Root/
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/mandatory.txt' for details. |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference |
69 | * Manual, Section 2. |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. |
71 | * |
72 | * Tidied up block list handling. Added '/proc/locks' interface. |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. |
74 | * |
75 | * Fixed deadlock condition for pathological code that mixes calls to |
76 | * flock() and fcntl(). |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. |
78 | * |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to |
81 | * guarantee sensible behaviour in the case where file system modules might |
82 | * be compiled with different options than the kernel itself. |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
84 | * |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. |
88 | * |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that |
91 | * have already been freed. |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. |
93 | * |
94 | * Made the block list a circular list to minimise searching in the list. |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. |
96 | * |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory |
98 | * locking. |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. |
100 | * |
101 | * Some adaptations for NFS support. |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, |
103 | * |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. |
106 | * |
107 | * Use slab allocator instead of kmalloc/kfree. |
108 | * Use generic list implementation from <linux/list.h>. |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. |
111 | * |
112 | * Leases and LOCK_MAND |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. |
115 | */ |
116 | |
117 | #include <linux/capability.h> |
118 | #include <linux/file.h> |
119 | #include <linux/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/smp_lock.h> |
126 | #include <linux/syscalls.h> |
127 | #include <linux/time.h> |
128 | #include <linux/rcupdate.h> |
129 | #include <linux/pid_namespace.h> |
130 | |
131 | #include <asm/uaccess.h> |
132 | |
133 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) |
134 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) |
135 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) |
136 | |
137 | int leases_enable = 1; |
138 | int 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 | |
143 | static LIST_HEAD(file_lock_list); |
144 | static LIST_HEAD(blocked_list); |
145 | |
146 | static struct kmem_cache *filelock_cache __read_mostly; |
147 | |
148 | /* Allocate an empty lock structure. */ |
149 | static struct file_lock *locks_alloc_lock(void) |
150 | { |
151 | return kmem_cache_alloc(filelock_cache, GFP_KERNEL); |
152 | } |
153 | |
154 | 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 | EXPORT_SYMBOL_GPL(locks_release_private); |
169 | |
170 | /* Free a lock which is not in use. */ |
171 | static void locks_free_lock(struct file_lock *fl) |
172 | { |
173 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
174 | BUG_ON(!list_empty(&fl->fl_block)); |
175 | BUG_ON(!list_empty(&fl->fl_link)); |
176 | |
177 | locks_release_private(fl); |
178 | kmem_cache_free(filelock_cache, fl); |
179 | } |
180 | |
181 | void locks_init_lock(struct file_lock *fl) |
182 | { |
183 | INIT_LIST_HEAD(&fl->fl_link); |
184 | INIT_LIST_HEAD(&fl->fl_block); |
185 | init_waitqueue_head(&fl->fl_wait); |
186 | fl->fl_next = NULL; |
187 | fl->fl_fasync = NULL; |
188 | fl->fl_owner = NULL; |
189 | fl->fl_pid = 0; |
190 | fl->fl_nspid = NULL; |
191 | fl->fl_file = NULL; |
192 | fl->fl_flags = 0; |
193 | fl->fl_type = 0; |
194 | fl->fl_start = fl->fl_end = 0; |
195 | fl->fl_ops = NULL; |
196 | fl->fl_lmops = NULL; |
197 | } |
198 | |
199 | EXPORT_SYMBOL(locks_init_lock); |
200 | |
201 | /* |
202 | * Initialises the fields of the file lock which are invariant for |
203 | * free file_locks. |
204 | */ |
205 | static void init_once(void *foo) |
206 | { |
207 | struct file_lock *lock = (struct file_lock *) foo; |
208 | |
209 | locks_init_lock(lock); |
210 | } |
211 | |
212 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
213 | { |
214 | if (fl->fl_ops) { |
215 | if (fl->fl_ops->fl_copy_lock) |
216 | fl->fl_ops->fl_copy_lock(new, fl); |
217 | new->fl_ops = fl->fl_ops; |
218 | } |
219 | if (fl->fl_lmops) { |
220 | if (fl->fl_lmops->fl_copy_lock) |
221 | fl->fl_lmops->fl_copy_lock(new, fl); |
222 | new->fl_lmops = fl->fl_lmops; |
223 | } |
224 | } |
225 | |
226 | /* |
227 | * Initialize a new lock from an existing file_lock structure. |
228 | */ |
229 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
230 | { |
231 | new->fl_owner = fl->fl_owner; |
232 | new->fl_pid = fl->fl_pid; |
233 | new->fl_file = NULL; |
234 | new->fl_flags = fl->fl_flags; |
235 | new->fl_type = fl->fl_type; |
236 | new->fl_start = fl->fl_start; |
237 | new->fl_end = fl->fl_end; |
238 | new->fl_ops = NULL; |
239 | new->fl_lmops = NULL; |
240 | } |
241 | EXPORT_SYMBOL(__locks_copy_lock); |
242 | |
243 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) |
244 | { |
245 | locks_release_private(new); |
246 | |
247 | __locks_copy_lock(new, fl); |
248 | new->fl_file = fl->fl_file; |
249 | new->fl_ops = fl->fl_ops; |
250 | new->fl_lmops = fl->fl_lmops; |
251 | |
252 | locks_copy_private(new, fl); |
253 | } |
254 | |
255 | EXPORT_SYMBOL(locks_copy_lock); |
256 | |
257 | static inline int flock_translate_cmd(int cmd) { |
258 | if (cmd & LOCK_MAND) |
259 | return cmd & (LOCK_MAND | LOCK_RW); |
260 | switch (cmd) { |
261 | case LOCK_SH: |
262 | return F_RDLCK; |
263 | case LOCK_EX: |
264 | return F_WRLCK; |
265 | case LOCK_UN: |
266 | return F_UNLCK; |
267 | } |
268 | return -EINVAL; |
269 | } |
270 | |
271 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ |
272 | static int flock_make_lock(struct file *filp, struct file_lock **lock, |
273 | unsigned int cmd) |
274 | { |
275 | struct file_lock *fl; |
276 | int type = flock_translate_cmd(cmd); |
277 | if (type < 0) |
278 | return type; |
279 | |
280 | fl = locks_alloc_lock(); |
281 | if (fl == NULL) |
282 | return -ENOMEM; |
283 | |
284 | fl->fl_file = filp; |
285 | fl->fl_pid = current->tgid; |
286 | fl->fl_flags = FL_FLOCK; |
287 | fl->fl_type = type; |
288 | fl->fl_end = OFFSET_MAX; |
289 | |
290 | *lock = fl; |
291 | return 0; |
292 | } |
293 | |
294 | static int assign_type(struct file_lock *fl, int type) |
295 | { |
296 | switch (type) { |
297 | case F_RDLCK: |
298 | case F_WRLCK: |
299 | case F_UNLCK: |
300 | fl->fl_type = type; |
301 | break; |
302 | default: |
303 | return -EINVAL; |
304 | } |
305 | return 0; |
306 | } |
307 | |
308 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
309 | * style lock. |
310 | */ |
311 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, |
312 | struct flock *l) |
313 | { |
314 | off_t start, end; |
315 | |
316 | switch (l->l_whence) { |
317 | case SEEK_SET: |
318 | start = 0; |
319 | break; |
320 | case SEEK_CUR: |
321 | start = filp->f_pos; |
322 | break; |
323 | case SEEK_END: |
324 | start = i_size_read(filp->f_path.dentry->d_inode); |
325 | break; |
326 | default: |
327 | return -EINVAL; |
328 | } |
329 | |
330 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; |
331 | POSIX-2001 defines it. */ |
332 | start += l->l_start; |
333 | if (start < 0) |
334 | return -EINVAL; |
335 | fl->fl_end = OFFSET_MAX; |
336 | if (l->l_len > 0) { |
337 | end = start + l->l_len - 1; |
338 | fl->fl_end = end; |
339 | } else if (l->l_len < 0) { |
340 | end = start - 1; |
341 | fl->fl_end = end; |
342 | start += l->l_len; |
343 | if (start < 0) |
344 | return -EINVAL; |
345 | } |
346 | fl->fl_start = start; /* we record the absolute position */ |
347 | if (fl->fl_end < fl->fl_start) |
348 | return -EOVERFLOW; |
349 | |
350 | fl->fl_owner = current->files; |
351 | fl->fl_pid = current->tgid; |
352 | fl->fl_file = filp; |
353 | fl->fl_flags = FL_POSIX; |
354 | fl->fl_ops = NULL; |
355 | fl->fl_lmops = NULL; |
356 | |
357 | return assign_type(fl, l->l_type); |
358 | } |
359 | |
360 | #if BITS_PER_LONG == 32 |
361 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
362 | struct flock64 *l) |
363 | { |
364 | loff_t start; |
365 | |
366 | switch (l->l_whence) { |
367 | case SEEK_SET: |
368 | start = 0; |
369 | break; |
370 | case SEEK_CUR: |
371 | start = filp->f_pos; |
372 | break; |
373 | case SEEK_END: |
374 | start = i_size_read(filp->f_path.dentry->d_inode); |
375 | break; |
376 | default: |
377 | return -EINVAL; |
378 | } |
379 | |
380 | start += l->l_start; |
381 | if (start < 0) |
382 | return -EINVAL; |
383 | fl->fl_end = OFFSET_MAX; |
