blob: 6222e4b580e283728025181690b9528448e74e09 [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
Trond Myklebust47831f32006-03-20 13:44:05 -0500154static void locks_release_private(struct file_lock *fl)
155{
156 if (fl->fl_ops) {
157 if (fl->fl_ops->fl_release_private)
158 fl->fl_ops->fl_release_private(fl);
159 fl->fl_ops = NULL;
160 }
161 if (fl->fl_lmops) {
162 if (fl->fl_lmops->fl_release_private)
163 fl->fl_lmops->fl_release_private(fl);
164 fl->fl_lmops = NULL;
165 }
166
167}
168
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169/* Free a lock which is not in use. */
Matt Mackall33443c42006-01-08 01:05:22 -0800170static void locks_free_lock(struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171{
Miklos Szeredi5ce29642006-03-31 02:30:29 -0800172 BUG_ON(waitqueue_active(&fl->fl_wait));
173 BUG_ON(!list_empty(&fl->fl_block));
174 BUG_ON(!list_empty(&fl->fl_link));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175
Trond Myklebust47831f32006-03-20 13:44:05 -0500176 locks_release_private(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 kmem_cache_free(filelock_cache, fl);
178}
179
180void locks_init_lock(struct file_lock *fl)
181{
182 INIT_LIST_HEAD(&fl->fl_link);
183 INIT_LIST_HEAD(&fl->fl_block);
184 init_waitqueue_head(&fl->fl_wait);
185 fl->fl_next = NULL;
186 fl->fl_fasync = NULL;
187 fl->fl_owner = NULL;
188 fl->fl_pid = 0;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000189 fl->fl_nspid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 fl->fl_file = NULL;
191 fl->fl_flags = 0;
192 fl->fl_type = 0;
193 fl->fl_start = fl->fl_end = 0;
194 fl->fl_ops = NULL;
195 fl->fl_lmops = NULL;
196}
197
198EXPORT_SYMBOL(locks_init_lock);
199
200/*
201 * Initialises the fields of the file lock which are invariant for
202 * free file_locks.
203 */
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700204static void init_once(struct kmem_cache *cache, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205{
206 struct file_lock *lock = (struct file_lock *) foo;
207
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 locks_init_lock(lock);
209}
210
Trond Myklebust47831f32006-03-20 13:44:05 -0500211static void locks_copy_private(struct file_lock *new, struct file_lock *fl)
212{
213 if (fl->fl_ops) {
214 if (fl->fl_ops->fl_copy_lock)
215 fl->fl_ops->fl_copy_lock(new, fl);
216 new->fl_ops = fl->fl_ops;
217 }
218 if (fl->fl_lmops) {
219 if (fl->fl_lmops->fl_copy_lock)
220 fl->fl_lmops->fl_copy_lock(new, fl);
221 new->fl_lmops = fl->fl_lmops;
222 }
223}
224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225/*
226 * Initialize a new lock from an existing file_lock structure.
227 */
J. Bruce Fields1a747ee2008-04-24 10:08:22 -0400228void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229{
230 new->fl_owner = fl->fl_owner;
231 new->fl_pid = fl->fl_pid;
Trond Myklebust09969052006-03-20 13:44:38 -0500232 new->fl_file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 new->fl_flags = fl->fl_flags;
234 new->fl_type = fl->fl_type;
235 new->fl_start = fl->fl_start;
236 new->fl_end = fl->fl_end;
Trond Myklebust09969052006-03-20 13:44:38 -0500237 new->fl_ops = NULL;
238 new->fl_lmops = NULL;
239}
Roland Dreier3dd7b712008-04-25 15:32:51 -0700240EXPORT_SYMBOL(__locks_copy_lock);
Trond Myklebust09969052006-03-20 13:44:38 -0500241
242void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
243{
244 locks_release_private(new);
245
246 __locks_copy_lock(new, fl);
247 new->fl_file = fl->fl_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 new->fl_ops = fl->fl_ops;
249 new->fl_lmops = fl->fl_lmops;
Trond Myklebust47831f32006-03-20 13:44:05 -0500250
251 locks_copy_private(new, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252}
253
254EXPORT_SYMBOL(locks_copy_lock);
255
256static inline int flock_translate_cmd(int cmd) {
257 if (cmd & LOCK_MAND)
258 return cmd & (LOCK_MAND | LOCK_RW);
259 switch (cmd) {
260 case LOCK_SH:
261 return F_RDLCK;
262 case LOCK_EX:
263 return F_WRLCK;
264 case LOCK_UN:
265 return F_UNLCK;
266 }
267 return -EINVAL;
268}
269
270/* Fill in a file_lock structure with an appropriate FLOCK lock. */
271static int flock_make_lock(struct file *filp, struct file_lock **lock,
272 unsigned int cmd)
273{
274 struct file_lock *fl;
275 int type = flock_translate_cmd(cmd);
276 if (type < 0)
277 return type;
278
279 fl = locks_alloc_lock();
280 if (fl == NULL)
281 return -ENOMEM;
282
283 fl->fl_file = filp;
284 fl->fl_pid = current->tgid;
285 fl->fl_flags = FL_FLOCK;
286 fl->fl_type = type;
287 fl->fl_end = OFFSET_MAX;
288
289 *lock = fl;
290 return 0;
291}
292
293static int assign_type(struct file_lock *fl, int type)
294{
295 switch (type) {
296 case F_RDLCK:
297 case F_WRLCK:
298 case F_UNLCK:
299 fl->fl_type = type;
300 break;
301 default:
302 return -EINVAL;
303 }
304 return 0;
305}
306
307/* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX
308 * style lock.
309 */
310static int flock_to_posix_lock(struct file *filp, struct file_lock *fl,
311 struct flock *l)
312{
313 off_t start, end;
314
315 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700316 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 start = 0;
318 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700319 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 start = filp->f_pos;
321 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700322 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800323 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 break;
325 default:
326 return -EINVAL;
327 }
328
329 /* POSIX-1996 leaves the case l->l_len < 0 undefined;
330 POSIX-2001 defines it. */
331 start += l->l_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 if (start < 0)
333 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700334 fl->fl_end = OFFSET_MAX;
335 if (l->l_len > 0) {
336 end = start + l->l_len - 1;
337 fl->fl_end = end;
338 } else if (l->l_len < 0) {
339 end = start - 1;
340 fl->fl_end = end;
341 start += l->l_len;
342 if (start < 0)
343 return -EINVAL;
344 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700346 if (fl->fl_end < fl->fl_start)
347 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 fl->fl_owner = current->files;
350 fl->fl_pid = current->tgid;
351 fl->fl_file = filp;
352 fl->fl_flags = FL_POSIX;
353 fl->fl_ops = NULL;
354 fl->fl_lmops = NULL;
355
356 return assign_type(fl, l->l_type);
357}
358
359#if BITS_PER_LONG == 32
360static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl,
361 struct flock64 *l)
362{
363 loff_t start;
364
365 switch (l->l_whence) {
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700366 case SEEK_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 start = 0;
368 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700369 case SEEK_CUR:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 start = filp->f_pos;
371 break;
Josef 'Jeff' Sipekf5579f82006-09-30 23:27:35 -0700372 case SEEK_END:
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800373 start = i_size_read(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 break;
375 default:
376 return -EINVAL;
377 }
378
Trond Myklebust4c780a42005-10-18 14:20:21 -0700379 start += l->l_start;
380 if (start < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 return -EINVAL;
Trond Myklebust4c780a42005-10-18 14:20:21 -0700382 fl->fl_end = OFFSET_MAX;
383 if (l->l_len > 0) {
384 fl->fl_end = start + l->l_len - 1;
385 } else if (l->l_len < 0) {
386 fl->fl_end = start - 1;
387 start += l->l_len;
388 if (start < 0)
389 return -EINVAL;
390 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 fl->fl_start = start; /* we record the absolute position */
Trond Myklebust4c780a42005-10-18 14:20:21 -0700392 if (fl->fl_end < fl->fl_start)
393 return -EOVERFLOW;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
395 fl->fl_owner = current->files;
396 fl->fl_pid = current->tgid;
397 fl->fl_file = filp;
398 fl->fl_flags = FL_POSIX;
399 fl->fl_ops = NULL;
400 fl->fl_lmops = NULL;
401
402 switch (l->l_type) {
403 case F_RDLCK:
404 case F_WRLCK:
405 case F_UNLCK:
406 fl->fl_type = l->l_type;
407 break;
408 default:
409 return -EINVAL;
410 }
411
412 return (0);
413}
414#endif
415
416/* default lease lock manager operations */
417static void lease_break_callback(struct file_lock *fl)
418{
419 kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG);
420}
421
422static void lease_release_private_callback(struct file_lock *fl)
423{
424 if (!fl->fl_file)
425 return;
426
427 f_delown(fl->fl_file);
428 fl->fl_file->f_owner.signum = 0;
429}
430
Adrian Bunk75c96f82005-05-05 16:16:09 -0700431static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
433 return fl->fl_file == try->fl_file;
434}
435
Adrian Bunk75c96f82005-05-05 16:16:09 -0700436static struct lock_manager_operations lease_manager_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 .fl_break = lease_break_callback,
438 .fl_release_private = lease_release_private_callback,
439 .fl_mylease = lease_mylease_callback,
440 .fl_change = lease_modify,
441};
442
443/*
444 * Initialize a lease, use the default lock manager operations
445 */
446static int lease_init(struct file *filp, int type, struct file_lock *fl)
447 {
Trond Myklebust75dff552006-05-07 23:02:42 -0400448 if (assign_type(fl, type) != 0)
449 return -EINVAL;
450
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 fl->fl_owner = current->files;
452 fl->fl_pid = current->tgid;
453
454 fl->fl_file = filp;
455 fl->fl_flags = FL_LEASE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 fl->fl_start = 0;
457 fl->fl_end = OFFSET_MAX;
458 fl->fl_ops = NULL;
459 fl->fl_lmops = &lease_manager_ops;
460 return 0;
461}
462
463/* Allocate a file_lock initialised to this type of lease */
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500464static struct file_lock *lease_alloc(struct file *filp, int type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465{
466 struct file_lock *fl = locks_alloc_lock();
Trond Myklebust75dff552006-05-07 23:02:42 -0400467 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
469 if (fl == NULL)
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500470 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 error = lease_init(filp, type, fl);
Trond Myklebust75dff552006-05-07 23:02:42 -0400473 if (error) {
474 locks_free_lock(fl);
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500475 return ERR_PTR(error);
Trond Myklebust75dff552006-05-07 23:02:42 -0400476 }
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -0500477 return fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478}
479
480/* Check if two locks overlap each other.
