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. | |
395cf969 | 63 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
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> | |
1da177e4 LT |
125 | #include <linux/syscalls.h> |
126 | #include <linux/time.h> | |
4fb3a538 | 127 | #include <linux/rcupdate.h> |
ab1f1611 | 128 | #include <linux/pid_namespace.h> |
1da177e4 | 129 | |
1da177e4 LT |
130 | #include <asm/uaccess.h> |
131 | ||
132 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
133 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
134 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) | |
135 | ||
ab83fa4b BF |
136 | static bool lease_breaking(struct file_lock *fl) |
137 | { | |
778fc546 BF |
138 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
139 | } | |
140 | ||
141 | static int target_leasetype(struct file_lock *fl) | |
142 | { | |
143 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
144 | return F_UNLCK; | |
145 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
146 | return F_RDLCK; | |
147 | return fl->fl_type; | |
ab83fa4b BF |
148 | } |
149 | ||
1da177e4 LT |
150 | int leases_enable = 1; |
151 | int lease_break_time = 45; | |
152 | ||
153 | #define for_each_lock(inode, lockp) \ | |
154 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
155 | ||
26bcbf96 | 156 | static LIST_HEAD(file_lock_list); |
1da177e4 | 157 | static LIST_HEAD(blocked_list); |
72f98e72 | 158 | static DEFINE_SPINLOCK(file_lock_lock); |
1da177e4 | 159 | |
b89f4321 AB |
160 | /* |
161 | * Protects the two list heads above, plus the inode->i_flock list | |
b89f4321 AB |
162 | */ |
163 | void lock_flocks(void) | |
164 | { | |
72f98e72 | 165 | spin_lock(&file_lock_lock); |
b89f4321 AB |
166 | } |
167 | EXPORT_SYMBOL_GPL(lock_flocks); | |
168 | ||
169 | void unlock_flocks(void) | |
170 | { | |
72f98e72 | 171 | spin_unlock(&file_lock_lock); |
b89f4321 AB |
172 | } |
173 | EXPORT_SYMBOL_GPL(unlock_flocks); | |
174 | ||
e18b890b | 175 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 176 | |
ee19cc40 | 177 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 178 | { |
ee19cc40 MS |
179 | INIT_LIST_HEAD(&fl->fl_link); |
180 | INIT_LIST_HEAD(&fl->fl_block); | |
181 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
182 | } |
183 | ||
1da177e4 | 184 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 185 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 186 | { |
ee19cc40 | 187 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
188 | |
189 | if (fl) | |
ee19cc40 | 190 | locks_init_lock_heads(fl); |
a51cb91d MS |
191 | |
192 | return fl; | |
1da177e4 | 193 | } |
c5b1f0d9 | 194 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 195 | |
a9e61e25 | 196 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
197 | { |
198 | if (fl->fl_ops) { | |
199 | if (fl->fl_ops->fl_release_private) | |
200 | fl->fl_ops->fl_release_private(fl); | |
201 | fl->fl_ops = NULL; | |
202 | } | |
068535f1 | 203 | fl->fl_lmops = NULL; |
47831f35 TM |
204 | |
205 | } | |
a9e61e25 | 206 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 207 | |
1da177e4 | 208 | /* Free a lock which is not in use. */ |
05fa3135 | 209 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 210 | { |
5ce29646 MS |
211 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
212 | BUG_ON(!list_empty(&fl->fl_block)); | |
213 | BUG_ON(!list_empty(&fl->fl_link)); | |
1da177e4 | 214 | |
47831f35 | 215 | locks_release_private(fl); |
1da177e4 LT |
216 | kmem_cache_free(filelock_cache, fl); |
217 | } | |
05fa3135 | 218 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
219 | |
220 | void locks_init_lock(struct file_lock *fl) | |
221 | { | |
ee19cc40 MS |
222 | memset(fl, 0, sizeof(struct file_lock)); |
223 | locks_init_lock_heads(fl); | |
1da177e4 LT |
224 | } |
225 | ||
226 | EXPORT_SYMBOL(locks_init_lock); | |
227 | ||
47831f35 TM |
228 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
229 | { | |
230 | if (fl->fl_ops) { | |
231 | if (fl->fl_ops->fl_copy_lock) | |
232 | fl->fl_ops->fl_copy_lock(new, fl); | |
233 | new->fl_ops = fl->fl_ops; | |
234 | } | |
bb8430a2 | 235 | if (fl->fl_lmops) |
47831f35 | 236 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
237 | } |
238 | ||
1da177e4 LT |
239 | /* |
240 | * Initialize a new lock from an existing file_lock structure. | |
241 | */ | |
1a747ee0 | 242 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
243 | { |
244 | new->fl_owner = fl->fl_owner; | |
245 | new->fl_pid = fl->fl_pid; | |
0996905f | 246 | new->fl_file = NULL; |
1da177e4 LT |
247 | new->fl_flags = fl->fl_flags; |
248 | new->fl_type = fl->fl_type; | |
249 | new->fl_start = fl->fl_start; | |
250 | new->fl_end = fl->fl_end; | |
0996905f TM |
251 | new->fl_ops = NULL; |
252 | new->fl_lmops = NULL; | |
253 | } | |
3dd7b71c | 254 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
255 | |
256 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
257 | { | |
258 | locks_release_private(new); | |
259 | ||
260 | __locks_copy_lock(new, fl); | |
261 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
262 | new->fl_ops = fl->fl_ops; |
263 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
264 | |
265 | locks_copy_private(new, fl); | |
1da177e4 LT |
266 | } |
267 | ||
268 | EXPORT_SYMBOL(locks_copy_lock); | |
269 | ||
270 | static inline int flock_translate_cmd(int cmd) { | |
271 | if (cmd & LOCK_MAND) | |
272 | return cmd & (LOCK_MAND | LOCK_RW); | |
273 | switch (cmd) { | |
274 | case LOCK_SH: | |
275 | return F_RDLCK; | |
276 | case LOCK_EX: | |
277 | return F_WRLCK; | |
278 | case LOCK_UN: | |
279 | return F_UNLCK; | |
280 | } | |
281 | return -EINVAL; | |
282 | } | |
283 | ||
284 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
285 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
286 | unsigned int cmd) | |
287 | { | |
288 | struct file_lock *fl; | |
289 | int type = flock_translate_cmd(cmd); | |
290 | if (type < 0) | |
291 | return type; | |
292 | ||
293 | fl = locks_alloc_lock(); | |
294 | if (fl == NULL) | |
295 | return -ENOMEM; | |
296 | ||
297 | fl->fl_file = filp; | |
298 | fl->fl_pid = current->tgid; | |
299 | fl->fl_flags = FL_FLOCK; | |
300 | fl->fl_type = type; | |
301 | fl->fl_end = OFFSET_MAX; | |
302 | ||
303 | *lock = fl; | |
304 | return 0; | |
305 | } | |
306 | ||
0ec4f431 | 307 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
308 | { |
309 | switch (type) { | |
310 | case F_RDLCK: | |
311 | case F_WRLCK: | |
312 | case F_UNLCK: | |
313 | fl->fl_type = type; | |
314 | break; | |
315 | default: | |
316 | return -EINVAL; | |
317 | } | |
318 | return 0; | |
319 | } | |
320 | ||
321 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
322 | * style lock. | |
323 | */ | |
324 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
325 | struct flock *l) | |
326 | { | |
327 | off_t start, end; | |
328 | ||
329 | switch (l->l_whence) { | |
f5579f8c | 330 | case SEEK_SET: |
1da177e4 LT |
331 | start = 0; |
332 | break; | |
f5579f8c | 333 | case SEEK_CUR: |
1da177e4 LT |
334 | start = filp->f_pos; |
335 | break; | |
f5579f8c | 336 | case SEEK_END: |
496ad9aa | 337 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
338 | break; |
339 | default: | |
340 | return -EINVAL; | |
341 | } | |
342 | ||
343 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
344 | POSIX-2001 defines it. */ | |
345 | start += l->l_start; | |
4c780a46 TM |
346 | if (start < 0) |
347 | return -EINVAL; | |
348 | fl->fl_end = OFFSET_MAX; | |
349 | if (l->l_len > 0) { | |
350 | end = start + l->l_len - 1; | |
351 | fl->fl_end = end; | |
352 | } else if (l->l_len < 0) { | |
1da177e4 | 353 | end = start - 1; |
4c780a46 | 354 | fl->fl_end = end; |
1da177e4 | 355 | start += l->l_len; |
4c780a46 TM |
356 | if (start < 0) |
357 | return -EINVAL; | |
1da177e4 | 358 | } |
1da177e4 | 359 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
360 | if (fl->fl_end < fl->fl_start) |
361 | return -EOVERFLOW; | |
1da177e4 LT |
362 | |
363 | fl->fl_owner = current->files; | |
364 | fl->fl_pid = current->tgid; | |
365 | fl->fl_file = filp; | |
366 | fl->fl_flags = FL_POSIX; | |
367 | fl->fl_ops = NULL; | |
368 | fl->fl_lmops = NULL; | |
369 | ||
370 | return assign_type(fl, l->l_type); | |
371 | } | |
372 | ||
373 | #if BITS_PER_LONG == 32 | |
374 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
375 | struct flock64 *l) | |
376 | { | |
377 | loff_t start; | |
378 | ||
379 | switch (l->l_whence) { | |
f5579f8c | 380 | case SEEK_SET: |
1da177e4 LT |
381 | start = 0; |
382 | break; | |
f5579f8c | 383 | case SEEK_CUR: |
1da177e4 LT |
384 | start = filp->f_pos; |
385 | break; | |
f5579f8c | 386 | case SEEK_END: |
496ad9aa | 387 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
388 | break; |
389 | default: | |
390 | return -EINVAL; | |
391 | } | |
392 | ||
4c780a46 TM |
393 | start += l->l_start; |
394 | if (start < 0) | |
1da177e4 | 395 | return -EINVAL; |
4c780a46 TM |
396 | fl->fl_end = OFFSET_MAX; |
397 | if (l->l_len > 0) { | |
398 | fl->fl_end = start + l->l_len - 1; | |
399 | } else if (l->l_len < 0) { | |
400 | fl->fl_end = start - 1; | |
401 | start += l->l_len; | |
402 | if (start < 0) | |
403 | return -EINVAL; | |
404 | } | |
1da177e4 | 405 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
406 | if (fl->fl_end < fl->fl_start) |
407 | return -EOVERFLOW; | |
1da177e4 LT |
408 | |
409 | fl->fl_owner = current->files; | |
410 | fl->fl_pid = current->tgid; | |
411 | fl->fl_file = filp; | |
412 | fl->fl_flags = FL_POSIX; | |
413 | fl->fl_ops = NULL; | |
414 | fl->fl_lmops = NULL; | |
415 | ||
f32cb532 | 416 | return assign_type(fl, l->l_type); |
1da177e4 LT |
417 | } |
418 | #endif | |
419 | ||
420 | /* default lease lock manager operations */ | |