384 | if (l->l_len > 0) { |
385 | fl->fl_end = start + l->l_len - 1; |
386 | } else if (l->l_len < 0) { |
387 | fl->fl_end = start - 1; |
388 | start += l->l_len; |
389 | if (start < 0) |
390 | return -EINVAL; |
391 | } |
392 | fl->fl_start = start; /* we record the absolute position */ |
393 | if (fl->fl_end < fl->fl_start) |
394 | return -EOVERFLOW; |
395 | |
396 | fl->fl_owner = current->files; |
397 | fl->fl_pid = current->tgid; |
398 | fl->fl_file = filp; |
399 | fl->fl_flags = FL_POSIX; |
400 | fl->fl_ops = NULL; |
401 | fl->fl_lmops = NULL; |
402 | |
403 | switch (l->l_type) { |
404 | case F_RDLCK: |
405 | case F_WRLCK: |
406 | case F_UNLCK: |
407 | fl->fl_type = l->l_type; |
408 | break; |
409 | default: |
410 | return -EINVAL; |
411 | } |
412 | |
413 | return (0); |
414 | } |
415 | #endif |
416 | |
417 | /* default lease lock manager operations */ |
418 | static void lease_break_callback(struct file_lock *fl) |
419 | { |
420 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); |
421 | } |
422 | |
423 | static void lease_release_private_callback(struct file_lock *fl) |
424 | { |
425 | if (!fl->fl_file) |
426 | return; |
427 | |
428 | f_delown(fl->fl_file); |
429 | fl->fl_file->f_owner.signum = 0; |
430 | } |
431 | |
432 | static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try) |
433 | { |
434 | return fl->fl_file == try->fl_file; |
435 | } |
436 | |
437 | static const struct lock_manager_operations lease_manager_ops = { |
438 | .fl_break = lease_break_callback, |
439 | .fl_release_private = lease_release_private_callback, |
440 | .fl_mylease = lease_mylease_callback, |
441 | .fl_change = lease_modify, |
442 | }; |
443 | |
444 | /* |
445 | * Initialize a lease, use the default lock manager operations |
446 | */ |
447 | static int lease_init(struct file *filp, int type, struct file_lock *fl) |
448 | { |
449 | if (assign_type(fl, type) != 0) |
450 | return -EINVAL; |
451 | |
452 | fl->fl_owner = current->files; |
453 | fl->fl_pid = current->tgid; |
454 | |
455 | fl->fl_file = filp; |
456 | fl->fl_flags = FL_LEASE; |
457 | fl->fl_start = 0; |
458 | fl->fl_end = OFFSET_MAX; |
459 | fl->fl_ops = NULL; |
460 | fl->fl_lmops = &lease_manager_ops; |
461 | return 0; |
462 | } |
463 | |
464 | /* Allocate a file_lock initialised to this type of lease */ |
465 | static struct file_lock *lease_alloc(struct file *filp, int type) |
466 | { |
467 | struct file_lock *fl = locks_alloc_lock(); |
468 | int error = -ENOMEM; |
469 | |
470 | if (fl == NULL) |
471 | return ERR_PTR(error); |
472 | |
473 | error = lease_init(filp, type, fl); |
474 | if (error) { |
475 | locks_free_lock(fl); |
476 | return ERR_PTR(error); |
477 | } |
478 | return fl; |
479 | } |
480 | |
481 | /* Check if two locks overlap each other. |
482 | */ |
483 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) |
484 | { |
485 | return ((fl1->fl_end >= fl2->fl_start) && |
486 | (fl2->fl_end >= fl1->fl_start)); |
487 | } |
488 | |
489 | /* |
490 | * Check whether two locks have the same owner. |
491 | */ |
492 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
493 | { |
494 | if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner) |
495 | return fl2->fl_lmops == fl1->fl_lmops && |
496 | fl1->fl_lmops->fl_compare_owner(fl1, fl2); |
497 | return fl1->fl_owner == fl2->fl_owner; |
498 | } |
499 | |
500 | /* Remove waiter from blocker's block list. |
501 | * When blocker ends up pointing to itself then the list is empty. |
502 | */ |
503 | static void __locks_delete_block(struct file_lock *waiter) |
504 | { |
505 | list_del_init(&waiter->fl_block); |
506 | list_del_init(&waiter->fl_link); |
507 | waiter->fl_next = NULL; |
508 | } |
509 | |
510 | /* |
511 | */ |
512 | static void locks_delete_block(struct file_lock *waiter) |
513 | { |
514 | lock_kernel(); |
515 | __locks_delete_block(waiter); |
516 | unlock_kernel(); |
517 | } |
518 | |
519 | /* Insert waiter into blocker's block list. |
520 | * We use a circular list so that processes can be easily woken up in |
521 | * the order they blocked. The documentation doesn't require this but |
522 | * it seems like the reasonable thing to do. |
523 | */ |
524 | static void locks_insert_block(struct file_lock *blocker, |
525 | struct file_lock *waiter) |
526 | { |
527 | BUG_ON(!list_empty(&waiter->fl_block)); |
528 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
529 | waiter->fl_next = blocker; |
530 | if (IS_POSIX(blocker)) |
531 | list_add(&waiter->fl_link, &blocked_list); |
532 | } |
533 | |
534 | /* Wake up processes blocked waiting for blocker. |
535 | * If told to wait then schedule the processes until the block list |
536 | * is empty, otherwise empty the block list ourselves. |
537 | */ |
538 | static void locks_wake_up_blocks(struct file_lock *blocker) |
539 | { |
540 | while (!list_empty(&blocker->fl_block)) { |
541 | struct file_lock *waiter; |
542 | |
543 | waiter = list_first_entry(&blocker->fl_block, |
544 | struct file_lock, fl_block); |
545 | __locks_delete_block(waiter); |
546 | if (waiter->fl_lmops && waiter->fl_lmops->fl_notify) |
547 | waiter->fl_lmops->fl_notify(waiter); |
548 | else |
549 | wake_up(&waiter->fl_wait); |
550 | } |
551 | } |
552 | |
553 | /* Insert file lock fl into an inode's lock list at the position indicated |
554 | * by pos. At the same time add the lock to the global file lock list. |
555 | */ |
556 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) |
557 | { |
558 | list_add(&fl->fl_link, &file_lock_list); |
559 | |
560 | fl->fl_nspid = get_pid(task_tgid(current)); |
561 | |
562 | /* insert into file's list */ |
563 | fl->fl_next = *pos; |
564 | *pos = fl; |
565 | } |
566 | |
567 | /* |
568 | * Delete a lock and then free it. |
569 | * Wake up processes that are blocked waiting for this lock, |
570 | * notify the FS that the lock has been cleared and |
571 | * finally free the lock. |
572 | */ |
573 | static void locks_delete_lock(struct file_lock **thisfl_p) |
574 | { |
575 | struct file_lock *fl = *thisfl_p; |
576 | |
577 | *thisfl_p = fl->fl_next; |
578 | fl->fl_next = NULL; |
579 | list_del_init(&fl->fl_link); |
580 | |
581 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
582 | if (fl->fl_fasync != NULL) { |
583 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); |
584 | fl->fl_fasync = NULL; |
585 | } |
586 | |
587 | if (fl->fl_nspid) { |
588 | put_pid(fl->fl_nspid); |
589 | fl->fl_nspid = NULL; |
590 | } |
591 | |
592 | locks_wake_up_blocks(fl); |
593 | locks_free_lock(fl); |
594 | } |
595 | |
596 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality |
597 | * checks for shared/exclusive status of overlapping locks. |
598 | */ |
599 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
600 | { |
601 | if (sys_fl->fl_type == F_WRLCK) |
602 | return 1; |
603 | if (caller_fl->fl_type == F_WRLCK) |
604 | return 1; |
605 | return 0; |
606 | } |
607 | |
608 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific |
609 | * checking before calling the locks_conflict(). |
610 | */ |
611 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
612 | { |
613 | /* POSIX locks owned by the same process do not conflict with |
614 | * each other. |
615 | */ |
616 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) |
617 | return (0); |
618 | |
619 | /* Check whether they overlap */ |
620 | if (!locks_overlap(caller_fl, sys_fl)) |
621 | return 0; |
622 | |
623 | return (locks_conflict(caller_fl, sys_fl)); |
624 | } |
625 | |
626 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific |
627 | * checking before calling the locks_conflict(). |
628 | */ |
629 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) |
630 | { |
631 | /* FLOCK locks referring to the same filp do not conflict with |
632 | * each other. |
633 | */ |
634 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) |
635 | return (0); |
636 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) |
637 | return 0; |
638 | |
639 | return (locks_conflict(caller_fl, sys_fl)); |
640 | } |
641 | |
642 | void |
643 | posix_test_lock(struct file *filp, struct file_lock *fl) |
644 | { |
645 | struct file_lock *cfl; |
646 | |
647 | lock_kernel(); |
648 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
649 | if (!IS_POSIX(cfl)) |
650 | continue; |
651 | if (posix_locks_conflict(fl, cfl)) |
652 | break; |
653 | } |
654 | if (cfl) { |
655 | __locks_copy_lock(fl, cfl); |
656 | if (cfl->fl_nspid) |
657 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
658 | } else |
659 | fl->fl_type = F_UNLCK; |
660 | unlock_kernel(); |
661 | return; |
662 | } |
663 | EXPORT_SYMBOL(posix_test_lock); |
664 | |
665 | /* |
666 | * Deadlock detection: |
667 | * |