481 */
482static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2)
483{
484 return ((fl1->fl_end >= fl2->fl_start) &&
485 (fl2->fl_end >= fl1->fl_start));
486}
487
488/*
489 * Check whether two locks have the same owner.
490 */
Matt Mackall33443c42006-01-08 01:05:22 -0800491static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner)
494 return fl2->fl_lmops == fl1->fl_lmops &&
495 fl1->fl_lmops->fl_compare_owner(fl1, fl2);
496 return fl1->fl_owner == fl2->fl_owner;
497}
498
499/* Remove waiter from blocker's block list.
500 * When blocker ends up pointing to itself then the list is empty.
501 */
Matt Mackall33443c42006-01-08 01:05:22 -0800502static void __locks_delete_block(struct file_lock *waiter)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503{
504 list_del_init(&waiter->fl_block);
505 list_del_init(&waiter->fl_link);
506 waiter->fl_next = NULL;
507}
508
509/*
510 */
511static void locks_delete_block(struct file_lock *waiter)
512{
513 lock_kernel();
514 __locks_delete_block(waiter);
515 unlock_kernel();
516}
517
518/* Insert waiter into blocker's block list.
519 * We use a circular list so that processes can be easily woken up in
520 * the order they blocked. The documentation doesn't require this but
521 * it seems like the reasonable thing to do.
522 */
523static void locks_insert_block(struct file_lock *blocker,
524 struct file_lock *waiter)
525{
J. Bruce Fields6dc0fe82006-03-26 01:37:24 -0800526 BUG_ON(!list_empty(&waiter->fl_block));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 list_add_tail(&waiter->fl_block, &blocker->fl_block);
528 waiter->fl_next = blocker;
529 if (IS_POSIX(blocker))
530 list_add(&waiter->fl_link, &blocked_list);
531}
532
533/* Wake up processes blocked waiting for blocker.
534 * If told to wait then schedule the processes until the block list
535 * is empty, otherwise empty the block list ourselves.
536 */
537static void locks_wake_up_blocks(struct file_lock *blocker)
538{
539 while (!list_empty(&blocker->fl_block)) {
Pavel Emelyanovf0c1cd02007-09-19 16:44:07 +0400540 struct file_lock *waiter;
541
542 waiter = list_first_entry(&blocker->fl_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct file_lock, fl_block);
544 __locks_delete_block(waiter);
545 if (waiter->fl_lmops && waiter->fl_lmops->fl_notify)
546 waiter->fl_lmops->fl_notify(waiter);
547 else
548 wake_up(&waiter->fl_wait);
549 }
550}
551
552/* Insert file lock fl into an inode's lock list at the position indicated
553 * by pos. At the same time add the lock to the global file lock list.
554 */
555static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl)
556{
557 list_add(&fl->fl_link, &file_lock_list);
558
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000559 fl->fl_nspid = get_pid(task_tgid(current));
560
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 /* insert into file's list */
562 fl->fl_next = *pos;
563 *pos = fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564}
565
566/*
567 * Delete a lock and then free it.
568 * Wake up processes that are blocked waiting for this lock,
569 * notify the FS that the lock has been cleared and
570 * finally free the lock.
571 */
572static void locks_delete_lock(struct file_lock **thisfl_p)
573{
574 struct file_lock *fl = *thisfl_p;
575
576 *thisfl_p = fl->fl_next;
577 fl->fl_next = NULL;
578 list_del_init(&fl->fl_link);
579
580 fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync);
581 if (fl->fl_fasync != NULL) {
582 printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync);
583 fl->fl_fasync = NULL;
584 }
585
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000586 if (fl->fl_nspid) {
587 put_pid(fl->fl_nspid);
588 fl->fl_nspid = NULL;
589 }
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 locks_wake_up_blocks(fl);
592 locks_free_lock(fl);
593}
594
595/* Determine if lock sys_fl blocks lock caller_fl. Common functionality
596 * checks for shared/exclusive status of overlapping locks.
597 */
598static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
599{
600 if (sys_fl->fl_type == F_WRLCK)
601 return 1;
602 if (caller_fl->fl_type == F_WRLCK)
603 return 1;
604 return 0;
605}
606
607/* Determine if lock sys_fl blocks lock caller_fl. POSIX specific
608 * checking before calling the locks_conflict().
609 */
610static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
611{
612 /* POSIX locks owned by the same process do not conflict with
613 * each other.
614 */
615 if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl))
616 return (0);
617
618 /* Check whether they overlap */
619 if (!locks_overlap(caller_fl, sys_fl))
620 return 0;
621
622 return (locks_conflict(caller_fl, sys_fl));
623}
624
625/* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific
626 * checking before calling the locks_conflict().
627 */
628static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl)
629{
630 /* FLOCK locks referring to the same filp do not conflict with
631 * each other.
632 */
633 if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file))
634 return (0);
635 if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND))
636 return 0;
637
638 return (locks_conflict(caller_fl, sys_fl));
639}
640
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400641void
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500642posix_test_lock(struct file *filp, struct file_lock *fl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643{
644 struct file_lock *cfl;
645
646 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800647 for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 if (!IS_POSIX(cfl))
649 continue;
J. Bruce Fieldsb842e242007-05-10 19:02:07 -0400650 if (posix_locks_conflict(fl, cfl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 break;
652 }
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000653 if (cfl) {
Marc Eshel9d6a8c52007-02-21 00:55:18 -0500654 __locks_copy_lock(fl, cfl);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000655 if (cfl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800656 fl->fl_pid = pid_vnr(cfl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +0000657 } else
J. Bruce Fields129a84d2007-05-10 18:38:43 -0400658 fl->fl_type = F_UNLCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 unlock_kernel();
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400660 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662EXPORT_SYMBOL(posix_test_lock);
663
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400664/*
665 * Deadlock detection:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400667 * We attempt to detect deadlocks that are due purely to posix file
668 * locks.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400670 * We assume that a task can be waiting for at most one lock at a time.
671 * So for any acquired lock, the process holding that lock may be
672 * waiting on at most one other lock. That lock in turns may be held by
673 * someone waiting for at most one other lock. Given a requested lock
674 * caller_fl which is about to wait for a conflicting lock block_fl, we
675 * follow this chain of waiters to ensure we are not about to create a
676 * cycle.
J. Bruce Fields97855b42007-10-30 11:20:02 -0400677 *
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400678 * Since we do this before we ever put a process to sleep on a lock, we
679 * are ensured that there is never a cycle; that is what guarantees that
680 * the while() loop in posix_locks_deadlock() eventually completes.
681 *
682 * Note: the above assumption may not be true when handling lock
683 * requests from a broken NFS client. It may also fail in the presence
684 * of tasks (such as posix threads) sharing the same open file table.