421 | static void lease_break_callback(struct file_lock *fl) | |
422 | { | |
423 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
424 | } | |
425 | ||
7b021967 | 426 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 427 | .lm_break = lease_break_callback, |
8fb47a4f | 428 | .lm_change = lease_modify, |
1da177e4 LT |
429 | }; |
430 | ||
431 | /* | |
432 | * Initialize a lease, use the default lock manager operations | |
433 | */ | |
0ec4f431 | 434 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 435 | { |
75dff55a TM |
436 | if (assign_type(fl, type) != 0) |
437 | return -EINVAL; | |
438 | ||
1da177e4 LT |
439 | fl->fl_owner = current->files; |
440 | fl->fl_pid = current->tgid; | |
441 | ||
442 | fl->fl_file = filp; | |
443 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
444 | fl->fl_start = 0; |
445 | fl->fl_end = OFFSET_MAX; | |
446 | fl->fl_ops = NULL; | |
447 | fl->fl_lmops = &lease_manager_ops; | |
448 | return 0; | |
449 | } | |
450 | ||
451 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 452 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
453 | { |
454 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 455 | int error = -ENOMEM; |
1da177e4 LT |
456 | |
457 | if (fl == NULL) | |
e32b8ee2 | 458 | return ERR_PTR(error); |
1da177e4 LT |
459 | |
460 | error = lease_init(filp, type, fl); | |
75dff55a TM |
461 | if (error) { |
462 | locks_free_lock(fl); | |
e32b8ee2 | 463 | return ERR_PTR(error); |
75dff55a | 464 | } |
e32b8ee2 | 465 | return fl; |
1da177e4 LT |
466 | } |
467 | ||
468 | /* Check if two locks overlap each other. | |
469 | */ | |
470 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
471 | { | |
472 | return ((fl1->fl_end >= fl2->fl_start) && | |
473 | (fl2->fl_end >= fl1->fl_start)); | |
474 | } | |
475 | ||
476 | /* | |
477 | * Check whether two locks have the same owner. | |
478 | */ | |
33443c42 | 479 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 480 | { |
8fb47a4f | 481 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 482 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 483 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
484 | return fl1->fl_owner == fl2->fl_owner; |
485 | } | |
486 | ||
487 | /* Remove waiter from blocker's block list. | |
488 | * When blocker ends up pointing to itself then the list is empty. | |
489 | */ | |
33443c42 | 490 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 LT |
491 | { |
492 | list_del_init(&waiter->fl_block); | |
493 | list_del_init(&waiter->fl_link); | |
494 | waiter->fl_next = NULL; | |
495 | } | |
496 | ||
497 | /* | |
498 | */ | |
1a9e64a7 | 499 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 500 | { |
b89f4321 | 501 | lock_flocks(); |
1da177e4 | 502 | __locks_delete_block(waiter); |
b89f4321 | 503 | unlock_flocks(); |
1da177e4 LT |
504 | } |
505 | ||
506 | /* Insert waiter into blocker's block list. | |
507 | * We use a circular list so that processes can be easily woken up in | |
508 | * the order they blocked. The documentation doesn't require this but | |
509 | * it seems like the reasonable thing to do. | |
510 | */ | |
511 | static void locks_insert_block(struct file_lock *blocker, | |
512 | struct file_lock *waiter) | |
513 | { | |
6dc0fe8f | 514 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 LT |
515 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
516 | waiter->fl_next = blocker; | |
517 | if (IS_POSIX(blocker)) | |
518 | list_add(&waiter->fl_link, &blocked_list); | |
519 | } | |
520 | ||
521 | /* Wake up processes blocked waiting for blocker. | |
522 | * If told to wait then schedule the processes until the block list | |
523 | * is empty, otherwise empty the block list ourselves. | |
524 | */ | |
525 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
526 | { | |
527 | while (!list_empty(&blocker->fl_block)) { | |
f0c1cd0e PE |
528 | struct file_lock *waiter; |
529 | ||
530 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
531 | struct file_lock, fl_block); |
532 | __locks_delete_block(waiter); | |
8fb47a4f BF |
533 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
534 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
535 | else |
536 | wake_up(&waiter->fl_wait); | |
537 | } | |
538 | } | |
539 | ||
540 | /* Insert file lock fl into an inode's lock list at the position indicated | |
541 | * by pos. At the same time add the lock to the global file lock list. | |
542 | */ | |
543 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
544 | { | |
545 | list_add(&fl->fl_link, &file_lock_list); | |
546 | ||
ab1f1611 VG |
547 | fl->fl_nspid = get_pid(task_tgid(current)); |
548 | ||
1da177e4 LT |
549 | /* insert into file's list */ |
550 | fl->fl_next = *pos; | |
551 | *pos = fl; | |
1da177e4 LT |
552 | } |
553 | ||
554 | /* | |
555 | * Delete a lock and then free it. | |
556 | * Wake up processes that are blocked waiting for this lock, | |
557 | * notify the FS that the lock has been cleared and | |
558 | * finally free the lock. | |
559 | */ | |
560 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
561 | { | |
562 | struct file_lock *fl = *thisfl_p; | |
563 | ||
564 | *thisfl_p = fl->fl_next; | |
565 | fl->fl_next = NULL; | |
566 | list_del_init(&fl->fl_link); | |
567 | ||
ab1f1611 VG |
568 | if (fl->fl_nspid) { |
569 | put_pid(fl->fl_nspid); | |
570 | fl->fl_nspid = NULL; | |
571 | } | |
572 | ||
1da177e4 LT |
573 | locks_wake_up_blocks(fl); |
574 | locks_free_lock(fl); | |
575 | } | |
576 | ||
577 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
578 | * checks for shared/exclusive status of overlapping locks. | |
579 | */ | |
580 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
581 | { | |
582 | if (sys_fl->fl_type == F_WRLCK) | |
583 | return 1; | |
584 | if (caller_fl->fl_type == F_WRLCK) | |
585 | return 1; | |
586 | return 0; | |
587 | } | |
588 | ||
589 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
590 | * checking before calling the locks_conflict(). | |
591 | */ | |
592 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
593 | { | |
594 | /* POSIX locks owned by the same process do not conflict with | |
595 | * each other. | |
596 | */ | |
597 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
598 | return (0); | |
599 | ||
600 | /* Check whether they overlap */ | |
601 | if (!locks_overlap(caller_fl, sys_fl)) | |
602 | return 0; | |
603 | ||
604 | return (locks_conflict(caller_fl, sys_fl)); | |
605 | } | |
606 | ||
607 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
608 | * checking before calling the locks_conflict(). | |
609 | */ | |
610 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
611 | { | |
612 | /* FLOCK locks referring to the same filp do not conflict with | |
613 | * each other. | |
614 | */ | |
615 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
616 | return (0); | |
617 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
618 | return 0; | |
619 | ||
620 | return (locks_conflict(caller_fl, sys_fl)); | |
621 | } | |
622 | ||
6d34ac19 | 623 | void |
9d6a8c5c | 624 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
625 | { |
626 | struct file_lock *cfl; | |
627 | ||
b89f4321 | 628 | lock_flocks(); |
496ad9aa | 629 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
630 | if (!IS_POSIX(cfl)) |
631 | continue; | |
b842e240 | 632 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
633 | break; |
634 | } | |
ab1f1611 | 635 | if (cfl) { |
9d6a8c5c | 636 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 637 | if (cfl->fl_nspid) |
6c5f3e7b | 638 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 639 | } else |
129a84de | 640 | fl->fl_type = F_UNLCK; |
b89f4321 | 641 | unlock_flocks(); |
6d34ac19 | 642 | return; |
1da177e4 | 643 | } |
1da177e4 LT |
644 | EXPORT_SYMBOL(posix_test_lock); |
645 | ||
b533184f BF |
646 | /* |
647 | * Deadlock detection: | |
648 | * | |
649 | * We attempt to detect deadlocks that are due purely to posix file | |
650 | * locks. | |
1da177e4 | 651 | * |
b533184f BF |
652 | * We assume that a task can be waiting for at most one lock at a time. |
653 | * So for any acquired lock, the process holding that lock may be | |
654 | * waiting on at most one other lock. That lock in turns may be held by | |
655 | * someone waiting for at most one other lock. Given a requested lock | |
656 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
657 | * follow this chain of waiters to ensure we are not about to create a | |
658 | * cycle. | |
1da177e4 | 659 | * |
b533184f BF |
660 | * Since we do this before we ever put a process to sleep on a lock, we |
661 | * are ensured that there is never a cycle; that is what guarantees that | |
662 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 663 | * |
b533184f BF |
664 | * Note: the above assumption may not be true when handling lock |
665 | * requests from a broken NFS client. It may also fail in the presence | |
666 | * of tasks (such as posix threads) sharing the same open file table. | |
667 | * | |
668 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 669 | */ |
97855b49 BF |
670 | |
671 | #define MAX_DEADLK_ITERATIONS 10 | |
672 | ||
b533184f BF |
673 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
674 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
675 | { | |
676 | struct file_lock *fl; | |
677 | ||
678 | list_for_each_entry(fl, &blocked_list, fl_link) { | |
679 | if (posix_same_owner(fl, block_fl)) | |
680 | return fl->fl_next; | |
681 | } | |
682 | return NULL; | |
683 | } | |
684 | ||
b0904e14 | 685 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
686 | struct file_lock *block_fl) |
687 | { | |
97855b49 | 688 | int i = 0; |
1da177e4 | 689 | |
b533184f BF |
690 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
691 | if (i++ > MAX_DEADLK_ITERATIONS) | |
