Root/
Source at commit b386be689295730688885552666ea40b2e639b14 created 11 years 11 months ago. By Maarten ter Huurne, Revert "MIPS: JZ4740: reset: Initialize hibernate wakeup counters." | |
---|---|
1 | /* |
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/filesystems/mandatory-locking.txt' for details. |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference |
69 | * Manual, Section 2. |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. |
71 | * |
72 | * Tidied up block list handling. Added '/proc/locks' interface. |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. |
74 | * |
75 | * Fixed deadlock condition for pathological code that mixes calls to |
76 | * flock() and fcntl(). |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. |
78 | * |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to |
81 | * guarantee sensible behaviour in the case where file system modules might |
82 | * be compiled with different options than the kernel itself. |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
84 | * |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
88 | * |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that |
91 | * have already been freed. |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. |
93 | * |
94 | * Made the block list a circular list to minimise searching in the list. |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. |
96 | * |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory |
98 | * locking. |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. |
100 | * |
101 | * Some adaptations for NFS support. |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, |
103 | * |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. |
106 | * |
107 | * Use slab allocator instead of kmalloc/kfree. |
108 | * Use generic list implementation from <linux/list.h>. |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. |
111 | * |
112 | * Leases and LOCK_MAND |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. |
115 | */ |
116 | |
117 | #include <linux/capability.h> |
118 | #include <linux/file.h> |
119 | #include <linux/fdtable.h> |
120 | #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/syscalls.h> |
126 | #include <linux/time.h> |
127 | #include <linux/rcupdate.h> |
128 | #include <linux/pid_namespace.h> |
129 | |
130 | #include <asm/uaccess.h> |
131 | |
132 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) |
133 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) |
134 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) |
135 | |
136 | static bool lease_breaking(struct file_lock *fl) |
137 | { |
138 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
139 | } |
140 | |
141 | static int target_leasetype(struct file_lock *fl) |
142 | { |
143 | if (fl->fl_flags & FL_UNLOCK_PENDING) |
144 | return F_UNLCK; |
145 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) |
146 | return F_RDLCK; |
147 | return fl->fl_type; |
148 | } |
149 | |
150 | int leases_enable = 1; |
151 | int lease_break_time = 45; |
152 | |
153 | #define for_each_lock(inode, lockp) \ |
154 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) |
155 | |
156 | static LIST_HEAD(file_lock_list); |
157 | static LIST_HEAD(blocked_list); |
158 | static DEFINE_SPINLOCK(file_lock_lock); |
159 | |
160 | /* |
161 | * Protects the two list heads above, plus the inode->i_flock list |
162 | */ |
163 | void lock_flocks(void) |
164 | { |
165 | spin_lock(&file_lock_lock); |
166 | } |
167 | EXPORT_SYMBOL_GPL(lock_flocks); |
168 | |
169 | void unlock_flocks(void) |
170 | { |
171 | spin_unlock(&file_lock_lock); |
172 | } |
173 | EXPORT_SYMBOL_GPL(unlock_flocks); |
174 | |
175 | static struct kmem_cache *filelock_cache __read_mostly; |
176 | |
177 | static void locks_init_lock_heads(struct file_lock *fl) |
178 | { |
179 | INIT_LIST_HEAD(&fl->fl_link); |
180 | INIT_LIST_HEAD(&fl->fl_block); |
181 | init_waitqueue_head(&fl->fl_wait); |
182 | } |
183 | |
184 | /* Allocate an empty lock structure. */ |
185 | struct file_lock *locks_alloc_lock(void) |
186 | { |
187 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
188 | |
189 | if (fl) |
190 | locks_init_lock_heads(fl); |
191 | |
192 | return fl; |
193 | } |
194 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
195 | |
196 | void locks_release_private(struct file_lock *fl) |
197 | { |
198 | if (fl->fl_ops) { |
199 | if (fl->fl_ops->fl_release_private) |
200 | fl->fl_ops->fl_release_private(fl); |
201 | fl->fl_ops = NULL; |
202 | } |
203 | if (fl->fl_lmops) { |
204 | if (fl->fl_lmops->lm_release_private) |
205 | fl->fl_lmops->lm_release_private(fl); |
206 | fl->fl_lmops = NULL; |
207 | } |
208 | |
209 | } |
210 | EXPORT_SYMBOL_GPL(locks_release_private); |
211 | |
212 | /* Free a lock which is not in use. */ |
213 | void locks_free_lock(struct file_lock *fl) |
214 | { |
215 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
216 | BUG_ON(!list_empty(&fl->fl_block)); |
217 | BUG_ON(!list_empty(&fl->fl_link)); |
218 | |
219 | locks_release_private(fl); |
220 | kmem_cache_free(filelock_cache, fl); |
221 | } |
222 | EXPORT_SYMBOL(locks_free_lock); |
223 | |
224 | void locks_init_lock(struct file_lock *fl) |
225 | { |
226 | memset(fl, 0, sizeof(struct file_lock)); |
227 | locks_init_lock_heads(fl); |
228 | } |
229 | |
230 | EXPORT_SYMBOL(locks_init_lock); |
231 | |
232 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
233 | { |
234 | if (fl->fl_ops) { |
235 | if (fl->fl_ops->fl_copy_lock) |
236 | fl->fl_ops->fl_copy_lock(new, fl); |
237 | new->fl_ops = fl->fl_ops; |
238 | } |
239 | if (fl->fl_lmops) |
240 | new->fl_lmops = fl->fl_lmops; |
241 | } |
242 | |
243 | /* |
244 | * Initialize a new lock from an existing file_lock structure. |
245 | */ |
246 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
247 | { |
248 | new->fl_owner = fl->fl_owner; |
249 | new->fl_pid = fl->fl_pid; |
250 | new->fl_file = NULL; |
251 | new->fl_flags = fl->fl_flags; |
252 | new->fl_type = fl->fl_type; |
253 | new->fl_start = fl->fl_start; |
254 | new->fl_end = fl->fl_end; |
255 | new->fl_ops = NULL; |
256 | new->fl_lmops = NULL; |
257 | } |
258 | EXPORT_SYMBOL(__locks_copy_lock); |
259 | |
260 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) |
261 | { |
262 | locks_release_private(new); |
263 | |
264 | __locks_copy_lock(new, fl); |
265 | new->fl_file = fl->fl_file; |
266 | new->fl_ops = fl->fl_ops; |
267 | new->fl_lmops = fl->fl_lmops; |
268 | |
269 | locks_copy_private(new, fl); |
270 | } |
271 | |
272 | EXPORT_SYMBOL(locks_copy_lock); |
273 | |
274 | static inline int flock_translate_cmd(int cmd) { |
275 | if (cmd & LOCK_MAND) |
276 | return cmd & (LOCK_MAND | LOCK_RW); |
277 | switch (cmd) { |
278 | case LOCK_SH: |
279 | return F_RDLCK; |
280 | case LOCK_EX: |
281 | return F_WRLCK; |
282 | case LOCK_UN: |
283 | return F_UNLCK; |
284 | } |
285 | return -EINVAL; |
286 | } |
287 | |
288 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ |
289 | static int flock_make_lock(struct file *filp, struct file_lock **lock, |
290 | unsigned int cmd) |
291 | { |
292 | struct file_lock *fl; |
293 | int type = flock_translate_cmd(cmd); |
294 | if (type < 0) |
295 | return type; |
296 | |
297 | fl = locks_alloc_lock(); |
298 | if (fl == NULL) |
299 | return -ENOMEM; |
300 | |
301 | fl->fl_file = filp; |
302 | fl->fl_pid = current->tgid; |
303 | fl->fl_flags = FL_FLOCK; |
304 | fl->fl_type = type; |
305 | fl->fl_end = OFFSET_MAX; |
306 | |
307 | *lock = fl; |
308 | return 0; |
309 | } |
310 | |
311 | static int assign_type(struct file_lock *fl, int type) |
312 | { |
313 | switch (type) { |
314 | case F_RDLCK: |
315 | case F_WRLCK: |
316 | case F_UNLCK: |
317 | fl->fl_type = type; |
318 | break; |
319 | default: |
320 | return -EINVAL; |
321 | } |
322 | return 0; |
323 | } |
324 | |
325 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
326 | * style lock. |
327 | */ |
328 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, |
329 | struct flock *l) |
330 | { |
331 | off_t start, end; |
332 | |
333 | switch (l->l_whence) { |
334 | case SEEK_SET: |
335 | start = 0; |
336 | break; |
337 | case SEEK_CUR: |
338 | start = filp->f_pos; |
339 | break; |
340 | case SEEK_END: |
341 | start = i_size_read(filp->f_path.dentry->d_inode); |
342 | break; |
343 | default: |
344 | return -EINVAL; |
345 | } |
346 | |
347 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; |
348 | POSIX-2001 defines it. */ |
349 | start += l->l_start; |
350 | if (start < 0) |
351 | return -EINVAL; |
352 | fl->fl_end = OFFSET_MAX; |
353 | if (l->l_len > 0) { |
354 | end = start + l->l_len - 1; |
355 | fl->fl_end = end; |
356 | } else if (l->l_len < 0) { |
357 | end = start - 1; |
358 | fl->fl_end = end; |
359 | start += l->l_len; |
360 | if (start < 0) |
361 | return -EINVAL; |
362 | } |
363 | fl->fl_start = start; /* we record the absolute position */ |
364 | if (fl->fl_end < fl->fl_start) |
365 | return -EOVERFLOW; |
