Commit | Line | Data |
---|---|---|
1da177e4 LT |
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> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
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> | |
4fb3a538 | 128 | #include <linux/rcupdate.h> |
ab1f1611 | 129 | #include <linux/pid_namespace.h> |
1da177e4 | 130 | |
1da177e4 LT |
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 | ||
26bcbf96 | 143 | static LIST_HEAD(file_lock_list); |
1da177e4 LT |
144 | static LIST_HEAD(blocked_list); |
145 | ||
e18b890b | 146 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 LT |
147 | |
148 | /* Allocate an empty lock structure. */ | |
149 | static struct file_lock *locks_alloc_lock(void) | |
150 | { | |
e94b1766 | 151 | return kmem_cache_alloc(filelock_cache, GFP_KERNEL); |
1da177e4 LT |
152 | } |
153 | ||
47831f35 TM |
154 | static 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 | ||
1da177e4 | 169 | /* Free a lock which is not in use. */ |
33443c42 | 170 | static void locks_free_lock(struct file_lock *fl) |
1da177e4 | 171 | { |
5ce29646 MS |
172 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
173 | BUG_ON(!list_empty(&fl->fl_block)); | |
174 | BUG_ON(!list_empty(&fl->fl_link)); | |
1da177e4 | 175 | |
47831f35 | 176 | locks_release_private(fl); |
1da177e4 LT |
177 | kmem_cache_free(filelock_cache, fl); |
178 | } | |
179 | ||
180 | void locks_init_lock(struct file_lock *fl) | |
181 | { | |
182 | INIT_LIST_HEAD(&fl->fl_link); | |
183 | INIT_LIST_HEAD(&fl->fl_block); | |
184 | init_waitqueue_head(&fl->fl_wait); | |
185 | fl->fl_next = NULL; | |
186 | fl->fl_fasync = NULL; | |
187 | fl->fl_owner = NULL; | |
188 | fl->fl_pid = 0; | |
ab1f1611 | 189 | fl->fl_nspid = NULL; |
1da177e4 LT |
190 | fl->fl_file = NULL; |
191 | fl->fl_flags = 0; | |
192 | fl->fl_type = 0; | |
193 | fl->fl_start = fl->fl_end = 0; | |
194 | fl->fl_ops = NULL; | |
195 | fl->fl_lmops = NULL; | |
196 | } | |
197 | ||
198 | EXPORT_SYMBOL(locks_init_lock); | |
199 | ||
200 | /* | |
201 | * Initialises the fields of the file lock which are invariant for | |
202 | * free file_locks. | |
203 | */ | |
51cc5068 | 204 | static void init_once(void *foo) |
1da177e4 LT |
205 | { |
206 | struct file_lock *lock = (struct file_lock *) foo; | |
207 | ||
1da177e4 LT |
208 | locks_init_lock(lock); |
209 | } | |
210 | ||
47831f35 TM |
211 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
212 | { | |
213 | if (fl->fl_ops) { | |
214 | if (fl->fl_ops->fl_copy_lock) | |
215 | fl->fl_ops->fl_copy_lock(new, fl); | |
216 | new->fl_ops = fl->fl_ops; | |
217 | } | |
218 | if (fl->fl_lmops) { | |
219 | if (fl->fl_lmops->fl_copy_lock) | |
220 | fl->fl_lmops->fl_copy_lock(new, fl); | |
221 | new->fl_lmops = fl->fl_lmops; | |
222 | } | |
223 | } | |
224 | ||
1da177e4 LT |
225 | /* |
226 | * Initialize a new lock from an existing file_lock structure. | |
227 | */ | |
1a747ee0 | 228 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
229 | { |
230 | new->fl_owner = fl->fl_owner; | |
231 | new->fl_pid = fl->fl_pid; | |
0996905f | 232 | new->fl_file = NULL; |
1da177e4 LT |
233 | new->fl_flags = fl->fl_flags; |
234 | new->fl_type = fl->fl_type; | |
235 | new->fl_start = fl->fl_start; | |
236 | new->fl_end = fl->fl_end; | |
0996905f TM |
237 | new->fl_ops = NULL; |
238 | new->fl_lmops = NULL; | |
239 | } | |
3dd7b71c | 240 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
241 | |
242 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
243 | { | |
244 | locks_release_private(new); | |
245 | ||
246 | __locks_copy_lock(new, fl); | |
247 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
248 | new->fl_ops = fl->fl_ops; |
249 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
250 | |
251 | locks_copy_private(new, fl); | |
1da177e4 LT |
252 | } |
253 | ||
254 | EXPORT_SYMBOL(locks_copy_lock); | |
255 | ||
256 | static inline int flock_translate_cmd(int cmd) { | |
257 | if (cmd & LOCK_MAND) | |
258 | return cmd & (LOCK_MAND | LOCK_RW); | |
259 | switch (cmd) { | |
260 | case LOCK_SH: | |
261 | return F_RDLCK; | |
262 | case LOCK_EX: | |
263 | return F_WRLCK; | |
264 | case LOCK_UN: | |
265 | return F_UNLCK; | |
266 | } | |
267 | return -EINVAL; | |
268 | } | |
269 | ||
270 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
271 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
272 | unsigned int cmd) | |
273 | { | |
274 | struct file_lock *fl; | |
275 | int type = flock_translate_cmd(cmd); | |
276 | if (type < 0) | |
277 | return type; | |
278 | ||
279 | fl = locks_alloc_lock(); | |
280 | if (fl == NULL) | |
281 | return -ENOMEM; | |
282 | ||
283 | fl->fl_file = filp; | |
284 | fl->fl_pid = current->tgid; | |
285 | fl->fl_flags = FL_FLOCK; | |
286 | fl->fl_type = type; | |
287 | fl->fl_end = OFFSET_MAX; | |
288 | ||
289 | *lock = fl; | |
290 | return 0; | |
291 | } | |
292 | ||
293 | static int assign_type(struct file_lock *fl, int type) | |
294 | { | |
295 | switch (type) { | |
296 | case F_RDLCK: | |
297 | case F_WRLCK: | |
298 | case F_UNLCK: | |
299 | fl->fl_type = type; | |
300 | break; | |
301 | default: | |
302 | return -EINVAL; | |
303 | } | |
304 | return 0; | |
305 | } | |
306 | ||
307 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
308 | * style lock. | |
309 | */ | |
310 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
311 | struct flock *l) | |
312 | { | |
313 | off_t start, end; | |
314 | ||
315 | switch (l->l_whence) { | |
f5579f8c | 316 | case SEEK_SET: |
1da177e4 LT |
317 | start = 0; |
318 | break; | |
f5579f8c | 319 | case SEEK_CUR: |
1da177e4 LT |
320 | start = filp->f_pos; |
321 | break; | |
f5579f8c | 322 | case SEEK_END: |
0f7fc9e4 | 323 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
324 | break; |
325 | default: | |
326 | return -EINVAL; | |
327 | } | |
328 | ||
329 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
330 | POSIX-2001 defines it. */ | |
331 | start += l->l_start; | |
4c780a46 TM |
332 | if (start < 0) |
333 | return -EINVAL; | |
334 | fl->fl_end = OFFSET_MAX; | |
335 | if (l->l_len > 0) { | |
336 | end = start + l->l_len - 1; | |
337 | fl->fl_end = end; | |
338 | } else if (l->l_len < 0) { | |
1da177e4 | 339 | end = start - 1; |
4c780a46 | 340 | fl->fl_end = end; |
1da177e4 | 341 | start += l->l_len; |
4c780a46 TM |
342 | if (start < 0) |
343 | return -EINVAL; | |
1da177e4 | 344 | } |
1da177e4 | 345 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
346 | if (fl->fl_end < fl->fl_start) |
347 | return -EOVERFLOW; | |
1da177e4 LT |
348 | |
349 | fl->fl_owner = current->files; | |
350 | fl->fl_pid = current->tgid; | |
351 | fl->fl_file = filp; | |
352 | fl->fl_flags = FL_POSIX; | |
353 | fl->fl_ops = NULL; | |
354 | fl->fl_lmops = NULL; | |
355 | ||
356 | return assign_type(fl, l->l_type); | |
357 | } | |
358 | ||
359 | #if BITS_PER_LONG == 32 | |
360 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
361 | struct flock64 *l) | |
362 | { | |
363 | loff_t start; | |
364 | ||
365 | switch (l->l_whence) { | |
f5579f8c | 366 | case SEEK_SET: |
1da177e4 LT |
367 | start = 0; |
368 | break; | |
f5579f8c | 369 | case SEEK_CUR: |
1da177e4 LT |
370 | start = filp->f_pos; |
371 | break; | |
f5579f8c | 372 | case SEEK_END: |
0f7fc9e4 | 373 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
374 | break; |
375 | default: | |
376 | return -EINVAL; | |
377 | } | |
378 | ||
4c780a46 TM |
379 | start += l->l_start; |
380 | if (start < 0) | |
1da177e4 | 381 | return -EINVAL; |
4c780a46 TM |
382 | fl->fl_end = OFFSET_MAX; |
383 | if (l->l_len > 0) { | |
384 | fl->fl_end = start + l->l_len - 1; | |
385 | } else if (l->l_len < 0) { | |
386 | fl->fl_end = start - 1; | |
387 | start += l->l_len; | |
388 | if (start < 0) | |
389 | return -EINVAL; | |
390 | } | |
1da177e4 | 391 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
392 | if (fl->fl_end < fl->fl_start) |
393 | return -EOVERFLOW; | |
1da177e4 LT |
394 | |
395 | fl->fl_owner = current->files; | |
396 | fl->fl_pid = current->tgid; | |
397 | fl->fl_file = filp; | |
398 | fl->fl_flags = FL_POSIX; | |
399 | fl->fl_ops = NULL; | |
400 | fl->fl_lmops = NULL; | |
401 | ||
402 | switch (l->l_type) { | |
403 | case F_RDLCK: | |
404 | case F_WRLCK: | |
405 | case F_UNLCK: | |
406 | fl->fl_type = l->l_type; | |
407 | break; | |
408 | default: | |
409 | return -EINVAL; | |
410 | } | |
411 | ||
412 | return (0); | |
413 | } | |
414 | #endif | |
415 | ||
416 | /* default lease lock manager operations */ | |
417 | static void lease_break_callback(struct file_lock *fl) | |
418 | { | |
419 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
420 | } | |
421 | ||
422 | static void lease_release_private_callback(struct file_lock *fl) | |
423 | { | |
424 | if (!fl->fl_file) | |
425 | return; | |
426 | ||
427 | f_delown(fl->fl_file); | |
428 | fl->fl_file->f_owner.signum = 0; | |
429 | } | |
430 | ||
75c96f85 | 431 | static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try) |
1da177e4 LT |
432 | { |
433 | return fl->fl_file == try->fl_file; | |
434 | } | |
435 | ||
75c96f85 | 436 | static struct lock_manager_operations lease_manager_ops = { |
1da177e4 LT |
437 | .fl_break = lease_break_callback, |
438 | .fl_release_private = lease_release_private_callback, | |