668 | * We attempt to detect deadlocks that are due purely to posix file |
669 | * locks. |
670 | * |
671 | * We assume that a task can be waiting for at most one lock at a time. |
672 | * So for any acquired lock, the process holding that lock may be |
673 | * waiting on at most one other lock. That lock in turns may be held by |
674 | * someone waiting for at most one other lock. Given a requested lock |
675 | * caller_fl which is about to wait for a conflicting lock block_fl, we |
676 | * follow this chain of waiters to ensure we are not about to create a |
677 | * cycle. |
678 | * |
679 | * Since we do this before we ever put a process to sleep on a lock, we |
680 | * are ensured that there is never a cycle; that is what guarantees that |
681 | * the while() loop in posix_locks_deadlock() eventually completes. |
682 | * |
683 | * Note: the above assumption may not be true when handling lock |
684 | * requests from a broken NFS client. It may also fail in the presence |
685 | * of tasks (such as posix threads) sharing the same open file table. |
686 | * |
687 | * To handle those cases, we just bail out after a few iterations. |
688 | */ |
689 | |
690 | #define MAX_DEADLK_ITERATIONS 10 |
691 | |
692 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
693 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) |
694 | { |
695 | struct file_lock *fl; |
696 | |
697 | list_for_each_entry(fl, &blocked_list, fl_link) { |
698 | if (posix_same_owner(fl, block_fl)) |
699 | return fl->fl_next; |
700 | } |
701 | return NULL; |
702 | } |
703 | |
704 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
705 | struct file_lock *block_fl) |
706 | { |
707 | int i = 0; |
708 | |
709 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
710 | if (i++ > MAX_DEADLK_ITERATIONS) |
711 | return 0; |
712 | if (posix_same_owner(caller_fl, block_fl)) |
713 | return 1; |
714 | } |
715 | return 0; |
716 | } |
717 | |
718 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
719 | * after any leases, but before any posix locks. |
720 | * |
721 | * Note that if called with an FL_EXISTS argument, the caller may determine |
722 | * whether or not a lock was successfully freed by testing the return |
723 | * value for -ENOENT. |
724 | */ |
725 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
726 | { |
727 | struct file_lock *new_fl = NULL; |
728 | struct file_lock **before; |
729 | struct inode * inode = filp->f_path.dentry->d_inode; |
730 | int error = 0; |
731 | int found = 0; |
732 | |
733 | lock_kernel(); |
734 | if (request->fl_flags & FL_ACCESS) |
735 | goto find_conflict; |
736 | |
737 | if (request->fl_type != F_UNLCK) { |
738 | error = -ENOMEM; |
739 | new_fl = locks_alloc_lock(); |
740 | if (new_fl == NULL) |
741 | goto out; |
742 | error = 0; |
743 | } |
744 | |
745 | for_each_lock(inode, before) { |
746 | struct file_lock *fl = *before; |
747 | if (IS_POSIX(fl)) |
748 | break; |
749 | if (IS_LEASE(fl)) |
750 | continue; |
751 | if (filp != fl->fl_file) |
752 | continue; |
753 | if (request->fl_type == fl->fl_type) |
754 | goto out; |
755 | found = 1; |
756 | locks_delete_lock(before); |
757 | break; |
758 | } |
759 | |
760 | if (request->fl_type == F_UNLCK) { |
761 | if ((request->fl_flags & FL_EXISTS) && !found) |
762 | error = -ENOENT; |
763 | goto out; |
764 | } |
765 | |
766 | /* |
767 | * If a higher-priority process was blocked on the old file lock, |
768 | * give it the opportunity to lock the file. |
769 | */ |
770 | if (found) |
771 | cond_resched(); |
772 | |
773 | find_conflict: |
774 | for_each_lock(inode, before) { |
775 | struct file_lock *fl = *before; |
776 | if (IS_POSIX(fl)) |
777 | break; |
778 | if (IS_LEASE(fl)) |
779 | continue; |
780 | if (!flock_locks_conflict(request, fl)) |
781 | continue; |
782 | error = -EAGAIN; |
783 | if (!(request->fl_flags & FL_SLEEP)) |
784 | goto out; |
785 | error = FILE_LOCK_DEFERRED; |
786 | locks_insert_block(fl, request); |
787 | goto out; |
788 | } |
789 | if (request->fl_flags & FL_ACCESS) |
790 | goto out; |
791 | locks_copy_lock(new_fl, request); |
792 | locks_insert_lock(before, new_fl); |
793 | new_fl = NULL; |
794 | error = 0; |
795 | |
796 | out: |
797 | unlock_kernel(); |
798 | if (new_fl) |
799 | locks_free_lock(new_fl); |
800 | return error; |
801 | } |
802 | |
803 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
804 | { |
805 | struct file_lock *fl; |
806 | struct file_lock *new_fl = NULL; |
807 | struct file_lock *new_fl2 = NULL; |
808 | struct file_lock *left = NULL; |
809 | struct file_lock *right = NULL; |
810 | struct file_lock **before; |
811 | int error, added = 0; |
812 | |
813 | /* |
814 | * We may need two file_lock structures for this operation, |
815 | * so we get them in advance to avoid races. |
816 | * |
817 | * In some cases we can be sure, that no new locks will be needed |
818 | */ |
819 | if (!(request->fl_flags & FL_ACCESS) && |
820 | (request->fl_type != F_UNLCK || |
821 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { |
822 | new_fl = locks_alloc_lock(); |
823 | new_fl2 = locks_alloc_lock(); |
824 | } |
825 | |
826 | lock_kernel(); |
827 | if (request->fl_type != F_UNLCK) { |
828 | for_each_lock(inode, before) { |
829 | fl = *before; |
830 | if (!IS_POSIX(fl)) |
831 | continue; |
832 | if (!posix_locks_conflict(request, fl)) |
833 | continue; |
834 | if (conflock) |
835 | __locks_copy_lock(conflock, fl); |
836 | error = -EAGAIN; |
837 | if (!(request->fl_flags & FL_SLEEP)) |
838 | goto out; |
839 | error = -EDEADLK; |
840 | if (posix_locks_deadlock(request, fl)) |
841 | goto out; |
842 | error = FILE_LOCK_DEFERRED; |
843 | locks_insert_block(fl, request); |
844 | goto out; |
845 | } |
846 | } |
847 | |
848 | /* If we're just looking for a conflict, we're done. */ |
849 | error = 0; |
850 | if (request->fl_flags & FL_ACCESS) |
851 | goto out; |
852 | |
853 | /* |
854 | * Find the first old lock with the same owner as the new lock. |
855 | */ |
856 | |
857 | before = &inode->i_flock; |
858 | |
859 | /* First skip locks owned by other processes. */ |
860 | while ((fl = *before) && (!IS_POSIX(fl) || |
861 | !posix_same_owner(request, fl))) { |
862 | before = &fl->fl_next; |
863 | } |
864 | |
865 | /* Process locks with this owner. */ |
866 | while ((fl = *before) && posix_same_owner(request, fl)) { |
867 | /* Detect adjacent or overlapping regions (if same lock type) |
868 | */ |
869 | if (request->fl_type == fl->fl_type) { |
870 | /* In all comparisons of start vs end, use |
871 | * "start - 1" rather than "end + 1". If end |
872 | * is OFFSET_MAX, end + 1 will become negative. |
873 | */ |
874 | if (fl->fl_end < request->fl_start - 1) |
875 | goto next_lock; |
876 | /* If the next lock in the list has entirely bigger |
877 | * addresses than the new one, insert the lock here. |
878 | */ |
879 | if (fl->fl_start - 1 > request->fl_end) |
880 | break; |
881 | |
882 | /* If we come here, the new and old lock are of the |
883 | * same type and adjacent or overlapping. Make one |
884 | * lock yielding from the lower start address of both |
885 | * locks to the higher end address. |
886 | */ |
887 | if (fl->fl_start > request->fl_start) |
888 | fl->fl_start = request->fl_start; |
889 | else |
890 | request->fl_start = fl->fl_start; |
891 | if (fl->fl_end < request->fl_end) |
892 | fl->fl_end = request->fl_end; |
893 | else |
894 | request->fl_end = fl->fl_end; |
895 | if (added) { |
896 | locks_delete_lock(before); |
897 | continue; |
898 | } |
899 | request = fl; |
900 | added = 1; |
901 | } |
902 | else { |
903 | /* Processing for different lock types is a bit |
904 | * more complex. |
905 | */ |
906 | if (fl->fl_end < request->fl_start) |
907 | goto next_lock; |
908 | if (fl->fl_start > request->fl_end) |
909 | break; |
910 | if (request->fl_type == F_UNLCK) |
911 | added = 1; |
912 | if (fl->fl_start < request->fl_start) |
913 | left = fl; |
914 | /* If the next lock in the list has a higher end |
915 | * address than the new one, insert the new one here. |
916 | */ |
917 | if (fl->fl_end > request->fl_end) { |
918 | right = fl; |
919 | break; |
920 | } |
921 | if (fl->fl_start >= request->fl_start) { |
922 | /* The new lock completely replaces an old |
923 | * one (This may happen several times). |
924 | */ |
925 | if (added) { |
926 | locks_delete_lock(before); |
927 | continue; |
928 | } |
929 | /* Replace the old lock with the new one. |
930 | * Wake up anybody waiting for the old one, |
931 | * as the change in lock type might satisfy |
932 | * their needs. |
933 | */ |
934 | locks_wake_up_blocks(fl); |
935 | fl->fl_start = request->fl_start; |
936 | fl->fl_end = request->fl_end; |
937 | fl->fl_type = request->fl_type; |