685 *
686 * To handle those cases, we just bail out after a few iterations.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 */
J. Bruce Fields97855b42007-10-30 11:20:02 -0400688
689#define MAX_DEADLK_ITERATIONS 10
690
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400691/* Find a lock that the owner of the given block_fl is blocking on. */
692static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl)
693{
694 struct file_lock *fl;
695
696 list_for_each_entry(fl, &blocked_list, fl_link) {
697 if (posix_same_owner(fl, block_fl))
698 return fl->fl_next;
699 }
700 return NULL;
701}
702
Adrian Bunkb0904e12006-06-23 02:05:13 -0700703static int posix_locks_deadlock(struct file_lock *caller_fl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 struct file_lock *block_fl)
705{
J. Bruce Fields97855b42007-10-30 11:20:02 -0400706 int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
J. Bruce Fieldsb5331842007-10-26 18:05:40 -0400708 while ((block_fl = what_owner_is_waiting_for(block_fl))) {
709 if (i++ > MAX_DEADLK_ITERATIONS)
710 return 0;
711 if (posix_same_owner(caller_fl, block_fl))
712 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 }
714 return 0;
715}
716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
J. Bruce Fields02888f42007-09-12 15:45:07 -0400718 * after any leases, but before any posix locks.
Trond Myklebustf475ae92006-06-29 16:38:32 -0400719 *
720 * Note that if called with an FL_EXISTS argument, the caller may determine
721 * whether or not a lock was successfully freed by testing the return
722 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 */
Trond Myklebust993dfa82006-03-31 02:30:55 -0800724static int flock_lock_file(struct file *filp, struct file_lock *request)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725{
Trond Myklebust993dfa82006-03-31 02:30:55 -0800726 struct file_lock *new_fl = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 struct file_lock **before;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800728 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 int error = 0;
730 int found = 0;
731
732 lock_kernel();
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400733 if (request->fl_flags & FL_ACCESS)
734 goto find_conflict;
Pavel Emelyanov84d535a2007-09-11 16:38:13 +0400735
736 if (request->fl_type != F_UNLCK) {
737 error = -ENOMEM;
738 new_fl = locks_alloc_lock();
739 if (new_fl == NULL)
740 goto out;
741 error = 0;
742 }
743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 for_each_lock(inode, before) {
745 struct file_lock *fl = *before;
746 if (IS_POSIX(fl))
747 break;
748 if (IS_LEASE(fl))
749 continue;
750 if (filp != fl->fl_file)
751 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800752 if (request->fl_type == fl->fl_type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 goto out;
754 found = 1;
755 locks_delete_lock(before);
756 break;
757 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Trond Myklebustf475ae92006-06-29 16:38:32 -0400759 if (request->fl_type == F_UNLCK) {
760 if ((request->fl_flags & FL_EXISTS) && !found)
761 error = -ENOENT;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800762 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400763 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
765 /*
766 * If a higher-priority process was blocked on the old file lock,
767 * give it the opportunity to lock the file.
768 */
769 if (found)
Linus Torvaldsc3921ab2008-05-11 16:04:48 -0700770 cond_resched_bkl();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400772find_conflict:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 for_each_lock(inode, before) {
774 struct file_lock *fl = *before;
775 if (IS_POSIX(fl))
776 break;
777 if (IS_LEASE(fl))
778 continue;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800779 if (!flock_locks_conflict(request, fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 continue;
781 error = -EAGAIN;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700782 if (!(request->fl_flags & FL_SLEEP))
783 goto out;
784 error = FILE_LOCK_DEFERRED;
785 locks_insert_block(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 goto out;
787 }
Trond Myklebustf07f18d2006-06-29 16:38:37 -0400788 if (request->fl_flags & FL_ACCESS)
789 goto out;
Trond Myklebust993dfa82006-03-31 02:30:55 -0800790 locks_copy_lock(new_fl, request);
Pavel Emelyanov0e2f6db2007-09-11 15:24:01 -0700791 locks_insert_lock(before, new_fl);
Trond Myklebust993dfa82006-03-31 02:30:55 -0800792 new_fl = NULL;
Kirill Korotaev9cedc192006-06-14 17:59:35 +0400793 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
795out:
796 unlock_kernel();
Trond Myklebust993dfa82006-03-31 02:30:55 -0800797 if (new_fl)
798 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 return error;
800}
801
Marc Eshel150b3932007-01-18 16:15:35 -0500802static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
804 struct file_lock *fl;
Miklos Szeredi39005d02006-06-23 02:05:10 -0700805 struct file_lock *new_fl = NULL;
806 struct file_lock *new_fl2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 struct file_lock *left = NULL;
808 struct file_lock *right = NULL;
809 struct file_lock **before;
810 int error, added = 0;
811
812 /*
813 * We may need two file_lock structures for this operation,
814 * so we get them in advance to avoid races.
Miklos Szeredi39005d02006-06-23 02:05:10 -0700815 *
816 * In some cases we can be sure, that no new locks will be needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 */
Miklos Szeredi39005d02006-06-23 02:05:10 -0700818 if (!(request->fl_flags & FL_ACCESS) &&
819 (request->fl_type != F_UNLCK ||
820 request->fl_start != 0 || request->fl_end != OFFSET_MAX)) {
821 new_fl = locks_alloc_lock();
822 new_fl2 = locks_alloc_lock();
823 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824
825 lock_kernel();
826 if (request->fl_type != F_UNLCK) {
827 for_each_lock(inode, before) {
J. Bruce Fields526985b2006-11-14 16:54:36 -0500828 fl = *before;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 if (!IS_POSIX(fl))
830 continue;
831 if (!posix_locks_conflict(request, fl))
832 continue;
Andy Adamson5842add2006-03-26 01:37:26 -0800833 if (conflock)
J. Bruce Fields1a747ee2008-04-24 10:08:22 -0400834 __locks_copy_lock(conflock, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 error = -EAGAIN;
836 if (!(request->fl_flags & FL_SLEEP))
837 goto out;
838 error = -EDEADLK;
839 if (posix_locks_deadlock(request, fl))
840 goto out;
Miklos Szeredibde74e42008-07-25 01:48:57 -0700841 error = FILE_LOCK_DEFERRED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 locks_insert_block(fl, request);
843 goto out;
844 }
845 }
846
847 /* If we're just looking for a conflict, we're done. */
848 error = 0;
849 if (request->fl_flags & FL_ACCESS)
850 goto out;
851
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 * Find the first old lock with the same owner as the new lock.
854 */
855
856 before = &inode->i_flock;
857
858 /* First skip locks owned by other processes. */
859 while ((fl = *before) && (!IS_POSIX(fl) ||
860 !posix_same_owner(request, fl))) {
861 before = &fl->fl_next;
862 }
863
864 /* Process locks with this owner. */
865 while ((fl = *before) && posix_same_owner(request, fl)) {
866 /* Detect adjacent or overlapping regions (if same lock type)
867 */
868 if (request->fl_type == fl->fl_type) {
Olaf Kirch449231d2005-08-25 16:25:35 -0700869 /* In all comparisons of start vs end, use
870 * "start - 1" rather than "end + 1". If end
871 * is OFFSET_MAX, end + 1 will become negative.
872 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 if (fl->fl_end < request->fl_start - 1)
874 goto next_lock;
875 /* If the next lock in the list has entirely bigger
876 * addresses than the new one, insert the lock here.
877 */
Olaf Kirch449231d2005-08-25 16:25:35 -0700878 if (fl->fl_start - 1 > request->fl_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 break;
880
881 /* If we come here, the new and old lock are of the
882 * same type and adjacent or overlapping. Make one
883 * lock yielding from the lower start address of both
884 * locks to the higher end address.
885 */
886 if (fl->fl_start > request->fl_start)
887 fl->fl_start = request->fl_start;
888 else
889 request->fl_start = fl->fl_start;
890 if (fl->fl_end < request->fl_end)
891 fl->fl_end = request->fl_end;
892 else
893 request->fl_end = fl->fl_end;
894 if (added) {
895 locks_delete_lock(before);
896 continue;
897 }
898 request = fl;
899 added = 1;
900 }
901 else {
902 /* Processing for different lock types is a bit
903 * more complex.
904 */
905 if (fl->fl_end < request->fl_start)
906 goto next_lock;
907 if (fl->fl_start > request->fl_end)
908 break;
909 if (request->fl_type == F_UNLCK)
910 added = 1;
911 if (fl->fl_start < request->fl_start)
912 left = fl;
913 /* If the next lock in the list has a higher end
914 * address than the new one, insert the new one here.
915 */
916 if (fl->fl_end > request->fl_end) {
917 right = fl;
918 break;
919 }
920 if (fl->fl_start >= request->fl_start) {
921 /* The new lock completely replaces an old
922 * one (This may happen several times).
923 */
924 if (added) {
925 locks_delete_lock(before);
926 continue;
927 }
928 /* Replace the old lock with the new one.
929 * Wake up anybody waiting for the old one,
930 * as the change in lock type might satisfy
931 * their needs.
932 */
933 locks_wake_up_blocks(fl);
934 fl->fl_start = request->fl_start;
935 fl->fl_end = request->fl_end;
936 fl->fl_type = request->fl_type;
Trond Myklebust47831f32006-03-20 13:44:05 -0500937 locks_release_private(fl);
938 locks_copy_private(fl, request);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 request = fl;
940 added = 1;
941 }
942 }
943 /* Go on to next lock.
944 */
945 next_lock:
946 before = &fl->fl_next;
947 }
948
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700949 /*
950 * The above code only modifies existing locks in case of
951 * merging or replacing. If new lock(s) need to be inserted
952 * all modifications are done bellow this, so it's safe yet to
953 * bail out.
954 */
955 error = -ENOLCK; /* "no luck" */
956 if (right && left == right && !new_fl2)
957 goto out;
958
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 error = 0;
960 if (!added) {
Trond Myklebustf475ae92006-06-29 16:38:32 -0400961 if (request->fl_type == F_UNLCK) {
962 if (request->fl_flags & FL_EXISTS)
963 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 goto out;
Trond Myklebustf475ae92006-06-29 16:38:32 -0400965 }
Miklos Szeredi0d9a4902006-06-23 02:05:09 -0700966
967 if (!new_fl) {
968 error = -ENOLCK;
969 goto out;
970 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 locks_copy_lock(new_fl, request);
972 locks_insert_lock(before, new_fl);
973 new_fl = NULL;
974 }
975 if (right) {
976 if (left == right) {
977 /* The new lock breaks the old one in two pieces,
978 * so we have to use the second new lock.