439 | .fl_mylease = lease_mylease_callback, | |
440 | .fl_change = lease_modify, | |
441 | }; | |
442 | ||
443 | /* | |
444 | * Initialize a lease, use the default lock manager operations | |
445 | */ | |
446 | static int lease_init(struct file *filp, int type, struct file_lock *fl) | |
447 | { | |
75dff55a TM |
448 | if (assign_type(fl, type) != 0) |
449 | return -EINVAL; | |
450 | ||
1da177e4 LT |
451 | fl->fl_owner = current->files; |
452 | fl->fl_pid = current->tgid; | |
453 | ||
454 | fl->fl_file = filp; | |
455 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
456 | fl->fl_start = 0; |
457 | fl->fl_end = OFFSET_MAX; | |
458 | fl->fl_ops = NULL; | |
459 | fl->fl_lmops = &lease_manager_ops; | |
460 | return 0; | |
461 | } | |
462 | ||
463 | /* Allocate a file_lock initialised to this type of lease */ | |
e32b8ee2 | 464 | static struct file_lock *lease_alloc(struct file *filp, int type) |
1da177e4 LT |
465 | { |
466 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 467 | int error = -ENOMEM; |
1da177e4 LT |
468 | |
469 | if (fl == NULL) | |
e32b8ee2 | 470 | return ERR_PTR(error); |
1da177e4 LT |
471 | |
472 | error = lease_init(filp, type, fl); | |
75dff55a TM |
473 | if (error) { |
474 | locks_free_lock(fl); | |
e32b8ee2 | 475 | return ERR_PTR(error); |
75dff55a | 476 | } |
e32b8ee2 | 477 | return fl; |
1da177e4 LT |
478 | } |
479 | ||
480 | /* Check if two locks overlap each other. | |
481 | */ | |
482 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
483 | { | |
484 | return ((fl1->fl_end >= fl2->fl_start) && | |
485 | (fl2->fl_end >= fl1->fl_start)); | |
486 | } | |
487 | ||
488 | /* | |
489 | * Check whether two locks have the same owner. | |
490 | */ | |
33443c42 | 491 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 LT |
492 | { |
493 | if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner) | |
494 | return fl2->fl_lmops == fl1->fl_lmops && | |
495 | fl1->fl_lmops->fl_compare_owner(fl1, fl2); | |
496 | return fl1->fl_owner == fl2->fl_owner; | |
497 | } | |
498 | ||
499 | /* Remove waiter from blocker's block list. | |
500 | * When blocker ends up pointing to itself then the list is empty. | |
501 | */ | |
33443c42 | 502 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 LT |
503 | { |
504 | list_del_init(&waiter->fl_block); | |
505 | list_del_init(&waiter->fl_link); | |
506 | waiter->fl_next = NULL; | |
507 | } | |
508 | ||
509 | /* | |
510 | */ | |
511 | static void locks_delete_block(struct file_lock *waiter) | |
512 | { | |
513 | lock_kernel(); | |
514 | __locks_delete_block(waiter); | |
515 | unlock_kernel(); | |
516 | } | |
517 | ||
518 | /* Insert waiter into blocker's block list. | |
519 | * We use a circular list so that processes can be easily woken up in | |
520 | * the order they blocked. The documentation doesn't require this but | |
521 | * it seems like the reasonable thing to do. | |
522 | */ | |
523 | static void locks_insert_block(struct file_lock *blocker, | |
524 | struct file_lock *waiter) | |
525 | { | |
6dc0fe8f | 526 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 LT |
527 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
528 | waiter->fl_next = blocker; | |
529 | if (IS_POSIX(blocker)) | |
530 | list_add(&waiter->fl_link, &blocked_list); | |
531 | } | |
532 | ||
533 | /* Wake up processes blocked waiting for blocker. | |
534 | * If told to wait then schedule the processes until the block list | |
535 | * is empty, otherwise empty the block list ourselves. | |
536 | */ | |
537 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
538 | { | |
539 | while (!list_empty(&blocker->fl_block)) { | |
f0c1cd0e PE |
540 | struct file_lock *waiter; |
541 | ||
542 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
543 | struct file_lock, fl_block); |
544 | __locks_delete_block(waiter); | |
545 | if (waiter->fl_lmops && waiter->fl_lmops->fl_notify) | |
546 | waiter->fl_lmops->fl_notify(waiter); | |
547 | else | |
548 | wake_up(&waiter->fl_wait); | |
549 | } | |
550 | } | |
551 | ||
552 | /* Insert file lock fl into an inode's lock list at the position indicated | |
553 | * by pos. At the same time add the lock to the global file lock list. | |
554 | */ | |
555 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
556 | { | |
557 | list_add(&fl->fl_link, &file_lock_list); | |
558 | ||
ab1f1611 VG |
559 | fl->fl_nspid = get_pid(task_tgid(current)); |
560 | ||
1da177e4 LT |
561 | /* insert into file's list */ |
562 | fl->fl_next = *pos; | |
563 | *pos = fl; | |
1da177e4 LT |
564 | } |
565 | ||
566 | /* | |
567 | * Delete a lock and then free it. | |
568 | * Wake up processes that are blocked waiting for this lock, | |
569 | * notify the FS that the lock has been cleared and | |
570 | * finally free the lock. | |
571 | */ | |
572 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
573 | { | |
574 | struct file_lock *fl = *thisfl_p; | |
575 | ||
576 | *thisfl_p = fl->fl_next; | |
577 | fl->fl_next = NULL; | |
578 | list_del_init(&fl->fl_link); | |
579 | ||
580 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); | |
581 | if (fl->fl_fasync != NULL) { | |
582 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
583 | fl->fl_fasync = NULL; | |
584 | } | |
585 | ||
ab1f1611 VG |
586 | if (fl->fl_nspid) { |
587 | put_pid(fl->fl_nspid); | |
588 | fl->fl_nspid = NULL; | |
589 | } | |
590 | ||
1da177e4 LT |
591 | locks_wake_up_blocks(fl); |
592 | locks_free_lock(fl); | |
593 | } | |
594 | ||
595 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
596 | * checks for shared/exclusive status of overlapping locks. | |
597 | */ | |
598 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
599 | { | |
600 | if (sys_fl->fl_type == F_WRLCK) | |
601 | return 1; | |
602 | if (caller_fl->fl_type == F_WRLCK) | |
603 | return 1; | |
604 | return 0; | |
605 | } | |
606 | ||
607 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
608 | * checking before calling the locks_conflict(). | |
609 | */ | |
610 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
611 | { | |
612 | /* POSIX locks owned by the same process do not conflict with | |
613 | * each other. | |
614 | */ | |
615 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
616 | return (0); | |
617 | ||
618 | /* Check whether they overlap */ | |
619 | if (!locks_overlap(caller_fl, sys_fl)) | |
620 | return 0; | |
621 | ||
622 | return (locks_conflict(caller_fl, sys_fl)); | |
623 | } | |
624 | ||
625 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
626 | * checking before calling the locks_conflict(). | |
627 | */ | |
628 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
629 | { | |
630 | /* FLOCK locks referring to the same filp do not conflict with | |
631 | * each other. | |
632 | */ | |
633 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
634 | return (0); | |
635 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
636 | return 0; | |
637 | ||
638 | return (locks_conflict(caller_fl, sys_fl)); | |
639 | } | |
640 | ||
6d34ac19 | 641 | void |
9d6a8c5c | 642 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
643 | { |
644 | struct file_lock *cfl; | |
645 | ||
646 | lock_kernel(); | |
0f7fc9e4 | 647 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
648 | if (!IS_POSIX(cfl)) |
649 | continue; | |
b842e240 | 650 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
651 | break; |
652 | } | |
ab1f1611 | 653 | if (cfl) { |
9d6a8c5c | 654 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 655 | if (cfl->fl_nspid) |
6c5f3e7b | 656 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 657 | } else |
129a84de | 658 | fl->fl_type = F_UNLCK; |
1da177e4 | 659 | unlock_kernel(); |
6d34ac19 | 660 | return; |
1da177e4 | 661 | } |
1da177e4 LT |
662 | EXPORT_SYMBOL(posix_test_lock); |
663 | ||
b533184f BF |
664 | /* |
665 | * Deadlock detection: | |
666 | * | |
667 | * We attempt to detect deadlocks that are due purely to posix file | |
668 | * locks. | |
1da177e4 | 669 | * |
b533184f BF |
670 | * We assume that a task can be waiting for at most one lock at a time. |
671 | * So for any acquired lock, the process holding that lock may be | |
672 | * waiting on at most one other lock. That lock in turns may be held by | |
673 | * someone waiting for at most one other lock. Given a requested lock | |
674 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
675 | * follow this chain of waiters to ensure we are not about to create a | |
676 | * cycle. | |
1da177e4 | 677 | * |
b533184f BF |
678 | * Since we do this before we ever put a process to sleep on a lock, we |
679 | * are ensured that there is never a cycle; that is what guarantees that | |
680 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 681 | * |
b533184f BF |
682 | * Note: the above assumption may not be true when handling lock |
683 | * requests from a broken NFS client. It may also fail in the presence | |
684 | * of tasks (such as posix threads) sharing the same open file table. | |
685 | * | |
686 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 687 | */ |
97855b49 BF |
688 | |
689 | #define MAX_DEADLK_ITERATIONS 10 | |
690 | ||
b533184f BF |
691 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
692 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