692 | return 0; | |
693 | if (posix_same_owner(caller_fl, block_fl)) | |
694 | return 1; | |
1da177e4 LT |
695 | } |
696 | return 0; | |
697 | } | |
698 | ||
1da177e4 | 699 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 700 | * after any leases, but before any posix locks. |
f475ae95 TM |
701 | * |
702 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
703 | * whether or not a lock was successfully freed by testing the return | |
704 | * value for -ENOENT. | |
1da177e4 | 705 | */ |
993dfa87 | 706 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 707 | { |
993dfa87 | 708 | struct file_lock *new_fl = NULL; |
1da177e4 | 709 | struct file_lock **before; |
496ad9aa | 710 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
711 | int error = 0; |
712 | int found = 0; | |
713 | ||
b89f4321 | 714 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 715 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
716 | if (!new_fl) |
717 | return -ENOMEM; | |
84d535ad PE |
718 | } |
719 | ||
b89f4321 AB |
720 | lock_flocks(); |
721 | if (request->fl_flags & FL_ACCESS) | |
722 | goto find_conflict; | |
723 | ||
1da177e4 LT |
724 | for_each_lock(inode, before) { |
725 | struct file_lock *fl = *before; | |
726 | if (IS_POSIX(fl)) | |
727 | break; | |
728 | if (IS_LEASE(fl)) | |
729 | continue; | |
730 | if (filp != fl->fl_file) | |
731 | continue; | |
993dfa87 | 732 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
733 | goto out; |
734 | found = 1; | |
735 | locks_delete_lock(before); | |
736 | break; | |
737 | } | |
1da177e4 | 738 | |
f475ae95 TM |
739 | if (request->fl_type == F_UNLCK) { |
740 | if ((request->fl_flags & FL_EXISTS) && !found) | |
741 | error = -ENOENT; | |
993dfa87 | 742 | goto out; |
f475ae95 | 743 | } |
1da177e4 LT |
744 | |
745 | /* | |
746 | * If a higher-priority process was blocked on the old file lock, | |
747 | * give it the opportunity to lock the file. | |
748 | */ | |
b89f4321 AB |
749 | if (found) { |
750 | unlock_flocks(); | |
def01bc5 | 751 | cond_resched(); |
b89f4321 AB |
752 | lock_flocks(); |
753 | } | |
1da177e4 | 754 | |
f07f18dd | 755 | find_conflict: |
1da177e4 LT |
756 | for_each_lock(inode, before) { |
757 | struct file_lock *fl = *before; | |
758 | if (IS_POSIX(fl)) | |
759 | break; | |
760 | if (IS_LEASE(fl)) | |
761 | continue; | |
993dfa87 | 762 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
763 | continue; |
764 | error = -EAGAIN; | |
bde74e4b MS |
765 | if (!(request->fl_flags & FL_SLEEP)) |
766 | goto out; | |
767 | error = FILE_LOCK_DEFERRED; | |
768 | locks_insert_block(fl, request); | |
1da177e4 LT |
769 | goto out; |
770 | } | |
f07f18dd TM |
771 | if (request->fl_flags & FL_ACCESS) |
772 | goto out; | |
993dfa87 | 773 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 774 | locks_insert_lock(before, new_fl); |
993dfa87 | 775 | new_fl = NULL; |
9cedc194 | 776 | error = 0; |
1da177e4 LT |
777 | |
778 | out: | |
b89f4321 | 779 | unlock_flocks(); |
993dfa87 TM |
780 | if (new_fl) |
781 | locks_free_lock(new_fl); | |
1da177e4 LT |
782 | return error; |
783 | } | |
784 | ||
150b3934 | 785 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
786 | { |
787 | struct file_lock *fl; | |
39005d02 MS |
788 | struct file_lock *new_fl = NULL; |
789 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
790 | struct file_lock *left = NULL; |
791 | struct file_lock *right = NULL; | |
792 | struct file_lock **before; | |
793 | int error, added = 0; | |
794 | ||
795 | /* | |
796 | * We may need two file_lock structures for this operation, | |
797 | * so we get them in advance to avoid races. | |
39005d02 MS |
798 | * |
799 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 800 | */ |
39005d02 MS |
801 | if (!(request->fl_flags & FL_ACCESS) && |
802 | (request->fl_type != F_UNLCK || | |
803 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
804 | new_fl = locks_alloc_lock(); | |
805 | new_fl2 = locks_alloc_lock(); | |
806 | } | |
1da177e4 | 807 | |
b89f4321 | 808 | lock_flocks(); |
1da177e4 LT |
809 | if (request->fl_type != F_UNLCK) { |
810 | for_each_lock(inode, before) { | |
526985b9 | 811 | fl = *before; |
1da177e4 LT |
812 | if (!IS_POSIX(fl)) |
813 | continue; | |
814 | if (!posix_locks_conflict(request, fl)) | |
815 | continue; | |
5842add2 | 816 | if (conflock) |
1a747ee0 | 817 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
818 | error = -EAGAIN; |
819 | if (!(request->fl_flags & FL_SLEEP)) | |
820 | goto out; | |
821 | error = -EDEADLK; | |
822 | if (posix_locks_deadlock(request, fl)) | |
823 | goto out; | |
bde74e4b | 824 | error = FILE_LOCK_DEFERRED; |
1da177e4 LT |
825 | locks_insert_block(fl, request); |
826 | goto out; | |
827 | } | |
828 | } | |
829 | ||
830 | /* If we're just looking for a conflict, we're done. */ | |
831 | error = 0; | |
832 | if (request->fl_flags & FL_ACCESS) | |
833 | goto out; | |
834 | ||
1da177e4 | 835 | /* |
1da177e4 LT |
836 | * Find the first old lock with the same owner as the new lock. |
837 | */ | |
838 | ||
839 | before = &inode->i_flock; | |
840 | ||
841 | /* First skip locks owned by other processes. */ | |
842 | while ((fl = *before) && (!IS_POSIX(fl) || | |
843 | !posix_same_owner(request, fl))) { | |
844 | before = &fl->fl_next; | |
845 | } | |
846 | ||
847 | /* Process locks with this owner. */ | |
848 | while ((fl = *before) && posix_same_owner(request, fl)) { | |
849 | /* Detect adjacent or overlapping regions (if same lock type) | |
850 | */ | |
851 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
852 | /* In all comparisons of start vs end, use |
853 | * "start - 1" rather than "end + 1". If end | |
854 | * is OFFSET_MAX, end + 1 will become negative. | |
855 | */ | |
1da177e4 LT |
856 | if (fl->fl_end < request->fl_start - 1) |
857 | goto next_lock; | |
858 | /* If the next lock in the list has entirely bigger | |
859 | * addresses than the new one, insert the lock here. | |
860 | */ | |
449231d6 | 861 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
862 | break; |
863 | ||
864 | /* If we come here, the new and old lock are of the | |
865 | * same type and adjacent or overlapping. Make one | |
866 | * lock yielding from the lower start address of both | |
867 | * locks to the higher end address. | |
868 | */ | |
869 | if (fl->fl_start > request->fl_start) | |
870 | fl->fl_start = request->fl_start; | |
871 | else | |
872 | request->fl_start = fl->fl_start; | |
873 | if (fl->fl_end < request->fl_end) | |
874 | fl->fl_end = request->fl_end; | |
875 | else | |
876 | request->fl_end = fl->fl_end; | |
877 | if (added) { | |
878 | locks_delete_lock(before); | |
879 | continue; | |
880 | } | |
881 | request = fl; | |
882 | added = 1; | |
883 | } | |
884 | else { | |
885 | /* Processing for different lock types is a bit | |
886 | * more complex. | |
887 | */ | |
888 | if (fl->fl_end < request->fl_start) | |
889 | goto next_lock; | |
890 | if (fl->fl_start > request->fl_end) | |
891 | break; | |
892 | if (request->fl_type == F_UNLCK) | |
893 | added = 1; | |
894 | if (fl->fl_start < request->fl_start) | |
895 | left = fl; | |
896 | /* If the next lock in the list has a higher end | |
897 | * address than the new one, insert the new one here. | |
898 | */ | |
899 | if (fl->fl_end > request->fl_end) { | |
900 | right = fl; | |
901 | break; | |
902 | } | |
903 | if (fl->fl_start >= request->fl_start) { | |
904 | /* The new lock completely replaces an old | |
905 | * one (This may happen several times). | |
906 | */ | |
907 | if (added) { | |
908 | locks_delete_lock(before); | |
909 | continue; | |
910 | } | |
911 | /* Replace the old lock with the new one. | |
912 | * Wake up anybody waiting for the old one, | |
913 | * as the change in lock type might satisfy | |
914 | * their needs. | |
915 | */ | |
916 | locks_wake_up_blocks(fl); | |
917 | fl->fl_start = request->fl_start; | |
918 | fl->fl_end = request->fl_end; | |
919 | fl->fl_type = request->fl_type; | |
47831f35 TM |
920 | locks_release_private(fl); |
921 | locks_copy_private(fl, request); | |
1da177e4 LT |
922 | request = fl; |
923 | added = 1; | |
924 | } | |
925 | } | |
926 | /* Go on to next lock. | |
927 | */ | |
928 | next_lock: | |
929 | before = &fl->fl_next; | |
930 | } | |
931 | ||
0d9a490a MS |
932 | /* |
933 | * The above code only modifies existing locks in case of | |
934 | * merging or replacing. If new lock(s) need to be inserted | |
935 | * all modifications are done bellow this, so it's safe yet to | |
936 | * bail out. | |
937 | */ | |
938 | error = -ENOLCK; /* "no luck" */ | |
939 | if (right && left == right && !new_fl2) | |
940 | goto out; | |
941 | ||
1da177e4 LT |
942 | error = 0; |
943 | if (!added) { | |
f475ae95 TM |
944 | if (request->fl_type == F_UNLCK) { |
945 | if (request->fl_flags & FL_EXISTS) | |
946 | error = -ENOENT; | |
1da177e4 | 947 | goto out; |
f475ae95 | 948 | } |
0d9a490a MS |
949 | |
950 | if (!new_fl) { | |
951 | error = -ENOLCK; | |
952 | goto out; | |
953 | } | |
1da177e4 LT |
954 | locks_copy_lock(new_fl, request); |
955 | locks_insert_lock(before, new_fl); | |
956 | new_fl = NULL; | |
957 | } | |
958 | if (right) { | |
959 | if (left == right) { | |
960 | /* The new lock breaks the old one in two pieces, | |
961 | * so we have to use the second new lock. | |
962 | */ | |
963 | left = new_fl2; | |
964 | new_fl2 = NULL; | |
965 | locks_copy_lock(left, right); | |
966 | locks_insert_lock(before, left); | |
967 | } | |
968 | right->fl_start = request->fl_end + 1; | |
969 | locks_wake_up_blocks(right); | |
970 | } | |
971 | if (left) { | |
972 | left->fl_end = request->fl_start - 1; | |
973 | locks_wake_up_blocks(left); | |
974 | } | |
975 | out: | |
b89f4321 | 976 | unlock_flocks(); |
1da177e4 LT |
977 | /* |
978 | * Free any unused locks. | |
979 | */ | |