979 */
980 left = new_fl2;
981 new_fl2 = NULL;
982 locks_copy_lock(left, right);
983 locks_insert_lock(before, left);
984 }
985 right->fl_start = request->fl_end + 1;
986 locks_wake_up_blocks(right);
987 }
988 if (left) {
989 left->fl_end = request->fl_start - 1;
990 locks_wake_up_blocks(left);
991 }
992 out:
993 unlock_kernel();
994 /*
995 * Free any unused locks.
996 */
997 if (new_fl)
998 locks_free_lock(new_fl);
999 if (new_fl2)
1000 locks_free_lock(new_fl2);
1001 return error;
1002}
1003
1004/**
1005 * posix_lock_file - Apply a POSIX-style lock to a file
1006 * @filp: The file to apply the lock to
1007 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001008 * @conflock: Place to return a copy of the conflicting lock, if found.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 *
1010 * Add a POSIX style lock to a file.
1011 * We merge adjacent & overlapping locks whenever possible.
1012 * POSIX locks are sorted by owner task, then by starting address
Trond Myklebustf475ae92006-06-29 16:38:32 -04001013 *
1014 * Note that if called with an FL_EXISTS argument, the caller may determine
1015 * whether or not a lock was successfully freed by testing the return
1016 * value for -ENOENT.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 */
Marc Eshel150b3932007-01-18 16:15:35 -05001018int posix_lock_file(struct file *filp, struct file_lock *fl,
Andy Adamson5842add2006-03-26 01:37:26 -08001019 struct file_lock *conflock)
1020{
Marc Eshel150b3932007-01-18 16:15:35 -05001021 return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock);
Andy Adamson5842add2006-03-26 01:37:26 -08001022}
Marc Eshel150b3932007-01-18 16:15:35 -05001023EXPORT_SYMBOL(posix_lock_file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
1025/**
1026 * posix_lock_file_wait - Apply a POSIX-style lock to a file
1027 * @filp: The file to apply the lock to
1028 * @fl: The lock to be applied
1029 *
1030 * Add a POSIX style lock to a file.
1031 * We merge adjacent & overlapping locks whenever possible.
1032 * POSIX locks are sorted by owner task, then by starting address
1033 */
1034int posix_lock_file_wait(struct file *filp, struct file_lock *fl)
1035{
1036 int error;
1037 might_sleep ();
1038 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001039 error = posix_lock_file(filp, fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001040 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 break;
1042 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1043 if (!error)
1044 continue;
1045
1046 locks_delete_block(fl);
1047 break;
1048 }
1049 return error;
1050}
1051EXPORT_SYMBOL(posix_lock_file_wait);
1052
1053/**
1054 * locks_mandatory_locked - Check for an active lock
1055 * @inode: the file to check
1056 *
1057 * Searches the inode's list of locks to find any POSIX locks which conflict.
1058 * This function is called from locks_verify_locked() only.
1059 */
1060int locks_mandatory_locked(struct inode *inode)
1061{
1062 fl_owner_t owner = current->files;
1063 struct file_lock *fl;
1064
1065 /*
1066 * Search the lock list for this inode for any POSIX locks.
1067 */
1068 lock_kernel();
1069 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
1070 if (!IS_POSIX(fl))
1071 continue;
1072 if (fl->fl_owner != owner)
1073 break;
1074 }
1075 unlock_kernel();
1076 return fl ? -EAGAIN : 0;
1077}
1078
1079/**
1080 * locks_mandatory_area - Check for a conflicting lock
1081 * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ
1082 * for shared
1083 * @inode: the file to check
1084 * @filp: how the file was opened (if it was)
1085 * @offset: start of area to check
1086 * @count: length of area to check
1087 *
1088 * Searches the inode's list of locks to find any POSIX locks which conflict.
1089 * This function is called from rw_verify_area() and
1090 * locks_verify_truncate().
1091 */
1092int locks_mandatory_area(int read_write, struct inode *inode,
1093 struct file *filp, loff_t offset,
1094 size_t count)
1095{
1096 struct file_lock fl;
1097 int error;
1098
1099 locks_init_lock(&fl);
1100 fl.fl_owner = current->files;
1101 fl.fl_pid = current->tgid;
1102 fl.fl_file = filp;
1103 fl.fl_flags = FL_POSIX | FL_ACCESS;
1104 if (filp && !(filp->f_flags & O_NONBLOCK))
1105 fl.fl_flags |= FL_SLEEP;
1106 fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK;
1107 fl.fl_start = offset;
1108 fl.fl_end = offset + count - 1;
1109
1110 for (;;) {
Marc Eshel150b3932007-01-18 16:15:35 -05001111 error = __posix_lock_file(inode, &fl, NULL);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001112 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 break;
1114 error = wait_event_interruptible(fl.fl_wait, !fl.fl_next);
1115 if (!error) {
1116 /*
1117 * If we've been sleeping someone might have
1118 * changed the permissions behind our back.
1119 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001120 if (__mandatory_lock(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 continue;
1122 }
1123
1124 locks_delete_block(&fl);
1125 break;
1126 }
1127
1128 return error;
1129}
1130
1131EXPORT_SYMBOL(locks_mandatory_area);
1132
1133/* We already had a lease on this file; just change its type */
1134int lease_modify(struct file_lock **before, int arg)
1135{
1136 struct file_lock *fl = *before;
1137 int error = assign_type(fl, arg);
1138
1139 if (error)
1140 return error;
1141 locks_wake_up_blocks(fl);
1142 if (arg == F_UNLCK)
1143 locks_delete_lock(before);
1144 return 0;
1145}
1146
1147EXPORT_SYMBOL(lease_modify);
1148
1149static void time_out_leases(struct inode *inode)
1150{
1151 struct file_lock **before;
1152 struct file_lock *fl;
1153
1154 before = &inode->i_flock;
1155 while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) {
1156 if ((fl->fl_break_time == 0)
1157 || time_before(jiffies, fl->fl_break_time)) {
1158 before = &fl->fl_next;
1159 continue;
1160 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 lease_modify(before, fl->fl_type & ~F_INPROGRESS);
1162 if (fl == *before) /* lease_modify may have freed fl */
1163 before = &fl->fl_next;
1164 }
1165}
1166
1167/**
1168 * __break_lease - revoke all outstanding leases on file
1169 * @inode: the inode of the file to return
1170 * @mode: the open mode (read or write)
1171 *
david m. richter87250dd2007-05-09 16:10:27 -04001172 * break_lease (inlined for speed) has checked there already is at least
1173 * some kind of lock (maybe a lease) on this file. Leases are broken on
1174 * a call to open() or truncate(). This function can sleep unless you
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 * specified %O_NONBLOCK to your open().
1176 */
1177int __break_lease(struct inode *inode, unsigned int mode)
1178{
1179 int error = 0, future;
1180 struct file_lock *new_fl, *flock;
1181 struct file_lock *fl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 unsigned long break_time;
1183 int i_have_this_lease = 0;
1184
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001185 new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 lock_kernel();
1188
1189 time_out_leases(inode);
1190
1191 flock = inode->i_flock;
1192 if ((flock == NULL) || !IS_LEASE(flock))
1193 goto out;
1194
1195 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next)
1196 if (fl->fl_owner == current->files)
1197 i_have_this_lease = 1;
1198
1199 if (mode & FMODE_WRITE) {
1200 /* If we want write access, we have to revoke any lease. */
1201 future = F_UNLCK | F_INPROGRESS;
1202 } else if (flock->fl_type & F_INPROGRESS) {
1203 /* If the lease is already being broken, we just leave it */
1204 future = flock->fl_type;
1205 } else if (flock->fl_type & F_WRLCK) {
1206 /* Downgrade the exclusive lease to a read-only lease. */
1207 future = F_RDLCK | F_INPROGRESS;
1208 } else {
1209 /* the existing lease was read-only, so we can read too. */
1210 goto out;
1211 }
1212
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001213 if (IS_ERR(new_fl) && !i_have_this_lease
1214 && ((mode & O_NONBLOCK) == 0)) {
1215 error = PTR_ERR(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 goto out;
1217 }
1218
1219 break_time = 0;
1220 if (lease_break_time > 0) {
1221 break_time = jiffies + lease_break_time * HZ;
1222 if (break_time == 0)
1223 break_time++; /* so that 0 means no break time */
1224 }
1225
1226 for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) {
1227 if (fl->fl_type != future) {
1228 fl->fl_type = future;
1229 fl->fl_break_time = break_time;
1230 /* lease must have lmops break callback */
1231 fl->fl_lmops->fl_break(fl);
1232 }
1233 }
1234
1235 if (i_have_this_lease || (mode & O_NONBLOCK)) {
1236 error = -EWOULDBLOCK;
1237 goto out;
1238 }
1239
1240restart:
1241 break_time = flock->fl_break_time;
1242 if (break_time != 0) {
1243 break_time -= jiffies;
1244 if (break_time == 0)
1245 break_time++;
1246 }
Matthew Wilcox4321e012008-01-14 21:28:30 -07001247 locks_insert_block(flock, new_fl);
1248 error = wait_event_interruptible_timeout(new_fl->fl_wait,
1249 !new_fl->fl_next, break_time);
1250 __locks_delete_block(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 if (error >= 0) {
1252 if (error == 0)
1253 time_out_leases(inode);
1254 /* Wait for the next lease that has not been broken yet */
1255 for (flock = inode->i_flock; flock && IS_LEASE(flock);
1256 flock = flock->fl_next) {
1257 if (flock->fl_type & F_INPROGRESS)
1258 goto restart;
1259 }
1260 error = 0;
1261 }
1262
1263out:
1264 unlock_kernel();
J. Bruce Fieldse32b8ee2007-03-01 14:34:35 -05001265 if (!IS_ERR(new_fl))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 locks_free_lock(new_fl);
1267 return error;
1268}
1269
1270EXPORT_SYMBOL(__break_lease);
1271
1272/**
Randy Dunlapa6b91912008-03-19 17:01:00 -07001273 * lease_get_mtime - get the last modified time of an inode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 * @inode: the inode
1275 * @time: pointer to a timespec which will contain the last modified time
1276 *
1277 * This is to force NFS clients to flush their caches for files with
1278 * exclusive leases. The justification is that if someone has an
Randy Dunlapa6b91912008-03-19 17:01:00 -07001279 * exclusive lease, then they could be modifying it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 */
1281void lease_get_mtime(struct inode *inode, struct timespec *time)
1282{
1283 struct file_lock *flock = inode->i_flock;
1284 if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK))
1285 *time = current_fs_time(inode->i_sb);
1286 else
1287 *time = inode->i_mtime;
1288}
1289
1290EXPORT_SYMBOL(lease_get_mtime);
1291
1292/**
1293 * fcntl_getlease - Enquire what lease is currently active
1294 * @filp: the file
1295 *
1296 * The value returned by this function will be one of
1297 * (if no lease break is pending):
1298 *
1299 * %F_RDLCK to indicate a shared lease is held.