938 | locks_release_private(fl); |
939 | locks_copy_private(fl, request); |
940 | request = fl; |
941 | added = 1; |
942 | } |
943 | } |
944 | /* Go on to next lock. |
945 | */ |
946 | next_lock: |
947 | before = &fl->fl_next; |
948 | } |
949 | |
950 | /* |
951 | * The above code only modifies existing locks in case of |
952 | * merging or replacing. If new lock(s) need to be inserted |
953 | * all modifications are done bellow this, so it's safe yet to |
954 | * bail out. |
955 | */ |
956 | error = -ENOLCK; /* "no luck" */ |
957 | if (right && left == right && !new_fl2) |
958 | goto out; |
959 | |
960 | error = 0; |
961 | if (!added) { |
962 | if (request->fl_type == F_UNLCK) { |
963 | if (request->fl_flags & FL_EXISTS) |
964 | error = -ENOENT; |
965 | goto out; |
966 | } |
967 | |
968 | if (!new_fl) { |
969 | error = -ENOLCK; |
970 | goto out; |
971 | } |
972 | locks_copy_lock(new_fl, request); |
973 | locks_insert_lock(before, new_fl); |
974 | new_fl = NULL; |
975 | } |
976 | if (right) { |
977 | if (left == right) { |
978 | /* The new lock breaks the old one in two pieces, |
979 | * so we have to use the second new lock. |
980 | */ |
981 | left = new_fl2; |
982 | new_fl2 = NULL; |
983 | locks_copy_lock(left, right); |
984 | locks_insert_lock(before, left); |
985 | } |
986 | right->fl_start = request->fl_end + 1; |
987 | locks_wake_up_blocks(right); |
988 | } |
989 | if (left) { |
990 | left->fl_end = request->fl_start - 1; |
991 | locks_wake_up_blocks(left); |
992 | } |
993 | out: |
994 | unlock_kernel(); |
995 | /* |
996 | * Free any unused locks. |
997 | */ |
998 | if (new_fl) |
999 | locks_free_lock(new_fl); |
1000 | if (new_fl2) |
1001 | locks_free_lock(new_fl2); |
1002 | return error; |
1003 | } |
1004 | |
1005 | /** |
1006 | * posix_lock_file - Apply a POSIX-style lock to a file |
1007 | * @filp: The file to apply the lock to |
1008 | * @fl: The lock to be applied |
1009 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1010 | * |
1011 | * Add a POSIX style lock to a file. |
1012 | * We merge adjacent & overlapping locks whenever possible. |
1013 | * POSIX locks are sorted by owner task, then by starting address |
1014 | * |
1015 | * Note that if called with an FL_EXISTS argument, the caller may determine |
1016 | * whether or not a lock was successfully freed by testing the return |
1017 | * value for -ENOENT. |
1018 | */ |
1019 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
1020 | struct file_lock *conflock) |
1021 | { |
1022 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
1023 | } |
1024 | EXPORT_SYMBOL(posix_lock_file); |
1025 | |
1026 | /** |
1027 | * posix_lock_file_wait - Apply a POSIX-style lock to a file |
1028 | * @filp: The file to apply the lock to |
1029 | * @fl: The lock to be applied |
1030 | * |
1031 | * Add a POSIX style lock to a file. |
1032 | * We merge adjacent & overlapping locks whenever possible. |
1033 | * POSIX locks are sorted by owner task, then by starting address |
1034 | */ |
1035 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) |
1036 | { |
1037 | int error; |
1038 | might_sleep (); |
1039 | for (;;) { |
1040 | error = posix_lock_file(filp, fl, NULL); |
1041 | if (error != FILE_LOCK_DEFERRED) |
1042 | break; |
1043 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1044 | if (!error) |
1045 | continue; |
1046 | |
1047 | locks_delete_block(fl); |
1048 | break; |
1049 | } |
1050 | return error; |
1051 | } |
1052 | EXPORT_SYMBOL(posix_lock_file_wait); |
1053 | |
1054 | /** |
1055 | * locks_mandatory_locked - Check for an active lock |
1056 | * @inode: the file to check |
1057 | * |
1058 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
1059 | * This function is called from locks_verify_locked() only. |
1060 | */ |
1061 | int locks_mandatory_locked(struct inode *inode) |
1062 | { |
1063 | fl_owner_t owner = current->files; |
1064 | struct file_lock *fl; |
1065 | |
1066 | /* |
1067 | * Search the lock list for this inode for any POSIX locks. |
1068 | */ |
1069 | lock_kernel(); |
1070 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1071 | if (!IS_POSIX(fl)) |
1072 | continue; |
1073 | if (fl->fl_owner != owner) |
1074 | break; |
1075 | } |
1076 | unlock_kernel(); |
1077 | return fl ? -EAGAIN : 0; |
1078 | } |
1079 | |
1080 | /** |
1081 | * locks_mandatory_area - Check for a conflicting lock |
1082 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ |
1083 | * for shared |
1084 | * @inode: the file to check |
1085 | * @filp: how the file was opened (if it was) |
1086 | * @offset: start of area to check |
1087 | * @count: length of area to check |
1088 | * |
1089 | * Searches the inode's list of locks to find any POSIX locks which conflict. |
1090 | * This function is called from rw_verify_area() and |
1091 | * locks_verify_truncate(). |
1092 | */ |
1093 | int locks_mandatory_area(int read_write, struct inode *inode, |
1094 | struct file *filp, loff_t offset, |
1095 | size_t count) |
1096 | { |
1097 | struct file_lock fl; |
1098 | int error; |
1099 | |
1100 | locks_init_lock(&fl); |
1101 | fl.fl_owner = current->files; |
1102 | fl.fl_pid = current->tgid; |
1103 | fl.fl_file = filp; |
1104 | fl.fl_flags = FL_POSIX | FL_ACCESS; |
1105 | if (filp && !(filp->f_flags & O_NONBLOCK)) |
1106 | fl.fl_flags |= FL_SLEEP; |
1107 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1108 | fl.fl_start = offset; |
1109 | fl.fl_end = offset + count - 1; |
1110 | |
1111 | for (;;) { |
1112 | error = __posix_lock_file(inode, &fl, NULL); |
1113 | if (error != FILE_LOCK_DEFERRED) |
1114 | break; |
1115 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); |
1116 | if (!error) { |
1117 | /* |
1118 | * If we've been sleeping someone might have |
1119 | * changed the permissions behind our back. |
1120 | */ |
1121 | if (__mandatory_lock(inode)) |
1122 | continue; |
1123 | } |
1124 | |
1125 | locks_delete_block(&fl); |
1126 | break; |
1127 | } |
1128 | |
1129 | return error; |
1130 | } |
1131 | |
1132 | EXPORT_SYMBOL(locks_mandatory_area); |
1133 | |
1134 | /* We already had a lease on this file; just change its type */ |
1135 | int lease_modify(struct file_lock **before, int arg) |
1136 | { |
1137 | struct file_lock *fl = *before; |
1138 | int error = assign_type(fl, arg); |
1139 | |
1140 | if (error) |
1141 | return error; |
1142 | locks_wake_up_blocks(fl); |
1143 | if (arg == F_UNLCK) |
1144 | locks_delete_lock(before); |
1145 | return 0; |
1146 | } |
1147 | |
1148 | EXPORT_SYMBOL(lease_modify); |
1149 | |
1150 | static void time_out_leases(struct inode *inode) |
1151 | { |
1152 | struct file_lock **before; |
1153 | struct file_lock *fl; |
1154 | |
1155 | before = &inode->i_flock; |
1156 | while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) { |
1157 | if ((fl->fl_break_time == 0) |
1158 | || time_before(jiffies, fl->fl_break_time)) { |
1159 | before = &fl->fl_next; |
1160 | continue; |
1161 | } |
1162 | lease_modify(before, fl->fl_type & ~F_INPROGRESS); |
1163 | if (fl == *before) /* lease_modify may have freed fl */ |
1164 | before = &fl->fl_next; |
1165 | } |
1166 | } |
1167 | |
1168 | /** |
1169 | * __break_lease - revoke all outstanding leases on file |
1170 | * @inode: the inode of the file to return |
1171 | * @mode: the open mode (read or write) |
1172 | * |
1173 | * break_lease (inlined for speed) has checked there already is at least |
1174 | * some kind of lock (maybe a lease) on this file. Leases are broken on |
1175 | * a call to open() or truncate(). This function can sleep unless you |
1176 | * specified %O_NONBLOCK to your open(). |
1177 | */ |
1178 | int __break_lease(struct inode *inode, unsigned int mode) |
1179 | { |
1180 | int error = 0, future; |
1181 | struct file_lock *new_fl, *flock; |
1182 | struct file_lock *fl; |
1183 | unsigned long break_time; |
1184 | int i_have_this_lease = 0; |
1185 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1186 | |
1187 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
1188 | |
1189 | lock_kernel(); |
1190 | |
1191 | time_out_leases(inode); |
1192 | |
1193 | flock = inode->i_flock; |
1194 | if ((flock == NULL) || !IS_LEASE(flock)) |
1195 | goto out; |
1196 | |
1197 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
1198 | if (fl->fl_owner == current->files) |
1199 | i_have_this_lease = 1; |
1200 | |
1201 | if (want_write) { |
1202 | /* If we want write access, we have to revoke any lease. */ |
1203 | future = F_UNLCK | F_INPROGRESS; |
1204 | } else if (flock->fl_type & F_INPROGRESS) { |
1205 | /* If the lease is already being broken, we just leave it */ |
1206 | future = flock->fl_type; |
1207 | } else if (flock->fl_type & F_WRLCK) { |
1208 | /* Downgrade the exclusive lease to a read-only lease. */ |