366 | |
367 | fl->fl_owner = current->files; |
368 | fl->fl_pid = current->tgid; |
369 | fl->fl_file = filp; |
370 | fl->fl_flags = FL_POSIX; |
371 | fl->fl_ops = NULL; |
372 | fl->fl_lmops = NULL; |
373 | |
374 | return assign_type(fl, l->l_type); |
375 | } |
376 | |
377 | #if BITS_PER_LONG == 32 |
378 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
379 | struct flock64 *l) |
380 | { |
381 | loff_t start; |
382 | |
383 | switch (l->l_whence) { |
384 | case SEEK_SET: |
385 | start = 0; |
386 | break; |
387 | case SEEK_CUR: |
388 | start = filp->f_pos; |
389 | break; |
390 | case SEEK_END: |
391 | start = i_size_read(filp->f_path.dentry->d_inode); |
392 | break; |
393 | default: |
394 | return -EINVAL; |
395 | } |
396 | |
397 | start += l->l_start; |
398 | if (start < 0) |
399 | return -EINVAL; |
400 | fl->fl_end = OFFSET_MAX; |
401 | if (l->l_len > 0) { |
402 | fl->fl_end = start + l->l_len - 1; |
403 | } else if (l->l_len < 0) { |
404 | fl->fl_end = start - 1; |
405 | start += l->l_len; |
406 | if (start < 0) |
407 | return -EINVAL; |
408 | } |
409 | fl->fl_start = start; /* we record the absolute position */ |
410 | if (fl->fl_end < fl->fl_start) |
411 | return -EOVERFLOW; |
412 | |
413 | fl->fl_owner = current->files; |
414 | fl->fl_pid = current->tgid; |
415 | fl->fl_file = filp; |
416 | fl->fl_flags = FL_POSIX; |
417 | fl->fl_ops = NULL; |
418 | fl->fl_lmops = NULL; |
419 | |
420 | return assign_type(fl, l->l_type); |
421 | } |
422 | #endif |
423 | |
424 | /* default lease lock manager operations */ |
425 | static void lease_break_callback(struct file_lock *fl) |
426 | { |
427 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); |
428 | } |
429 | |
430 | static void lease_release_private_callback(struct file_lock *fl) |
431 | { |
432 | if (!fl->fl_file) |
433 | return; |
434 | |
435 | f_delown(fl->fl_file); |
436 | fl->fl_file->f_owner.signum = 0; |
437 | } |
438 | |
439 | static const struct lock_manager_operations lease_manager_ops = { |
440 | .lm_break = lease_break_callback, |
441 | .lm_release_private = lease_release_private_callback, |
442 | .lm_change = lease_modify, |
443 | }; |
444 | |
445 | /* |
446 | * Initialize a lease, use the default lock manager operations |
447 | */ |
448 | static int lease_init(struct file *filp, int type, struct file_lock *fl) |
449 | { |
450 | if (assign_type(fl, type) != 0) |
451 | return -EINVAL; |
452 | |
453 | fl->fl_owner = current->files; |
454 | fl->fl_pid = current->tgid; |
455 | |
456 | fl->fl_file = filp; |
457 | fl->fl_flags = FL_LEASE; |
458 | fl->fl_start = 0; |
459 | fl->fl_end = OFFSET_MAX; |
460 | fl->fl_ops = NULL; |
461 | fl->fl_lmops = &lease_manager_ops; |
462 | return 0; |
463 | } |
464 | |
465 | /* Allocate a file_lock initialised to this type of lease */ |
466 | static struct file_lock *lease_alloc(struct file *filp, int type) |
467 | { |
468 | struct file_lock *fl = locks_alloc_lock(); |
469 | int error = -ENOMEM; |
470 | |
471 | if (fl == NULL) |
472 | return ERR_PTR(error); |
473 | |
474 | error = lease_init(filp, type, fl); |
475 | if (error) { |
476 | locks_free_lock(fl); |
477 | return ERR_PTR(error); |
478 | } |
479 | return fl; |
480 | } |
481 | |
482 | /* Check if two locks overlap each other. |
483 | */ |
484 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) |
485 | { |
486 | return ((fl1->fl_end >= fl2->fl_start) && |
487 | (fl2->fl_end >= fl1->fl_start)); |
488 | } |
489 | |
490 | /* |
491 | * Check whether two locks have the same owner. |
492 | */ |
493 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
494 | { |
495 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
496 | return fl2->fl_lmops == fl1->fl_lmops && |
497 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
498 | return fl1->fl_owner == fl2->fl_owner; |
499 | } |
500 | |
501 | /* Remove waiter from blocker's block list. |
502 | * When blocker ends up pointing to itself then the list is empty. |
503 | */ |
504 | static void __locks_delete_block(struct file_lock *waiter) |
505 | { |
506 | list_del_init(&waiter->fl_block); |
507 | list_del_init(&waiter->fl_link); |
508 | waiter->fl_next = NULL; |
509 | } |
510 | |
511 | /* |
512 | */ |
513 | static void locks_delete_block(struct file_lock *waiter) |
514 | { |
515 | lock_flocks(); |
516 | __locks_delete_block(waiter); |
517 | unlock_flocks(); |
518 | } |
519 | |
520 | /* Insert waiter into blocker's block list. |
521 | * We use a circular list so that processes can be easily woken up in |
522 | * the order they blocked. The documentation doesn't require this but |
523 | * it seems like the reasonable thing to do. |
524 | */ |
525 | static void locks_insert_block(struct file_lock *blocker, |
526 | struct file_lock *waiter) |
527 | { |
528 | BUG_ON(!list_empty(&waiter->fl_block)); |
529 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
530 | waiter->fl_next = blocker; |
531 | if (IS_POSIX(blocker)) |
532 | list_add(&waiter->fl_link, &blocked_list); |
533 | } |
534 | |
535 | /* Wake up processes blocked waiting for blocker. |
536 | * If told to wait then schedule the processes until the block list |
537 | * is empty, otherwise empty the block list ourselves. |
538 | */ |
539 | static void locks_wake_up_blocks(struct file_lock *blocker) |
540 | { |
541 | while (!list_empty(&blocker->fl_block)) { |
542 | struct file_lock *waiter; |
543 | |
544 | waiter = list_first_entry(&blocker->fl_block, |
545 | struct file_lock, fl_block); |
546 | __locks_delete_block(waiter); |
547 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
548 | waiter->fl_lmops->lm_notify(waiter); |
549 | else |
550 | wake_up(&waiter->fl_wait); |
551 | } |
552 | } |
553 | |
554 | /* Insert file lock fl into an inode's lock list at the position indicated |
555 | * by pos. At the same time add the lock to the global file lock list. |
556 | */ |
557 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) |
558 | { |
559 | list_add(&fl->fl_link, &file_lock_list); |
560 | |
561 | fl->fl_nspid = get_pid(task_tgid(current)); |
562 | |
563 | /* insert into file's list */ |
564 | fl->fl_next = *pos; |
565 | *pos = fl; |
566 | } |
567 | |
568 | /* |
569 | * Delete a lock and then free it. |
570 | * Wake up processes that are blocked waiting for this lock, |
571 | * notify the FS that the lock has been cleared and |
572 | * finally free the lock. |
573 | */ |
574 | static void locks_delete_lock(struct file_lock **thisfl_p) |
575 | { |
576 | struct file_lock *fl = *thisfl_p; |
577 | |
578 | *thisfl_p = fl->fl_next; |
579 | fl->fl_next = NULL; |
580 | list_del_init(&fl->fl_link); |
581 | |
582 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
583 | if (fl->fl_fasync != NULL) { |
584 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); |
585 | fl->fl_fasync = NULL; |
586 | } |
587 | |
588 | if (fl->fl_nspid) { |
589 | put_pid(fl->fl_nspid); |
590 | fl->fl_nspid = NULL; |
591 | } |
592 | |
593 | locks_wake_up_blocks(fl); |
594 | locks_free_lock(fl); |
595 | } |
596 | |
597 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality |
598 | * checks for shared/exclusive status of overlapping locks. |
599 | */ |
600 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
601 | { |
602 | if (sys_fl->fl_type == F_WRLCK) |
603 | return 1; |
604 | if (caller_fl->fl_type == F_WRLCK) |
605 | return 1; |
606 | return 0; |
607 | } |
608 | |
609 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific |
610 | * checking before calling the locks_conflict(). |
611 | */ |
612 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
613 | { |
614 | /* POSIX locks owned by the same process do not conflict with |
615 | * each other. |
616 | */ |
617 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) |
618 | return (0); |
619 | |
620 | /* Check whether they overlap */ |
621 | if (!locks_overlap(caller_fl, sys_fl)) |
622 | return 0; |
623 | |
624 | return (locks_conflict(caller_fl, sys_fl)); |
625 | } |
626 | |
627 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific |
628 | * checking before calling the locks_conflict(). |
629 | */ |
630 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
631 | { |
632 | /* FLOCK locks referring to the same filp do not conflict with |
633 | * each other. |
634 | */ |
635 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) |
636 | return (0); |
637 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
638 | return 0; |
639 | |
640 | return (locks_conflict(caller_fl, sys_fl)); |
641 | } |
642 | |
643 | void |
644 | posix_test_lock(struct file *filp, struct file_lock *fl) |
645 | { |
646 | struct file_lock *cfl; |
647 | |
648 | lock_flocks(); |
649 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
650 | if (!IS_POSIX(cfl)) |
651 | continue; |
652 | if (posix_locks_conflict(fl, cfl)) |
653 | break; |
654 | } |
655 | if (cfl) { |
656 | __locks_copy_lock(fl, cfl); |
657 | if (cfl->fl_nspid) |
658 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