1300 *
1301 * %F_WRLCK to indicate an exclusive lease is held.
1302 *
1303 * %F_UNLCK to indicate no lease is held.
1304 *
1305 * (if a lease break is pending):
1306 *
1307 * %F_RDLCK to indicate an exclusive lease needs to be
1308 * changed to a shared lease (or removed).
1309 *
1310 * %F_UNLCK to indicate the lease needs to be removed.
1311 *
1312 * XXX: sfr & willy disagree over whether F_INPROGRESS
1313 * should be returned to userspace.
1314 */
1315int fcntl_getlease(struct file *filp)
1316{
1317 struct file_lock *fl;
1318 int type = F_UNLCK;
1319
1320 lock_kernel();
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001321 time_out_leases(filp->f_path.dentry->d_inode);
1322 for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 fl = fl->fl_next) {
1324 if (fl->fl_file == filp) {
1325 type = fl->fl_type & ~F_INPROGRESS;
1326 break;
1327 }
1328 }
1329 unlock_kernel();
1330 return type;
1331}
1332
1333/**
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001334 * generic_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 * @filp: file pointer
1336 * @arg: type of lease to obtain
1337 * @flp: input - file_lock to use, output - file_lock inserted
1338 *
1339 * The (input) flp->fl_lmops->fl_break function is required
1340 * by break_lease().
1341 *
1342 * Called with kernel lock held.
1343 */
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001344int generic_setlease(struct file *filp, long arg, struct file_lock **flp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345{
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001346 struct file_lock *fl, **before, **my_before = NULL, *lease;
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001347 struct file_lock *new_fl = NULL;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001348 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 struct inode *inode = dentry->d_inode;
1350 int error, rdlease_count = 0, wrlease_count = 0;
1351
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001352 if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE))
1353 return -EACCES;
1354 if (!S_ISREG(inode->i_mode))
1355 return -EINVAL;
1356 error = security_file_lock(filp, arg);
1357 if (error)
1358 return error;
1359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 time_out_leases(inode);
1361
J. Bruce Fieldsd2ab0b0c2007-06-30 12:40:32 -04001362 BUG_ON(!(*flp)->fl_lmops->fl_break);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
KAMBAROV, ZAUR7eaae282005-07-07 17:57:06 -07001364 lease = *flp;
1365
David M. Richter288b2fd2008-04-23 16:29:00 -04001366 if (arg != F_UNLCK) {
David M. Richter19085552008-04-23 16:29:01 -04001367 error = -ENOMEM;
1368 new_fl = locks_alloc_lock();
1369 if (new_fl == NULL)
1370 goto out;
1371
David M. Richter288b2fd2008-04-23 16:29:00 -04001372 error = -EAGAIN;
1373 if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0))
1374 goto out;
1375 if ((arg == F_WRLCK)
1376 && ((atomic_read(&dentry->d_count) > 1)
1377 || (atomic_read(&inode->i_count) > 1)))
1378 goto out;
David M. Richter288b2fd2008-04-23 16:29:00 -04001379 }
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 /*
1382 * At this point, we know that if there is an exclusive
1383 * lease on this file, then we hold it on this filp
1384 * (otherwise our open of this file would have blocked).
1385 * And if we are trying to acquire an exclusive lease,
1386 * then the file is not open by anyone (including us)
1387 * except for this filp.
1388 */
1389 for (before = &inode->i_flock;
1390 ((fl = *before) != NULL) && IS_LEASE(fl);
1391 before = &fl->fl_next) {
1392 if (lease->fl_lmops->fl_mylease(fl, lease))
1393 my_before = before;
1394 else if (fl->fl_type == (F_INPROGRESS | F_UNLCK))
1395 /*
1396 * Someone is in the process of opening this
1397 * file for writing so we may not take an
1398 * exclusive lease on it.
1399 */
1400 wrlease_count++;
1401 else
1402 rdlease_count++;
1403 }
1404
David M. Richter5fcc60c2008-04-23 16:28:59 -04001405 error = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 if ((arg == F_RDLCK && (wrlease_count > 0)) ||
1407 (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0)))
1408 goto out;
1409
1410 if (my_before != NULL) {
Trond Myklebust75dff552006-05-07 23:02:42 -04001411 *flp = *my_before;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 error = lease->fl_lmops->fl_change(my_before, arg);
1413 goto out;
1414 }
1415
1416 error = 0;
1417 if (arg == F_UNLCK)
1418 goto out;
1419
1420 error = -EINVAL;
1421 if (!leases_enable)
1422 goto out;
1423
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001424 locks_copy_lock(new_fl, lease);
1425 locks_insert_lock(before, new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001427 *flp = new_fl;
1428 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430out:
Pavel Emelyanov85c59582007-09-20 12:45:02 +04001431 if (new_fl != NULL)
1432 locks_free_lock(new_fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 return error;
1434}
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001435EXPORT_SYMBOL(generic_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436
1437 /**
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001438 * vfs_setlease - sets a lease on an open file
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 * @filp: file pointer
1440 * @arg: type of lease to obtain
1441 * @lease: file_lock to use
1442 *
1443 * Call this to establish a lease on the file.
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001444 * The (*lease)->fl_lmops->fl_break operation must be set; if not,
1445 * break_lease will oops!
1446 *
1447 * This will call the filesystem's setlease file method, if
1448 * defined. Note that there is no getlease method; instead, the
1449 * filesystem setlease method should call back to setlease() to
1450 * add a lease to the inode's lease list, where fcntl_getlease() can
1451 * find it. Since fcntl_getlease() only reports whether the current
1452 * task holds a lease, a cluster filesystem need only do this for
1453 * leases held by processes on this node.
1454 *
1455 * There is also no break_lease method; filesystems that
1456 * handle their own leases shoud break leases themselves from the
1457 * filesystem's open, create, and (on truncate) setattr methods.
1458 *
1459 * Warning: the only current setlease methods exist only to disable
1460 * leases in certain cases. More vfs changes may be required to
1461 * allow a full filesystem lease implementation.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 */
1463
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001464int vfs_setlease(struct file *filp, long arg, struct file_lock **lease)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 int error;
1467
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 lock_kernel();
J. Bruce Fieldsf9ffed22006-11-14 15:51:40 -05001469 if (filp->f_op && filp->f_op->setlease)
1470 error = filp->f_op->setlease(filp, arg, lease);
1471 else
Christoph Hellwig0af1a452007-07-31 00:39:22 -07001472 error = generic_setlease(filp, arg, lease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 unlock_kernel();
1474
1475 return error;
1476}
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001477EXPORT_SYMBOL_GPL(vfs_setlease);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
1479/**
1480 * fcntl_setlease - sets a lease on an open file
1481 * @fd: open file descriptor
1482 * @filp: file pointer
1483 * @arg: type of lease to obtain
1484 *
1485 * Call this fcntl to establish a lease on the file.
1486 * Note that you also need to call %F_SETSIG to
1487 * receive a signal when the lease is broken.