693 | { | |
694 | struct file_lock *fl; | |
695 | ||
696 | list_for_each_entry(fl, &blocked_list, fl_link) { | |
697 | if (posix_same_owner(fl, block_fl)) | |
698 | return fl->fl_next; | |
699 | } | |
700 | return NULL; | |
701 | } | |
702 | ||
b0904e14 | 703 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
704 | struct file_lock *block_fl) |
705 | { | |
97855b49 | 706 | int i = 0; |
1da177e4 | 707 | |
b533184f BF |
708 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
709 | if (i++ > MAX_DEADLK_ITERATIONS) | |
710 | return 0; | |
711 | if (posix_same_owner(caller_fl, block_fl)) | |
712 | return 1; | |
1da177e4 LT |
713 | } |
714 | return 0; | |
715 | } | |
716 | ||
1da177e4 | 717 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 718 | * after any leases, but before any posix locks. |
f475ae95 TM |
719 | * |
720 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
721 | * whether or not a lock was successfully freed by testing the return | |
722 | * value for -ENOENT. | |
1da177e4 | 723 | */ |
993dfa87 | 724 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 725 | { |
993dfa87 | 726 | struct file_lock *new_fl = NULL; |
1da177e4 | 727 | struct file_lock **before; |
0f7fc9e4 | 728 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
729 | int error = 0; |
730 | int found = 0; | |
731 | ||
732 | lock_kernel(); | |
f07f18dd TM |
733 | if (request->fl_flags & FL_ACCESS) |
734 | goto find_conflict; | |
84d535ad PE |
735 | |
736 | if (request->fl_type != F_UNLCK) { | |
737 | error = -ENOMEM; | |
738 | new_fl = locks_alloc_lock(); | |
739 | if (new_fl == NULL) | |
740 | goto out; | |
741 | error = 0; | |
742 | } | |
743 | ||
1da177e4 LT |
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; | |
993dfa87 | 752 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
753 | goto out; |
754 | found = 1; | |
755 | locks_delete_lock(before); | |
756 | break; | |
757 | } | |
1da177e4 | 758 | |
f475ae95 TM |
759 | if (request->fl_type == F_UNLCK) { |
760 | if ((request->fl_flags & FL_EXISTS) && !found) | |
761 | error = -ENOENT; | |
993dfa87 | 762 | goto out; |
f475ae95 | 763 | } |
1da177e4 LT |
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) | |
c3921ab7 | 770 | cond_resched_bkl(); |
1da177e4 | 771 | |
f07f18dd | 772 | find_conflict: |
1da177e4 LT |
773 | for_each_lock(inode, before) { |
774 | struct file_lock *fl = *before; | |
775 | if (IS_POSIX(fl)) | |
776 | break; | |
777 | if (IS_LEASE(fl)) | |
778 | continue; | |
993dfa87 | 779 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
780 | continue; |
781 | error = -EAGAIN; | |
bde74e4b MS |
782 | if (!(request->fl_flags & FL_SLEEP)) |
783 | goto out; | |
784 | error = FILE_LOCK_DEFERRED; | |
785 | locks_insert_block(fl, request); | |
1da177e4 LT |
786 | goto out; |
787 | } | |
f07f18dd TM |
788 | if (request->fl_flags & FL_ACCESS) |
789 | goto out; | |
993dfa87 | 790 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 791 | locks_insert_lock(before, new_fl); |
993dfa87 | 792 | new_fl = NULL; |
9cedc194 | 793 | error = 0; |
1da177e4 LT |
794 | |
795 | out: | |
796 | unlock_kernel(); | |
993dfa87 TM |
797 | if (new_fl) |
798 | locks_free_lock(new_fl); | |
1da177e4 LT |
799 | return error; |
800 | } | |
801 | ||
150b3934 | 802 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
803 | { |
804 | struct file_lock *fl; | |
39005d02 MS |
805 | struct file_lock *new_fl = NULL; |
806 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
807 | struct file_lock *left = NULL; |
808 | struct file_lock *right = NULL; | |
809 | struct file_lock **before; | |
810 | int error, added = 0; | |
811 | ||
812 | /* | |
813 | * We may need two file_lock structures for this operation, | |
814 | * so we get them in advance to avoid races. | |
39005d02 MS |
815 | * |
816 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 817 | */ |
39005d02 MS |
818 | if (!(request->fl_flags & FL_ACCESS) && |
819 | (request->fl_type != F_UNLCK || | |
820 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
821 | new_fl = locks_alloc_lock(); | |
822 | new_fl2 = locks_alloc_lock(); | |
823 | } | |
1da177e4 LT |
824 | |
825 | lock_kernel(); | |
826 | if (request->fl_type != F_UNLCK) { | |
827 | for_each_lock(inode, before) { | |
526985b9 | 828 | fl = *before; |
1da177e4 LT |
829 | if (!IS_POSIX(fl)) |
830 | continue; | |
831 | if (!posix_locks_conflict(request, fl)) | |
832 | continue; | |
5842add2 | 833 | if (conflock) |
1a747ee0 | 834 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
835 | error = -EAGAIN; |
836 | if (!(request->fl_flags & FL_SLEEP)) | |
837 | goto out; | |
838 | error = -EDEADLK; | |
839 | if (posix_locks_deadlock(request, fl)) | |
840 | goto out; | |
bde74e4b | 841 | error = FILE_LOCK_DEFERRED; |
1da177e4 LT |
842 | locks_insert_block(fl, request); |
843 | goto out; | |
844 | } | |
845 | } | |
846 | ||
847 | /* If we're just looking for a conflict, we're done. */ | |
848 | error = 0; | |
849 | if (request->fl_flags & FL_ACCESS) | |
850 | goto out; | |
851 | ||
1da177e4 | 852 | /* |
1da177e4 LT |
853 | * Find the first old lock with the same owner as the new lock. |
854 | */ | |
855 | ||
856 | before = &inode->i_flock; | |
857 | ||
858 | /* First skip locks owned by other processes. */ | |
859 | while ((fl = *before) && (!IS_POSIX(fl) || | |
860 | !posix_same_owner(request, fl))) { | |
861 | before = &fl->fl_next; | |
862 | } | |
863 | ||
864 | /* Process locks with this owner. */ | |
865 | while ((fl = *before) && posix_same_owner(request, fl)) { | |
866 | /* Detect adjacent or overlapping regions (if same lock type) | |
867 | */ | |
868 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
869 | /* In all comparisons of start vs end, use |
870 | * "start - 1" rather than "end + 1". If end | |
871 | * is OFFSET_MAX, end + 1 will become negative. | |
872 | */ | |
1da177e4 LT |
873 | if (fl->fl_end < request->fl_start - 1) |
874 | goto next_lock; | |
875 | /* If the next lock in the list has entirely bigger | |
876 | * addresses than the new one, insert the lock here. | |
877 | */ | |
449231d6 | 878 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
879 | break; |
880 | ||
881 | /* If we come here, the new and old lock are of the | |
882 | * same type and adjacent or overlapping. Make one | |
883 | * lock yielding from the lower start address of both | |
884 | * locks to the higher end address. | |
885 | */ | |
886 | if (fl->fl_start > request->fl_start) | |
887 | fl->fl_start = request->fl_start; | |
888 | else | |
889 | request->fl_start = fl->fl_start; | |
890 | if (fl->fl_end < request->fl_end) | |
891 | fl->fl_end = request->fl_end; | |
892 | else | |
893 | request->fl_end = fl->fl_end; | |
894 | if (added) { | |
895 | locks_delete_lock(before); | |
896 | continue; | |
897 | } | |
898 | request = fl; | |
899 | added = 1; | |
900 | } | |
901 | else { | |
902 | /* Processing for different lock types is a bit | |
903 | * more complex. | |
904 | */ | |
905 | if (fl->fl_end < request->fl_start) | |
906 | goto next_lock; | |
907 | if (fl->fl_start > request->fl_end) | |
908 | break; | |
909 | if (request->fl_type == F_UNLCK) | |
910 | added = 1; | |
911 | if (fl->fl_start < request->fl_start) | |
912 | left = fl; | |
913 | /* If the next lock in the list has a higher end | |
914 | * address than the new one, insert the new one here. | |
915 | */ | |
916 | if (fl->fl_end > request->fl_end) { | |
917 | right = fl; | |
918 | break; | |
919 | } | |
920 | if (fl->fl_start >= request->fl_start) { | |
921 | /* The new lock completely replaces an old | |
922 | * one (This may happen several times). | |
923 | */ | |
924 | if (added) { | |
925 | locks_delete_lock(before); | |
926 | continue; | |
927 | } | |
928 | /* Replace the old lock with the new one. | |
929 | * Wake up anybody waiting for the old one, | |
930 | * as the change in lock type might satisfy | |
931 | * their needs. | |
932 | */ | |
933 | locks_wake_up_blocks(fl); | |
934 | fl->fl_start = request->fl_start; | |
935 | fl->fl_end = request->fl_end; | |
936 | fl->fl_type = request->fl_type; | |
47831f35 TM |
937 | locks_release_private(fl); |
938 | locks_copy_private(fl, request); | |
1da177e4 LT |
939 | request = fl; |
940 | added = 1; | |
941 | } | |
942 | } | |
943 | /* Go on to next lock. | |
944 | */ | |
945 | next_lock: | |
946 | before = &fl->fl_next; | |
947 | } | |
948 | ||
0d9a490a MS |
949 | /* |
950 | * The above code only modifies existing locks in case of | |
951 | * merging or replacing. If new lock(s) need to be inserted | |
952 | * all modifications are done bellow this, so it's safe yet to | |
953 | * bail out. | |
954 | */ | |
955 | error = -ENOLCK; /* "no luck" */ | |
956 | if (right && left == right && !new_fl2) | |
957 | goto out; | |
958 | ||
1da177e4 LT |
959 | error = 0; |
960 | if (!added) { | |
f475ae95 TM |
961 | if (request->fl_type == F_UNLCK) { |
962 | if (request->fl_flags & FL_EXISTS) | |
963 | error = -ENOENT; | |
1da177e4 | 964 | goto out; |
f475ae95 | 965 | } |
0d9a490a MS |
966 | |
967 | if (!new_fl) { | |
968 | error = -ENOLCK; | |
969 | goto out; | |
970 | } | |
1da177e4 LT |
971 | locks_copy_lock(new_fl, request); |
972 | locks_insert_lock(before, new_fl); | |