659 | } else |
660 | fl->fl_type = F_UNLCK; |
661 | unlock_flocks(); |
662 | return; |
663 | } |
664 | EXPORT_SYMBOL(posix_test_lock); |
665 | |
666 | /* |
667 | * Deadlock detection: |
668 | * |
669 | * We attempt to detect deadlocks that are due purely to posix file |
670 | * locks. |
671 | * |
672 | * We assume that a task can be waiting for at most one lock at a time. |
673 | * So for any acquired lock, the process holding that lock may be |
674 | * waiting on at most one other lock. That lock in turns may be held by |
675 | * someone waiting for at most one other lock. Given a requested lock |
676 | * caller_fl which is about to wait for a conflicting lock block_fl, we |
677 | * follow this chain of waiters to ensure we are not about to create a |
678 | * cycle. |
679 | * |
680 | * Since we do this before we ever put a process to sleep on a lock, we |
681 | * are ensured that there is never a cycle; that is what guarantees that |
682 | * the while() loop in posix_locks_deadlock() eventually completes. |
683 | * |
684 | * Note: the above assumption may not be true when handling lock |
685 | * requests from a broken NFS client. It may also fail in the presence |
686 | * of tasks (such as posix threads) sharing the same open file table. |
687 | * |
688 | * To handle those cases, we just bail out after a few iterations. |
689 | */ |
690 | |
691 | #define MAX_DEADLK_ITERATIONS 10 |
692 | |
693 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
694 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) |
695 | { |
696 | struct file_lock *fl; |
697 | |
698 | list_for_each_entry(fl, &blocked_list, fl_link) { |
699 | if (posix_same_owner(fl, block_fl)) |
700 | return fl->fl_next; |
701 | } |
702 | return NULL; |
703 | } |
704 | |
705 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
706 | struct file_lock *block_fl) |
707 | { |
708 | int i = 0; |
709 | |
710 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
711 | if (i++ > MAX_DEADLK_ITERATIONS) |
712 | return 0; |
713 | if (posix_same_owner(caller_fl, block_fl)) |
714 | return 1; |
715 | } |
716 | return 0; |
717 | } |
718 | |
719 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
720 | * after any leases, but before any posix locks. |
721 | * |
722 | * Note that if called with an FL_EXISTS argument, the caller may determine |
723 | * whether or not a lock was successfully freed by testing the return |
724 | * value for -ENOENT. |
725 | */ |
726 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
727 | { |
728 | struct file_lock *new_fl = NULL; |
729 | struct file_lock **before; |
730 | struct inode * inode = filp->f_path.dentry->d_inode; |
731 | int error = 0; |
732 | int found = 0; |
733 | |
734 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
735 | new_fl = locks_alloc_lock(); |
736 | if (!new_fl) |
737 | return -ENOMEM; |
738 | } |
739 | |
740 | lock_flocks(); |
741 | if (request->fl_flags & FL_ACCESS) |
742 | goto find_conflict; |
743 | |
744 | 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; |
752 | if (request->fl_type == fl->fl_type) |
753 | goto out; |
754 | found = 1; |
755 | locks_delete_lock(before); |
756 | break; |
757 | } |
758 | |
759 | if (request->fl_type == F_UNLCK) { |
760 | if ((request->fl_flags & FL_EXISTS) && !found) |
761 | error = -ENOENT; |
762 | goto out; |
763 | } |
764 | |
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) { |
770 | unlock_flocks(); |
771 | cond_resched(); |
772 | lock_flocks(); |
773 | } |
774 | |
775 | find_conflict: |
776 | for_each_lock(inode, before) { |
777 | struct file_lock *fl = *before; |
778 | if (IS_POSIX(fl)) |
779 | break; |
780 | if (IS_LEASE(fl)) |
781 | continue; |
782 | if (!flock_locks_conflict(request, fl)) |
783 | continue; |
784 | error = -EAGAIN; |
785 | if (!(request->fl_flags & FL_SLEEP)) |
786 | goto out; |
787 | error = FILE_LOCK_DEFERRED; |
788 | locks_insert_block(fl, request); |
789 | goto out; |
790 | } |
791 | if (request->fl_flags & FL_ACCESS) |
792 | goto out; |
793 | locks_copy_lock(new_fl, request); |
794 | locks_insert_lock(before, new_fl); |
795 | new_fl = NULL; |
796 | error = 0; |
797 | |
798 | out: |
799 | unlock_flocks(); |
800 | if (new_fl) |
801 | locks_free_lock(new_fl); |
802 | return error; |
803 | } |
804 | |
805 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
806 | { |
807 | struct file_lock *fl; |
808 | struct file_lock *new_fl = NULL; |
809 | struct file_lock *new_fl2 = NULL; |
810 | struct file_lock *left = NULL; |
811 | struct file_lock *right = NULL; |
812 | struct file_lock **before; |
813 | int error, added = 0; |
814 | |
815 | /* |
816 | * We may need two file_lock structures for this operation, |
817 | * so we get them in advance to avoid races. |
818 | * |
819 | * In some cases we can be sure, that no new locks will be needed |
820 | */ |
821 | if (!(request->fl_flags & FL_ACCESS) && |
822 | (request->fl_type != F_UNLCK || |
823 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { |
824 | new_fl = locks_alloc_lock(); |
825 | new_fl2 = locks_alloc_lock(); |
826 | } |
827 | |
828 | lock_flocks(); |
829 | if (request->fl_type != F_UNLCK) { |
830 | for_each_lock(inode, before) { |
831 | fl = *before; |
832 | if (!IS_POSIX(fl)) |
833 | continue; |
834 | if (!posix_locks_conflict(request, fl)) |
835 | continue; |
836 | if (conflock) |
837 | __locks_copy_lock(conflock, fl); |
838 | error = -EAGAIN; |
839 | if (!(request->fl_flags & FL_SLEEP)) |
840 | goto out; |
841 | error = -EDEADLK; |
842 | if (posix_locks_deadlock(request, fl)) |
843 | goto out; |
844 | error = FILE_LOCK_DEFERRED; |
845 | locks_insert_block(fl, request); |
846 | goto out; |
847 | } |
848 | } |
849 | |
850 | /* If we're just looking for a conflict, we're done. */ |
851 | error = 0; |
852 | if (request->fl_flags & FL_ACCESS) |
853 | goto out; |
854 | |
855 | /* |
856 | * Find the first old lock with the same owner as the new lock. |
857 | */ |
858 | |
859 | before = &inode->i_flock; |
860 | |
861 | /* First skip locks owned by other processes. */ |
862 | while ((fl = *before) && (!IS_POSIX(fl) || |
863 | !posix_same_owner(request, fl))) { |
864 | before = &fl->fl_next; |
865 | } |
866 | |
867 | /* Process locks with this owner. */ |
868 | while ((fl = *before) && posix_same_owner(request, fl)) { |
869 | /* Detect adjacent or overlapping regions (if same lock type) |
870 | */ |
871 | if (request->fl_type == fl->fl_type) { |
872 | /* In all comparisons of start vs end, use |
873 | * "start - 1" rather than "end + 1". If end |
874 | * is OFFSET_MAX, end + 1 will become negative. |
875 | */ |
876 | if (fl->fl_end < request->fl_start - 1) |
877 | goto next_lock; |
878 | /* If the next lock in the list has entirely bigger |
879 | * addresses than the new one, insert the lock here. |
880 | */ |
881 | if (fl->fl_start - 1 > request->fl_end) |
882 | break; |
883 | |
884 | /* If we come here, the new and old lock are of the |
885 | * same type and adjacent or overlapping. Make one |
886 | * lock yielding from the lower start address of both |
887 | * locks to the higher end address. |
888 | */ |
889 | if (fl->fl_start > request->fl_start) |
890 | fl->fl_start = request->fl_start; |
891 | else |
892 | request->fl_start = fl->fl_start; |
893 | if (fl->fl_end < request->fl_end) |
894 | fl->fl_end = request->fl_end; |
895 | else |
896 | request->fl_end = fl->fl_end; |
897 | if (added) { |
898 | locks_delete_lock(before); |
899 | continue; |
900 | } |
901 | request = fl; |
902 | added = 1; |
903 | } |
904 | else { |
905 | /* Processing for different lock types is a bit |
906 | * more complex. |
907 | */ |
908 | if (fl->fl_end < request->fl_start) |
909 | goto next_lock; |
910 | if (fl->fl_start > request->fl_end) |
911 | break; |
912 | if (request->fl_type == F_UNLCK) |
913 | added = 1; |
914 | if (fl->fl_start < request->fl_start) |
915 | left = fl; |
916 | /* If the next lock in the list has a higher end |
917 | * address than the new one, insert the new one here. |
918 | */ |
919 | if (fl->fl_end > request->fl_end) { |
920 | right = fl; |
921 | break; |
922 | } |
923 | if (fl->fl_start >= request->fl_start) { |
924 | /* The new lock completely replaces an old |
925 | * one (This may happen several times). |
926 | */ |
927 | if (added) { |
928 | locks_delete_lock(before); |
929 | continue; |
930 | } |
931 | /* Replace the old lock with the new one. |
932 | * Wake up anybody waiting for the old one, |
933 | * as the change in lock type might satisfy |
934 | * their needs. |
935 | */ |
936 | locks_wake_up_blocks(fl); |
937 | fl->fl_start = request->fl_start; |
938 | fl->fl_end = request->fl_end; |
939 | fl->fl_type = request->fl_type; |
940 | locks_release_private(fl); |
941 | locks_copy_private(fl, request); |
942 | request = fl; |
943 | added = 1; |
944 | } |
945 | } |
946 | /* Go on to next lock. |
947 | */ |
948 | next_lock: |
949 | before = &fl->fl_next; |
950 | } |
951 | |
952 | /* |