1488 */
1489int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1490{
1491 struct file_lock fl, *flp = &fl;
David M. Richter9d91cdc2008-04-23 16:29:02 -04001492 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 int error;
1494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 locks_init_lock(&fl);
1496 error = lease_init(filp, arg, &fl);
1497 if (error)
1498 return error;
1499
1500 lock_kernel();
1501
J. Bruce Fieldsa9933ce2007-06-07 17:09:49 -04001502 error = vfs_setlease(filp, arg, &flp);
J. Bruce Fieldsdc15ae12005-11-10 19:08:00 -05001503 if (error || arg == F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 goto out_unlock;
1505
1506 error = fasync_helper(fd, filp, 1, &flp->fl_fasync);
1507 if (error < 0) {
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001508 /* remove lease just inserted by setlease */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 flp->fl_type = F_UNLCK | F_INPROGRESS;
J. Bruce Fields6d5e8b02007-05-31 17:03:46 -04001510 flp->fl_break_time = jiffies - 10;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 time_out_leases(inode);
1512 goto out_unlock;
1513 }
1514
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001515 error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516out_unlock:
1517 unlock_kernel();
1518 return error;
1519}
1520
1521/**
1522 * flock_lock_file_wait - Apply a FLOCK-style lock to a file
1523 * @filp: The file to apply the lock to
1524 * @fl: The lock to be applied
1525 *
1526 * Add a FLOCK style lock to a file.
1527 */
1528int flock_lock_file_wait(struct file *filp, struct file_lock *fl)
1529{
1530 int error;
1531 might_sleep();
1532 for (;;) {
1533 error = flock_lock_file(filp, fl);
Miklos Szeredibde74e42008-07-25 01:48:57 -07001534 if (error != FILE_LOCK_DEFERRED)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 break;
1536 error = wait_event_interruptible(fl->fl_wait, !fl->fl_next);
1537 if (!error)
1538 continue;
1539
1540 locks_delete_block(fl);
1541 break;
1542 }
1543 return error;
1544}
1545
1546EXPORT_SYMBOL(flock_lock_file_wait);
1547
1548/**
1549 * sys_flock: - flock() system call.
1550 * @fd: the file descriptor to lock.
1551 * @cmd: the type of lock to apply.
1552 *
1553 * Apply a %FL_FLOCK style lock to an open file descriptor.
1554 * The @cmd can be one of
1555 *
1556 * %LOCK_SH -- a shared lock.
1557 *
1558 * %LOCK_EX -- an exclusive lock.
1559 *
1560 * %LOCK_UN -- remove an existing lock.
1561 *
1562 * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes.
1563 *
1564 * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other
1565 * processes read and write access respectively.
1566 */
1567asmlinkage long sys_flock(unsigned int fd, unsigned int cmd)
1568{
1569 struct file *filp;
1570 struct file_lock *lock;
1571 int can_sleep, unlock;
1572 int error;
1573
1574 error = -EBADF;
1575 filp = fget(fd);
1576 if (!filp)
1577 goto out;
1578
1579 can_sleep = !(cmd & LOCK_NB);
1580 cmd &= ~LOCK_NB;
1581 unlock = (cmd == LOCK_UN);
1582
1583 if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3))
1584 goto out_putf;
1585
1586 error = flock_make_lock(filp, &lock, cmd);
1587 if (error)
1588 goto out_putf;
1589 if (can_sleep)
1590 lock->fl_flags |= FL_SLEEP;
1591
1592 error = security_file_lock(filp, cmd);
1593 if (error)
1594 goto out_free;
1595
1596 if (filp->f_op && filp->f_op->flock)
1597 error = filp->f_op->flock(filp,
1598 (can_sleep) ? F_SETLKW : F_SETLK,
1599 lock);
1600 else
1601 error = flock_lock_file_wait(filp, lock);
1602
1603 out_free:
Trond Myklebust993dfa82006-03-31 02:30:55 -08001604 locks_free_lock(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
1606 out_putf:
1607 fput(filp);
1608 out:
1609 return error;
1610}
1611
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001612/**
1613 * vfs_test_lock - test file byte range lock
1614 * @filp: The file to test lock for
J. Bruce Fields6924c552007-05-11 16:22:50 -04001615 * @fl: The lock to test; also used to hold result
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001616 *
1617 * Returns -ERRNO on failure. Indicates presence of conflicting lock by
1618 * setting conf->fl_type to something other than F_UNLCK.
1619 */
1620int vfs_test_lock(struct file *filp, struct file_lock *fl)
1621{
1622 if (filp->f_op && filp->f_op->lock)
1623 return filp->f_op->lock(filp, F_GETLK, fl);
1624 posix_test_lock(filp, fl);
1625 return 0;
1626}
1627EXPORT_SYMBOL_GPL(vfs_test_lock);
1628
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001629static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl)
1630{
1631 flock->l_pid = fl->fl_pid;
1632#if BITS_PER_LONG == 32
1633 /*
1634 * Make sure we can represent the posix lock via
1635 * legacy 32bit flock.
1636 */
1637 if (fl->fl_start > OFFT_OFFSET_MAX)
1638 return -EOVERFLOW;
1639 if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX)
1640 return -EOVERFLOW;
1641#endif
1642 flock->l_start = fl->fl_start;
1643 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1644 fl->fl_end - fl->fl_start + 1;
1645 flock->l_whence = 0;
J. Bruce Fields129a84d2007-05-10 18:38:43 -04001646 flock->l_type = fl->fl_type;
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001647 return 0;
1648}
1649
1650#if BITS_PER_LONG == 32
1651static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl)
1652{
1653 flock->l_pid = fl->fl_pid;
1654 flock->l_start = fl->fl_start;
1655 flock->l_len = fl->fl_end == OFFSET_MAX ? 0 :
1656 fl->fl_end - fl->fl_start + 1;
1657 flock->l_whence = 0;
1658 flock->l_type = fl->fl_type;
1659}
1660#endif
1661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662/* Report the first existing lock that would conflict with l.
1663 * This implements the F_GETLK command of fcntl().
1664 */
1665int fcntl_getlk(struct file *filp, struct flock __user *l)
1666{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001667 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 struct flock flock;
1669 int error;
1670
1671 error = -EFAULT;
1672 if (copy_from_user(&flock, l, sizeof(flock)))
1673 goto out;
1674 error = -EINVAL;
1675 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1676 goto out;
1677
1678 error = flock_to_posix_lock(filp, &file_lock, &flock);
1679 if (error)
1680 goto out;
1681
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001682 error = vfs_test_lock(filp, &file_lock);
1683 if (error)
1684 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001686 flock.l_type = file_lock.fl_type;
1687 if (file_lock.fl_type != F_UNLCK) {
1688 error = posix_lock_to_flock(&flock, &file_lock);
J. Bruce Fieldsc2fa1b82007-02-20 16:10:11 -05001689 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 }
1692 error = -EFAULT;
1693 if (!copy_to_user(l, &flock, sizeof(flock)))
1694 error = 0;
1695out:
1696 return error;
1697}
1698
Marc Eshel7723ec92007-01-18 15:08:55 -05001699/**
1700 * vfs_lock_file - file byte range lock
1701 * @filp: The file to apply the lock to
1702 * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.)
1703 * @fl: The lock to be applied
Marc Eshel150b3932007-01-18 16:15:35 -05001704 * @conf: Place to return a copy of the conflicting lock, if found.
1705 *
1706 * A caller that doesn't care about the conflicting lock may pass NULL
1707 * as the final argument.
1708 *
1709 * If the filesystem defines a private ->lock() method, then @conf will
1710 * be left unchanged; so a caller that cares should initialize it to
1711 * some acceptable default.
Marc Eshel2beb6612006-12-05 23:31:28 -05001712 *
1713 * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX
1714 * locks, the ->lock() interface may return asynchronously, before the lock has
1715 * been granted or denied by the underlying filesystem, if (and only if)
1716 * fl_grant is set. Callers expecting ->lock() to return asynchronously
1717 * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if)
1718 * the request is for a blocking lock. When ->lock() does return asynchronously,
Miklos Szeredibde74e42008-07-25 01:48:57 -07001719 * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock
Marc Eshel2beb6612006-12-05 23:31:28 -05001720 * request completes.
1721 * If the request is for non-blocking lock the file system should return
Miklos Szeredibde74e42008-07-25 01:48:57 -07001722 * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine
1723 * with the result. If the request timed out the callback routine will return a
Marc Eshel2beb6612006-12-05 23:31:28 -05001724 * nonzero return code and the file system should release the lock. The file
1725 * system is also responsible to keep a corresponding posix lock when it
1726 * grants a lock so the VFS can find out which locks are locally held and do
1727 * the correct lock cleanup when required.
1728 * The underlying filesystem must not drop the kernel lock or call
Miklos Szeredibde74e42008-07-25 01:48:57 -07001729 * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED
Marc Eshel2beb6612006-12-05 23:31:28 -05001730 * return code.