980 | if (new_fl) | |
981 | locks_free_lock(new_fl); | |
982 | if (new_fl2) | |
983 | locks_free_lock(new_fl2); | |
984 | return error; | |
985 | } | |
986 | ||
987 | /** | |
988 | * posix_lock_file - Apply a POSIX-style lock to a file | |
989 | * @filp: The file to apply the lock to | |
990 | * @fl: The lock to be applied | |
150b3934 | 991 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
992 | * |
993 | * Add a POSIX style lock to a file. | |
994 | * We merge adjacent & overlapping locks whenever possible. | |
995 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
996 | * |
997 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
998 | * whether or not a lock was successfully freed by testing the return | |
999 | * value for -ENOENT. | |
1da177e4 | 1000 | */ |
150b3934 | 1001 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1002 | struct file_lock *conflock) |
1003 | { | |
496ad9aa | 1004 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1005 | } |
150b3934 | 1006 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1007 | |
1008 | /** | |
1009 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1010 | * @filp: The file to apply the lock to | |
1011 | * @fl: The lock to be applied | |
1012 | * | |
1013 | * Add a POSIX style lock to a file. | |
1014 | * We merge adjacent & overlapping locks whenever possible. | |
1015 | * POSIX locks are sorted by owner task, then by starting address | |
1016 | */ | |
1017 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1018 | { | |
1019 | int error; | |
1020 | might_sleep (); | |
1021 | for (;;) { | |
150b3934 | 1022 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1023 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1024 | break; |
1025 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1026 | if (!error) | |
1027 | continue; | |
1028 | ||
1029 | locks_delete_block(fl); | |
1030 | break; | |
1031 | } | |
1032 | return error; | |
1033 | } | |
1034 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1035 | ||
1036 | /** | |
1037 | * locks_mandatory_locked - Check for an active lock | |
1038 | * @inode: the file to check | |
1039 | * | |
1040 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1041 | * This function is called from locks_verify_locked() only. | |
1042 | */ | |
1043 | int locks_mandatory_locked(struct inode *inode) | |
1044 | { | |
1045 | fl_owner_t owner = current->files; | |
1046 | struct file_lock *fl; | |
1047 | ||
1048 | /* | |
1049 | * Search the lock list for this inode for any POSIX locks. | |
1050 | */ | |
b89f4321 | 1051 | lock_flocks(); |
1da177e4 LT |
1052 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1053 | if (!IS_POSIX(fl)) | |
1054 | continue; | |
1055 | if (fl->fl_owner != owner) | |
1056 | break; | |
1057 | } | |
b89f4321 | 1058 | unlock_flocks(); |
1da177e4 LT |
1059 | return fl ? -EAGAIN : 0; |
1060 | } | |
1061 | ||
1062 | /** | |
1063 | * locks_mandatory_area - Check for a conflicting lock | |
1064 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1065 | * for shared | |
1066 | * @inode: the file to check | |
1067 | * @filp: how the file was opened (if it was) | |
1068 | * @offset: start of area to check | |
1069 | * @count: length of area to check | |
1070 | * | |
1071 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1072 | * This function is called from rw_verify_area() and | |
1073 | * locks_verify_truncate(). | |
1074 | */ | |
1075 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1076 | struct file *filp, loff_t offset, | |
1077 | size_t count) | |
1078 | { | |
1079 | struct file_lock fl; | |
1080 | int error; | |
1081 | ||
1082 | locks_init_lock(&fl); | |
1083 | fl.fl_owner = current->files; | |
1084 | fl.fl_pid = current->tgid; | |
1085 | fl.fl_file = filp; | |
1086 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1087 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1088 | fl.fl_flags |= FL_SLEEP; | |
1089 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1090 | fl.fl_start = offset; | |
1091 | fl.fl_end = offset + count - 1; | |
1092 | ||
1093 | for (;;) { | |
150b3934 | 1094 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1095 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1096 | break; |
1097 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1098 | if (!error) { | |
1099 | /* | |
1100 | * If we've been sleeping someone might have | |
1101 | * changed the permissions behind our back. | |
1102 | */ | |
a16877ca | 1103 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1104 | continue; |
1105 | } | |
1106 | ||
1107 | locks_delete_block(&fl); | |
1108 | break; | |
1109 | } | |
1110 | ||
1111 | return error; | |
1112 | } | |
1113 | ||
1114 | EXPORT_SYMBOL(locks_mandatory_area); | |
1115 | ||
778fc546 BF |
1116 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1117 | { | |
1118 | switch (arg) { | |
1119 | case F_UNLCK: | |
1120 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1121 | /* fall through: */ | |
1122 | case F_RDLCK: | |
1123 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1124 | } | |
1125 | } | |
1126 | ||
1da177e4 LT |
1127 | /* We already had a lease on this file; just change its type */ |
1128 | int lease_modify(struct file_lock **before, int arg) | |
1129 | { | |
1130 | struct file_lock *fl = *before; | |
1131 | int error = assign_type(fl, arg); | |
1132 | ||
1133 | if (error) | |
1134 | return error; | |
778fc546 | 1135 | lease_clear_pending(fl, arg); |
1da177e4 | 1136 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1137 | if (arg == F_UNLCK) { |
1138 | struct file *filp = fl->fl_file; | |
1139 | ||
1140 | f_delown(filp); | |
1141 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1142 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1143 | if (fl->fl_fasync != NULL) { | |
1144 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1145 | fl->fl_fasync = NULL; | |
1146 | } | |
1da177e4 | 1147 | locks_delete_lock(before); |
3b6e2723 | 1148 | } |
1da177e4 LT |
1149 | return 0; |
1150 | } | |
1151 | ||
1152 | EXPORT_SYMBOL(lease_modify); | |
1153 | ||
778fc546 BF |
1154 | static bool past_time(unsigned long then) |
1155 | { | |
1156 | if (!then) | |
1157 | /* 0 is a special value meaning "this never expires": */ | |
1158 | return false; | |
1159 | return time_after(jiffies, then); | |
1160 | } | |
1161 | ||
1da177e4 LT |
1162 | static void time_out_leases(struct inode *inode) |
1163 | { | |
1164 | struct file_lock **before; | |
1165 | struct file_lock *fl; | |
1166 | ||
1167 | before = &inode->i_flock; | |
ab83fa4b | 1168 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
778fc546 BF |
1169 | if (past_time(fl->fl_downgrade_time)) |
1170 | lease_modify(before, F_RDLCK); | |
1171 | if (past_time(fl->fl_break_time)) | |
1172 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1173 | if (fl == *before) /* lease_modify may have freed fl */ |
1174 | before = &fl->fl_next; | |
1175 | } | |
1176 | } | |
1177 | ||
1178 | /** | |
1179 | * __break_lease - revoke all outstanding leases on file | |
1180 | * @inode: the inode of the file to return | |
1181 | * @mode: the open mode (read or write) | |
1182 | * | |
87250dd2 | 1183 | * break_lease (inlined for speed) has checked there already is at least |
1184 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1185 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1186 | * specified %O_NONBLOCK to your open(). |
1187 | */ | |
1188 | int __break_lease(struct inode *inode, unsigned int mode) | |
1189 | { | |
778fc546 | 1190 | int error = 0; |
1da177e4 LT |
1191 | struct file_lock *new_fl, *flock; |
1192 | struct file_lock *fl; | |
1da177e4 LT |
1193 | unsigned long break_time; |
1194 | int i_have_this_lease = 0; | |
8737c930 | 1195 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1196 | |
8737c930 | 1197 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1198 | if (IS_ERR(new_fl)) |
1199 | return PTR_ERR(new_fl); | |
1da177e4 | 1200 | |
b89f4321 | 1201 | lock_flocks(); |
1da177e4 LT |
1202 | |
1203 | time_out_leases(inode); | |
1204 | ||
1205 | flock = inode->i_flock; | |
1206 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1207 | goto out; | |
1208 | ||
778fc546 BF |
1209 | if (!locks_conflict(flock, new_fl)) |
1210 | goto out; | |
1211 | ||
1da177e4 LT |
1212 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
1213 | if (fl->fl_owner == current->files) | |
1214 | i_have_this_lease = 1; | |
1215 | ||
1da177e4 LT |
1216 | break_time = 0; |
1217 | if (lease_break_time > 0) { | |
1218 | break_time = jiffies + lease_break_time * HZ; | |
1219 | if (break_time == 0) | |
1220 | break_time++; /* so that 0 means no break time */ | |
1221 | } | |
1222 | ||
1223 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
778fc546 BF |
1224 | if (want_write) { |
1225 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1226 | continue; | |
1227 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1228 | fl->fl_break_time = break_time; |
778fc546 BF |
1229 | } else { |
1230 | if (lease_breaking(flock)) | |
1231 | continue; | |
1232 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1233 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1234 | } |
778fc546 | 1235 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1236 | } |
1237 | ||
1238 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1239 | error = -EWOULDBLOCK; | |
1240 | goto out; | |
1241 | } | |
1242 | ||
1243 | restart: | |
1244 | break_time = flock->fl_break_time; | |
1245 | if (break_time != 0) { | |
1246 | break_time -= jiffies; | |
1247 | if (break_time == 0) | |
1248 | break_time++; | |
1249 | } | |
4321e01e | 1250 | locks_insert_block(flock, new_fl); |
b89f4321 | 1251 | unlock_flocks(); |
4321e01e MW |
1252 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1253 | !new_fl->fl_next, break_time); | |