973 | new_fl = NULL; | |
974 | } | |
975 | if (right) { | |
976 | if (left == right) { | |
977 | /* The new lock breaks the old one in two pieces, | |
978 | * so we have to use the second new lock. | |
979 | */ | |
980 | left = new_fl2; | |
981 | new_fl2 = NULL; | |
982 | locks_copy_lock(left, right); | |
983 | locks_insert_lock(before, left); | |
984 | } | |
985 | right->fl_start = request->fl_end + 1; | |
986 | locks_wake_up_blocks(right); | |
987 | } | |
988 | if (left) { | |
989 | left->fl_end = request->fl_start - 1; | |
990 | locks_wake_up_blocks(left); | |
991 | } | |
992 | out: | |
993 | unlock_kernel(); | |
994 | /* | |
995 | * Free any unused locks. | |
996 | */ | |
997 | if (new_fl) | |
998 | locks_free_lock(new_fl); | |
999 | if (new_fl2) | |
1000 | locks_free_lock(new_fl2); | |
1001 | return error; | |
1002 | } | |
1003 | ||
1004 | /** | |
1005 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1006 | * @filp: The file to apply the lock to | |
1007 | * @fl: The lock to be applied | |
150b3934 | 1008 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1009 | * |
1010 | * Add a POSIX style lock to a file. | |
1011 | * We merge adjacent & overlapping locks whenever possible. | |
1012 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1013 | * |
1014 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1015 | * whether or not a lock was successfully freed by testing the return | |
1016 | * value for -ENOENT. | |
1da177e4 | 1017 | */ |
150b3934 | 1018 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1019 | struct file_lock *conflock) |
1020 | { | |
150b3934 | 1021 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
1da177e4 | 1022 | } |
150b3934 | 1023 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1024 | |
1025 | /** | |
1026 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1027 | * @filp: The file to apply the lock to | |
1028 | * @fl: The lock to be applied | |
1029 | * | |
1030 | * Add a POSIX style lock to a file. | |
1031 | * We merge adjacent & overlapping locks whenever possible. | |
1032 | * POSIX locks are sorted by owner task, then by starting address | |
1033 | */ | |
1034 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1035 | { | |
1036 | int error; | |
1037 | might_sleep (); | |
1038 | for (;;) { | |
150b3934 | 1039 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1040 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1041 | break; |
1042 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1043 | if (!error) | |
1044 | continue; | |
1045 | ||
1046 | locks_delete_block(fl); | |
1047 | break; | |
1048 | } | |
1049 | return error; | |
1050 | } | |
1051 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1052 | ||
1053 | /** | |
1054 | * locks_mandatory_locked - Check for an active lock | |
1055 | * @inode: the file to check | |
1056 | * | |
1057 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1058 | * This function is called from locks_verify_locked() only. | |
1059 | */ | |
1060 | int locks_mandatory_locked(struct inode *inode) | |
1061 | { | |
1062 | fl_owner_t owner = current->files; | |
1063 | struct file_lock *fl; | |
1064 | ||
1065 | /* | |
1066 | * Search the lock list for this inode for any POSIX locks. | |
1067 | */ | |
1068 | lock_kernel(); | |
1069 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { | |
1070 | if (!IS_POSIX(fl)) | |
1071 | continue; | |
1072 | if (fl->fl_owner != owner) | |
1073 | break; | |
1074 | } | |
1075 | unlock_kernel(); | |
1076 | return fl ? -EAGAIN : 0; | |
1077 | } | |
1078 | ||
1079 | /** | |
1080 | * locks_mandatory_area - Check for a conflicting lock | |
1081 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1082 | * for shared | |
1083 | * @inode: the file to check | |
1084 | * @filp: how the file was opened (if it was) | |
1085 | * @offset: start of area to check | |
1086 | * @count: length of area to check | |
1087 | * | |
1088 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1089 | * This function is called from rw_verify_area() and | |
1090 | * locks_verify_truncate(). | |
1091 | */ | |
1092 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1093 | struct file *filp, loff_t offset, | |
1094 | size_t count) | |
1095 | { | |
1096 | struct file_lock fl; | |
1097 | int error; | |
1098 | ||
1099 | locks_init_lock(&fl); | |
1100 | fl.fl_owner = current->files; | |
1101 | fl.fl_pid = current->tgid; | |
1102 | fl.fl_file = filp; | |
1103 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1104 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1105 | fl.fl_flags |= FL_SLEEP; | |
1106 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1107 | fl.fl_start = offset; | |
1108 | fl.fl_end = offset + count - 1; | |
1109 | ||
1110 | for (;;) { | |
150b3934 | 1111 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1112 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1113 | break; |
1114 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1115 | if (!error) { | |
1116 | /* | |
1117 | * If we've been sleeping someone might have | |
1118 | * changed the permissions behind our back. | |
1119 | */ | |
a16877ca | 1120 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1121 | continue; |
1122 | } | |
1123 | ||
1124 | locks_delete_block(&fl); | |
1125 | break; | |
1126 | } | |
1127 | ||
1128 | return error; | |
1129 | } | |
1130 | ||
1131 | EXPORT_SYMBOL(locks_mandatory_area); | |
1132 | ||
1133 | /* We already had a lease on this file; just change its type */ | |
1134 | int lease_modify(struct file_lock **before, int arg) | |
1135 | { | |
1136 | struct file_lock *fl = *before; | |
1137 | int error = assign_type(fl, arg); | |
1138 | ||
1139 | if (error) | |
1140 | return error; | |
1141 | locks_wake_up_blocks(fl); | |
1142 | if (arg == F_UNLCK) | |
1143 | locks_delete_lock(before); | |
1144 | return 0; | |
1145 | } | |
1146 | ||
1147 | EXPORT_SYMBOL(lease_modify); | |
1148 | ||
1149 | static void time_out_leases(struct inode *inode) | |
1150 | { | |
1151 | struct file_lock **before; | |
1152 | struct file_lock *fl; | |
1153 | ||
1154 | before = &inode->i_flock; | |
1155 | while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) { | |
1156 | if ((fl->fl_break_time == 0) | |
1157 | || time_before(jiffies, fl->fl_break_time)) { | |
1158 | before = &fl->fl_next; | |
1159 | continue; | |
1160 | } | |
1da177e4 LT |
1161 | lease_modify(before, fl->fl_type & ~F_INPROGRESS); |
1162 | if (fl == *before) /* lease_modify may have freed fl */ | |
1163 | before = &fl->fl_next; | |
1164 | } | |
1165 | } | |
1166 | ||
1167 | /** | |
1168 | * __break_lease - revoke all outstanding leases on file | |
1169 | * @inode: the inode of the file to return | |
1170 | * @mode: the open mode (read or write) | |
1171 | * | |
87250dd2 | 1172 | * break_lease (inlined for speed) has checked there already is at least |
1173 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1174 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1175 | * specified %O_NONBLOCK to your open(). |
1176 | */ | |
1177 | int __break_lease(struct inode *inode, unsigned int mode) | |
1178 | { | |
1179 | int error = 0, future; | |
1180 | struct file_lock *new_fl, *flock; | |
1181 | struct file_lock *fl; | |
1da177e4 LT |
1182 | unsigned long break_time; |
1183 | int i_have_this_lease = 0; | |
1184 | ||
e32b8ee2 | 1185 | new_fl = lease_alloc(NULL, mode & FMODE_WRITE ? F_WRLCK : F_RDLCK); |
1da177e4 LT |
1186 | |
1187 | lock_kernel(); | |
1188 | ||
1189 | time_out_leases(inode); | |
1190 | ||
1191 | flock = inode->i_flock; | |
1192 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1193 | goto out; | |
1194 | ||
1195 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) | |
1196 | if (fl->fl_owner == current->files) | |
1197 | i_have_this_lease = 1; | |
1198 | ||
1199 | if (mode & FMODE_WRITE) { | |
1200 | /* If we want write access, we have to revoke any lease. */ | |
1201 | future = F_UNLCK | F_INPROGRESS; | |
1202 | } else if (flock->fl_type & F_INPROGRESS) { | |
1203 | /* If the lease is already being broken, we just leave it */ | |
1204 | future = flock->fl_type; | |
1205 | } else if (flock->fl_type & F_WRLCK) { | |
1206 | /* Downgrade the exclusive lease to a read-only lease. */ | |
1207 | future = F_RDLCK | F_INPROGRESS; | |
1208 | } else { | |
1209 | /* the existing lease was read-only, so we can read too. */ | |
1210 | goto out; | |
1211 | } | |
1212 | ||
e32b8ee2 BF |
1213 | if (IS_ERR(new_fl) && !i_have_this_lease |
1214 | && ((mode & O_NONBLOCK) == 0)) { | |
1215 | error = PTR_ERR(new_fl); | |
1da177e4 LT |
1216 | goto out; |
1217 | } | |
1218 | ||
1219 | break_time = 0; | |
1220 | if (lease_break_time > 0) { | |
1221 | break_time = jiffies + lease_break_time * HZ; | |
1222 | if (break_time == 0) | |
1223 | break_time++; /* so that 0 means no break time */ | |
1224 | } | |
1225 | ||
1226 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1227 | if (fl->fl_type != future) { | |
1228 | fl->fl_type = future; | |
1229 | fl->fl_break_time = break_time; | |
1230 | /* lease must have lmops break callback */ | |
1231 | fl->fl_lmops->fl_break(fl); | |
1232 | } | |
1233 | } | |
1234 | ||
1235 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1236 | error = -EWOULDBLOCK; | |