953 | * The above code only modifies existing locks in case of |
954 | * merging or replacing. If new lock(s) need to be inserted |
955 | * all modifications are done bellow this, so it's safe yet to |
956 | * bail out. |
957 | */ |
958 | error = -ENOLCK; /* "no luck" */ |
959 | if (right && left == right && !new_fl2) |
960 | goto out; |
961 | |
962 | error = 0; |
963 | if (!added) { |
964 | if (request->fl_type == F_UNLCK) { |
965 | if (request->fl_flags & FL_EXISTS) |
966 | error = -ENOENT; |
967 | goto out; |
968 | } |
969 | |
970 | if (!new_fl) { |
971 | error = -ENOLCK; |
972 | goto out; |
973 | } |
974 | locks_copy_lock(new_fl, request); |
975 | locks_insert_lock(before, new_fl); |
976 | new_fl = NULL; |
977 | } |
978 | if (right) { |
979 | if (left == right) { |
980 | /* The new lock breaks the old one in two pieces, |
981 | * so we have to use the second new lock. |
982 | */ |
983 | left = new_fl2; |
984 | new_fl2 = NULL; |
985 | locks_copy_lock(left, right); |
986 | locks_insert_lock(before, left); |
987 | } |
988 | right->fl_start = request->fl_end + 1; |
989 | locks_wake_up_blocks(right); |
990 | } |
991 | if (left) { |
992 | left->fl_end = request->fl_start - 1; |
993 | locks_wake_up_blocks(left); |
994 | } |
995 | out: |
996 | unlock_flocks(); |
997 | /* |
998 | * Free any unused locks. |
999 | */ |
1000 | if (new_fl) |
1001 | locks_free_lock(new_fl); |
1002 | if (new_fl2) |
1003 | locks_free_lock(new_fl2); |
1004 | return error; |
1005 | } |
1006 | |
1007 | /** |
1008 | * posix_lock_file - Apply a POSIX-style lock to a file |
1009 | * @filp: The file to apply the lock to |
1010 | * @fl: The lock to be applied |
1011 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1012 | * |
1013 | * Add a POSIX style lock to a file. |
1014 | * We merge adjacent & overlapping locks whenever possible. |
1015 | * POSIX locks are sorted by owner task, then by starting address |
1016 | * |
1017 | * Note that if called with an FL_EXISTS argument, the caller may determine |
1018 | * whether or not a lock was successfully freed by testing the return |
1019 | * value for -ENOENT. |
1020 | */ |
1021 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
1022 | struct file_lock *conflock) |
1023 | { |
1024 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
1025 | } |
1026 | EXPORT_SYMBOL(posix_lock_file); |
1027 | |
1028 | /** |
1029 | * posix_lock_file_wait - Apply a POSIX-style lock to a file |
1030 | * @filp: The file to apply the lock to |
1031 | * @fl: The lock to be applied |
1032 | * |
1033 | * Add a POSIX style lock to a file. |
1034 | * We merge adjacent & overlapping locks whenever possible. |
1035 | * POSIX locks are sorted by owner task, then by starting address |
1036 | */ |
1037 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) |
1038 | { |
1039 | int error; |
1040 | might_sleep (); |
1041 | for (;;) { |
1042 | error = posix_lock_file(filp, fl, NULL); |
1043 | if (error != FILE_LOCK_DEFERRED) |
1044 | break; |
1045 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1046 | if (!error) |
1047 | continue; |
1048 | |
1049 | locks_delete_block(fl); |
1050 | break; |
1051 | } |
1052 | return error; |
1053 | } |
1054 | EXPORT_SYMBOL(posix_lock_file_wait); |
1055 | |
1056 | /** |
1057 | * locks_mandatory_locked - Check for an active lock |
1058 | * @inode: the file to check |
1059 | * |
1060 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
1061 | * This function is called from locks_verify_locked() only. |
1062 | */ |
1063 | int locks_mandatory_locked(struct inode *inode) |
1064 | { |
1065 | fl_owner_t owner = current->files; |
1066 | struct file_lock *fl; |
1067 | |
1068 | /* |
1069 | * Search the lock list for this inode for any POSIX locks. |
1070 | */ |
1071 | lock_flocks(); |
1072 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1073 | if (!IS_POSIX(fl)) |
1074 | continue; |
1075 | if (fl->fl_owner != owner) |
1076 | break; |
1077 | } |
1078 | unlock_flocks(); |
1079 | return fl ? -EAGAIN : 0; |
1080 | } |
1081 | |
1082 | /** |
1083 | * locks_mandatory_area - Check for a conflicting lock |
1084 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ |
1085 | * for shared |
1086 | * @inode: the file to check |
1087 | * @filp: how the file was opened (if it was) |
1088 | * @offset: start of area to check |
1089 | * @count: length of area to check |
1090 | * |
1091 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
1092 | * This function is called from rw_verify_area() and |
1093 | * locks_verify_truncate(). |
1094 | */ |
1095 | int locks_mandatory_area(int read_write, struct inode *inode, |
1096 | struct file *filp, loff_t offset, |
1097 | size_t count) |
1098 | { |
1099 | struct file_lock fl; |
1100 | int error; |
1101 | |
1102 | locks_init_lock(&fl); |
1103 | fl.fl_owner = current->files; |
1104 | fl.fl_pid = current->tgid; |
1105 | fl.fl_file = filp; |
1106 | fl.fl_flags = FL_POSIX | FL_ACCESS; |
1107 | if (filp && !(filp->f_flags & O_NONBLOCK)) |
1108 | fl.fl_flags |= FL_SLEEP; |
1109 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1110 | fl.fl_start = offset; |
1111 | fl.fl_end = offset + count - 1; |
1112 | |
1113 | for (;;) { |
1114 | error = __posix_lock_file(inode, &fl, NULL); |
1115 | if (error != FILE_LOCK_DEFERRED) |
1116 | break; |
1117 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); |
1118 | if (!error) { |
1119 | /* |
1120 | * If we've been sleeping someone might have |
1121 | * changed the permissions behind our back. |
1122 | */ |
1123 | if (__mandatory_lock(inode)) |
1124 | continue; |
1125 | } |
1126 | |
1127 | locks_delete_block(&fl); |
1128 | break; |
1129 | } |
1130 | |
1131 | return error; |
1132 | } |
1133 | |
1134 | EXPORT_SYMBOL(locks_mandatory_area); |
1135 | |
1136 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1137 | { |
1138 | switch (arg) { |
1139 | case F_UNLCK: |
1140 | fl->fl_flags &= ~FL_UNLOCK_PENDING; |
1141 | /* fall through: */ |
1142 | case F_RDLCK: |
1143 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; |
1144 | } |
1145 | } |
1146 | |
1147 | /* We already had a lease on this file; just change its type */ |
1148 | int lease_modify(struct file_lock **before, int arg) |
1149 | { |
1150 | struct file_lock *fl = *before; |
1151 | int error = assign_type(fl, arg); |
1152 | |
1153 | if (error) |
1154 | return error; |
1155 | lease_clear_pending(fl, arg); |
1156 | locks_wake_up_blocks(fl); |
1157 | if (arg == F_UNLCK) |
1158 | locks_delete_lock(before); |
1159 | return 0; |
1160 | } |
1161 | |
1162 | EXPORT_SYMBOL(lease_modify); |
1163 | |
1164 | static bool past_time(unsigned long then) |
1165 | { |
1166 | if (!then) |
1167 | /* 0 is a special value meaning "this never expires": */ |
1168 | return false; |
1169 | return time_after(jiffies, then); |
1170 | } |
1171 | |
1172 | static void time_out_leases(struct inode *inode) |
1173 | { |
1174 | struct file_lock **before; |
1175 | struct file_lock *fl; |
1176 | |
1177 | before = &inode->i_flock; |
1178 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
1179 | if (past_time(fl->fl_downgrade_time)) |
1180 | lease_modify(before, F_RDLCK); |
1181 | if (past_time(fl->fl_break_time)) |
1182 | lease_modify(before, F_UNLCK); |
1183 | if (fl == *before) /* lease_modify may have freed fl */ |
1184 | before = &fl->fl_next; |
1185 | } |
1186 | } |
1187 | |
1188 | /** |
1189 | * __break_lease - revoke all outstanding leases on file |
1190 | * @inode: the inode of the file to return |
1191 | * @mode: the open mode (read or write) |
1192 | * |
1193 | * break_lease (inlined for speed) has checked there already is at least |
1194 | * some kind of lock (maybe a lease) on this file. Leases are broken on |
1195 | * a call to open() or truncate(). This function can sleep unless you |
1196 | * specified %O_NONBLOCK to your open(). |
1197 | */ |
1198 | int __break_lease(struct inode *inode, unsigned int mode) |
1199 | { |
1200 | int error = 0; |
1201 | struct file_lock *new_fl, *flock; |
1202 | struct file_lock *fl; |
1203 | unsigned long break_time; |
1204 | int i_have_this_lease = 0; |
1205 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1206 | |
1207 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
1208 | if (IS_ERR(new_fl)) |
1209 | return PTR_ERR(new_fl); |
1210 | |
1211 | lock_flocks(); |
1212 | |
1213 | time_out_leases(inode); |
1214 | |
1215 | flock = inode->i_flock; |
1216 | if ((flock == NULL) || !IS_LEASE(flock)) |
1217 | goto out; |
1218 | |
1219 | if (!locks_conflict(flock, new_fl)) |
1220 | goto out; |
1221 | |
1222 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
1223 | if (fl->fl_owner == current->files) |
1224 | i_have_this_lease = 1; |
1225 | |
1226 | break_time = 0; |
1227 | if (lease_break_time > 0) { |
1228 | break_time = jiffies + lease_break_time * HZ; |
1229 | if (break_time == 0) |
1230 | break_time++; /* so that 0 means no break time */ |
1231 | } |
1232 | |