1209 | future = F_RDLCK | F_INPROGRESS; |
1210 | } else { |
1211 | /* the existing lease was read-only, so we can read too. */ |
1212 | goto out; |
1213 | } |
1214 | |
1215 | if (IS_ERR(new_fl) && !i_have_this_lease |
1216 | && ((mode & O_NONBLOCK) == 0)) { |
1217 | error = PTR_ERR(new_fl); |
1218 | goto out; |
1219 | } |
1220 | |
1221 | break_time = 0; |
1222 | if (lease_break_time > 0) { |
1223 | break_time = jiffies + lease_break_time * HZ; |
1224 | if (break_time == 0) |
1225 | break_time++; /* so that 0 means no break time */ |
1226 | } |
1227 | |
1228 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1229 | if (fl->fl_type != future) { |
1230 | fl->fl_type = future; |
1231 | fl->fl_break_time = break_time; |
1232 | /* lease must have lmops break callback */ |
1233 | fl->fl_lmops->fl_break(fl); |
1234 | } |
1235 | } |
1236 | |
1237 | if (i_have_this_lease || (mode & O_NONBLOCK)) { |
1238 | error = -EWOULDBLOCK; |
1239 | goto out; |
1240 | } |
1241 | |
1242 | restart: |
1243 | break_time = flock->fl_break_time; |
1244 | if (break_time != 0) { |
1245 | break_time -= jiffies; |
1246 | if (break_time == 0) |
1247 | break_time++; |
1248 | } |
1249 | locks_insert_block(flock, new_fl); |
1250 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1251 | !new_fl->fl_next, break_time); |
1252 | __locks_delete_block(new_fl); |
1253 | if (error >= 0) { |
1254 | if (error == 0) |
1255 | time_out_leases(inode); |
1256 | /* Wait for the next lease that has not been broken yet */ |
1257 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1258 | flock = flock->fl_next) { |
1259 | if (flock->fl_type & F_INPROGRESS) |
1260 | goto restart; |
1261 | } |
1262 | error = 0; |
1263 | } |
1264 | |
1265 | out: |
1266 | unlock_kernel(); |
1267 | if (!IS_ERR(new_fl)) |
1268 | locks_free_lock(new_fl); |
1269 | return error; |
1270 | } |
1271 | |
1272 | EXPORT_SYMBOL(__break_lease); |
1273 | |
1274 | /** |
1275 | * lease_get_mtime - get the last modified time of an inode |
1276 | * @inode: the inode |
1277 | * @time: pointer to a timespec which will contain the last modified time |
1278 | * |
1279 | * This is to force NFS clients to flush their caches for files with |
1280 | * exclusive leases. The justification is that if someone has an |
1281 | * exclusive lease, then they could be modifying it. |
1282 | */ |
1283 | void lease_get_mtime(struct inode *inode, struct timespec *time) |
1284 | { |
1285 | struct file_lock *flock = inode->i_flock; |
1286 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) |
1287 | *time = current_fs_time(inode->i_sb); |
1288 | else |
1289 | *time = inode->i_mtime; |
1290 | } |
1291 | |
1292 | EXPORT_SYMBOL(lease_get_mtime); |
1293 | |
1294 | /** |
1295 | * fcntl_getlease - Enquire what lease is currently active |
1296 | * @filp: the file |
1297 | * |
1298 | * The value returned by this function will be one of |
1299 | * (if no lease break is pending): |
1300 | * |
1301 | * %F_RDLCK to indicate a shared lease is held. |
1302 | * |
1303 | * %F_WRLCK to indicate an exclusive lease is held. |
1304 | * |
1305 | * %F_UNLCK to indicate no lease is held. |
1306 | * |
1307 | * (if a lease break is pending): |
1308 | * |
1309 | * %F_RDLCK to indicate an exclusive lease needs to be |
1310 | * changed to a shared lease (or removed). |
1311 | * |
1312 | * %F_UNLCK to indicate the lease needs to be removed. |
1313 | * |
1314 | * XXX: sfr & willy disagree over whether F_INPROGRESS |
1315 | * should be returned to userspace. |
1316 | */ |
1317 | int fcntl_getlease(struct file *filp) |
1318 | { |
1319 | struct file_lock *fl; |
1320 | int type = F_UNLCK; |
1321 | |
1322 | lock_kernel(); |
1323 | time_out_leases(filp->f_path.dentry->d_inode); |
1324 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); |
1325 | fl = fl->fl_next) { |
1326 | if (fl->fl_file == filp) { |
1327 | type = fl->fl_type & ~F_INPROGRESS; |
1328 | break; |
1329 | } |
1330 | } |
1331 | unlock_kernel(); |
1332 | return type; |
1333 | } |
1334 | |
1335 | /** |
1336 | * generic_setlease - sets a lease on an open file |
1337 | * @filp: file pointer |
1338 | * @arg: type of lease to obtain |
1339 | * @flp: input - file_lock to use, output - file_lock inserted |
1340 | * |
1341 | * The (input) flp->fl_lmops->fl_break function is required |
1342 | * by break_lease(). |
1343 | * |
1344 | * Called with kernel lock held. |
1345 | */ |
1346 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
1347 | { |
1348 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
1349 | struct file_lock *new_fl = NULL; |
1350 | struct dentry *dentry = filp->f_path.dentry; |
1351 | struct inode *inode = dentry->d_inode; |
1352 | int error, rdlease_count = 0, wrlease_count = 0; |
1353 | |
1354 | if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE)) |
1355 | return -EACCES; |
1356 | if (!S_ISREG(inode->i_mode)) |
1357 | return -EINVAL; |
1358 | error = security_file_lock(filp, arg); |
1359 | if (error) |
1360 | return error; |
1361 | |
1362 | time_out_leases(inode); |
1363 | |
1364 | BUG_ON(!(*flp)->fl_lmops->fl_break); |
1365 | |
1366 | lease = *flp; |
1367 | |
1368 | if (arg != F_UNLCK) { |
1369 | error = -ENOMEM; |
1370 | new_fl = locks_alloc_lock(); |
1371 | if (new_fl == NULL) |
1372 | goto out; |
1373 | |
1374 | error = -EAGAIN; |
1375 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) |
1376 | goto out; |
1377 | if ((arg == F_WRLCK) |
1378 | && ((atomic_read(&dentry->d_count) > 1) |
1379 | || (atomic_read(&inode->i_count) > 1))) |
1380 | goto out; |
1381 | } |
1382 | |
1383 | /* |
1384 | * At this point, we know that if there is an exclusive |
1385 | * lease on this file, then we hold it on this filp |
1386 | * (otherwise our open of this file would have blocked). |
1387 | * And if we are trying to acquire an exclusive lease, |
1388 | * then the file is not open by anyone (including us) |
1389 | * except for this filp. |
1390 | */ |
1391 | for (before = &inode->i_flock; |
1392 | ((fl = *before) != NULL) && IS_LEASE(fl); |
1393 | before = &fl->fl_next) { |
1394 | if (lease->fl_lmops->fl_mylease(fl, lease)) |
1395 | my_before = before; |
1396 | else if (fl->fl_type == (F_INPROGRESS | F_UNLCK)) |
1397 | /* |
1398 | * Someone is in the process of opening this |
1399 | * file for writing so we may not take an |
1400 | * exclusive lease on it. |
1401 | */ |
1402 | wrlease_count++; |
1403 | else |
1404 | rdlease_count++; |
1405 | } |
1406 | |
1407 | error = -EAGAIN; |
1408 | if ((arg == F_RDLCK && (wrlease_count > 0)) || |
1409 | (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0))) |
1410 | goto out; |
1411 | |
1412 | if (my_before != NULL) { |
1413 | *flp = *my_before; |
1414 | error = lease->fl_lmops->fl_change(my_before, arg); |
1415 | goto out; |
1416 | } |
1417 | |
1418 | error = 0; |
1419 | if (arg == F_UNLCK) |
1420 | goto out; |
1421 | |
1422 | error = -EINVAL; |
1423 | if (!leases_enable) |
1424 | goto out; |
1425 | |
1426 | locks_copy_lock(new_fl, lease); |
1427 | locks_insert_lock(before, new_fl); |
1428 | |
1429 | *flp = new_fl; |
1430 | return 0; |
1431 | |
1432 | out: |
1433 | if (new_fl != NULL) |
1434 | locks_free_lock(new_fl); |
1435 | return error; |
1436 | } |
1437 | EXPORT_SYMBOL(generic_setlease); |
1438 | |
1439 | /** |
1440 | * vfs_setlease - sets a lease on an open file |
1441 | * @filp: file pointer |
1442 | * @arg: type of lease to obtain |
1443 | * @lease: file_lock to use |
1444 | * |
1445 | * Call this to establish a lease on the file. |
1446 | * The (*lease)->fl_lmops->fl_break operation must be set; if not, |
1447 | * break_lease will oops! |
1448 | * |
1449 | * This will call the filesystem's setlease file method, if |
1450 | * defined. Note that there is no getlease method; instead, the |
1451 | * filesystem setlease method should call back to setlease() to |
1452 | * add a lease to the inode's lease list, where fcntl_getlease() can |
1453 | * find it. Since fcntl_getlease() only reports whether the current |
1454 | * task holds a lease, a cluster filesystem need only do this for |
1455 | * leases held by processes on this node. |
1456 | * |
1457 | * There is also no break_lease method; filesystems that |
1458 | * handle their own leases should break leases themselves from the |
1459 | * filesystem's open, create, and (on truncate) setattr methods. |
1460 | * |
1461 | * Warning: the only current setlease methods exist only to disable |
1462 | * leases in certain cases. More vfs changes may be required to |
1463 | * allow a full filesystem lease implementation. |
1464 | */ |
1465 | |
1466 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1467 | { |
1468 | int error; |
1469 | |