1237 | goto out; | |
1238 | } | |
1239 | ||
1240 | restart: | |
1241 | break_time = flock->fl_break_time; | |
1242 | if (break_time != 0) { | |
1243 | break_time -= jiffies; | |
1244 | if (break_time == 0) | |
1245 | break_time++; | |
1246 | } | |
4321e01e MW |
1247 | locks_insert_block(flock, new_fl); |
1248 | error = wait_event_interruptible_timeout(new_fl->fl_wait, | |
1249 | !new_fl->fl_next, break_time); | |
1250 | __locks_delete_block(new_fl); | |
1da177e4 LT |
1251 | if (error >= 0) { |
1252 | if (error == 0) | |
1253 | time_out_leases(inode); | |
1254 | /* Wait for the next lease that has not been broken yet */ | |
1255 | for (flock = inode->i_flock; flock && IS_LEASE(flock); | |
1256 | flock = flock->fl_next) { | |
1257 | if (flock->fl_type & F_INPROGRESS) | |
1258 | goto restart; | |
1259 | } | |
1260 | error = 0; | |
1261 | } | |
1262 | ||
1263 | out: | |
1264 | unlock_kernel(); | |
e32b8ee2 | 1265 | if (!IS_ERR(new_fl)) |
1da177e4 LT |
1266 | locks_free_lock(new_fl); |
1267 | return error; | |
1268 | } | |
1269 | ||
1270 | EXPORT_SYMBOL(__break_lease); | |
1271 | ||
1272 | /** | |
a6b91919 | 1273 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1274 | * @inode: the inode |
1275 | * @time: pointer to a timespec which will contain the last modified time | |
1276 | * | |
1277 | * This is to force NFS clients to flush their caches for files with | |
1278 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1279 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1280 | */ |
1281 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1282 | { | |
1283 | struct file_lock *flock = inode->i_flock; | |
1284 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) | |
1285 | *time = current_fs_time(inode->i_sb); | |
1286 | else | |
1287 | *time = inode->i_mtime; | |
1288 | } | |
1289 | ||
1290 | EXPORT_SYMBOL(lease_get_mtime); | |
1291 | ||
1292 | /** | |
1293 | * fcntl_getlease - Enquire what lease is currently active | |
1294 | * @filp: the file | |
1295 | * | |
1296 | * The value returned by this function will be one of | |
1297 | * (if no lease break is pending): | |
1298 | * | |
1299 | * %F_RDLCK to indicate a shared lease is held. | |
1300 | * | |
1301 | * %F_WRLCK to indicate an exclusive lease is held. | |
1302 | * | |
1303 | * %F_UNLCK to indicate no lease is held. | |
1304 | * | |
1305 | * (if a lease break is pending): | |
1306 | * | |
1307 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1308 | * changed to a shared lease (or removed). | |
1309 | * | |
1310 | * %F_UNLCK to indicate the lease needs to be removed. | |
1311 | * | |
1312 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1313 | * should be returned to userspace. | |
1314 | */ | |
1315 | int fcntl_getlease(struct file *filp) | |
1316 | { | |
1317 | struct file_lock *fl; | |
1318 | int type = F_UNLCK; | |
1319 | ||
1320 | lock_kernel(); | |
0f7fc9e4 JJS |
1321 | time_out_leases(filp->f_path.dentry->d_inode); |
1322 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1323 | fl = fl->fl_next) { |
1324 | if (fl->fl_file == filp) { | |
1325 | type = fl->fl_type & ~F_INPROGRESS; | |
1326 | break; | |
1327 | } | |
1328 | } | |
1329 | unlock_kernel(); | |
1330 | return type; | |
1331 | } | |
1332 | ||
1333 | /** | |
0af1a450 | 1334 | * generic_setlease - sets a lease on an open file |
1da177e4 LT |
1335 | * @filp: file pointer |
1336 | * @arg: type of lease to obtain | |
1337 | * @flp: input - file_lock to use, output - file_lock inserted | |
1338 | * | |
1339 | * The (input) flp->fl_lmops->fl_break function is required | |
1340 | * by break_lease(). | |
1341 | * | |
1342 | * Called with kernel lock held. | |
1343 | */ | |
0af1a450 | 1344 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1345 | { |
7eaae282 | 1346 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
85c59580 | 1347 | struct file_lock *new_fl = NULL; |
0f7fc9e4 | 1348 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 LT |
1349 | struct inode *inode = dentry->d_inode; |
1350 | int error, rdlease_count = 0, wrlease_count = 0; | |
1351 | ||
6d5e8b05 BF |
1352 | if ((current->fsuid != inode->i_uid) && !capable(CAP_LEASE)) |
1353 | return -EACCES; | |
1354 | if (!S_ISREG(inode->i_mode)) | |
1355 | return -EINVAL; | |
1356 | error = security_file_lock(filp, arg); | |
1357 | if (error) | |
1358 | return error; | |
1359 | ||
1da177e4 LT |
1360 | time_out_leases(inode); |
1361 | ||
d2ab0b0c | 1362 | BUG_ON(!(*flp)->fl_lmops->fl_break); |
1da177e4 | 1363 | |
7eaae282 KZ |
1364 | lease = *flp; |
1365 | ||
288b2fd8 | 1366 | if (arg != F_UNLCK) { |
19085557 DR |
1367 | error = -ENOMEM; |
1368 | new_fl = locks_alloc_lock(); | |
1369 | if (new_fl == NULL) | |
1370 | goto out; | |
1371 | ||
288b2fd8 DR |
1372 | error = -EAGAIN; |
1373 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1374 | goto out; | |
1375 | if ((arg == F_WRLCK) | |
1376 | && ((atomic_read(&dentry->d_count) > 1) | |
1377 | || (atomic_read(&inode->i_count) > 1))) | |
1378 | goto out; | |
288b2fd8 | 1379 | } |
85c59580 | 1380 | |
1da177e4 LT |
1381 | /* |
1382 | * At this point, we know that if there is an exclusive | |
1383 | * lease on this file, then we hold it on this filp | |
1384 | * (otherwise our open of this file would have blocked). | |
1385 | * And if we are trying to acquire an exclusive lease, | |
1386 | * then the file is not open by anyone (including us) | |
1387 | * except for this filp. | |
1388 | */ | |
1389 | for (before = &inode->i_flock; | |
1390 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1391 | before = &fl->fl_next) { | |
1392 | if (lease->fl_lmops->fl_mylease(fl, lease)) | |
1393 | my_before = before; | |
1394 | else if (fl->fl_type == (F_INPROGRESS | F_UNLCK)) | |
1395 | /* | |
1396 | * Someone is in the process of opening this | |
1397 | * file for writing so we may not take an | |
1398 | * exclusive lease on it. | |
1399 | */ | |
1400 | wrlease_count++; | |
1401 | else | |
1402 | rdlease_count++; | |
1403 | } | |
1404 | ||
5fcc60c3 | 1405 | error = -EAGAIN; |
1da177e4 LT |
1406 | if ((arg == F_RDLCK && (wrlease_count > 0)) || |
1407 | (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0))) | |
1408 | goto out; | |
1409 | ||
1410 | if (my_before != NULL) { | |
75dff55a | 1411 | *flp = *my_before; |
1da177e4 LT |
1412 | error = lease->fl_lmops->fl_change(my_before, arg); |
1413 | goto out; | |
1414 | } | |
1415 | ||
1416 | error = 0; | |
1417 | if (arg == F_UNLCK) | |
1418 | goto out; | |
1419 | ||
1420 | error = -EINVAL; | |
1421 | if (!leases_enable) | |
1422 | goto out; | |
1423 | ||
85c59580 PE |
1424 | locks_copy_lock(new_fl, lease); |
1425 | locks_insert_lock(before, new_fl); | |
1da177e4 | 1426 | |
85c59580 PE |
1427 | *flp = new_fl; |
1428 | return 0; | |
1da177e4 | 1429 | |
1da177e4 | 1430 | out: |
85c59580 PE |
1431 | if (new_fl != NULL) |
1432 | locks_free_lock(new_fl); | |
1da177e4 LT |
1433 | return error; |
1434 | } | |
0af1a450 | 1435 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 LT |
1436 | |
1437 | /** | |
a9933cea | 1438 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1439 | * @filp: file pointer |
1440 | * @arg: type of lease to obtain | |
1441 | * @lease: file_lock to use | |
1442 | * | |
1443 | * Call this to establish a lease on the file. | |
f9ffed26 BF |
1444 | * The (*lease)->fl_lmops->fl_break operation must be set; if not, |
1445 | * break_lease will oops! | |
1446 | * | |
1447 | * This will call the filesystem's setlease file method, if | |
1448 | * defined. Note that there is no getlease method; instead, the | |
1449 | * filesystem setlease method should call back to setlease() to | |
1450 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1451 | * find it. Since fcntl_getlease() only reports whether the current | |
1452 | * task holds a lease, a cluster filesystem need only do this for | |
1453 | * leases held by processes on this node. | |
1454 | * | |
1455 | * There is also no break_lease method; filesystems that | |
1456 | * handle their own leases shoud break leases themselves from the | |
1457 | * filesystem's open, create, and (on truncate) setattr methods. | |
1458 | * | |
1459 | * Warning: the only current setlease methods exist only to disable | |
1460 | * leases in certain cases. More vfs changes may be required to | |
1461 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1462 | */ |
1463 | ||
a9933cea | 1464 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1465 | { |
1da177e4 LT |
1466 | int error; |
1467 | ||
1da177e4 | 1468 | lock_kernel(); |
f9ffed26 BF |
1469 | if (filp->f_op && filp->f_op->setlease) |
1470 | error = filp->f_op->setlease(filp, arg, lease); | |
1471 | else | |
0af1a450 | 1472 | error = generic_setlease(filp, arg, lease); |
1da177e4 LT |
1473 | unlock_kernel(); |
1474 | ||
1475 | return error; | |
1476 | } | |
a9933cea | 1477 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 LT |
1478 | |
1479 | /** | |
1480 | * fcntl_setlease - sets a lease on an open file | |
1481 | * @fd: open file descriptor | |
1482 | * @filp: file pointer | |
1483 | * @arg: type of lease to obtain | |
1484 | * | |
1485 | * Call this fcntl to establish a lease on the file. | |
1486 | * Note that you also need to call %F_SETSIG to | |