Marc Eshel7723ec92007-01-18 15:08:55 -05001731 */
Marc Eshel150b3932007-01-18 16:15:35 -05001732int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf)
Marc Eshel7723ec92007-01-18 15:08:55 -05001733{
1734 if (filp->f_op && filp->f_op->lock)
1735 return filp->f_op->lock(filp, cmd, fl);
1736 else
Marc Eshel150b3932007-01-18 16:15:35 -05001737 return posix_lock_file(filp, fl, conf);
Marc Eshel7723ec92007-01-18 15:08:55 -05001738}
1739EXPORT_SYMBOL_GPL(vfs_lock_file);
1740
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001741static int do_lock_file_wait(struct file *filp, unsigned int cmd,
1742 struct file_lock *fl)
1743{
1744 int error;
1745
1746 error = security_file_lock(filp, fl->fl_type);
1747 if (error)
1748 return error;
1749
1750 if (filp->f_op && filp->f_op->lock != NULL)
1751 error = filp->f_op->lock(filp, cmd, fl);
1752 else {
1753 for (;;) {
1754 error = posix_lock_file(filp, fl, NULL);
1755 if (error != FILE_LOCK_DEFERRED)
1756 break;
1757 error = wait_event_interruptible(fl->fl_wait,
1758 !fl->fl_next);
1759 if (!error)
1760 continue;
1761
1762 locks_delete_block(fl);
1763 break;
1764 }
1765 }
1766
1767 return error;
1768}
1769
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770/* Apply the lock described by l to an open file descriptor.
1771 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1772 */
Peter Staubachc2936212005-07-27 11:45:09 -07001773int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd,
1774 struct flock __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775{
1776 struct file_lock *file_lock = locks_alloc_lock();
1777 struct flock flock;
1778 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001779 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 int error;
1781
1782 if (file_lock == NULL)
1783 return -ENOLCK;
1784
1785 /*
1786 * This might block, so we do it before checking the inode.
1787 */
1788 error = -EFAULT;
1789 if (copy_from_user(&flock, l, sizeof(flock)))
1790 goto out;
1791
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001792 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793
1794 /* Don't allow mandatory locks on files that may be memory mapped
1795 * and shared.
1796 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001797 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 error = -EAGAIN;
1799 goto out;
1800 }
1801
Peter Staubachc2936212005-07-27 11:45:09 -07001802again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 error = flock_to_posix_lock(filp, file_lock, &flock);
1804 if (error)
1805 goto out;
1806 if (cmd == F_SETLKW) {
1807 file_lock->fl_flags |= FL_SLEEP;
1808 }
1809
1810 error = -EBADF;
1811 switch (flock.l_type) {
1812 case F_RDLCK:
1813 if (!(filp->f_mode & FMODE_READ))
1814 goto out;
1815 break;
1816 case F_WRLCK:
1817 if (!(filp->f_mode & FMODE_WRITE))
1818 goto out;
1819 break;
1820 case F_UNLCK:
1821 break;
1822 default:
1823 error = -EINVAL;
1824 goto out;
1825 }
1826
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001827 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828
Peter Staubachc2936212005-07-27 11:45:09 -07001829 /*
1830 * Attempt to detect a close/fcntl race and recover by
1831 * releasing the lock that was just acquired.
1832 */
Al Viro0b2bac22008-05-06 13:58:34 -04001833 /*
1834 * we need that spin_lock here - it prevents reordering between
1835 * update of inode->i_flock and check for it done in close().
1836 * rcu_read_lock() wouldn't do.
1837 */
1838 spin_lock(&current->files->file_lock);
1839 f = fcheck(fd);
1840 spin_unlock(&current->files->file_lock);
1841 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001842 flock.l_type = F_UNLCK;
1843 goto again;
1844 }
1845
1846out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 locks_free_lock(file_lock);
1848 return error;
1849}
1850
1851#if BITS_PER_LONG == 32
1852/* Report the first existing lock that would conflict with l.
1853 * This implements the F_GETLK command of fcntl().
1854 */
1855int fcntl_getlk64(struct file *filp, struct flock64 __user *l)
1856{
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001857 struct file_lock file_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 struct flock64 flock;
1859 int error;
1860
1861 error = -EFAULT;
1862 if (copy_from_user(&flock, l, sizeof(flock)))
1863 goto out;
1864 error = -EINVAL;
1865 if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK))
1866 goto out;
1867
1868 error = flock64_to_posix_lock(filp, &file_lock, &flock);
1869 if (error)
1870 goto out;
1871
J. Bruce Fields3ee17ab2007-02-21 00:58:50 -05001872 error = vfs_test_lock(filp, &file_lock);
1873 if (error)
1874 goto out;
1875
Marc Eshel9d6a8c52007-02-21 00:55:18 -05001876 flock.l_type = file_lock.fl_type;
1877 if (file_lock.fl_type != F_UNLCK)
1878 posix_lock_to_flock64(&flock, &file_lock);
1879
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 error = -EFAULT;
1881 if (!copy_to_user(l, &flock, sizeof(flock)))
1882 error = 0;
1883
1884out:
1885 return error;
1886}
1887
1888/* Apply the lock described by l to an open file descriptor.
1889 * This implements both the F_SETLK and F_SETLKW commands of fcntl().
1890 */
Peter Staubachc2936212005-07-27 11:45:09 -07001891int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd,
1892 struct flock64 __user *l)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893{
1894 struct file_lock *file_lock = locks_alloc_lock();
1895 struct flock64 flock;
1896 struct inode *inode;
Al Viro0b2bac22008-05-06 13:58:34 -04001897 struct file *f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 int error;
1899
1900 if (file_lock == NULL)
1901 return -ENOLCK;
1902
1903 /*
1904 * This might block, so we do it before checking the inode.
1905 */
1906 error = -EFAULT;
1907 if (copy_from_user(&flock, l, sizeof(flock)))
1908 goto out;
1909
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001910 inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
1912 /* Don't allow mandatory locks on files that may be memory mapped
1913 * and shared.
1914 */
Pavel Emelyanova16877c2007-10-01 14:41:11 -07001915 if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 error = -EAGAIN;
1917 goto out;
1918 }
1919
Peter Staubachc2936212005-07-27 11:45:09 -07001920again:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921 error = flock64_to_posix_lock(filp, file_lock, &flock);
1922 if (error)
1923 goto out;
1924 if (cmd == F_SETLKW64) {
1925 file_lock->fl_flags |= FL_SLEEP;
1926 }
1927
1928 error = -EBADF;
1929 switch (flock.l_type) {
1930 case F_RDLCK:
1931 if (!(filp->f_mode & FMODE_READ))
1932 goto out;
1933 break;
1934 case F_WRLCK:
1935 if (!(filp->f_mode & FMODE_WRITE))
1936 goto out;
1937 break;
1938 case F_UNLCK:
1939 break;
1940 default:
1941 error = -EINVAL;
1942 goto out;
1943 }
1944
Miklos Szeredib648a6d2008-07-25 01:48:57 -07001945 error = do_lock_file_wait(filp, cmd, file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946
Peter Staubachc2936212005-07-27 11:45:09 -07001947 /*
1948 * Attempt to detect a close/fcntl race and recover by
1949 * releasing the lock that was just acquired.
1950 */
Al Viro0b2bac22008-05-06 13:58:34 -04001951 spin_lock(&current->files->file_lock);
1952 f = fcheck(fd);
1953 spin_unlock(&current->files->file_lock);
1954 if (!error && f != filp && flock.l_type != F_UNLCK) {
Peter Staubachc2936212005-07-27 11:45:09 -07001955 flock.l_type = F_UNLCK;
1956 goto again;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 }
1958
1959out:
1960 locks_free_lock(file_lock);
1961 return error;
1962}
1963#endif /* BITS_PER_LONG == 32 */
1964
1965/*
1966 * This function is called when the file is being removed
1967 * from the task's fd array. POSIX locks belonging to this task
1968 * are deleted at this time.
1969 */
1970void locks_remove_posix(struct file *filp, fl_owner_t owner)
1971{
Miklos Szerediff7b86b2006-06-23 02:05:11 -07001972 struct file_lock lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973
1974 /*
1975 * If there are no locks held on this file, we don't need to call
1976 * posix_lock_file(). Another process could be setting a lock on this
1977 * file at the same time, but we wouldn't remove that lock anyway.
1978 */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001979 if (!filp->f_path.dentry->d_inode->i_flock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 return;
1981
1982 lock.fl_type = F_UNLCK;
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -07001983 lock.fl_flags = FL_POSIX | FL_CLOSE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 lock.fl_start = 0;
1985 lock.fl_end = OFFSET_MAX;
1986 lock.fl_owner = owner;
1987 lock.fl_pid = current->tgid;
1988 lock.fl_file = filp;
1989 lock.fl_ops = NULL;
1990 lock.fl_lmops = NULL;
1991
Marc Eshel150b3932007-01-18 16:15:35 -05001992 vfs_lock_file(filp, F_SETLK, &lock, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 if (lock.fl_ops && lock.fl_ops->fl_release_private)
1995 lock.fl_ops->fl_release_private(&lock);
1996}
1997
1998EXPORT_SYMBOL(locks_remove_posix);
1999
2000/*
2001 * This function is called on the last close of an open file.