1470 | lock_kernel(); |
1471 | if (filp->f_op && filp->f_op->setlease) |
1472 | error = filp->f_op->setlease(filp, arg, lease); |
1473 | else |
1474 | error = generic_setlease(filp, arg, lease); |
1475 | unlock_kernel(); |
1476 | |
1477 | return error; |
1478 | } |
1479 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1480 | |
1481 | /** |
1482 | * fcntl_setlease - sets a lease on an open file |
1483 | * @fd: open file descriptor |
1484 | * @filp: file pointer |
1485 | * @arg: type of lease to obtain |
1486 | * |
1487 | * Call this fcntl to establish a lease on the file. |
1488 | * Note that you also need to call %F_SETSIG to |
1489 | * receive a signal when the lease is broken. |
1490 | */ |
1491 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) |
1492 | { |
1493 | struct file_lock fl, *flp = &fl; |
1494 | struct inode *inode = filp->f_path.dentry->d_inode; |
1495 | int error; |
1496 | |
1497 | locks_init_lock(&fl); |
1498 | error = lease_init(filp, arg, &fl); |
1499 | if (error) |
1500 | return error; |
1501 | |
1502 | lock_kernel(); |
1503 | |
1504 | error = vfs_setlease(filp, arg, &flp); |
1505 | if (error || arg == F_UNLCK) |
1506 | goto out_unlock; |
1507 | |
1508 | error = fasync_helper(fd, filp, 1, &flp->fl_fasync); |
1509 | if (error < 0) { |
1510 | /* remove lease just inserted by setlease */ |
1511 | flp->fl_type = F_UNLCK | F_INPROGRESS; |
1512 | flp->fl_break_time = jiffies - 10; |
1513 | time_out_leases(inode); |
1514 | goto out_unlock; |
1515 | } |
1516 | |
1517 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1518 | out_unlock: |
1519 | unlock_kernel(); |
1520 | return error; |
1521 | } |
1522 | |
1523 | /** |
1524 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file |
1525 | * @filp: The file to apply the lock to |
1526 | * @fl: The lock to be applied |
1527 | * |
1528 | * Add a FLOCK style lock to a file. |
1529 | */ |
1530 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) |
1531 | { |
1532 | int error; |
1533 | might_sleep(); |
1534 | for (;;) { |
1535 | error = flock_lock_file(filp, fl); |
1536 | if (error != FILE_LOCK_DEFERRED) |
1537 | break; |
1538 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1539 | if (!error) |
1540 | continue; |
1541 | |
1542 | locks_delete_block(fl); |
1543 | break; |
1544 | } |
1545 | return error; |
1546 | } |
1547 | |
1548 | EXPORT_SYMBOL(flock_lock_file_wait); |
1549 | |
1550 | /** |
1551 | * sys_flock: - flock() system call. |
1552 | * @fd: the file descriptor to lock. |
1553 | * @cmd: the type of lock to apply. |
1554 | * |
1555 | * Apply a %FL_FLOCK style lock to an open file descriptor. |
1556 | * The @cmd can be one of |
1557 | * |
1558 | * %LOCK_SH -- a shared lock. |
1559 | * |
1560 | * %LOCK_EX -- an exclusive lock. |
1561 | * |
1562 | * %LOCK_UN -- remove an existing lock. |
1563 | * |
1564 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. |
1565 | * |
1566 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other |
1567 | * processes read and write access respectively. |
1568 | */ |
1569 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1570 | { |
1571 | struct file *filp; |
1572 | struct file_lock *lock; |
1573 | int can_sleep, unlock; |
1574 | int error; |
1575 | |
1576 | error = -EBADF; |
1577 | filp = fget(fd); |
1578 | if (!filp) |
1579 | goto out; |
1580 | |
1581 | can_sleep = !(cmd & LOCK_NB); |
1582 | cmd &= ~LOCK_NB; |
1583 | unlock = (cmd == LOCK_UN); |
1584 | |
1585 | if (!unlock && !(cmd & LOCK_MAND) && |
1586 | !(filp->f_mode & (FMODE_READ|FMODE_WRITE))) |
1587 | goto out_putf; |
1588 | |
1589 | error = flock_make_lock(filp, &lock, cmd); |
1590 | if (error) |
1591 | goto out_putf; |
1592 | if (can_sleep) |
1593 | lock->fl_flags |= FL_SLEEP; |
1594 | |
1595 | error = security_file_lock(filp, lock->fl_type); |
1596 | if (error) |
1597 | goto out_free; |
1598 | |
1599 | if (filp->f_op && filp->f_op->flock) |
1600 | error = filp->f_op->flock(filp, |
1601 | (can_sleep) ? F_SETLKW : F_SETLK, |
1602 | lock); |
1603 | else |
1604 | error = flock_lock_file_wait(filp, lock); |
1605 | |
1606 | out_free: |
1607 | locks_free_lock(lock); |
1608 | |
1609 | out_putf: |
1610 | fput(filp); |
1611 | out: |
1612 | return error; |
1613 | } |
1614 | |
1615 | /** |
1616 | * vfs_test_lock - test file byte range lock |
1617 | * @filp: The file to test lock for |
1618 | * @fl: The lock to test; also used to hold result |
1619 | * |
1620 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by |
1621 | * setting conf->fl_type to something other than F_UNLCK. |
1622 | */ |
1623 | int vfs_test_lock(struct file *filp, struct file_lock *fl) |
1624 | { |
1625 | if (filp->f_op && filp->f_op->lock) |
1626 | return filp->f_op->lock(filp, F_GETLK, fl); |
1627 | posix_test_lock(filp, fl); |
1628 | return 0; |
1629 | } |
1630 | EXPORT_SYMBOL_GPL(vfs_test_lock); |
1631 | |
1632 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1633 | { |
1634 | flock->l_pid = fl->fl_pid; |
1635 | #if BITS_PER_LONG == 32 |
1636 | /* |
1637 | * Make sure we can represent the posix lock via |
1638 | * legacy 32bit flock. |
1639 | */ |
1640 | if (fl->fl_start > OFFT_OFFSET_MAX) |
1641 | return -EOVERFLOW; |
1642 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) |
1643 | return -EOVERFLOW; |
1644 | #endif |
1645 | flock->l_start = fl->fl_start; |
1646 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
1647 | fl->fl_end - fl->fl_start + 1; |
1648 | flock->l_whence = 0; |
1649 | flock->l_type = fl->fl_type; |
1650 | return 0; |
1651 | } |
1652 | |
1653 | #if BITS_PER_LONG == 32 |
1654 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) |
1655 | { |
1656 | flock->l_pid = fl->fl_pid; |
1657 | flock->l_start = fl->fl_start; |
1658 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : |
1659 | fl->fl_end - fl->fl_start + 1; |
1660 | flock->l_whence = 0; |
1661 | flock->l_type = fl->fl_type; |
1662 | } |
1663 | #endif |
1664 | |
1665 | /* Report the first existing lock that would conflict with l. |
1666 | * This implements the F_GETLK command of fcntl(). |
1667 | */ |
1668 | int fcntl_getlk(struct file *filp, struct flock __user *l) |
1669 | { |
1670 | struct file_lock file_lock; |
1671 | struct flock flock; |
1672 | int error; |
1673 | |
1674 | error = -EFAULT; |
1675 | if (copy_from_user(&flock, l, sizeof(flock))) |
1676 | goto out; |
1677 | error = -EINVAL; |
1678 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
1679 | goto out; |
1680 | |
1681 | error = flock_to_posix_lock(filp, &file_lock, &flock); |
1682 | if (error) |
1683 | goto out; |
1684 | |
1685 | error = vfs_test_lock(filp, &file_lock); |
1686 | if (error) |
1687 | goto out; |
1688 | |
1689 | flock.l_type = file_lock.fl_type; |
1690 | if (file_lock.fl_type != F_UNLCK) { |
1691 | error = posix_lock_to_flock(&flock, &file_lock); |
1692 | if (error) |
1693 | goto out; |
1694 | } |
1695 | error = -EFAULT; |
1696 | if (!copy_to_user(l, &flock, sizeof(flock))) |
1697 | error = 0; |
1698 | out: |
1699 | return error; |
1700 | } |
1701 | |
1702 | /** |
1703 | * vfs_lock_file - file byte range lock |
1704 | * @filp: The file to apply the lock to |
1705 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) |
1706 | * @fl: The lock to be applied |
1707 | * @conf: Place to return a copy of the conflicting lock, if found. |
1708 | * |
1709 | * A caller that doesn't care about the conflicting lock may pass NULL |
1710 | * as the final argument. |
1711 | * |
1712 | * If the filesystem defines a private ->lock() method, then @conf will |
1713 | * be left unchanged; so a caller that cares should initialize it to |
1714 | * some acceptable default. |
1715 | * |
1716 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX |
1717 | * locks, the ->lock() interface may return asynchronously, before the lock has |
1718 | * been granted or denied by the underlying filesystem, if (and only if) |
1719 | * fl_grant is set. Callers expecting ->lock() to return asynchronously |
1720 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1721 | * the request is for a blocking lock. When ->lock() does return asynchronously, |
1722 | * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock |
1723 | * request completes. |
1724 | * If the request is for non-blocking lock the file system should return |
1725 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1726 | * with the result. If the request timed out the callback routine will return a |
1727 | * nonzero return code and the file system should release the lock. The file |
1728 | * system is also responsible to keep a corresponding posix lock when it |
1729 | * grants a lock so the VFS can find out which locks are locally held and do |