1487 | * receive a signal when the lease is broken. | |
1488 | */ | |
1489 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1490 | { | |
1491 | struct file_lock fl, *flp = &fl; | |
9d91cdcc | 1492 | struct inode *inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1493 | int error; |
1494 | ||
1da177e4 LT |
1495 | locks_init_lock(&fl); |
1496 | error = lease_init(filp, arg, &fl); | |
1497 | if (error) | |
1498 | return error; | |
1499 | ||
1500 | lock_kernel(); | |
1501 | ||
a9933cea | 1502 | error = vfs_setlease(filp, arg, &flp); |
dc15ae14 | 1503 | if (error || arg == F_UNLCK) |
1da177e4 LT |
1504 | goto out_unlock; |
1505 | ||
1506 | error = fasync_helper(fd, filp, 1, &flp->fl_fasync); | |
1507 | if (error < 0) { | |
6d5e8b05 | 1508 | /* remove lease just inserted by setlease */ |
1da177e4 | 1509 | flp->fl_type = F_UNLCK | F_INPROGRESS; |
6d5e8b05 | 1510 | flp->fl_break_time = jiffies - 10; |
1da177e4 LT |
1511 | time_out_leases(inode); |
1512 | goto out_unlock; | |
1513 | } | |
1514 | ||
609d7fa9 | 1515 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1da177e4 LT |
1516 | out_unlock: |
1517 | unlock_kernel(); | |
1518 | return error; | |
1519 | } | |
1520 | ||
1521 | /** | |
1522 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1523 | * @filp: The file to apply the lock to | |
1524 | * @fl: The lock to be applied | |
1525 | * | |
1526 | * Add a FLOCK style lock to a file. | |
1527 | */ | |
1528 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1529 | { | |
1530 | int error; | |
1531 | might_sleep(); | |
1532 | for (;;) { | |
1533 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1534 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1535 | break; |
1536 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1537 | if (!error) | |
1538 | continue; | |
1539 | ||
1540 | locks_delete_block(fl); | |
1541 | break; | |
1542 | } | |
1543 | return error; | |
1544 | } | |
1545 | ||
1546 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1547 | ||
1548 | /** | |
1549 | * sys_flock: - flock() system call. | |
1550 | * @fd: the file descriptor to lock. | |
1551 | * @cmd: the type of lock to apply. | |
1552 | * | |
1553 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1554 | * The @cmd can be one of | |
1555 | * | |
1556 | * %LOCK_SH -- a shared lock. | |
1557 | * | |
1558 | * %LOCK_EX -- an exclusive lock. | |
1559 | * | |
1560 | * %LOCK_UN -- remove an existing lock. | |
1561 | * | |
1562 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1563 | * | |
1564 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1565 | * processes read and write access respectively. | |
1566 | */ | |
1567 | asmlinkage long sys_flock(unsigned int fd, unsigned int cmd) | |
1568 | { | |
1569 | struct file *filp; | |
1570 | struct file_lock *lock; | |
1571 | int can_sleep, unlock; | |
1572 | int error; | |
1573 | ||
1574 | error = -EBADF; | |
1575 | filp = fget(fd); | |
1576 | if (!filp) | |
1577 | goto out; | |
1578 | ||
1579 | can_sleep = !(cmd & LOCK_NB); | |
1580 | cmd &= ~LOCK_NB; | |
1581 | unlock = (cmd == LOCK_UN); | |
1582 | ||
1583 | if (!unlock && !(cmd & LOCK_MAND) && !(filp->f_mode & 3)) | |
1584 | goto out_putf; | |
1585 | ||
1586 | error = flock_make_lock(filp, &lock, cmd); | |
1587 | if (error) | |
1588 | goto out_putf; | |
1589 | if (can_sleep) | |
1590 | lock->fl_flags |= FL_SLEEP; | |
1591 | ||
1592 | error = security_file_lock(filp, cmd); | |
1593 | if (error) | |
1594 | goto out_free; | |
1595 | ||
1596 | if (filp->f_op && filp->f_op->flock) | |
1597 | error = filp->f_op->flock(filp, | |
1598 | (can_sleep) ? F_SETLKW : F_SETLK, | |
1599 | lock); | |
1600 | else | |
1601 | error = flock_lock_file_wait(filp, lock); | |
1602 | ||
1603 | out_free: | |
993dfa87 | 1604 | locks_free_lock(lock); |
1da177e4 LT |
1605 | |
1606 | out_putf: | |
1607 | fput(filp); | |
1608 | out: | |
1609 | return error; | |
1610 | } | |
1611 | ||
3ee17abd BF |
1612 | /** |
1613 | * vfs_test_lock - test file byte range lock | |
1614 | * @filp: The file to test lock for | |
6924c554 | 1615 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1616 | * |
1617 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1618 | * setting conf->fl_type to something other than F_UNLCK. | |
1619 | */ | |
1620 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1621 | { | |
1622 | if (filp->f_op && filp->f_op->lock) | |
1623 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1624 | posix_test_lock(filp, fl); | |
1625 | return 0; | |
1626 | } | |
1627 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1628 | ||
c2fa1b8a BF |
1629 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1630 | { | |
1631 | flock->l_pid = fl->fl_pid; | |
1632 | #if BITS_PER_LONG == 32 | |
1633 | /* | |
1634 | * Make sure we can represent the posix lock via | |
1635 | * legacy 32bit flock. | |
1636 | */ | |
1637 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1638 | return -EOVERFLOW; | |
1639 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1640 | return -EOVERFLOW; | |
1641 | #endif | |
1642 | flock->l_start = fl->fl_start; | |
1643 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1644 | fl->fl_end - fl->fl_start + 1; | |
1645 | flock->l_whence = 0; | |
129a84de | 1646 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1647 | return 0; |
1648 | } | |
1649 | ||
1650 | #if BITS_PER_LONG == 32 | |
1651 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1652 | { | |
1653 | flock->l_pid = fl->fl_pid; | |
1654 | flock->l_start = fl->fl_start; | |
1655 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1656 | fl->fl_end - fl->fl_start + 1; | |
1657 | flock->l_whence = 0; | |
1658 | flock->l_type = fl->fl_type; | |
1659 | } | |
1660 | #endif | |
1661 | ||
1da177e4 LT |
1662 | /* Report the first existing lock that would conflict with l. |
1663 | * This implements the F_GETLK command of fcntl(). | |
1664 | */ | |
1665 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1666 | { | |
9d6a8c5c | 1667 | struct file_lock file_lock; |
1da177e4 LT |
1668 | struct flock flock; |
1669 | int error; | |
1670 | ||
1671 | error = -EFAULT; | |
1672 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1673 | goto out; | |
1674 | error = -EINVAL; | |
1675 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1676 | goto out; | |
1677 | ||
1678 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1679 | if (error) | |
1680 | goto out; | |
1681 | ||
3ee17abd BF |
1682 | error = vfs_test_lock(filp, &file_lock); |
1683 | if (error) | |
1684 | goto out; | |
1da177e4 | 1685 | |
9d6a8c5c ME |
1686 | flock.l_type = file_lock.fl_type; |
1687 | if (file_lock.fl_type != F_UNLCK) { | |
1688 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1689 | if (error) |
1da177e4 | 1690 | goto out; |
1da177e4 LT |
1691 | } |
1692 | error = -EFAULT; | |
1693 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1694 | error = 0; | |
1695 | out: | |
1696 | return error; | |
1697 | } | |
1698 | ||
7723ec97 ME |
1699 | /** |
1700 | * vfs_lock_file - file byte range lock | |
1701 | * @filp: The file to apply the lock to | |
1702 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1703 | * @fl: The lock to be applied | |
150b3934 ME |
1704 | * @conf: Place to return a copy of the conflicting lock, if found. |
1705 | * | |
1706 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1707 | * as the final argument. | |
1708 | * | |
1709 | * If the filesystem defines a private ->lock() method, then @conf will | |
1710 | * be left unchanged; so a caller that cares should initialize it to | |
1711 | * some acceptable default. | |
2beb6614 ME |
1712 | * |
1713 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1714 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1715 | * been granted or denied by the underlying filesystem, if (and only if) | |
1716 | * fl_grant is set. Callers expecting ->lock() to return asynchronously | |
1717 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) | |
1718 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
bde74e4b | 1719 | * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock |
2beb6614 ME |
1720 | * request completes. |
1721 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1722 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1723 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1724 | * nonzero return code and the file system should release the lock. The file |
1725 | * system is also responsible to keep a corresponding posix lock when it | |
1726 | * grants a lock so the VFS can find out which locks are locally held and do | |
1727 | * the correct lock cleanup when required. | |
1728 | * The underlying filesystem must not drop the kernel lock or call | |
bde74e4b | 1729 | * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1730 | * return code. |
7723ec97 | 1731 | */ |
150b3934 | 1732 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1733 | { |
1734 | if (filp->f_op && filp->f_op->lock) | |
1735 | return filp->f_op->lock(filp, cmd, fl); | |
1736 | else | |
150b3934 | 1737 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1738 | } |
1739 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1740 | ||
b648a6de MS |