b89f4321 | 1254 | lock_flocks(); |
4321e01e | 1255 | __locks_delete_block(new_fl); |
1da177e4 LT |
1256 | if (error >= 0) { |
1257 | if (error == 0) | |
1258 | time_out_leases(inode); | |
778fc546 BF |
1259 | /* |
1260 | * Wait for the next conflicting lease that has not been | |
1261 | * broken yet | |
1262 | */ | |
1da177e4 LT |
1263 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1264 | flock = flock->fl_next) { | |
778fc546 | 1265 | if (locks_conflict(new_fl, flock)) |
1da177e4 LT |
1266 | goto restart; |
1267 | } | |
1268 | error = 0; | |
1269 | } | |
1270 | ||
1271 | out: | |
b89f4321 | 1272 | unlock_flocks(); |
6d4b9e38 | 1273 | locks_free_lock(new_fl); |
1da177e4 LT |
1274 | return error; |
1275 | } | |
1276 | ||
1277 | EXPORT_SYMBOL(__break_lease); | |
1278 | ||
1279 | /** | |
a6b91919 | 1280 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1281 | * @inode: the inode |
1282 | * @time: pointer to a timespec which will contain the last modified time | |
1283 | * | |
1284 | * This is to force NFS clients to flush their caches for files with | |
1285 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1286 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1287 | */ |
1288 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1289 | { | |
1290 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1291 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1292 | *time = current_fs_time(inode->i_sb); |
1293 | else | |
1294 | *time = inode->i_mtime; | |
1295 | } | |
1296 | ||
1297 | EXPORT_SYMBOL(lease_get_mtime); | |
1298 | ||
1299 | /** | |
1300 | * fcntl_getlease - Enquire what lease is currently active | |
1301 | * @filp: the file | |
1302 | * | |
1303 | * The value returned by this function will be one of | |
1304 | * (if no lease break is pending): | |
1305 | * | |
1306 | * %F_RDLCK to indicate a shared lease is held. | |
1307 | * | |
1308 | * %F_WRLCK to indicate an exclusive lease is held. | |
1309 | * | |
1310 | * %F_UNLCK to indicate no lease is held. | |
1311 | * | |
1312 | * (if a lease break is pending): | |
1313 | * | |
1314 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1315 | * changed to a shared lease (or removed). | |
1316 | * | |
1317 | * %F_UNLCK to indicate the lease needs to be removed. | |
1318 | * | |
1319 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1320 | * should be returned to userspace. | |
1321 | */ | |
1322 | int fcntl_getlease(struct file *filp) | |
1323 | { | |
1324 | struct file_lock *fl; | |
1325 | int type = F_UNLCK; | |
1326 | ||
b89f4321 | 1327 | lock_flocks(); |
496ad9aa AV |
1328 | time_out_leases(file_inode(filp)); |
1329 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1330 | fl = fl->fl_next) { |
1331 | if (fl->fl_file == filp) { | |
778fc546 | 1332 | type = target_leasetype(fl); |
1da177e4 LT |
1333 | break; |
1334 | } | |
1335 | } | |
b89f4321 | 1336 | unlock_flocks(); |
1da177e4 LT |
1337 | return type; |
1338 | } | |
1339 | ||
d4f22d19 | 1340 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1341 | { |
7eaae282 | 1342 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1343 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1344 | struct inode *inode = dentry->d_inode; |
c1f24ef4 | 1345 | int error; |
1da177e4 | 1346 | |
096657b6 BF |
1347 | lease = *flp; |
1348 | ||
8335ebd9 BF |
1349 | error = -EAGAIN; |
1350 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
096657b6 | 1351 | goto out; |
8335ebd9 BF |
1352 | if ((arg == F_WRLCK) |
1353 | && ((dentry->d_count > 1) | |
1354 | || (atomic_read(&inode->i_count) > 1))) | |
096657b6 | 1355 | goto out; |
6d5e8b05 | 1356 | |
1da177e4 LT |
1357 | /* |
1358 | * At this point, we know that if there is an exclusive | |
1359 | * lease on this file, then we hold it on this filp | |
1360 | * (otherwise our open of this file would have blocked). | |
1361 | * And if we are trying to acquire an exclusive lease, | |
1362 | * then the file is not open by anyone (including us) | |
1363 | * except for this filp. | |
1364 | */ | |
c1f24ef4 | 1365 | error = -EAGAIN; |
1da177e4 LT |
1366 | for (before = &inode->i_flock; |
1367 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1368 | before = &fl->fl_next) { | |
c1f24ef4 | 1369 | if (fl->fl_file == filp) { |
1da177e4 | 1370 | my_before = before; |
c1f24ef4 BF |
1371 | continue; |
1372 | } | |
1373 | /* | |
1374 | * No exclusive leases if someone else has a lease on | |
1375 | * this file: | |
1376 | */ | |
1377 | if (arg == F_WRLCK) | |
1378 | goto out; | |
1379 | /* | |
1380 | * Modifying our existing lease is OK, but no getting a | |
1381 | * new lease if someone else is opening for write: | |
1382 | */ | |
1383 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1384 | goto out; | |
1da177e4 LT |
1385 | } |
1386 | ||
1da177e4 | 1387 | if (my_before != NULL) { |
8fb47a4f | 1388 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1389 | if (!error) |
1390 | *flp = *my_before; | |
1da177e4 LT |
1391 | goto out; |
1392 | } | |
1393 | ||
1da177e4 LT |
1394 | error = -EINVAL; |
1395 | if (!leases_enable) | |
1396 | goto out; | |
1397 | ||
c5b1f0d9 | 1398 | locks_insert_lock(before, lease); |
85c59580 | 1399 | return 0; |
1da177e4 | 1400 | |
1da177e4 LT |
1401 | out: |
1402 | return error; | |
1403 | } | |
8335ebd9 | 1404 | |
d4f22d19 | 1405 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1406 | { |
1407 | struct file_lock *fl, **before; | |
1408 | struct dentry *dentry = filp->f_path.dentry; | |
1409 | struct inode *inode = dentry->d_inode; | |
1410 | ||
1411 | for (before = &inode->i_flock; | |
1412 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1413 | before = &fl->fl_next) { | |
1414 | if (fl->fl_file != filp) | |
1415 | continue; | |
1416 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1417 | } | |
1418 | return -EAGAIN; | |
1419 | } | |
1420 | ||
1421 | /** | |
1422 | * generic_setlease - sets a lease on an open file | |
1423 | * @filp: file pointer | |
1424 | * @arg: type of lease to obtain | |
1425 | * @flp: input - file_lock to use, output - file_lock inserted | |
1426 | * | |
1427 | * The (input) flp->fl_lmops->lm_break function is required | |
1428 | * by break_lease(). | |
1429 | * | |
1430 | * Called with file_lock_lock held. | |
1431 | */ | |
1432 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1433 | { | |
1434 | struct dentry *dentry = filp->f_path.dentry; | |
1435 | struct inode *inode = dentry->d_inode; | |
1436 | int error; | |
1437 | ||
8e96e3b7 | 1438 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1439 | return -EACCES; |
1440 | if (!S_ISREG(inode->i_mode)) | |
1441 | return -EINVAL; | |
1442 | error = security_file_lock(filp, arg); | |
1443 | if (error) | |
1444 | return error; | |
1445 | ||
1446 | time_out_leases(inode); | |
1447 | ||
1448 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1449 | ||
1450 | switch (arg) { | |
1451 | case F_UNLCK: | |
1452 | return generic_delete_lease(filp, flp); | |
1453 | case F_RDLCK: | |
1454 | case F_WRLCK: | |
1455 | return generic_add_lease(filp, arg, flp); | |
1456 | default: | |
8d657eb3 | 1457 | return -EINVAL; |
8335ebd9 BF |
1458 | } |
1459 | } | |
0af1a450 | 1460 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1461 | |
b89f4321 AB |
1462 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1463 | { | |
1464 | if (filp->f_op && filp->f_op->setlease) | |
1465 | return filp->f_op->setlease(filp, arg, lease); | |
1466 | else | |
1467 | return generic_setlease(filp, arg, lease); | |
1468 | } | |
1469 | ||
1470 | /** | |
a9933cea | 1471 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1472 | * @filp: file pointer |
1473 | * @arg: type of lease to obtain | |
1474 | * @lease: file_lock to use | |
1475 | * | |
1476 | * Call this to establish a lease on the file. | |
8fb47a4f | 1477 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1478 | * break_lease will oops! |
1479 | * | |
1480 | * This will call the filesystem's setlease file method, if | |
1481 | * defined. Note that there is no getlease method; instead, the | |
1482 | * filesystem setlease method should call back to setlease() to | |
1483 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1484 | * find it. Since fcntl_getlease() only reports whether the current | |
1485 | * task holds a lease, a cluster filesystem need only do this for | |
1486 | * leases held by processes on this node. | |
1487 | * | |
1488 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1489 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1490 | * filesystem's open, create, and (on truncate) setattr methods. |
1491 | * | |
1492 | * Warning: the only current setlease methods exist only to disable | |
1493 | * leases in certain cases. More vfs changes may be required to | |
1494 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1495 | */ |
1496 | ||
a9933cea | 1497 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1498 | { |
1da177e4 LT |
1499 | int error; |
1500 | ||
b89f4321 AB |
1501 | lock_flocks(); |
1502 | error = __vfs_setlease(filp, arg, lease); | |
1503 | unlock_flocks(); | |
1da177e4 LT |
1504 | |
1505 | return error; | |
1506 | } | |
a9933cea | 1507 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1508 | |
0ceaf6c7 BF |
1509 | static int do_fcntl_delete_lease(struct file *filp) |
1510 | { | |
1511 | struct file_lock fl, *flp = &fl; | |
1512 | ||
1513 | lease_init(filp, F_UNLCK, flp); | |
1514 | ||
1515 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1516 | } | |
1517 | ||
1518 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1519 | { |
3df057ac | 1520 | struct file_lock *fl, *ret; |
f7347ce4 | 1521 | struct fasync_struct *new; |
1da177e4 LT |
1522 | int error; |
1523 | ||
c5b1f0d9 AB |