1730 | * the correct lock cleanup when required. |
1731 | * The underlying filesystem must not drop the kernel lock or call |
1732 | * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
1733 | * return code. |
1734 | */ |
1735 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
1736 | { |
1737 | if (filp->f_op && filp->f_op->lock) |
1738 | return filp->f_op->lock(filp, cmd, fl); |
1739 | else |
1740 | return posix_lock_file(filp, fl, conf); |
1741 | } |
1742 | EXPORT_SYMBOL_GPL(vfs_lock_file); |
1743 | |
1744 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1745 | struct file_lock *fl) |
1746 | { |
1747 | int error; |
1748 | |
1749 | error = security_file_lock(filp, fl->fl_type); |
1750 | if (error) |
1751 | return error; |
1752 | |
1753 | for (;;) { |
1754 | error = vfs_lock_file(filp, cmd, fl, NULL); |
1755 | if (error != FILE_LOCK_DEFERRED) |
1756 | break; |
1757 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1758 | if (!error) |
1759 | continue; |
1760 | |
1761 | locks_delete_block(fl); |
1762 | break; |
1763 | } |
1764 | |
1765 | return error; |
1766 | } |
1767 | |
1768 | /* Apply the lock described by l to an open file descriptor. |
1769 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
1770 | */ |
1771 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1772 | struct flock __user *l) |
1773 | { |
1774 | struct file_lock *file_lock = locks_alloc_lock(); |
1775 | struct flock flock; |
1776 | struct inode *inode; |
1777 | struct file *f; |
1778 | int error; |
1779 | |
1780 | if (file_lock == NULL) |
1781 | return -ENOLCK; |
1782 | |
1783 | /* |
1784 | * This might block, so we do it before checking the inode. |
1785 | */ |
1786 | error = -EFAULT; |
1787 | if (copy_from_user(&flock, l, sizeof(flock))) |
1788 | goto out; |
1789 | |
1790 | inode = filp->f_path.dentry->d_inode; |
1791 | |
1792 | /* Don't allow mandatory locks on files that may be memory mapped |
1793 | * and shared. |
1794 | */ |
1795 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1796 | error = -EAGAIN; |
1797 | goto out; |
1798 | } |
1799 | |
1800 | again: |
1801 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1802 | if (error) |
1803 | goto out; |
1804 | if (cmd == F_SETLKW) { |
1805 | file_lock->fl_flags |= FL_SLEEP; |
1806 | } |
1807 | |
1808 | error = -EBADF; |
1809 | switch (flock.l_type) { |
1810 | case F_RDLCK: |
1811 | if (!(filp->f_mode & FMODE_READ)) |
1812 | goto out; |
1813 | break; |
1814 | case F_WRLCK: |
1815 | if (!(filp->f_mode & FMODE_WRITE)) |
1816 | goto out; |
1817 | break; |
1818 | case F_UNLCK: |
1819 | break; |
1820 | default: |
1821 | error = -EINVAL; |
1822 | goto out; |
1823 | } |
1824 | |
1825 | error = do_lock_file_wait(filp, cmd, file_lock); |
1826 | |
1827 | /* |
1828 | * Attempt to detect a close/fcntl race and recover by |
1829 | * releasing the lock that was just acquired. |
1830 | */ |
1831 | /* |
1832 | * we need that spin_lock here - it prevents reordering between |
1833 | * update of inode->i_flock and check for it done in close(). |
1834 | * rcu_read_lock() wouldn't do. |
1835 | */ |
1836 | spin_lock(¤t->files->file_lock); |
1837 | f = fcheck(fd); |
1838 | spin_unlock(¤t->files->file_lock); |
1839 | if (!error && f != filp && flock.l_type != F_UNLCK) { |
1840 | flock.l_type = F_UNLCK; |
1841 | goto again; |
1842 | } |
1843 | |
1844 | out: |
1845 | locks_free_lock(file_lock); |
1846 | return error; |
1847 | } |
1848 | |
1849 | #if BITS_PER_LONG == 32 |
1850 | /* Report the first existing lock that would conflict with l. |
1851 | * This implements the F_GETLK command of fcntl(). |
1852 | */ |
1853 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) |
1854 | { |
1855 | struct file_lock file_lock; |
1856 | struct flock64 flock; |
1857 | int error; |
1858 | |
1859 | error = -EFAULT; |
1860 | if (copy_from_user(&flock, l, sizeof(flock))) |
1861 | goto out; |
1862 | error = -EINVAL; |
1863 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) |
1864 | goto out; |
1865 | |
1866 | error = flock64_to_posix_lock(filp, &file_lock, &flock); |
1867 | if (error) |
1868 | goto out; |
1869 | |
1870 | error = vfs_test_lock(filp, &file_lock); |
1871 | if (error) |
1872 | goto out; |
1873 | |
1874 | flock.l_type = file_lock.fl_type; |
1875 | if (file_lock.fl_type != F_UNLCK) |
1876 | posix_lock_to_flock64(&flock, &file_lock); |
1877 | |
1878 | error = -EFAULT; |
1879 | if (!copy_to_user(l, &flock, sizeof(flock))) |
1880 | error = 0; |
1881 | |
1882 | out: |
1883 | return error; |
1884 | } |
1885 | |
1886 | /* Apply the lock described by l to an open file descriptor. |
1887 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). |
1888 | */ |
1889 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1890 | struct flock64 __user *l) |
1891 | { |
1892 | struct file_lock *file_lock = locks_alloc_lock(); |
1893 | struct flock64 flock; |
1894 | struct inode *inode; |
1895 | struct file *f; |
1896 | int error; |
1897 | |
1898 | if (file_lock == NULL) |
1899 | return -ENOLCK; |
1900 | |
1901 | /* |
1902 | * This might block, so we do it before checking the inode. |
1903 | */ |
1904 | error = -EFAULT; |
1905 | if (copy_from_user(&flock, l, sizeof(flock))) |
1906 | goto out; |
1907 | |
1908 | inode = filp->f_path.dentry->d_inode; |
1909 | |
1910 | /* Don't allow mandatory locks on files that may be memory mapped |
1911 | * and shared. |
1912 | */ |
1913 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1914 | error = -EAGAIN; |
1915 | goto out; |
1916 | } |
1917 | |
1918 | again: |
1919 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1920 | if (error) |
1921 | goto out; |
1922 | if (cmd == F_SETLKW64) { |
1923 | file_lock->fl_flags |= FL_SLEEP; |
1924 | } |
1925 | |
1926 | error = -EBADF; |
1927 | switch (flock.l_type) { |
1928 | case F_RDLCK: |
1929 | if (!(filp->f_mode & FMODE_READ)) |
1930 | goto out; |
1931 | break; |
1932 | case F_WRLCK: |
1933 | if (!(filp->f_mode & FMODE_WRITE)) |
1934 | goto out; |
1935 | break; |
1936 | case F_UNLCK: |
1937 | break; |
1938 | default: |
1939 | error = -EINVAL; |
1940 | goto out; |
1941 | } |
1942 | |
1943 | error = do_lock_file_wait(filp, cmd, file_lock); |
1944 | |
1945 | /* |
1946 | * Attempt to detect a close/fcntl race and recover by |
1947 | * releasing the lock that was just acquired. |
1948 | */ |
1949 | spin_lock(¤t->files->file_lock); |
1950 | f = fcheck(fd); |
1951 | spin_unlock(¤t->files->file_lock); |
1952 | if (!error && f != filp && flock.l_type != F_UNLCK) { |
1953 | flock.l_type = F_UNLCK; |
1954 | goto again; |
1955 | } |
1956 | |
1957 | out: |
1958 | locks_free_lock(file_lock); |
1959 | return error; |
1960 | } |
1961 | #endif /* BITS_PER_LONG == 32 */ |
1962 | |
1963 | /* |
1964 | * This function is called when the file is being removed |
1965 | * from the task's fd array. POSIX locks belonging to this task |
1966 | * are deleted at this time. |
1967 | */ |
1968 | void locks_remove_posix(struct file *filp, fl_owner_t owner) |
1969 | { |
1970 | struct file_lock lock; |
1971 | |
1972 | /* |
1973 | * If there are no locks held on this file, we don't need to call |
1974 | * posix_lock_file(). Another process could be setting a lock on this |
1975 | * file at the same time, but we wouldn't remove that lock anyway. |
1976 | */ |
1977 | if (!filp->f_path.dentry->d_inode->i_flock) |
1978 | return; |
1979 | |
1980 | lock.fl_type = F_UNLCK; |
1981 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1982 | lock.fl_start = 0; |
1983 | lock.fl_end = OFFSET_MAX; |
1984 | lock.fl_owner = owner; |
1985 | lock.fl_pid = current->tgid; |
1986 | lock.fl_file = filp; |
1987 | lock.fl_ops = NULL; |
1988 | lock.fl_lmops = NULL; |
1989 | |
1990 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1991 | |
1992 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
1993 | lock.fl_ops->fl_release_private(&lock); |
1994 | } |
1995 | |
1996 | EXPORT_SYMBOL(locks_remove_posix); |
1997 | |
1998 | /* |
1999 | * This function is called on the last close of an open file. |
2000 | */ |
2001 | void locks_remove_flock(struct file *filp) |
2002 | { |
2003 | struct inode * inode = filp->f_path.dentry->d_inode; |
2004 | struct file_lock *fl; |
2005 | struct file_lock **before; |
2006 | |
2007 | if (!inode->i_flock) |
2008 | return; |
2009 | |
2010 | if (filp->f_op && filp->f_op->flock) { |
2011 | struct file_lock fl = { |
2012 | .fl_pid = current->tgid, |
2013 | .fl_file = filp, |
2014 | .fl_flags = FL_FLOCK, |
2015 | .fl_type = F_UNLCK, |
2016 | .fl_end = OFFSET_MAX, |
2017 | }; |
2018 | filp->f_op->flock(filp, F_SETLKW, &fl); |
2019 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2020 | fl.fl_ops->fl_release_private(&fl); |
2021 | } |
2022 | |
2023 | lock_kernel(); |