2002 */
2003void locks_remove_flock(struct file *filp)
2004{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002005 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 struct file_lock *fl;
2007 struct file_lock **before;
2008
2009 if (!inode->i_flock)
2010 return;
2011
2012 if (filp->f_op && filp->f_op->flock) {
2013 struct file_lock fl = {
2014 .fl_pid = current->tgid,
2015 .fl_file = filp,
2016 .fl_flags = FL_FLOCK,
2017 .fl_type = F_UNLCK,
2018 .fl_end = OFFSET_MAX,
2019 };
2020 filp->f_op->flock(filp, F_SETLKW, &fl);
Trond Myklebust80fec4c2005-06-22 17:16:31 +00002021 if (fl.fl_ops && fl.fl_ops->fl_release_private)
2022 fl.fl_ops->fl_release_private(&fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 }
2024
2025 lock_kernel();
2026 before = &inode->i_flock;
2027
2028 while ((fl = *before) != NULL) {
2029 if (fl->fl_file == filp) {
Peter Staubachc2936212005-07-27 11:45:09 -07002030 if (IS_FLOCK(fl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 locks_delete_lock(before);
2032 continue;
2033 }
2034 if (IS_LEASE(fl)) {
2035 lease_modify(before, F_UNLCK);
2036 continue;
2037 }
2038 /* What? */
2039 BUG();
2040 }
2041 before = &fl->fl_next;
2042 }
2043 unlock_kernel();
2044}
2045
2046/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 * posix_unblock_lock - stop waiting for a file lock
2048 * @filp: how the file was opened
2049 * @waiter: the lock which was waiting
2050 *
2051 * lockd needs to block waiting for locks.
2052 */
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002053int
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054posix_unblock_lock(struct file *filp, struct file_lock *waiter)
2055{
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002056 int status = 0;
2057
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 lock_kernel();
J. Bruce Fields5996a292006-01-03 09:55:44 +01002059 if (waiter->fl_next)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060 __locks_delete_block(waiter);
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002061 else
2062 status = -ENOENT;
J. Bruce Fields5996a292006-01-03 09:55:44 +01002063 unlock_kernel();
J. Bruce Fields64a318e2006-01-03 09:55:46 +01002064 return status;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065}
2066
2067EXPORT_SYMBOL(posix_unblock_lock);
2068
Marc Eshel9b9d2ab2007-01-18 17:52:58 -05002069/**
2070 * vfs_cancel_lock - file byte range unblock lock
2071 * @filp: The file to apply the unblock to
2072 * @fl: The lock to be unblocked
2073 *
2074 * Used by lock managers to cancel blocked requests
2075 */
2076int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
2077{
2078 if (filp->f_op && filp->f_op->lock)
2079 return filp->f_op->lock(filp, F_CANCELLK, fl);
2080 return 0;
2081}
2082
2083EXPORT_SYMBOL_GPL(vfs_cancel_lock);
2084
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002085#ifdef CONFIG_PROC_FS
2086#include <linux/seq_file.h>
2087
2088static void lock_get_status(struct seq_file *f, struct file_lock *fl,
2089 int id, char *pfx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
2091 struct inode *inode = NULL;
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002092 unsigned int fl_pid;
2093
2094 if (fl->fl_nspid)
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08002095 fl_pid = pid_vnr(fl->fl_nspid);
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002096 else
2097 fl_pid = fl->fl_pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098
2099 if (fl->fl_file != NULL)
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08002100 inode = fl->fl_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002102 seq_printf(f, "%d:%s ", id, pfx);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 if (IS_POSIX(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002104 seq_printf(f, "%6s %s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ",
2106 (inode == NULL) ? "*NOINODE*" :
Pavel Emelyanova16877c2007-10-01 14:41:11 -07002107 mandatory_lock(inode) ? "MANDATORY" : "ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 } else if (IS_FLOCK(fl)) {
2109 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002110 seq_printf(f, "FLOCK MSNFS ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002112 seq_printf(f, "FLOCK ADVISORY ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 }
2114 } else if (IS_LEASE(fl)) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002115 seq_printf(f, "LEASE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 if (fl->fl_type & F_INPROGRESS)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002117 seq_printf(f, "BREAKING ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 else if (fl->fl_file)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002119 seq_printf(f, "ACTIVE ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002121 seq_printf(f, "BREAKER ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002123 seq_printf(f, "UNKNOWN UNKNOWN ");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124 }
2125 if (fl->fl_type & LOCK_MAND) {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002126 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 (fl->fl_type & LOCK_READ)
2128 ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ "
2129 : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE ");
2130 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002131 seq_printf(f, "%s ",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 (fl->fl_type & F_INPROGRESS)
2133 ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ "
2134 : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ ");
2135 }
2136 if (inode) {
2137#ifdef WE_CAN_BREAK_LSLK_NOW
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002138 seq_printf(f, "%d %s:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 inode->i_sb->s_id, inode->i_ino);
2140#else
2141 /* userspace relies on this representation of dev_t ;-( */
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002142 seq_printf(f, "%d %02x:%02x:%ld ", fl_pid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 MAJOR(inode->i_sb->s_dev),
2144 MINOR(inode->i_sb->s_dev), inode->i_ino);
2145#endif
2146 } else {
Vitaliy Gusevab1f1612008-01-17 00:07:08 +00002147 seq_printf(f, "%d <none>:0 ", fl_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148 }
2149 if (IS_POSIX(fl)) {
2150 if (fl->fl_end == OFFSET_MAX)
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002151 seq_printf(f, "%Ld EOF\n", fl->fl_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 else
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002153 seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 } else {
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002155 seq_printf(f, "0 EOF\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 }
2157}
2158
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002159static int locks_show(struct seq_file *f, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160{
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002161 struct file_lock *fl, *bfl;
2162
2163 fl = list_entry(v, struct file_lock, fl_link);
2164
2165 lock_get_status(f, fl, (long)f->private, "");
2166
2167 list_for_each_entry(bfl, &fl->fl_block, fl_block)
2168 lock_get_status(f, bfl, (long)f->private, " ->");
2169
2170 f->private++;
2171 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002172}
2173
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002174static void *locks_start(struct seq_file *f, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 lock_kernel();
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002177 f->private = (void *)1;
2178 return seq_list_start(&file_lock_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179}
2180
Pavel Emelyanov7f8ada92007-10-01 14:41:15 -07002181static void *locks_next(struct seq_file *f, void *v, loff_t *pos)
2182{
2183 return seq_list_next(v, &file_lock_list, pos);
2184}
2185
2186static void locks_stop(struct seq_file *f, void *v)
2187{
2188 unlock_kernel();
2189}
2190
2191struct seq_operations locks_seq_operations = {
2192 .start = locks_start,
2193 .next = locks_next,
2194 .stop = locks_stop,
2195 .show = locks_show,
2196};
2197#endif
2198
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199/**
2200 * lock_may_read - checks that the region is free of locks
2201 * @inode: the inode that is being read
2202 * @start: the first byte to read
2203 * @len: the number of bytes to read
2204 *
2205 * Emulates Windows locking requirements. Whole-file
2206 * mandatory locks (share modes) can prohibit a read and
2207 * byte-range POSIX locks can prohibit a read if they overlap.
2208 *
2209 * N.B. this function is only ever called
2210 * from knfsd and ownership of locks is never checked.
2211 */
2212int lock_may_read(struct inode *inode, loff_t start, unsigned long len)
2213{
2214 struct file_lock *fl;
2215 int result = 1;
2216 lock_kernel();
2217 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2218 if (IS_POSIX(fl)) {
2219 if (fl->fl_type == F_RDLCK)
2220 continue;
2221 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2222 continue;
2223 } else if (IS_FLOCK(fl)) {
2224 if (!(fl->fl_type & LOCK_MAND))
2225 continue;
2226 if (fl->fl_type & LOCK_READ)
2227 continue;
2228 } else
2229 continue;
2230 result = 0;
2231 break;
2232 }
2233 unlock_kernel();
2234 return result;
2235}
2236
2237EXPORT_SYMBOL(lock_may_read);
2238
2239/**
2240 * lock_may_write - checks that the region is free of locks
2241 * @inode: the inode that is being written
2242 * @start: the first byte to write
2243 * @len: the number of bytes to write
2244 *
2245 * Emulates Windows locking requirements. Whole-file
2246 * mandatory locks (share modes) can prohibit a write and
2247 * byte-range POSIX locks can prohibit a write if they overlap.
2248 *
2249 * N.B. this function is only ever called
2250 * from knfsd and ownership of locks is never checked.
2251 */
2252int lock_may_write(struct inode *inode, loff_t start, unsigned long len)
2253{
2254 struct file_lock *fl;
2255 int result = 1;
2256 lock_kernel();
2257 for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
2258 if (IS_POSIX(fl)) {
2259 if ((fl->fl_end < start) || (fl->fl_start > (start + len)))
2260 continue;
2261 } else if (IS_FLOCK(fl)) {
2262 if (!(fl->fl_type & LOCK_MAND))
2263 continue;
2264 if (fl->fl_type & LOCK_WRITE)
2265 continue;
2266 } else
2267 continue;
2268 result = 0;
2269 break;
2270 }
2271 unlock_kernel();
2272 return result;
2273}
2274
2275EXPORT_SYMBOL(lock_may_write);
2276
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277static int __init filelock_init(void)
2278{
2279 filelock_cache = kmem_cache_create("file_lock_cache",
2280 sizeof(struct file_lock), 0, SLAB_PANIC,
Paul Mundt20c2df82007-07-20 10:11:58 +09002281 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282 return 0;
2283}
2284
2285core_initcall(filelock_init);