1524 | fl = lease_alloc(filp, arg); |
1525 | if (IS_ERR(fl)) | |
1526 | return PTR_ERR(fl); | |
1da177e4 | 1527 | |
f7347ce4 LT |
1528 | new = fasync_alloc(); |
1529 | if (!new) { | |
1530 | locks_free_lock(fl); | |
1531 | return -ENOMEM; | |
1532 | } | |
3df057ac | 1533 | ret = fl; |
b89f4321 | 1534 | lock_flocks(); |
8896b93f | 1535 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 CH |
1536 | if (error) { |
1537 | unlock_flocks(); | |
1538 | locks_free_lock(fl); | |
1539 | goto out_free_fasync; | |
1540 | } | |
3df057ac BF |
1541 | if (ret != fl) |
1542 | locks_free_lock(fl); | |
1da177e4 | 1543 | |
f7347ce4 LT |
1544 | /* |
1545 | * fasync_insert_entry() returns the old entry if any. | |
1546 | * If there was no old entry, then it used 'new' and | |
1547 | * inserted it into the fasync list. Clear new so that | |
1548 | * we don't release it here. | |
1549 | */ | |
3df057ac | 1550 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1551 | new = NULL; |
1552 | ||
8896b93f | 1553 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
b89f4321 | 1554 | unlock_flocks(); |
51ee4b84 CH |
1555 | |
1556 | out_free_fasync: | |
f7347ce4 LT |
1557 | if (new) |
1558 | fasync_free(new); | |
1da177e4 LT |
1559 | return error; |
1560 | } | |
1561 | ||
0ceaf6c7 BF |
1562 | /** |
1563 | * fcntl_setlease - sets a lease on an open file | |
1564 | * @fd: open file descriptor | |
1565 | * @filp: file pointer | |
1566 | * @arg: type of lease to obtain | |
1567 | * | |
1568 | * Call this fcntl to establish a lease on the file. | |
1569 | * Note that you also need to call %F_SETSIG to | |
1570 | * receive a signal when the lease is broken. | |
1571 | */ | |
1572 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1573 | { | |
1574 | if (arg == F_UNLCK) | |
1575 | return do_fcntl_delete_lease(filp); | |
1576 | return do_fcntl_add_lease(fd, filp, arg); | |
1577 | } | |
1578 | ||
1da177e4 LT |
1579 | /** |
1580 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1581 | * @filp: The file to apply the lock to | |
1582 | * @fl: The lock to be applied | |
1583 | * | |
1584 | * Add a FLOCK style lock to a file. | |
1585 | */ | |
1586 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1587 | { | |
1588 | int error; | |
1589 | might_sleep(); | |
1590 | for (;;) { | |
1591 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1592 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1593 | break; |
1594 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1595 | if (!error) | |
1596 | continue; | |
1597 | ||
1598 | locks_delete_block(fl); | |
1599 | break; | |
1600 | } | |
1601 | return error; | |
1602 | } | |
1603 | ||
1604 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1605 | ||
1606 | /** | |
1607 | * sys_flock: - flock() system call. | |
1608 | * @fd: the file descriptor to lock. | |
1609 | * @cmd: the type of lock to apply. | |
1610 | * | |
1611 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1612 | * The @cmd can be one of | |
1613 | * | |
1614 | * %LOCK_SH -- a shared lock. | |
1615 | * | |
1616 | * %LOCK_EX -- an exclusive lock. | |
1617 | * | |
1618 | * %LOCK_UN -- remove an existing lock. | |
1619 | * | |
1620 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1621 | * | |
1622 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1623 | * processes read and write access respectively. | |
1624 | */ | |
002c8976 | 1625 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1626 | { |
2903ff01 | 1627 | struct fd f = fdget(fd); |
1da177e4 LT |
1628 | struct file_lock *lock; |
1629 | int can_sleep, unlock; | |
1630 | int error; | |
1631 | ||
1632 | error = -EBADF; | |
2903ff01 | 1633 | if (!f.file) |
1da177e4 LT |
1634 | goto out; |
1635 | ||
1636 | can_sleep = !(cmd & LOCK_NB); | |
1637 | cmd &= ~LOCK_NB; | |
1638 | unlock = (cmd == LOCK_UN); | |
1639 | ||
aeb5d727 | 1640 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1641 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1642 | goto out_putf; |
1643 | ||
2903ff01 | 1644 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1645 | if (error) |
1646 | goto out_putf; | |
1647 | if (can_sleep) | |
1648 | lock->fl_flags |= FL_SLEEP; | |
1649 | ||
2903ff01 | 1650 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1651 | if (error) |
1652 | goto out_free; | |
1653 | ||
2903ff01 AV |
1654 | if (f.file->f_op && f.file->f_op->flock) |
1655 | error = f.file->f_op->flock(f.file, | |
1da177e4 LT |
1656 | (can_sleep) ? F_SETLKW : F_SETLK, |
1657 | lock); | |
1658 | else | |
2903ff01 | 1659 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1660 | |
1661 | out_free: | |
993dfa87 | 1662 | locks_free_lock(lock); |
1da177e4 LT |
1663 | |
1664 | out_putf: | |
2903ff01 | 1665 | fdput(f); |
1da177e4 LT |
1666 | out: |
1667 | return error; | |
1668 | } | |
1669 | ||
3ee17abd BF |
1670 | /** |
1671 | * vfs_test_lock - test file byte range lock | |
1672 | * @filp: The file to test lock for | |
6924c554 | 1673 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1674 | * |
1675 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1676 | * setting conf->fl_type to something other than F_UNLCK. | |
1677 | */ | |
1678 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1679 | { | |
1680 | if (filp->f_op && filp->f_op->lock) | |
1681 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1682 | posix_test_lock(filp, fl); | |
1683 | return 0; | |
1684 | } | |
1685 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1686 | ||
c2fa1b8a BF |
1687 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1688 | { | |
1689 | flock->l_pid = fl->fl_pid; | |
1690 | #if BITS_PER_LONG == 32 | |
1691 | /* | |
1692 | * Make sure we can represent the posix lock via | |
1693 | * legacy 32bit flock. | |
1694 | */ | |
1695 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1696 | return -EOVERFLOW; | |
1697 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1698 | return -EOVERFLOW; | |
1699 | #endif | |
1700 | flock->l_start = fl->fl_start; | |
1701 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1702 | fl->fl_end - fl->fl_start + 1; | |
1703 | flock->l_whence = 0; | |
129a84de | 1704 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1705 | return 0; |
1706 | } | |
1707 | ||
1708 | #if BITS_PER_LONG == 32 | |
1709 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1710 | { | |
1711 | flock->l_pid = fl->fl_pid; | |
1712 | flock->l_start = fl->fl_start; | |
1713 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1714 | fl->fl_end - fl->fl_start + 1; | |
1715 | flock->l_whence = 0; | |
1716 | flock->l_type = fl->fl_type; | |
1717 | } | |
1718 | #endif | |
1719 | ||
1da177e4 LT |
1720 | /* Report the first existing lock that would conflict with l. |
1721 | * This implements the F_GETLK command of fcntl(). | |
1722 | */ | |
1723 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1724 | { | |
9d6a8c5c | 1725 | struct file_lock file_lock; |
1da177e4 LT |
1726 | struct flock flock; |
1727 | int error; | |
1728 | ||
1729 | error = -EFAULT; | |
1730 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1731 | goto out; | |
1732 | error = -EINVAL; | |
1733 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1734 | goto out; | |
1735 | ||
1736 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1737 | if (error) | |
1738 | goto out; | |
1739 | ||
3ee17abd BF |
1740 | error = vfs_test_lock(filp, &file_lock); |
1741 | if (error) | |
1742 | goto out; | |
1da177e4 | 1743 | |
9d6a8c5c ME |
1744 | flock.l_type = file_lock.fl_type; |
1745 | if (file_lock.fl_type != F_UNLCK) { | |
1746 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1747 | if (error) |
1da177e4 | 1748 | goto out; |
1da177e4 LT |
1749 | } |
1750 | error = -EFAULT; | |
1751 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1752 | error = 0; | |
1753 | out: | |
1754 | return error; | |
1755 | } | |
1756 | ||
7723ec97 ME |
1757 | /** |
1758 | * vfs_lock_file - file byte range lock | |
1759 | * @filp: The file to apply the lock to | |
1760 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1761 | * @fl: The lock to be applied | |
150b3934 ME |
1762 | * @conf: Place to return a copy of the conflicting lock, if found. |
1763 | * | |
1764 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1765 | * as the final argument. | |
1766 | * | |
1767 | * If the filesystem defines a private ->lock() method, then @conf will | |
1768 | * be left unchanged; so a caller that cares should initialize it to | |
1769 | * some acceptable default. | |
2beb6614 ME |
1770 | * |
1771 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1772 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1773 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1774 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1775 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1776 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1777 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1778 | * request completes. |
1779 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1780 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1781 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1782 | * nonzero return code and the file system should release the lock. The file |
1783 | * system is also responsible to keep a corresponding posix lock when it | |
1784 | * grants a lock so the VFS can find out which locks are locally held and do | |
1785 | * the correct lock cleanup when required. | |
1786 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1787 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1788 | * return code. |
7723ec97 | 1789 | */ |
150b3934 | 1790 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1791 | { |
1792 | if (filp->f_op && filp->f_op->lock) | |
1793 | return filp->f_op->lock(filp, cmd, fl); | |