1233 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1234 | if (want_write) { |
1235 | if (fl->fl_flags & FL_UNLOCK_PENDING) |
1236 | continue; |
1237 | fl->fl_flags |= FL_UNLOCK_PENDING; |
1238 | fl->fl_break_time = break_time; |
1239 | } else { |
1240 | if (lease_breaking(flock)) |
1241 | continue; |
1242 | fl->fl_flags |= FL_DOWNGRADE_PENDING; |
1243 | fl->fl_downgrade_time = break_time; |
1244 | } |
1245 | fl->fl_lmops->lm_break(fl); |
1246 | } |
1247 | |
1248 | if (i_have_this_lease || (mode & O_NONBLOCK)) { |
1249 | error = -EWOULDBLOCK; |
1250 | goto out; |
1251 | } |
1252 | |
1253 | restart: |
1254 | break_time = flock->fl_break_time; |
1255 | if (break_time != 0) { |
1256 | break_time -= jiffies; |
1257 | if (break_time == 0) |
1258 | break_time++; |
1259 | } |
1260 | locks_insert_block(flock, new_fl); |
1261 | unlock_flocks(); |
1262 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1263 | !new_fl->fl_next, break_time); |
1264 | lock_flocks(); |
1265 | __locks_delete_block(new_fl); |
1266 | if (error >= 0) { |
1267 | if (error == 0) |
1268 | time_out_leases(inode); |
1269 | /* |
1270 | * Wait for the next conflicting lease that has not been |
1271 | * broken yet |
1272 | */ |
1273 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1274 | flock = flock->fl_next) { |
1275 | if (locks_conflict(new_fl, flock)) |
1276 | goto restart; |
1277 | } |
1278 | error = 0; |
1279 | } |
1280 | |
1281 | out: |
1282 | unlock_flocks(); |
1283 | locks_free_lock(new_fl); |
1284 | return error; |
1285 | } |
1286 | |
1287 | EXPORT_SYMBOL(__break_lease); |
1288 | |
1289 | /** |
1290 | * lease_get_mtime - get the last modified time of an inode |
1291 | * @inode: the inode |
1292 | * @time: pointer to a timespec which will contain the last modified time |
1293 | * |
1294 | * This is to force NFS clients to flush their caches for files with |
1295 | * exclusive leases. The justification is that if someone has an |
1296 | * exclusive lease, then they could be modifying it. |
1297 | */ |
1298 | void lease_get_mtime(struct inode *inode, struct timespec *time) |
1299 | { |
1300 | struct file_lock *flock = inode->i_flock; |
1301 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) |
1302 | *time = current_fs_time(inode->i_sb); |
1303 | else |
1304 | *time = inode->i_mtime; |
1305 | } |
1306 | |
1307 | EXPORT_SYMBOL(lease_get_mtime); |
1308 | |
1309 | /** |
1310 | * fcntl_getlease - Enquire what lease is currently active |
1311 | * @filp: the file |
1312 | * |
1313 | * The value returned by this function will be one of |
1314 | * (if no lease break is pending): |
1315 | * |
1316 | * %F_RDLCK to indicate a shared lease is held. |
1317 | * |
1318 | * %F_WRLCK to indicate an exclusive lease is held. |
1319 | * |
1320 | * %F_UNLCK to indicate no lease is held. |
1321 | * |
1322 | * (if a lease break is pending): |
1323 | * |
1324 | * %F_RDLCK to indicate an exclusive lease needs to be |
1325 | * changed to a shared lease (or removed). |
1326 | * |
1327 | * %F_UNLCK to indicate the lease needs to be removed. |
1328 | * |
1329 | * XXX: sfr & willy disagree over whether F_INPROGRESS |
1330 | * should be returned to userspace. |
1331 | */ |
1332 | int fcntl_getlease(struct file *filp) |
1333 | { |
1334 | struct file_lock *fl; |
1335 | int type = F_UNLCK; |
1336 | |
1337 | lock_flocks(); |
1338 | time_out_leases(filp->f_path.dentry->d_inode); |
1339 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); |
1340 | fl = fl->fl_next) { |
1341 | if (fl->fl_file == filp) { |
1342 | type = target_leasetype(fl); |
1343 | break; |
1344 | } |
1345 | } |
1346 | unlock_flocks(); |
1347 | return type; |
1348 | } |
1349 | |
1350 | int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1351 | { |
1352 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
1353 | struct dentry *dentry = filp->f_path.dentry; |
1354 | struct inode *inode = dentry->d_inode; |
1355 | int error; |
1356 | |
1357 | lease = *flp; |
1358 | |
1359 | error = -EAGAIN; |
1360 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
1361 | goto out; |
1362 | if ((arg == F_WRLCK) |
1363 | && ((dentry->d_count > 1) |
1364 | || (atomic_read(&inode->i_count) > 1))) |
1365 | goto out; |
1366 | |
1367 | /* |
1368 | * At this point, we know that if there is an exclusive |
1369 | * lease on this file, then we hold it on this filp |
1370 | * (otherwise our open of this file would have blocked). |
1371 | * And if we are trying to acquire an exclusive lease, |
1372 | * then the file is not open by anyone (including us) |
1373 | * except for this filp. |
1374 | */ |
1375 | error = -EAGAIN; |
1376 | for (before = &inode->i_flock; |
1377 | ((fl = *before) != NULL) && IS_LEASE(fl); |
1378 | before = &fl->fl_next) { |
1379 | if (fl->fl_file == filp) { |
1380 | my_before = before; |
1381 | continue; |
1382 | } |
1383 | /* |
1384 | * No exclusive leases if someone else has a lease on |
1385 | * this file: |
1386 | */ |
1387 | if (arg == F_WRLCK) |
1388 | goto out; |
1389 | /* |
1390 | * Modifying our existing lease is OK, but no getting a |
1391 | * new lease if someone else is opening for write: |
1392 | */ |
1393 | if (fl->fl_flags & FL_UNLOCK_PENDING) |
1394 | goto out; |
1395 | } |
1396 | |
1397 | if (my_before != NULL) { |
1398 | error = lease->fl_lmops->lm_change(my_before, arg); |
1399 | if (!error) |
1400 | *flp = *my_before; |
1401 | goto out; |
1402 | } |
1403 | |
1404 | error = -EINVAL; |
1405 | if (!leases_enable) |
1406 | goto out; |
1407 | |
1408 | locks_insert_lock(before, lease); |
1409 | return 0; |
1410 | |
1411 | out: |
1412 | return error; |
1413 | } |
1414 | |
1415 | int generic_delete_lease(struct file *filp, struct file_lock **flp) |
1416 | { |
1417 | struct file_lock *fl, **before; |
1418 | struct dentry *dentry = filp->f_path.dentry; |
1419 | struct inode *inode = dentry->d_inode; |
1420 | |
1421 | for (before = &inode->i_flock; |
1422 | ((fl = *before) != NULL) && IS_LEASE(fl); |
1423 | before = &fl->fl_next) { |
1424 | if (fl->fl_file != filp) |
1425 | continue; |
1426 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); |
1427 | } |
1428 | return -EAGAIN; |
1429 | } |
1430 | |
1431 | /** |
1432 | * generic_setlease - sets a lease on an open file |
1433 | * @filp: file pointer |
1434 | * @arg: type of lease to obtain |
1435 | * @flp: input - file_lock to use, output - file_lock inserted |
1436 | * |
1437 | * The (input) flp->fl_lmops->lm_break function is required |
1438 | * by break_lease(). |
1439 | * |
1440 | * Called with file_lock_lock held. |
1441 | */ |
1442 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
1443 | { |
1444 | struct dentry *dentry = filp->f_path.dentry; |
1445 | struct inode *inode = dentry->d_inode; |
1446 | int error; |
1447 | |
1448 | if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE)) |
1449 | return -EACCES; |
1450 | if (!S_ISREG(inode->i_mode)) |
1451 | return -EINVAL; |
1452 | error = security_file_lock(filp, arg); |
1453 | if (error) |
1454 | return error; |
1455 | |
1456 | time_out_leases(inode); |
1457 | |
1458 | BUG_ON(!(*flp)->fl_lmops->lm_break); |
1459 | |
1460 | switch (arg) { |
1461 | case F_UNLCK: |
1462 | return generic_delete_lease(filp, flp); |
1463 | case F_RDLCK: |
1464 | case F_WRLCK: |
1465 | return generic_add_lease(filp, arg, flp); |
1466 | default: |
1467 | BUG(); |
1468 | } |
1469 | } |
1470 | EXPORT_SYMBOL(generic_setlease); |
1471 | |
1472 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1473 | { |
1474 | if (filp->f_op && filp->f_op->setlease) |
1475 | return filp->f_op->setlease(filp, arg, lease); |
1476 | else |
1477 | return generic_setlease(filp, arg, lease); |
1478 | } |
1479 | |
1480 | /** |
1481 | * vfs_setlease - sets a lease on an open file |
1482 | * @filp: file pointer |
1483 | * @arg: type of lease to obtain |
1484 | * @lease: file_lock to use |
1485 | * |
1486 | * Call this to establish a lease on the file. |
1487 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
1488 | * break_lease will oops! |
1489 | * |
1490 | * This will call the filesystem's setlease file method, if |
1491 | * defined. Note that there is no getlease method; instead, the |
1492 | * filesystem setlease method should call back to setlease() to |
1493 | * add a lease to the inode's lease list, where fcntl_getlease() can |
1494 | * find it. Since fcntl_getlease() only reports whether the current |
1495 | * task holds a lease, a cluster filesystem need only do this for |
1496 | * leases held by processes on this node. |
1497 | * |
1498 | * There is also no break_lease method; filesystems that |
1499 | * handle their own leases should break leases themselves from the |
1500 | * filesystem's open, create, and (on truncate) setattr methods. |
1501 | * |
1502 | * Warning: the only current setlease methods exist only to disable |
1503 | * leases in certain cases. More vfs changes may be required to |
1504 | * allow a full filesystem lease implementation. |
1505 | */ |