1741 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1742 | struct file_lock *fl) | |
1743 | { | |
1744 | int error; | |
1745 | ||
1746 | error = security_file_lock(filp, fl->fl_type); | |
1747 | if (error) | |
1748 | return error; | |
1749 | ||
764c76b3 MS |
1750 | for (;;) { |
1751 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1752 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1753 | break; |
764c76b3 MS |
1754 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1755 | if (!error) | |
1756 | continue; | |
1757 | ||
1758 | locks_delete_block(fl); | |
1759 | break; | |
b648a6de MS |
1760 | } |
1761 | ||
1762 | return error; | |
1763 | } | |
1764 | ||
1da177e4 LT |
1765 | /* Apply the lock described by l to an open file descriptor. |
1766 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1767 | */ | |
c293621b PS |
1768 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1769 | struct flock __user *l) | |
1da177e4 LT |
1770 | { |
1771 | struct file_lock *file_lock = locks_alloc_lock(); | |
1772 | struct flock flock; | |
1773 | struct inode *inode; | |
0b2bac2f | 1774 | struct file *f; |
1da177e4 LT |
1775 | int error; |
1776 | ||
1777 | if (file_lock == NULL) | |
1778 | return -ENOLCK; | |
1779 | ||
1780 | /* | |
1781 | * This might block, so we do it before checking the inode. | |
1782 | */ | |
1783 | error = -EFAULT; | |
1784 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1785 | goto out; | |
1786 | ||
0f7fc9e4 | 1787 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1788 | |
1789 | /* Don't allow mandatory locks on files that may be memory mapped | |
1790 | * and shared. | |
1791 | */ | |
a16877ca | 1792 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1793 | error = -EAGAIN; |
1794 | goto out; | |
1795 | } | |
1796 | ||
c293621b | 1797 | again: |
1da177e4 LT |
1798 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1799 | if (error) | |
1800 | goto out; | |
1801 | if (cmd == F_SETLKW) { | |
1802 | file_lock->fl_flags |= FL_SLEEP; | |
1803 | } | |
1804 | ||
1805 | error = -EBADF; | |
1806 | switch (flock.l_type) { | |
1807 | case F_RDLCK: | |
1808 | if (!(filp->f_mode & FMODE_READ)) | |
1809 | goto out; | |
1810 | break; | |
1811 | case F_WRLCK: | |
1812 | if (!(filp->f_mode & FMODE_WRITE)) | |
1813 | goto out; | |
1814 | break; | |
1815 | case F_UNLCK: | |
1816 | break; | |
1817 | default: | |
1818 | error = -EINVAL; | |
1819 | goto out; | |
1820 | } | |
1821 | ||
b648a6de | 1822 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1823 | |
c293621b PS |
1824 | /* |
1825 | * Attempt to detect a close/fcntl race and recover by | |
1826 | * releasing the lock that was just acquired. | |
1827 | */ | |
0b2bac2f AV |
1828 | /* |
1829 | * we need that spin_lock here - it prevents reordering between | |
1830 | * update of inode->i_flock and check for it done in close(). | |
1831 | * rcu_read_lock() wouldn't do. | |
1832 | */ | |
1833 | spin_lock(¤t->files->file_lock); | |
1834 | f = fcheck(fd); | |
1835 | spin_unlock(¤t->files->file_lock); | |
1836 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1837 | flock.l_type = F_UNLCK; |
1838 | goto again; | |
1da177e4 LT |
1839 | } |
1840 | ||
c293621b | 1841 | out: |
1da177e4 LT |
1842 | locks_free_lock(file_lock); |
1843 | return error; | |
1844 | } | |
1845 | ||
1846 | #if BITS_PER_LONG == 32 | |
1847 | /* Report the first existing lock that would conflict with l. | |
1848 | * This implements the F_GETLK command of fcntl(). | |
1849 | */ | |
1850 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
1851 | { | |
9d6a8c5c | 1852 | struct file_lock file_lock; |
1da177e4 LT |
1853 | struct flock64 flock; |
1854 | int error; | |
1855 | ||
1856 | error = -EFAULT; | |
1857 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1858 | goto out; | |
1859 | error = -EINVAL; | |
1860 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1861 | goto out; | |
1862 | ||
1863 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
1864 | if (error) | |
1865 | goto out; | |
1866 | ||
3ee17abd BF |
1867 | error = vfs_test_lock(filp, &file_lock); |
1868 | if (error) | |
1869 | goto out; | |
1870 | ||
9d6a8c5c ME |
1871 | flock.l_type = file_lock.fl_type; |
1872 | if (file_lock.fl_type != F_UNLCK) | |
1873 | posix_lock_to_flock64(&flock, &file_lock); | |
1874 | ||
1da177e4 LT |
1875 | error = -EFAULT; |
1876 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1877 | error = 0; | |
1878 | ||
1879 | out: | |
1880 | return error; | |
1881 | } | |
1882 | ||
1883 | /* Apply the lock described by l to an open file descriptor. | |
1884 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1885 | */ | |
c293621b PS |
1886 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1887 | struct flock64 __user *l) | |
1da177e4 LT |
1888 | { |
1889 | struct file_lock *file_lock = locks_alloc_lock(); | |
1890 | struct flock64 flock; | |
1891 | struct inode *inode; | |
0b2bac2f | 1892 | struct file *f; |
1da177e4 LT |
1893 | int error; |
1894 | ||
1895 | if (file_lock == NULL) | |
1896 | return -ENOLCK; | |
1897 | ||
1898 | /* | |
1899 | * This might block, so we do it before checking the inode. | |
1900 | */ | |
1901 | error = -EFAULT; | |
1902 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1903 | goto out; | |
1904 | ||
0f7fc9e4 | 1905 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1906 | |
1907 | /* Don't allow mandatory locks on files that may be memory mapped | |
1908 | * and shared. | |
1909 | */ | |
a16877ca | 1910 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1911 | error = -EAGAIN; |
1912 | goto out; | |
1913 | } | |
1914 | ||
c293621b | 1915 | again: |
1da177e4 LT |
1916 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1917 | if (error) | |
1918 | goto out; | |
1919 | if (cmd == F_SETLKW64) { | |
1920 | file_lock->fl_flags |= FL_SLEEP; | |
1921 | } | |
1922 | ||
1923 | error = -EBADF; | |
1924 | switch (flock.l_type) { | |
1925 | case F_RDLCK: | |
1926 | if (!(filp->f_mode & FMODE_READ)) | |
1927 | goto out; | |
1928 | break; | |
1929 | case F_WRLCK: | |
1930 | if (!(filp->f_mode & FMODE_WRITE)) | |
1931 | goto out; | |
1932 | break; | |
1933 | case F_UNLCK: | |
1934 | break; | |
1935 | default: | |
1936 | error = -EINVAL; | |
1937 | goto out; | |
1938 | } | |
1939 | ||
b648a6de | 1940 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1941 | |
c293621b PS |
1942 | /* |
1943 | * Attempt to detect a close/fcntl race and recover by | |
1944 | * releasing the lock that was just acquired. | |
1945 | */ | |
0b2bac2f AV |
1946 | spin_lock(¤t->files->file_lock); |
1947 | f = fcheck(fd); | |
1948 | spin_unlock(¤t->files->file_lock); | |
1949 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1950 | flock.l_type = F_UNLCK; |
1951 | goto again; | |
1da177e4 LT |
1952 | } |
1953 | ||
1954 | out: | |
1955 | locks_free_lock(file_lock); | |
1956 | return error; | |
1957 | } | |
1958 | #endif /* BITS_PER_LONG == 32 */ | |
1959 | ||
1960 | /* | |
1961 | * This function is called when the file is being removed | |
1962 | * from the task's fd array. POSIX locks belonging to this task | |
1963 | * are deleted at this time. | |
1964 | */ | |
1965 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
1966 | { | |
ff7b86b8 | 1967 | struct file_lock lock; |
1da177e4 LT |
1968 | |
1969 | /* | |
1970 | * If there are no locks held on this file, we don't need to call | |
1971 | * posix_lock_file(). Another process could be setting a lock on this | |
1972 | * file at the same time, but we wouldn't remove that lock anyway. | |
1973 | */ | |
0f7fc9e4 | 1974 | if (!filp->f_path.dentry->d_inode->i_flock) |
1da177e4 LT |
1975 | return; |
1976 | ||
1977 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 1978 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
1979 | lock.fl_start = 0; |
1980 | lock.fl_end = OFFSET_MAX; | |
1981 | lock.fl_owner = owner; | |
1982 | lock.fl_pid = current->tgid; | |
1983 | lock.fl_file = filp; | |
1984 | lock.fl_ops = NULL; | |
1985 | lock.fl_lmops = NULL; | |
1986 | ||
150b3934 | 1987 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 1988 | |
1da177e4 LT |
1989 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
1990 | lock.fl_ops->fl_release_private(&lock); | |
1991 | } | |
1992 | ||
1993 | EXPORT_SYMBOL(locks_remove_posix); | |
1994 | ||
1995 | /* | |
1996 | * This function is called on the last close of an open file. | |
1997 | */ | |
1998 | void locks_remove_flock(struct file *filp) | |
1999 | { | |
0f7fc9e4 | 2000 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
2001 | struct file_lock *fl; |
2002 | struct file_lock **before; | |
2003 | ||
2004 | if (!inode->i_flock) | |
2005 | return; | |
2006 | ||
2007 | if (filp->f_op && filp->f_op->flock) { | |
2008 | struct file_lock fl = { | |
2009 | .fl_pid = current->tgid, | |
2010 | .fl_file = filp, | |
2011 | .fl_flags = FL_FLOCK, | |
2012 | .fl_type = F_UNLCK, | |
2013 | .fl_end = OFFSET_MAX, | |
2014 | }; | |
2015 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2016 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2017 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2018 | } |
2019 | ||
2020 | lock_kernel(); | |
2021 | before = &inode->i_flock; | |
2022 | ||