1794 | else | |
150b3934 | 1795 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1796 | } |
1797 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1798 | ||
b648a6de MS |
1799 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1800 | struct file_lock *fl) | |
1801 | { | |
1802 | int error; | |
1803 | ||
1804 | error = security_file_lock(filp, fl->fl_type); | |
1805 | if (error) | |
1806 | return error; | |
1807 | ||
764c76b3 MS |
1808 | for (;;) { |
1809 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1810 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1811 | break; |
764c76b3 MS |
1812 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1813 | if (!error) | |
1814 | continue; | |
1815 | ||
1816 | locks_delete_block(fl); | |
1817 | break; | |
b648a6de MS |
1818 | } |
1819 | ||
1820 | return error; | |
1821 | } | |
1822 | ||
1da177e4 LT |
1823 | /* Apply the lock described by l to an open file descriptor. |
1824 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1825 | */ | |
c293621b PS |
1826 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1827 | struct flock __user *l) | |
1da177e4 LT |
1828 | { |
1829 | struct file_lock *file_lock = locks_alloc_lock(); | |
1830 | struct flock flock; | |
1831 | struct inode *inode; | |
0b2bac2f | 1832 | struct file *f; |
1da177e4 LT |
1833 | int error; |
1834 | ||
1835 | if (file_lock == NULL) | |
1836 | return -ENOLCK; | |
1837 | ||
1838 | /* | |
1839 | * This might block, so we do it before checking the inode. | |
1840 | */ | |
1841 | error = -EFAULT; | |
1842 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1843 | goto out; | |
1844 | ||
496ad9aa | 1845 | inode = file_inode(filp); |
1da177e4 LT |
1846 | |
1847 | /* Don't allow mandatory locks on files that may be memory mapped | |
1848 | * and shared. | |
1849 | */ | |
a16877ca | 1850 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1851 | error = -EAGAIN; |
1852 | goto out; | |
1853 | } | |
1854 | ||
c293621b | 1855 | again: |
1da177e4 LT |
1856 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1857 | if (error) | |
1858 | goto out; | |
1859 | if (cmd == F_SETLKW) { | |
1860 | file_lock->fl_flags |= FL_SLEEP; | |
1861 | } | |
1862 | ||
1863 | error = -EBADF; | |
1864 | switch (flock.l_type) { | |
1865 | case F_RDLCK: | |
1866 | if (!(filp->f_mode & FMODE_READ)) | |
1867 | goto out; | |
1868 | break; | |
1869 | case F_WRLCK: | |
1870 | if (!(filp->f_mode & FMODE_WRITE)) | |
1871 | goto out; | |
1872 | break; | |
1873 | case F_UNLCK: | |
1874 | break; | |
1875 | default: | |
1876 | error = -EINVAL; | |
1877 | goto out; | |
1878 | } | |
1879 | ||
b648a6de | 1880 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1881 | |
c293621b PS |
1882 | /* |
1883 | * Attempt to detect a close/fcntl race and recover by | |
1884 | * releasing the lock that was just acquired. | |
1885 | */ | |
0b2bac2f AV |
1886 | /* |
1887 | * we need that spin_lock here - it prevents reordering between | |
1888 | * update of inode->i_flock and check for it done in close(). | |
1889 | * rcu_read_lock() wouldn't do. | |
1890 | */ | |
1891 | spin_lock(¤t->files->file_lock); | |
1892 | f = fcheck(fd); | |
1893 | spin_unlock(¤t->files->file_lock); | |
1894 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1895 | flock.l_type = F_UNLCK; |
1896 | goto again; | |
1da177e4 LT |
1897 | } |
1898 | ||
c293621b | 1899 | out: |
1da177e4 LT |
1900 | locks_free_lock(file_lock); |
1901 | return error; | |
1902 | } | |
1903 | ||
1904 | #if BITS_PER_LONG == 32 | |
1905 | /* Report the first existing lock that would conflict with l. | |
1906 | * This implements the F_GETLK command of fcntl(). | |
1907 | */ | |
1908 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
1909 | { | |
9d6a8c5c | 1910 | struct file_lock file_lock; |
1da177e4 LT |
1911 | struct flock64 flock; |
1912 | int error; | |
1913 | ||
1914 | error = -EFAULT; | |
1915 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1916 | goto out; | |
1917 | error = -EINVAL; | |
1918 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1919 | goto out; | |
1920 | ||
1921 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
1922 | if (error) | |
1923 | goto out; | |
1924 | ||
3ee17abd BF |
1925 | error = vfs_test_lock(filp, &file_lock); |
1926 | if (error) | |
1927 | goto out; | |
1928 | ||
9d6a8c5c ME |
1929 | flock.l_type = file_lock.fl_type; |
1930 | if (file_lock.fl_type != F_UNLCK) | |
1931 | posix_lock_to_flock64(&flock, &file_lock); | |
1932 | ||
1da177e4 LT |
1933 | error = -EFAULT; |
1934 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1935 | error = 0; | |
1936 | ||
1937 | out: | |
1938 | return error; | |
1939 | } | |
1940 | ||
1941 | /* Apply the lock described by l to an open file descriptor. | |
1942 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1943 | */ | |
c293621b PS |
1944 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1945 | struct flock64 __user *l) | |
1da177e4 LT |
1946 | { |
1947 | struct file_lock *file_lock = locks_alloc_lock(); | |
1948 | struct flock64 flock; | |
1949 | struct inode *inode; | |
0b2bac2f | 1950 | struct file *f; |
1da177e4 LT |
1951 | int error; |
1952 | ||
1953 | if (file_lock == NULL) | |
1954 | return -ENOLCK; | |
1955 | ||
1956 | /* | |
1957 | * This might block, so we do it before checking the inode. | |
1958 | */ | |
1959 | error = -EFAULT; | |
1960 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1961 | goto out; | |
1962 | ||
496ad9aa | 1963 | inode = file_inode(filp); |
1da177e4 LT |
1964 | |
1965 | /* Don't allow mandatory locks on files that may be memory mapped | |
1966 | * and shared. | |
1967 | */ | |
a16877ca | 1968 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1969 | error = -EAGAIN; |
1970 | goto out; | |
1971 | } | |
1972 | ||
c293621b | 1973 | again: |
1da177e4 LT |
1974 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1975 | if (error) | |
1976 | goto out; | |
1977 | if (cmd == F_SETLKW64) { | |
1978 | file_lock->fl_flags |= FL_SLEEP; | |
1979 | } | |
1980 | ||
1981 | error = -EBADF; | |
1982 | switch (flock.l_type) { | |
1983 | case F_RDLCK: | |
1984 | if (!(filp->f_mode & FMODE_READ)) | |
1985 | goto out; | |
1986 | break; | |
1987 | case F_WRLCK: | |
1988 | if (!(filp->f_mode & FMODE_WRITE)) | |
1989 | goto out; | |
1990 | break; | |
1991 | case F_UNLCK: | |
1992 | break; | |
1993 | default: | |
1994 | error = -EINVAL; | |
1995 | goto out; | |
1996 | } | |
1997 | ||
b648a6de | 1998 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1999 | |
c293621b PS |
2000 | /* |
2001 | * Attempt to detect a close/fcntl race and recover by | |
2002 | * releasing the lock that was just acquired. | |
2003 | */ | |
0b2bac2f AV |
2004 | spin_lock(¤t->files->file_lock); |
2005 | f = fcheck(fd); | |
2006 | spin_unlock(¤t->files->file_lock); | |
2007 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2008 | flock.l_type = F_UNLCK; |
2009 | goto again; | |
1da177e4 LT |
2010 | } |
2011 | ||
2012 | out: | |
2013 | locks_free_lock(file_lock); | |
2014 | return error; | |
2015 | } | |
2016 | #endif /* BITS_PER_LONG == 32 */ | |
2017 | ||
2018 | /* | |
2019 | * This function is called when the file is being removed | |
2020 | * from the task's fd array. POSIX locks belonging to this task | |
2021 | * are deleted at this time. | |
2022 | */ | |
2023 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2024 | { | |
ff7b86b8 | 2025 | struct file_lock lock; |
1da177e4 LT |
2026 | |
2027 | /* | |
2028 | * If there are no locks held on this file, we don't need to call | |
2029 | * posix_lock_file(). Another process could be setting a lock on this | |
2030 | * file at the same time, but we wouldn't remove that lock anyway. | |
2031 | */ | |
496ad9aa | 2032 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2033 | return; |
2034 | ||
2035 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2036 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2037 | lock.fl_start = 0; |
2038 | lock.fl_end = OFFSET_MAX; | |
2039 | lock.fl_owner = owner; | |
2040 | lock.fl_pid = current->tgid; | |
2041 | lock.fl_file = filp; | |
2042 | lock.fl_ops = NULL; | |
2043 | lock.fl_lmops = NULL; | |
2044 | ||
150b3934 | 2045 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2046 | |
1da177e4 LT |
2047 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2048 | lock.fl_ops->fl_release_private(&lock); | |
2049 | } | |
2050 | ||
2051 | EXPORT_SYMBOL(locks_remove_posix); | |
2052 | ||
2053 | /* | |
2054 | * This function is called on the last close of an open file. | |
2055 | */ | |
2056 | void locks_remove_flock(struct file *filp) | |
2057 | { | |
496ad9aa | 2058 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2059 | struct file_lock *fl; |
2060 | struct file_lock **before; | |
2061 | ||
2062 | if (!inode->i_flock) | |
2063 | return; | |
2064 | ||
2065 | if (filp->f_op && filp->f_op->flock) { | |
2066 | struct file_lock fl = { | |
2067 | .fl_pid = current->tgid, | |
2068 | .fl_file = filp, | |
2069 | .fl_flags = FL_FLOCK, | |
2070 | .fl_type = F_UNLCK, | |
2071 | .fl_end = OFFSET_MAX, | |
2072 | }; | |
2073 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2074 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2075 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2076 | } |
2077 | ||
b89f4321 | 2078 | lock_flocks(); |
1da177e4 LT |
2079 | before = &inode->i_flock; |
2080 | ||
2081 | while ((fl = *before) != NULL) { | |
2082 | if (fl->fl_file == filp) { | |
c293621b | 2083 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2084 | locks_delete_lock(before); |
2085 | continue; | |
2086 | } | |
2087 | if (IS_LEASE(fl)) { | |
2088 | lease_modify(before, F_UNLCK); | |
2089 | continue; | |
2090 | } | |
2091 | /* What? */ | |
2092 | BUG(); | |
2093 | } | |