1506 | |
1507 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1508 | { |
1509 | int error; |
1510 | |
1511 | lock_flocks(); |
1512 | error = __vfs_setlease(filp, arg, lease); |
1513 | unlock_flocks(); |
1514 | |
1515 | return error; |
1516 | } |
1517 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1518 | |
1519 | static int do_fcntl_delete_lease(struct file *filp) |
1520 | { |
1521 | struct file_lock fl, *flp = &fl; |
1522 | |
1523 | lease_init(filp, F_UNLCK, flp); |
1524 | |
1525 | return vfs_setlease(filp, F_UNLCK, &flp); |
1526 | } |
1527 | |
1528 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1529 | { |
1530 | struct file_lock *fl, *ret; |
1531 | struct fasync_struct *new; |
1532 | int error; |
1533 | |
1534 | fl = lease_alloc(filp, arg); |
1535 | if (IS_ERR(fl)) |
1536 | return PTR_ERR(fl); |
1537 | |
1538 | new = fasync_alloc(); |
1539 | if (!new) { |
1540 | locks_free_lock(fl); |
1541 | return -ENOMEM; |
1542 | } |
1543 | ret = fl; |
1544 | lock_flocks(); |
1545 | error = __vfs_setlease(filp, arg, &ret); |
1546 | if (error) { |
1547 | unlock_flocks(); |
1548 | locks_free_lock(fl); |
1549 | goto out_free_fasync; |
1550 | } |
1551 | if (ret != fl) |
1552 | locks_free_lock(fl); |
1553 | |
1554 | /* |
1555 | * fasync_insert_entry() returns the old entry if any. |
1556 | * If there was no old entry, then it used 'new' and |
1557 | * inserted it into the fasync list. Clear new so that |
1558 | * we don't release it here. |
1559 | */ |
1560 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
1561 | new = NULL; |
1562 | |
1563 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1564 | unlock_flocks(); |
1565 | |
1566 | out_free_fasync: |
1567 | if (new) |
1568 | fasync_free(new); |
1569 | return error; |
1570 | } |
1571 | |
1572 | /** |
1573 | * fcntl_setlease - sets a lease on an open file |
1574 | * @fd: open file descriptor |
1575 | * @filp: file pointer |
1576 | * @arg: type of lease to obtain |
1577 | * |
1578 | * Call this fcntl to establish a lease on the file. |
1579 | * Note that you also need to call %F_SETSIG to |
1580 | * receive a signal when the lease is broken. |
1581 | */ |
1582 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) |
1583 | { |
1584 | if (arg == F_UNLCK) |
1585 | return do_fcntl_delete_lease(filp); |
1586 | return do_fcntl_add_lease(fd, filp, arg); |
1587 | } |
1588 | |
1589 | /** |
1590 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file |
1591 | * @filp: The file to apply the lock to |
1592 | * @fl: The lock to be applied |
1593 | * |
1594 | * Add a FLOCK style lock to a file. |
1595 | */ |
1596 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) |
1597 | { |
1598 | int error; |
1599 | might_sleep(); |
1600 | for (;;) { |
1601 | error = flock_lock_file(filp, fl); |
1602 | if (error != FILE_LOCK_DEFERRED) |
1603 | break; |
1604 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1605 | if (!error) |
1606 | continue; |
1607 | |
1608 | locks_delete_block(fl); |
1609 | break; |
1610 | } |
1611 | return error; |
1612 | } |
1613 | |
1614 | EXPORT_SYMBOL(flock_lock_file_wait); |
1615 | |
1616 | /** |
1617 | * sys_flock: - flock() system call. |
1618 | * @fd: the file descriptor to lock. |
1619 | * @cmd: the type of lock to apply. |
1620 | * |
1621 | * Apply a %FL_FLOCK style lock to an open file descriptor. |
1622 | * The @cmd can be one of |
1623 | * |
1624 | * %LOCK_SH -- a shared lock. |
1625 | * |
1626 | * %LOCK_EX -- an exclusive lock. |
1627 | * |
1628 | * %LOCK_UN -- remove an existing lock. |
1629 | * |
1630 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. |
1631 | * |
1632 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other |
1633 | * processes read and write access respectively. |
1634 | */ |
1635 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1636 | { |
1637 | struct file *filp; |
1638 | struct file_lock *lock; |
1639 | int can_sleep, unlock; |
1640 | int error; |
1641 | |
1642 | error = -EBADF; |
1643 | filp = fget(fd); |
1644 | if (!filp) |
1645 | goto out; |
1646 | |
1647 | can_sleep = !(cmd & LOCK_NB); |
1648 | cmd &= ~LOCK_NB; |
1649 | unlock = (cmd == LOCK_UN); |
1650 | |
1651 | if (!unlock && !(cmd & LOCK_MAND) && |
1652 | !(filp->f_mode & (FMODE_READ|FMODE_WRITE))) |
1653 | goto out_putf; |
1654 | |
1655 | error = flock_make_lock(filp, &lock, cmd); |
1656 | if (error) |
1657 | goto out_putf; |
1658 | if (can_sleep) |
1659 | lock->fl_flags |= FL_SLEEP; |
1660 | |
1661 | error = security_file_lock(filp, lock->fl_type); |
1662 | if (error) |
1663 | goto out_free; |
1664 | |
1665 | if (filp->f_op && filp->f_op->flock) |
1666 | error = filp->f_op->flock(filp, |
1667 | (can_sleep) ? F_SETLKW : F_SETLK, |
1668 | lock); |
1669 | else |
1670 | error = flock_lock_file_wait(filp, lock); |
1671 | |
1672 | out_free: |
1673 | locks_free_lock(lock); |
1674 | |
1675 | out_putf: |
1676 | fput(filp); |
1677 | out: |
1678 | return error; |
1679 | } |
1680 | |
1681 | /** |
1682 | * vfs_test_lock - test file byte range lock |
1683 | * @filp: The file to test lock for |
1684 | * @fl: The lock to test; also used to hold result |
1685 | * |
1686 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by |
1687 | * setting conf->fl_type to something other than F_UNLCK. |
1688 | */ |
1689 | int vfs_test_lock(struct file *filp, struct file_lock *fl) |
1690 | { |
1691 | if (filp->f_op && filp->f_op->lock) |
1692 | return filp->f_op->lock(filp, F_GETLK, fl); |
1693 | posix_test_lock(filp, fl); |
1694 | return 0; |
1695 | } |
1696 | EXPORT_SYMBOL_GPL(vfs_test_lock); |
1697 | |
1698 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1699 | { |
1700 | flock->l_pid = fl->fl_pid; |
1701 | #if BITS_PER_LONG == 32 |
1702 | /* |
1703 | * Make sure we can represent the posix lock via |
1704 | * legacy 32bit flock. |
1705 | */ |
1706 | if (fl->fl_start > OFFT_OFFSET_MAX) |
1707 | return -EOVERFLOW; |
1708 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) |
1709 | return -EOVERFLOW; |
1710 | #endif |
1711 | flock->l_start = fl->fl_start; |
1712 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
1713 | fl->fl_end - fl->fl_start + 1; |
1714 | flock->l_whence = 0; |
1715 | flock->l_type = fl->fl_type; |
1716 | return 0; |
1717 | } |
1718 | |
1719 | #if BITS_PER_LONG == 32 |
1720 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) |
1721 | { |
1722 | flock->l_pid = fl->fl_pid; |
1723 | flock->l_start = fl->fl_start; |
1724 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
1725 | fl->fl_end - fl->fl_start + 1; |
1726 | flock->l_whence = 0; |
1727 | flock->l_type = fl->fl_type; |
1728 | } |
1729 | #endif |
1730 | |
1731 | /* Report the first existing lock that would conflict with l. |
1732 | * This implements the F_GETLK command of fcntl(). |
1733 | */ |
1734 | int fcntl_getlk(struct file *filp, struct flock __user *l) |
1735 | { |
1736 | struct file_lock file_lock; |
1737 | struct flock flock; |
1738 | int error; |
1739 | |
1740 | error = -EFAULT; |
1741 | if (copy_from_user(&flock, l, sizeof(flock))) |
1742 | goto out; |
1743 | error = -EINVAL; |
1744 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
1745 | goto out; |
1746 | |
1747 | error = flock_to_posix_lock(filp, &file_lock, &flock); |
1748 | if (error) |
1749 | goto out; |
1750 | |
1751 | error = vfs_test_lock(filp, &file_lock); |
1752 | if (error) |
1753 | goto out; |
1754 | |
1755 | flock.l_type = file_lock.fl_type; |
1756 | if (file_lock.fl_type != F_UNLCK) { |
1757 | error = posix_lock_to_flock(&flock, &file_lock); |
1758 | if (error) |
1759 | goto out; |
1760 | } |
1761 | error = -EFAULT; |
1762 | if (!copy_to_user(l, &flock, sizeof(flock))) |
1763 | error = 0; |
1764 | out: |
1765 | return error; |
1766 | } |
1767 | |
1768 | /** |
1769 | * vfs_lock_file - file byte range lock |
1770 | * @filp: The file to apply the lock to |
1771 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) |
1772 | * @fl: The lock to be applied |
1773 | * @conf: Place to return a copy of the conflicting lock, if found. |
1774 | * |
1775 | * A caller that doesn't care about the conflicting lock may pass NULL |
1776 | * as the final argument. |
1777 | * |
1778 | * If the filesystem defines a private ->lock() method, then @conf will |
1779 | * be left unchanged; so a caller that cares should initialize it to |
1780 | * some acceptable default. |
1781 | * |
1782 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX |
1783 | * locks, the ->lock() interface may return asynchronously, before the lock has |
1784 | * been granted or denied by the underlying filesystem, if (and only if) |
1785 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
1786 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1787 | * the request is for a blocking lock. When ->lock() does return asynchronously, |
1788 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
1789 | * request completes. |
1790 | * If the request is for non-blocking lock the file system should return |