2024 | before = &inode->i_flock; |
2025 | |
2026 | while ((fl = *before) != NULL) { |
2027 | if (fl->fl_file == filp) { |
2028 | if (IS_FLOCK(fl)) { |
2029 | locks_delete_lock(before); |
2030 | continue; |
2031 | } |
2032 | if (IS_LEASE(fl)) { |
2033 | lease_modify(before, F_UNLCK); |
2034 | continue; |
2035 | } |
2036 | /* What? */ |
2037 | BUG(); |
2038 | } |
2039 | before = &fl->fl_next; |
2040 | } |
2041 | unlock_kernel(); |
2042 | } |
2043 | |
2044 | /** |
2045 | * posix_unblock_lock - stop waiting for a file lock |
2046 | * @filp: how the file was opened |
2047 | * @waiter: the lock which was waiting |
2048 | * |
2049 | * lockd needs to block waiting for locks. |
2050 | */ |
2051 | int |
2052 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
2053 | { |
2054 | int status = 0; |
2055 | |
2056 | lock_kernel(); |
2057 | if (waiter->fl_next) |
2058 | __locks_delete_block(waiter); |
2059 | else |
2060 | status = -ENOENT; |
2061 | unlock_kernel(); |
2062 | return status; |
2063 | } |
2064 | |
2065 | EXPORT_SYMBOL(posix_unblock_lock); |
2066 | |
2067 | /** |
2068 | * vfs_cancel_lock - file byte range unblock lock |
2069 | * @filp: The file to apply the unblock to |
2070 | * @fl: The lock to be unblocked |
2071 | * |
2072 | * Used by lock managers to cancel blocked requests |
2073 | */ |
2074 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) |
2075 | { |
2076 | if (filp->f_op && filp->f_op->lock) |
2077 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2078 | return 0; |
2079 | } |
2080 | |
2081 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); |
2082 | |
2083 | #ifdef CONFIG_PROC_FS |
2084 | #include <linux/proc_fs.h> |
2085 | #include <linux/seq_file.h> |
2086 | |
2087 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
2088 | int id, char *pfx) |
2089 | { |
2090 | struct inode *inode = NULL; |
2091 | unsigned int fl_pid; |
2092 | |
2093 | if (fl->fl_nspid) |
2094 | fl_pid = pid_vnr(fl->fl_nspid); |
2095 | else |
2096 | fl_pid = fl->fl_pid; |
2097 | |
2098 | if (fl->fl_file != NULL) |
2099 | inode = fl->fl_file->f_path.dentry->d_inode; |
2100 | |
2101 | seq_printf(f, "%d:%s ", id, pfx); |
2102 | if (IS_POSIX(fl)) { |
2103 | seq_printf(f, "%6s %s ", |
2104 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2105 | (inode == NULL) ? "*NOINODE*" : |
2106 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
2107 | } else if (IS_FLOCK(fl)) { |
2108 | if (fl->fl_type & LOCK_MAND) { |
2109 | seq_printf(f, "FLOCK MSNFS "); |
2110 | } else { |
2111 | seq_printf(f, "FLOCK ADVISORY "); |
2112 | } |
2113 | } else if (IS_LEASE(fl)) { |
2114 | seq_printf(f, "LEASE "); |
2115 | if (fl->fl_type & F_INPROGRESS) |
2116 | seq_printf(f, "BREAKING "); |
2117 | else if (fl->fl_file) |
2118 | seq_printf(f, "ACTIVE "); |
2119 | else |
2120 | seq_printf(f, "BREAKER "); |
2121 | } else { |
2122 | seq_printf(f, "UNKNOWN UNKNOWN "); |
2123 | } |
2124 | if (fl->fl_type & LOCK_MAND) { |
2125 | seq_printf(f, "%s ", |
2126 | (fl->fl_type & LOCK_READ) |
2127 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " |
2128 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); |
2129 | } else { |
2130 | seq_printf(f, "%s ", |
2131 | (fl->fl_type & F_INPROGRESS) |
2132 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " |
2133 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); |
2134 | } |
2135 | if (inode) { |
2136 | #ifdef WE_CAN_BREAK_LSLK_NOW |
2137 | seq_printf(f, "%d %s:%ld ", fl_pid, |
2138 | inode->i_sb->s_id, inode->i_ino); |
2139 | #else |
2140 | /* userspace relies on this representation of dev_t ;-( */ |
2141 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
2142 | MAJOR(inode->i_sb->s_dev), |
2143 | MINOR(inode->i_sb->s_dev), inode->i_ino); |
2144 | #endif |
2145 | } else { |
2146 | seq_printf(f, "%d <none>:0 ", fl_pid); |
2147 | } |
2148 | if (IS_POSIX(fl)) { |
2149 | if (fl->fl_end == OFFSET_MAX) |
2150 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
2151 | else |
2152 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
2153 | } else { |
2154 | seq_printf(f, "0 EOF\n"); |
2155 | } |
2156 | } |
2157 | |
2158 | static int locks_show(struct seq_file *f, void *v) |
2159 | { |
2160 | struct file_lock *fl, *bfl; |
2161 | |
2162 | fl = list_entry(v, struct file_lock, fl_link); |
2163 | |
2164 | lock_get_status(f, fl, (long)f->private, ""); |
2165 | |
2166 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
2167 | lock_get_status(f, bfl, (long)f->private, " ->"); |
2168 | |
2169 | f->private++; |
2170 | return 0; |
2171 | } |
2172 | |
2173 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2174 | { |
2175 | lock_kernel(); |
2176 | f->private = (void *)1; |
2177 | return seq_list_start(&file_lock_list, *pos); |
2178 | } |
2179 | |
2180 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2181 | { |
2182 | return seq_list_next(v, &file_lock_list, pos); |
2183 | } |
2184 | |
2185 | static void locks_stop(struct seq_file *f, void *v) |
2186 | { |
2187 | unlock_kernel(); |
2188 | } |
2189 | |
2190 | static const struct seq_operations locks_seq_operations = { |
2191 | .start = locks_start, |
2192 | .next = locks_next, |
2193 | .stop = locks_stop, |
2194 | .show = locks_show, |
2195 | }; |
2196 | |
2197 | static int locks_open(struct inode *inode, struct file *filp) |
2198 | { |
2199 | return seq_open(filp, &locks_seq_operations); |
2200 | } |
2201 | |
2202 | static const struct file_operations proc_locks_operations = { |
2203 | .open = locks_open, |
2204 | .read = seq_read, |
2205 | .llseek = seq_lseek, |
2206 | .release = seq_release, |
2207 | }; |
2208 | |
2209 | static int __init proc_locks_init(void) |
2210 | { |
2211 | proc_create("locks", 0, NULL, &proc_locks_operations); |
2212 | return 0; |
2213 | } |
2214 | module_init(proc_locks_init); |
2215 | #endif |
2216 | |
2217 | /** |
2218 | * lock_may_read - checks that the region is free of locks |
2219 | * @inode: the inode that is being read |
2220 | * @start: the first byte to read |
2221 | * @len: the number of bytes to read |
2222 | * |
2223 | * Emulates Windows locking requirements. Whole-file |
2224 | * mandatory locks (share modes) can prohibit a read and |
2225 | * byte-range POSIX locks can prohibit a read if they overlap. |
2226 | * |
2227 | * N.B. this function is only ever called |
2228 | * from knfsd and ownership of locks is never checked. |
2229 | */ |
2230 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) |
2231 | { |
2232 | struct file_lock *fl; |
2233 | int result = 1; |
2234 | lock_kernel(); |
2235 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2236 | if (IS_POSIX(fl)) { |
2237 | if (fl->fl_type == F_RDLCK) |
2238 | continue; |
2239 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
2240 | continue; |
2241 | } else if (IS_FLOCK(fl)) { |
2242 | if (!(fl->fl_type & LOCK_MAND)) |
2243 | continue; |
2244 | if (fl->fl_type & LOCK_READ) |
2245 | continue; |
2246 | } else |
2247 | continue; |
2248 | result = 0; |
2249 | break; |
2250 | } |
2251 | unlock_kernel(); |
2252 | return result; |
2253 | } |
2254 | |
2255 | EXPORT_SYMBOL(lock_may_read); |
2256 | |
2257 | /** |
2258 | * lock_may_write - checks that the region is free of locks |
2259 | * @inode: the inode that is being written |
2260 | * @start: the first byte to write |
2261 | * @len: the number of bytes to write |
2262 | * |
2263 | * Emulates Windows locking requirements. Whole-file |
2264 | * mandatory locks (share modes) can prohibit a write and |
2265 | * byte-range POSIX locks can prohibit a write if they overlap. |
2266 | * |
2267 | * N.B. this function is only ever called |
2268 | * from knfsd and ownership of locks is never checked. |
2269 | */ |
2270 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) |
2271 | { |
2272 | struct file_lock *fl; |
2273 | int result = 1; |
2274 | lock_kernel(); |
2275 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2276 | if (IS_POSIX(fl)) { |
2277 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) |
2278 | continue; |
2279 | } else if (IS_FLOCK(fl)) { |
2280 | if (!(fl->fl_type & LOCK_MAND)) |
2281 | continue; |
2282 | if (fl->fl_type & LOCK_WRITE) |
2283 | continue; |
2284 | } else |
2285 | continue; |
2286 | result = 0; |
2287 | break; |
2288 | } |
2289 | unlock_kernel(); |
2290 | return result; |
2291 | } |
2292 | |
2293 | EXPORT_SYMBOL(lock_may_write); |
2294 | |
2295 | static int __init filelock_init(void) |
2296 | { |
2297 | filelock_cache = kmem_cache_create("file_lock_cache", |
2298 | sizeof(struct file_lock), 0, SLAB_PANIC, |
2299 | init_once); |
2300 | return 0; |
2301 | } |
2302 | |
2303 | core_initcall(filelock_init); |
2304 |
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