2094 | before = &fl->fl_next; | |
2095 | } | |
b89f4321 | 2096 | unlock_flocks(); |
1da177e4 LT |
2097 | } |
2098 | ||
1da177e4 LT |
2099 | /** |
2100 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2101 | * @waiter: the lock which was waiting |
2102 | * | |
2103 | * lockd needs to block waiting for locks. | |
2104 | */ | |
64a318ee | 2105 | int |
f891a29f | 2106 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2107 | { |
64a318ee BF |
2108 | int status = 0; |
2109 | ||
b89f4321 | 2110 | lock_flocks(); |
5996a298 | 2111 | if (waiter->fl_next) |
1da177e4 | 2112 | __locks_delete_block(waiter); |
64a318ee BF |
2113 | else |
2114 | status = -ENOENT; | |
b89f4321 | 2115 | unlock_flocks(); |
64a318ee | 2116 | return status; |
1da177e4 | 2117 | } |
1da177e4 LT |
2118 | EXPORT_SYMBOL(posix_unblock_lock); |
2119 | ||
9b9d2ab4 ME |
2120 | /** |
2121 | * vfs_cancel_lock - file byte range unblock lock | |
2122 | * @filp: The file to apply the unblock to | |
2123 | * @fl: The lock to be unblocked | |
2124 | * | |
2125 | * Used by lock managers to cancel blocked requests | |
2126 | */ | |
2127 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2128 | { | |
2129 | if (filp->f_op && filp->f_op->lock) | |
2130 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2131 | return 0; | |
2132 | } | |
2133 | ||
2134 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2135 | ||
7f8ada98 | 2136 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2137 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2138 | #include <linux/seq_file.h> |
2139 | ||
2140 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
99dc8292 | 2141 | loff_t id, char *pfx) |
1da177e4 LT |
2142 | { |
2143 | struct inode *inode = NULL; | |
ab1f1611 VG |
2144 | unsigned int fl_pid; |
2145 | ||
2146 | if (fl->fl_nspid) | |
6c5f3e7b | 2147 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2148 | else |
2149 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2150 | |
2151 | if (fl->fl_file != NULL) | |
496ad9aa | 2152 | inode = file_inode(fl->fl_file); |
1da177e4 | 2153 | |
99dc8292 | 2154 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2155 | if (IS_POSIX(fl)) { |
7f8ada98 | 2156 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2157 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2158 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2159 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2160 | } else if (IS_FLOCK(fl)) { |
2161 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2162 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2163 | } else { |
7f8ada98 | 2164 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2165 | } |
2166 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2167 | seq_printf(f, "LEASE "); |
ab83fa4b | 2168 | if (lease_breaking(fl)) |
7f8ada98 | 2169 | seq_printf(f, "BREAKING "); |
1da177e4 | 2170 | else if (fl->fl_file) |
7f8ada98 | 2171 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2172 | else |
7f8ada98 | 2173 | seq_printf(f, "BREAKER "); |
1da177e4 | 2174 | } else { |
7f8ada98 | 2175 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2176 | } |
2177 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2178 | seq_printf(f, "%s ", |
1da177e4 LT |
2179 | (fl->fl_type & LOCK_READ) |
2180 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2181 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2182 | } else { | |
7f8ada98 | 2183 | seq_printf(f, "%s ", |
ab83fa4b | 2184 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2185 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2186 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2187 | } |
2188 | if (inode) { | |
2189 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2190 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2191 | inode->i_sb->s_id, inode->i_ino); |
2192 | #else | |
2193 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2194 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2195 | MAJOR(inode->i_sb->s_dev), |
2196 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2197 | #endif | |
2198 | } else { | |
ab1f1611 | 2199 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2200 | } |
2201 | if (IS_POSIX(fl)) { | |
2202 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2203 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2204 | else |
7f8ada98 | 2205 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2206 | } else { |
7f8ada98 | 2207 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2208 | } |
2209 | } | |
2210 | ||
7f8ada98 | 2211 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2212 | { |
7f8ada98 | 2213 | struct file_lock *fl, *bfl; |
1da177e4 | 2214 | |
7f8ada98 | 2215 | fl = list_entry(v, struct file_lock, fl_link); |
1da177e4 | 2216 | |
99dc8292 | 2217 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
1da177e4 | 2218 | |
7f8ada98 | 2219 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
99dc8292 | 2220 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
094f2825 | 2221 | |
7f8ada98 PE |
2222 | return 0; |
2223 | } | |
1da177e4 | 2224 | |
7f8ada98 PE |
2225 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2226 | { | |
99dc8292 JM |
2227 | loff_t *p = f->private; |
2228 | ||
b89f4321 | 2229 | lock_flocks(); |
99dc8292 | 2230 | *p = (*pos + 1); |
7f8ada98 PE |
2231 | return seq_list_start(&file_lock_list, *pos); |
2232 | } | |
1da177e4 | 2233 | |
7f8ada98 PE |
2234 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2235 | { | |
99dc8292 JM |
2236 | loff_t *p = f->private; |
2237 | ++*p; | |
7f8ada98 PE |
2238 | return seq_list_next(v, &file_lock_list, pos); |
2239 | } | |
1da177e4 | 2240 | |
7f8ada98 PE |
2241 | static void locks_stop(struct seq_file *f, void *v) |
2242 | { | |
b89f4321 | 2243 | unlock_flocks(); |
1da177e4 LT |
2244 | } |
2245 | ||
d8ba7a36 | 2246 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2247 | .start = locks_start, |
2248 | .next = locks_next, | |
2249 | .stop = locks_stop, | |
2250 | .show = locks_show, | |
2251 | }; | |
d8ba7a36 AD |
2252 | |
2253 | static int locks_open(struct inode *inode, struct file *filp) | |
2254 | { | |
99dc8292 | 2255 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
d8ba7a36 AD |
2256 | } |
2257 | ||
2258 | static const struct file_operations proc_locks_operations = { | |
2259 | .open = locks_open, | |
2260 | .read = seq_read, | |
2261 | .llseek = seq_lseek, | |
99dc8292 | 2262 | .release = seq_release_private, |
d8ba7a36 AD |
2263 | }; |
2264 | ||
2265 | static int __init proc_locks_init(void) | |
2266 | { | |
2267 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2268 | return 0; | |
2269 | } | |
2270 | module_init(proc_locks_init); | |
7f8ada98 PE |
2271 | #endif |
2272 | ||
1da177e4 LT |
2273 | /** |
2274 | * lock_may_read - checks that the region is free of locks | |
2275 | * @inode: the inode that is being read | |
2276 | * @start: the first byte to read | |
2277 | * @len: the number of bytes to read | |
2278 | * | |
2279 | * Emulates Windows locking requirements. Whole-file | |
2280 | * mandatory locks (share modes) can prohibit a read and | |
2281 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2282 | * | |
2283 | * N.B. this function is only ever called | |
2284 | * from knfsd and ownership of locks is never checked. | |
2285 | */ | |
2286 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2287 | { | |
2288 | struct file_lock *fl; | |
2289 | int result = 1; | |
b89f4321 | 2290 | lock_flocks(); |
1da177e4 LT |
2291 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2292 | if (IS_POSIX(fl)) { | |
2293 | if (fl->fl_type == F_RDLCK) | |
2294 | continue; | |
2295 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2296 | continue; | |
2297 | } else if (IS_FLOCK(fl)) { | |
2298 | if (!(fl->fl_type & LOCK_MAND)) | |
2299 | continue; | |
2300 | if (fl->fl_type & LOCK_READ) | |
2301 | continue; | |
2302 | } else | |
2303 | continue; | |
2304 | result = 0; | |
2305 | break; | |
2306 | } | |
b89f4321 | 2307 | unlock_flocks(); |
1da177e4 LT |
2308 | return result; |
2309 | } | |
2310 | ||
2311 | EXPORT_SYMBOL(lock_may_read); | |
2312 | ||
2313 | /** | |
2314 | * lock_may_write - checks that the region is free of locks | |
2315 | * @inode: the inode that is being written | |
2316 | * @start: the first byte to write | |
2317 | * @len: the number of bytes to write | |
2318 | * | |
2319 | * Emulates Windows locking requirements. Whole-file | |
2320 | * mandatory locks (share modes) can prohibit a write and | |
2321 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2322 | * | |
2323 | * N.B. this function is only ever called | |
2324 | * from knfsd and ownership of locks is never checked. | |
2325 | */ | |
2326 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2327 | { | |
2328 | struct file_lock *fl; | |
2329 | int result = 1; | |
b89f4321 | 2330 | lock_flocks(); |
1da177e4 LT |
2331 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2332 | if (IS_POSIX(fl)) { | |
2333 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2334 | continue; | |
2335 | } else if (IS_FLOCK(fl)) { | |
2336 | if (!(fl->fl_type & LOCK_MAND)) | |
2337 | continue; | |
2338 | if (fl->fl_type & LOCK_WRITE) | |
2339 | continue; | |
2340 | } else | |
2341 | continue; | |
2342 | result = 0; | |
2343 | break; | |
2344 | } | |
b89f4321 | 2345 | unlock_flocks(); |
1da177e4 LT |
2346 | return result; |
2347 | } | |
2348 | ||
2349 | EXPORT_SYMBOL(lock_may_write); | |
2350 | ||
1da177e4 LT |
2351 | static int __init filelock_init(void) |
2352 | { | |
2353 | filelock_cache = kmem_cache_create("file_lock_cache", | |
ee19cc40 MS |
2354 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2355 | ||
1da177e4 LT |
2356 | return 0; |
2357 | } | |
2358 | ||
2359 | core_initcall(filelock_init); |