1791 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1792 | * with the result. If the request timed out the callback routine will return a |
1793 | * nonzero return code and the file system should release the lock. The file |
1794 | * system is also responsible to keep a corresponding posix lock when it |
1795 | * grants a lock so the VFS can find out which locks are locally held and do |
1796 | * the correct lock cleanup when required. |
1797 | * The underlying filesystem must not drop the kernel lock or call |
1798 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
1799 | * return code. |
1800 | */ |
1801 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
1802 | { |
1803 | if (filp->f_op && filp->f_op->lock) |
1804 | return filp->f_op->lock(filp, cmd, fl); |
1805 | else |
1806 | return posix_lock_file(filp, fl, conf); |
1807 | } |
1808 | EXPORT_SYMBOL_GPL(vfs_lock_file); |
1809 | |
1810 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1811 | struct file_lock *fl) |
1812 | { |
1813 | int error; |
1814 | |
1815 | error = security_file_lock(filp, fl->fl_type); |
1816 | if (error) |
1817 | return error; |
1818 | |
1819 | for (;;) { |
1820 | error = vfs_lock_file(filp, cmd, fl, NULL); |
1821 | if (error != FILE_LOCK_DEFERRED) |
1822 | break; |
1823 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1824 | if (!error) |
1825 | continue; |
1826 | |
1827 | locks_delete_block(fl); |
1828 | break; |
1829 | } |
1830 | |
1831 | return error; |
1832 | } |
1833 | |
1834 | /* Apply the lock described by l to an open file descriptor. |
1835 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
1836 | */ |
1837 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1838 | struct flock __user *l) |
1839 | { |
1840 | struct file_lock *file_lock = locks_alloc_lock(); |
1841 | struct flock flock; |
1842 | struct inode *inode; |
1843 | struct file *f; |
1844 | int error; |
1845 | |
1846 | if (file_lock == NULL) |
1847 | return -ENOLCK; |
1848 | |
1849 | /* |
1850 | * This might block, so we do it before checking the inode. |
1851 | */ |
1852 | error = -EFAULT; |
1853 | if (copy_from_user(&flock, l, sizeof(flock))) |
1854 | goto out; |
1855 | |
1856 | inode = filp->f_path.dentry->d_inode; |
1857 | |
1858 | /* Don't allow mandatory locks on files that may be memory mapped |
1859 | * and shared. |
1860 | */ |
1861 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1862 | error = -EAGAIN; |
1863 | goto out; |
1864 | } |
1865 | |
1866 | again: |
1867 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1868 | if (error) |
1869 | goto out; |
1870 | if (cmd == F_SETLKW) { |
1871 | file_lock->fl_flags |= FL_SLEEP; |
1872 | } |
1873 | |
1874 | error = -EBADF; |
1875 | switch (flock.l_type) { |
1876 | case F_RDLCK: |
1877 | if (!(filp->f_mode & FMODE_READ)) |
1878 | goto out; |
1879 | break; |
1880 | case F_WRLCK: |
1881 | if (!(filp->f_mode & FMODE_WRITE)) |
1882 | goto out; |
1883 | break; |
1884 | case F_UNLCK: |
1885 | break; |
1886 | default: |
1887 | error = -EINVAL; |
1888 | goto out; |
1889 | } |
1890 | |
1891 | error = do_lock_file_wait(filp, cmd, file_lock); |
1892 | |
1893 | /* |
1894 | * Attempt to detect a close/fcntl race and recover by |
1895 | * releasing the lock that was just acquired. |
1896 | */ |
1897 | /* |
1898 | * we need that spin_lock here - it prevents reordering between |
1899 | * update of inode->i_flock and check for it done in close(). |
1900 | * rcu_read_lock() wouldn't do. |
1901 | */ |
1902 | spin_lock(¤t->files->file_lock); |
1903 | f = fcheck(fd); |
1904 | spin_unlock(¤t->files->file_lock); |
1905 | if (!error && f != filp && flock.l_type != F_UNLCK) { |
1906 | flock.l_type = F_UNLCK; |
1907 | goto again; |
1908 | } |
1909 | |
1910 | out: |
1911 | locks_free_lock(file_lock); |
1912 | return error; |
1913 | } |
1914 | |
1915 | #if BITS_PER_LONG == 32 |
1916 | /* Report the first existing lock that would conflict with l. |
1917 | * This implements the F_GETLK command of fcntl(). |
1918 | */ |
1919 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) |
1920 | { |
1921 | struct file_lock file_lock; |
1922 | struct flock64 flock; |
1923 | int error; |
1924 | |
1925 | error = -EFAULT; |
1926 | if (copy_from_user(&flock, l, sizeof(flock))) |
1927 | goto out; |
1928 | error = -EINVAL; |
1929 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
1930 | goto out; |
1931 | |
1932 | error = flock64_to_posix_lock(filp, &file_lock, &flock); |
1933 | if (error) |
1934 | goto out; |
1935 | |
1936 | error = vfs_test_lock(filp, &file_lock); |
1937 | if (error) |
1938 | goto out; |
1939 | |
1940 | flock.l_type = file_lock.fl_type; |
1941 | if (file_lock.fl_type != F_UNLCK) |
1942 | posix_lock_to_flock64(&flock, &file_lock); |
1943 | |
1944 | error = -EFAULT; |
1945 | if (!copy_to_user(l, &flock, sizeof(flock))) |
1946 | error = 0; |
1947 | |
1948 | out: |
1949 | return error; |
1950 | } |
1951 | |
1952 | /* Apply the lock described by l to an open file descriptor. |
1953 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
1954 | */ |
1955 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1956 | struct flock64 __user *l) |
1957 | { |
1958 | struct file_lock *file_lock = locks_alloc_lock(); |
1959 | struct flock64 flock; |
1960 | struct inode *inode; |
1961 | struct file *f; |
1962 | int error; |
1963 | |
1964 | if (file_lock == NULL) |
1965 | return -ENOLCK; |
1966 | |
1967 | /* |
1968 | * This might block, so we do it before checking the inode. |
1969 | */ |
1970 | error = -EFAULT; |
1971 | if (copy_from_user(&flock, l, sizeof(flock))) |
1972 | goto out; |
1973 | |
1974 | inode = filp->f_path.dentry->d_inode; |
1975 | |
1976 | /* Don't allow mandatory locks on files that may be memory mapped |
1977 | * and shared. |
1978 | */ |
1979 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1980 | error = -EAGAIN; |
1981 | goto out; |
1982 | } |
1983 | |
1984 | again: |
1985 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1986 | if (error) |
1987 | goto out; |
1988 | if (cmd == F_SETLKW64) { |
1989 | file_lock->fl_flags |= FL_SLEEP; |
1990 | } |
1991 | |
1992 | error = -EBADF; |
1993 | switch (flock.l_type) { |
1994 | case F_RDLCK: |
1995 | if (!(filp->f_mode & FMODE_READ)) |
1996 | goto out; |
1997 | break; |
1998 | case F_WRLCK: |
1999 | if (!(filp->f_mode & FMODE_WRITE)) |
2000 | goto out; |
2001 | break; |
2002 | case F_UNLCK: |
2003 | break; |
2004 | default: |
2005 | error = -EINVAL; |
2006 | goto out; |
2007 | } |
2008 | |
2009 | error = do_lock_file_wait(filp, cmd, file_lock); |
2010 | |
2011 | /* |
2012 | * Attempt to detect a close/fcntl race and recover by |
2013 | * releasing the lock that was just acquired. |
2014 | */ |
2015 | spin_lock(¤t->files->file_lock); |
2016 | f = fcheck(fd); |
2017 | spin_unlock(¤t->files->file_lock); |
2018 | if (!error && f != filp && flock.l_type != F_UNLCK) { |
2019 | flock.l_type = F_UNLCK; |
2020 | goto again; |
2021 | } |
2022 | |
2023 | out: |
2024 | locks_free_lock(file_lock); |
2025 | return error; |
2026 | } |
2027 | #endif /* BITS_PER_LONG == 32 */ |
2028 | |
2029 | /* |
2030 | * This function is called when the file is being removed |
2031 | * from the task's fd array. POSIX locks belonging to this task |
2032 | * are deleted at this time. |
2033 | */ |
2034 | void locks_remove_posix(struct file *filp, fl_owner_t owner) |
2035 | { |
2036 | struct file_lock lock; |
2037 | |
2038 | /* |
2039 | * If there are no locks held on this file, we don't need to call |
2040 | * posix_lock_file(). Another process could be setting a lock on this |
2041 | * file at the same time, but we wouldn't remove that lock anyway. |
2042 | */ |
2043 | if (!filp->f_path.dentry->d_inode->i_flock) |
2044 | return; |
2045 | |
2046 | lock.fl_type = F_UNLCK; |
2047 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
2048 | lock.fl_start = 0; |
2049 | lock.fl_end = OFFSET_MAX; |
2050 | lock.fl_owner = owner; |
2051 | lock.fl_pid = current->tgid; |
2052 | lock.fl_file = filp; |
2053 | lock.fl_ops = NULL; |
2054 | lock.fl_lmops = NULL; |
2055 | |
2056 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
2057 | |
2058 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2059 | lock.fl_ops->fl_release_private(&lock); |
2060 | } |
2061 | |
2062 | EXPORT_SYMBOL(locks_remove_posix); |
2063 | |
2064 | /* |
2065 | * This function is called on the last close of an open file. |
2066 | */ |
2067 | void locks_remove_flock(struct file *filp) |
2068 | { |
2069 | struct inode * inode = filp->f_path.dentry->d_inode; |
2070 | struct file_lock *fl; |
2071 | struct file_lock **before; |
2072 | |
2073 | if (!inode->i_flock) |
2074 | return; |
2075 | |
2076 | if (filp->f_op && filp->f_op->flock) { |
2077 | struct file_lock fl = { |
2078 | .fl_pid = current->tgid, |
2079 | .fl_file = filp, |
2080 | .fl_flags = FL_FLOCK, |
2081 | .fl_type = F_UNLCK, |
2082 | .fl_end = OFFSET_MAX, |
2083 | }; |