2023 | while ((fl = *before) != NULL) { | |
2024 | if (fl->fl_file == filp) { | |
c293621b | 2025 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2026 | locks_delete_lock(before); |
2027 | continue; | |
2028 | } | |
2029 | if (IS_LEASE(fl)) { | |
2030 | lease_modify(before, F_UNLCK); | |
2031 | continue; | |
2032 | } | |
2033 | /* What? */ | |
2034 | BUG(); | |
2035 | } | |
2036 | before = &fl->fl_next; | |
2037 | } | |
2038 | unlock_kernel(); | |
2039 | } | |
2040 | ||
1da177e4 LT |
2041 | /** |
2042 | * posix_unblock_lock - stop waiting for a file lock | |
2043 | * @filp: how the file was opened | |
2044 | * @waiter: the lock which was waiting | |
2045 | * | |
2046 | * lockd needs to block waiting for locks. | |
2047 | */ | |
64a318ee | 2048 | int |
1da177e4 LT |
2049 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
2050 | { | |
64a318ee BF |
2051 | int status = 0; |
2052 | ||
1da177e4 | 2053 | lock_kernel(); |
5996a298 | 2054 | if (waiter->fl_next) |
1da177e4 | 2055 | __locks_delete_block(waiter); |
64a318ee BF |
2056 | else |
2057 | status = -ENOENT; | |
5996a298 | 2058 | unlock_kernel(); |
64a318ee | 2059 | return status; |
1da177e4 LT |
2060 | } |
2061 | ||
2062 | EXPORT_SYMBOL(posix_unblock_lock); | |
2063 | ||
9b9d2ab4 ME |
2064 | /** |
2065 | * vfs_cancel_lock - file byte range unblock lock | |
2066 | * @filp: The file to apply the unblock to | |
2067 | * @fl: The lock to be unblocked | |
2068 | * | |
2069 | * Used by lock managers to cancel blocked requests | |
2070 | */ | |
2071 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2072 | { | |
2073 | if (filp->f_op && filp->f_op->lock) | |
2074 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2075 | return 0; | |
2076 | } | |
2077 | ||
2078 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2079 | ||
7f8ada98 PE |
2080 | #ifdef CONFIG_PROC_FS |
2081 | #include <linux/seq_file.h> | |
2082 | ||
2083 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
2084 | int id, char *pfx) | |
1da177e4 LT |
2085 | { |
2086 | struct inode *inode = NULL; | |
ab1f1611 VG |
2087 | unsigned int fl_pid; |
2088 | ||
2089 | if (fl->fl_nspid) | |
6c5f3e7b | 2090 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2091 | else |
2092 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2093 | |
2094 | if (fl->fl_file != NULL) | |
0f7fc9e4 | 2095 | inode = fl->fl_file->f_path.dentry->d_inode; |
1da177e4 | 2096 | |
7f8ada98 | 2097 | seq_printf(f, "%d:%s ", id, pfx); |
1da177e4 | 2098 | if (IS_POSIX(fl)) { |
7f8ada98 | 2099 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2100 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2101 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2102 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2103 | } else if (IS_FLOCK(fl)) { |
2104 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2105 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2106 | } else { |
7f8ada98 | 2107 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2108 | } |
2109 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2110 | seq_printf(f, "LEASE "); |
1da177e4 | 2111 | if (fl->fl_type & F_INPROGRESS) |
7f8ada98 | 2112 | seq_printf(f, "BREAKING "); |
1da177e4 | 2113 | else if (fl->fl_file) |
7f8ada98 | 2114 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2115 | else |
7f8ada98 | 2116 | seq_printf(f, "BREAKER "); |
1da177e4 | 2117 | } else { |
7f8ada98 | 2118 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2119 | } |
2120 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2121 | seq_printf(f, "%s ", |
1da177e4 LT |
2122 | (fl->fl_type & LOCK_READ) |
2123 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2124 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2125 | } else { | |
7f8ada98 | 2126 | seq_printf(f, "%s ", |
1da177e4 LT |
2127 | (fl->fl_type & F_INPROGRESS) |
2128 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " | |
2129 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); | |
2130 | } | |
2131 | if (inode) { | |
2132 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2133 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2134 | inode->i_sb->s_id, inode->i_ino); |
2135 | #else | |
2136 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2137 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2138 | MAJOR(inode->i_sb->s_dev), |
2139 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2140 | #endif | |
2141 | } else { | |
ab1f1611 | 2142 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2143 | } |
2144 | if (IS_POSIX(fl)) { | |
2145 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2146 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2147 | else |
7f8ada98 | 2148 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2149 | } else { |
7f8ada98 | 2150 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2151 | } |
2152 | } | |
2153 | ||
7f8ada98 | 2154 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2155 | { |
7f8ada98 | 2156 | struct file_lock *fl, *bfl; |
1da177e4 | 2157 | |
7f8ada98 | 2158 | fl = list_entry(v, struct file_lock, fl_link); |
1da177e4 | 2159 | |
7f8ada98 | 2160 | lock_get_status(f, fl, (long)f->private, ""); |
1da177e4 | 2161 | |
7f8ada98 PE |
2162 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
2163 | lock_get_status(f, bfl, (long)f->private, " ->"); | |
094f2825 | 2164 | |
7f8ada98 PE |
2165 | f->private++; |
2166 | return 0; | |
2167 | } | |
1da177e4 | 2168 | |
7f8ada98 PE |
2169 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2170 | { | |
2171 | lock_kernel(); | |
2172 | f->private = (void *)1; | |
2173 | return seq_list_start(&file_lock_list, *pos); | |
2174 | } | |
1da177e4 | 2175 | |
7f8ada98 PE |
2176 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2177 | { | |
2178 | return seq_list_next(v, &file_lock_list, pos); | |
2179 | } | |
1da177e4 | 2180 | |
7f8ada98 PE |
2181 | static void locks_stop(struct seq_file *f, void *v) |
2182 | { | |
1da177e4 | 2183 | unlock_kernel(); |
1da177e4 LT |
2184 | } |
2185 | ||
7f8ada98 PE |
2186 | struct seq_operations locks_seq_operations = { |
2187 | .start = locks_start, | |
2188 | .next = locks_next, | |
2189 | .stop = locks_stop, | |
2190 | .show = locks_show, | |
2191 | }; | |
2192 | #endif | |
2193 | ||
1da177e4 LT |
2194 | /** |
2195 | * lock_may_read - checks that the region is free of locks | |
2196 | * @inode: the inode that is being read | |
2197 | * @start: the first byte to read | |
2198 | * @len: the number of bytes to read | |
2199 | * | |
2200 | * Emulates Windows locking requirements. Whole-file | |
2201 | * mandatory locks (share modes) can prohibit a read and | |
2202 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2203 | * | |
2204 | * N.B. this function is only ever called | |
2205 | * from knfsd and ownership of locks is never checked. | |
2206 | */ | |
2207 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2208 | { | |
2209 | struct file_lock *fl; | |
2210 | int result = 1; | |
2211 | lock_kernel(); | |
2212 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { | |
2213 | if (IS_POSIX(fl)) { | |
2214 | if (fl->fl_type == F_RDLCK) | |
2215 | continue; | |
2216 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2217 | continue; | |
2218 | } else if (IS_FLOCK(fl)) { | |
2219 | if (!(fl->fl_type & LOCK_MAND)) | |
2220 | continue; | |
2221 | if (fl->fl_type & LOCK_READ) | |
2222 | continue; | |
2223 | } else | |
2224 | continue; | |
2225 | result = 0; | |
2226 | break; | |
2227 | } | |
2228 | unlock_kernel(); | |
2229 | return result; | |
2230 | } | |
2231 | ||
2232 | EXPORT_SYMBOL(lock_may_read); | |
2233 | ||
2234 | /** | |
2235 | * lock_may_write - checks that the region is free of locks | |
2236 | * @inode: the inode that is being written | |
2237 | * @start: the first byte to write | |
2238 | * @len: the number of bytes to write | |
2239 | * | |
2240 | * Emulates Windows locking requirements. Whole-file | |
2241 | * mandatory locks (share modes) can prohibit a write and | |
2242 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2243 | * | |
2244 | * N.B. this function is only ever called | |
2245 | * from knfsd and ownership of locks is never checked. | |
2246 | */ | |
2247 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2248 | { | |
2249 | struct file_lock *fl; | |
2250 | int result = 1; | |
2251 | lock_kernel(); | |
2252 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { | |
2253 | if (IS_POSIX(fl)) { | |
2254 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2255 | continue; | |
2256 | } else if (IS_FLOCK(fl)) { | |
2257 | if (!(fl->fl_type & LOCK_MAND)) | |
2258 | continue; | |
2259 | if (fl->fl_type & LOCK_WRITE) | |
2260 | continue; | |
2261 | } else | |
2262 | continue; | |
2263 | result = 0; | |
2264 | break; | |
2265 | } | |
2266 | unlock_kernel(); | |
2267 | return result; | |
2268 | } | |
2269 | ||
2270 | EXPORT_SYMBOL(lock_may_write); | |
2271 | ||
1da177e4 LT |
2272 | static int __init filelock_init(void) |
2273 | { | |
2274 | filelock_cache = kmem_cache_create("file_lock_cache", | |
2275 | sizeof(struct file_lock), 0, SLAB_PANIC, | |
20c2df83 | 2276 | init_once); |
1da177e4 LT |
2277 | return 0; |
2278 | } | |
2279 | ||
2280 | core_initcall(filelock_init); |