file *f, void *v)
2186{
2187 unlock_kernel();
2188}
2189
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002190static const struct seq_operations locks_seq_operations = {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002191 .start = locks_start,
2192 .next = locks_next,
2193 .stop = locks_stop,
2194 .show = locks_show,
2195};
Alexey Dobriyand8ba7a32008-10-04 22:34:18 +04002196
2197static int locks_open(struct inode *inode, struct file *filp)
2198{
2199 return seq_open(filp, &locks_seq_operations);
2200}
2201
2202static const struct file_operations proc_locks_operations = {
2203 .open = locks_open,
2204 .read = seq_read,
2205 .llseek = seq_lseek,
2206 .release = seq_release,
2207};
2208
2209static int __init proc_locks_init(void)
2210{
2211 proc_create("locks", 0, NULL, &proc_locks_operations);
2212 return 0;
2213}
2214module_init(proc_locks_init);
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002215#endif
2216
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217/**
2218 * lock_may_read - checks that the region is free of locks
2219 * @inode: the inode that is being read
2220 * @start: the first byte to read
2221 * @len: the number of bytes to read
2222 *
2223 * Emulates Windows locking requirements. Whole-file
2224 * mandatory locks (share modes) can prohibit a read and
2225 * byte-range POSIX locks can prohibit a read if they overlap.
2226 *
2227 * N.B. this function is only ever called
2228 * from knfsd and ownership of locks is never checked.
2229 */
2230int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2231{
2232 struct file_lock *fl;
2233 int result = 1;
2234 lock_kernel();
2235 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2236 if (IS_POSIX(fl)) {
2237 if (fl->fl_type == F_RDLCK)
2238 continue;
2239 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2240 continue;
2241 } else if (IS_FLOCK(fl)) {
2242 if (!(fl->fl_type & LOCK_MAND))
2243 continue;
2244 if (fl->fl_type & LOCK_READ)
2245 continue;
2246 } else
2247 continue;
2248 result = 0;
2249 break;
2250 }
2251 unlock_kernel();
2252 return result;
2253}
2254
2255EXPORT_SYMBOL(lock_may_read);
2256
2257/**
2258 * lock_may_write - checks that the region is free of locks
2259 * @inode: the inode that is being written
2260 * @start: the first byte to write
2261 * @len: the number of bytes to write
2262 *
2263 * Emulates Windows locking requirements. Whole-file
2264 * mandatory locks (share modes) can prohibit a write and
2265 * byte-range POSIX locks can prohibit a write if they overlap.
2266 *
2267 * N.B. this function is only ever called
2268 * from knfsd and ownership of locks is never checked.
2269 */
2270int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2271{
2272 struct file_lock *fl;
2273 int result = 1;
2274 lock_kernel();
2275 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2276 if (IS_POSIX(fl)) {
2277 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2278 continue;
2279 } else if (IS_FLOCK(fl)) {
2280 if (!(fl->fl_type & LOCK_MAND))
2281 continue;
2282 if (fl->fl_type & LOCK_WRITE)
2283 continue;
2284 } else
2285 continue;
2286 result = 0;
2287 break;
2288 }
2289 unlock_kernel();
2290 return result;
2291}
2292
2293EXPORT_SYMBOL(lock_may_write);
2294
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295static int __init filelock_init(void)
2296{
2297 filelock_cache = kmem_cache_create("file_lock_cache",
2298 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002299 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002300 return 0;
2301}
2302
2303core_initcall(filelock_init);