2084 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2085 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2086 | fl.fl_ops->fl_release_private(&fl); |
2087 | } |
2088 | |
2089 | lock_flocks(); |
2090 | before = &inode->i_flock; |
2091 | |
2092 | while ((fl = *before) != NULL) { |
2093 | if (fl->fl_file == filp) { |
2094 | if (IS_FLOCK(fl)) { |
2095 | locks_delete_lock(before); |
2096 | continue; |
2097 | } |
2098 | if (IS_LEASE(fl)) { |
2099 | lease_modify(before, F_UNLCK); |
2100 | continue; |
2101 | } |
2102 | /* What? */ |
2103 | BUG(); |
2104 | } |
2105 | before = &fl->fl_next; |
2106 | } |
2107 | unlock_flocks(); |
2108 | } |
2109 | |
2110 | /** |
2111 | * posix_unblock_lock - stop waiting for a file lock |
2112 | * @filp: how the file was opened |
2113 | * @waiter: the lock which was waiting |
2114 | * |
2115 | * lockd needs to block waiting for locks. |
2116 | */ |
2117 | int |
2118 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
2119 | { |
2120 | int status = 0; |
2121 | |
2122 | lock_flocks(); |
2123 | if (waiter->fl_next) |
2124 | __locks_delete_block(waiter); |
2125 | else |
2126 | status = -ENOENT; |
2127 | unlock_flocks(); |
2128 | return status; |
2129 | } |
2130 | |
2131 | EXPORT_SYMBOL(posix_unblock_lock); |
2132 | |
2133 | /** |
2134 | * vfs_cancel_lock - file byte range unblock lock |
2135 | * @filp: The file to apply the unblock to |
2136 | * @fl: The lock to be unblocked |
2137 | * |
2138 | * Used by lock managers to cancel blocked requests |
2139 | */ |
2140 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) |
2141 | { |
2142 | if (filp->f_op && filp->f_op->lock) |
2143 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2144 | return 0; |
2145 | } |
2146 | |
2147 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
2148 | |
2149 | #ifdef CONFIG_PROC_FS |
2150 | #include <linux/proc_fs.h> |
2151 | #include <linux/seq_file.h> |
2152 | |
2153 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
2154 | loff_t id, char *pfx) |
2155 | { |
2156 | struct inode *inode = NULL; |
2157 | unsigned int fl_pid; |
2158 | |
2159 | if (fl->fl_nspid) |
2160 | fl_pid = pid_vnr(fl->fl_nspid); |
2161 | else |
2162 | fl_pid = fl->fl_pid; |
2163 | |
2164 | if (fl->fl_file != NULL) |
2165 | inode = fl->fl_file->f_path.dentry->d_inode; |
2166 | |
2167 | seq_printf(f, "%lld:%s ", id, pfx); |
2168 | if (IS_POSIX(fl)) { |
2169 | seq_printf(f, "%6s %s ", |
2170 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2171 | (inode == NULL) ? "*NOINODE*" : |
2172 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
2173 | } else if (IS_FLOCK(fl)) { |
2174 | if (fl->fl_type & LOCK_MAND) { |
2175 | seq_printf(f, "FLOCK MSNFS "); |
2176 | } else { |
2177 | seq_printf(f, "FLOCK ADVISORY "); |
2178 | } |
2179 | } else if (IS_LEASE(fl)) { |
2180 | seq_printf(f, "LEASE "); |
2181 | if (lease_breaking(fl)) |
2182 | seq_printf(f, "BREAKING "); |
2183 | else if (fl->fl_file) |
2184 | seq_printf(f, "ACTIVE "); |
2185 | else |
2186 | seq_printf(f, "BREAKER "); |
2187 | } else { |
2188 | seq_printf(f, "UNKNOWN UNKNOWN "); |
2189 | } |
2190 | if (fl->fl_type & LOCK_MAND) { |
2191 | seq_printf(f, "%s ", |
2192 | (fl->fl_type & LOCK_READ) |
2193 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " |
2194 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); |
2195 | } else { |
2196 | seq_printf(f, "%s ", |
2197 | (lease_breaking(fl)) |
2198 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " |
2199 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); |
2200 | } |
2201 | if (inode) { |
2202 | #ifdef WE_CAN_BREAK_LSLK_NOW |
2203 | seq_printf(f, "%d %s:%ld ", fl_pid, |
2204 | inode->i_sb->s_id, inode->i_ino); |
2205 | #else |
2206 | /* userspace relies on this representation of dev_t ;-( */ |
2207 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
2208 | MAJOR(inode->i_sb->s_dev), |
2209 | MINOR(inode->i_sb->s_dev), inode->i_ino); |
2210 | #endif |
2211 | } else { |
2212 | seq_printf(f, "%d <none>:0 ", fl_pid); |
2213 | } |
2214 | if (IS_POSIX(fl)) { |
2215 | if (fl->fl_end == OFFSET_MAX) |
2216 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
2217 | else |
2218 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
2219 | } else { |
2220 | seq_printf(f, "0 EOF\n"); |
2221 | } |
2222 | } |
2223 | |
2224 | static int locks_show(struct seq_file *f, void *v) |
2225 | { |
2226 | struct file_lock *fl, *bfl; |
2227 | |
2228 | fl = list_entry(v, struct file_lock, fl_link); |
2229 | |
2230 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
2231 | |
2232 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
2233 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
2234 | |
2235 | return 0; |
2236 | } |
2237 | |
2238 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2239 | { |
2240 | loff_t *p = f->private; |
2241 | |
2242 | lock_flocks(); |
2243 | *p = (*pos + 1); |
2244 | return seq_list_start(&file_lock_list, *pos); |
2245 | } |
2246 | |
2247 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2248 | { |
2249 | loff_t *p = f->private; |
2250 | ++*p; |
2251 | return seq_list_next(v, &file_lock_list, pos); |
2252 | } |
2253 | |
2254 | static void locks_stop(struct seq_file *f, void *v) |
2255 | { |
2256 | unlock_flocks(); |
2257 | } |
2258 | |
2259 | static const struct seq_operations locks_seq_operations = { |
2260 | .start = locks_start, |
2261 | .next = locks_next, |
2262 | .stop = locks_stop, |
2263 | .show = locks_show, |
2264 | }; |
2265 | |
2266 | static int locks_open(struct inode *inode, struct file *filp) |
2267 | { |
2268 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
2269 | } |
2270 | |
2271 | static const struct file_operations proc_locks_operations = { |
2272 | .open = locks_open, |
2273 | .read = seq_read, |
2274 | .llseek = seq_lseek, |
2275 | .release = seq_release_private, |
2276 | }; |
2277 | |
2278 | static int __init proc_locks_init(void) |
2279 | { |
2280 | proc_create("locks", 0, NULL, &proc_locks_operations); |
2281 | return 0; |
2282 | } |
2283 | module_init(proc_locks_init); |
2284 | #endif |
2285 | |
2286 | /** |
2287 | * lock_may_read - checks that the region is free of locks |
2288 | * @inode: the inode that is being read |
2289 | * @start: the first byte to read |
2290 | * @len: the number of bytes to read |
2291 | * |
2292 | * Emulates Windows locking requirements. Whole-file |
2293 | * mandatory locks (share modes) can prohibit a read and |
2294 | * byte-range POSIX locks can prohibit a read if they overlap. |
2295 | * |
2296 | * N.B. this function is only ever called |
2297 | * from knfsd and ownership of locks is never checked. |
2298 | */ |
2299 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) |
2300 | { |
2301 | struct file_lock *fl; |
2302 | int result = 1; |
2303 | lock_flocks(); |
2304 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2305 | if (IS_POSIX(fl)) { |
2306 | if (fl->fl_type == F_RDLCK) |
2307 | continue; |
2308 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
2309 | continue; |
2310 | } else if (IS_FLOCK(fl)) { |
2311 | if (!(fl->fl_type & LOCK_MAND)) |
2312 | continue; |
2313 | if (fl->fl_type & LOCK_READ) |
2314 | continue; |
2315 | } else |
2316 | continue; |
2317 | result = 0; |
2318 | break; |
2319 | } |
2320 | unlock_flocks(); |
2321 | return result; |
2322 | } |
2323 | |
2324 | EXPORT_SYMBOL(lock_may_read); |
2325 | |
2326 | /** |
2327 | * lock_may_write - checks that the region is free of locks |
2328 | * @inode: the inode that is being written |
2329 | * @start: the first byte to write |
2330 | * @len: the number of bytes to write |
2331 | * |
2332 | * Emulates Windows locking requirements. Whole-file |
2333 | * mandatory locks (share modes) can prohibit a write and |
2334 | * byte-range POSIX locks can prohibit a write if they overlap. |
2335 | * |
2336 | * N.B. this function is only ever called |
2337 | * from knfsd and ownership of locks is never checked. |
2338 | */ |
2339 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) |
2340 | { |
2341 | struct file_lock *fl; |
2342 | int result = 1; |
2343 | lock_flocks(); |
2344 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2345 | if (IS_POSIX(fl)) { |
2346 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
2347 | continue; |
2348 | } else if (IS_FLOCK(fl)) { |
2349 | if (!(fl->fl_type & LOCK_MAND)) |
2350 | continue; |
2351 | if (fl->fl_type & LOCK_WRITE) |
2352 | continue; |
2353 | } else |
2354 | continue; |
2355 | result = 0; |
2356 | break; |
2357 | } |
2358 | unlock_flocks(); |
2359 | return result; |
2360 | } |
2361 | |
2362 | EXPORT_SYMBOL(lock_may_write); |
2363 | |
2364 | static int __init filelock_init(void) |
2365 | { |
2366 | filelock_cache = kmem_cache_create("file_lock_cache", |
2367 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2368 | |
2369 | return 0; |
2370 | } |
2371 | |
2372 | core_initcall(filelock_init); |
2373 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9