]>
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> | |
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 | { | |
710b7216 | 138 | return fl->fl_flags & FL_INPROGRESS; |
ab83fa4b BF |
139 | } |
140 | ||
1da177e4 LT |
141 | int leases_enable = 1; |
142 | int lease_break_time = 45; | |
143 | ||
144 | #define for_each_lock(inode, lockp) \ | |
145 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
146 | ||
26bcbf96 | 147 | static LIST_HEAD(file_lock_list); |
1da177e4 | 148 | static LIST_HEAD(blocked_list); |
72f98e72 | 149 | static DEFINE_SPINLOCK(file_lock_lock); |
1da177e4 | 150 | |
b89f4321 AB |
151 | /* |
152 | * Protects the two list heads above, plus the inode->i_flock list | |
b89f4321 AB |
153 | */ |
154 | void lock_flocks(void) | |
155 | { | |
72f98e72 | 156 | spin_lock(&file_lock_lock); |
b89f4321 AB |
157 | } |
158 | EXPORT_SYMBOL_GPL(lock_flocks); | |
159 | ||
160 | void unlock_flocks(void) | |
161 | { | |
72f98e72 | 162 | spin_unlock(&file_lock_lock); |
b89f4321 AB |
163 | } |
164 | EXPORT_SYMBOL_GPL(unlock_flocks); | |
165 | ||
e18b890b | 166 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 167 | |
ee19cc40 | 168 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 169 | { |
ee19cc40 MS |
170 | INIT_LIST_HEAD(&fl->fl_link); |
171 | INIT_LIST_HEAD(&fl->fl_block); | |
172 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
173 | } |
174 | ||
1da177e4 | 175 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 176 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 177 | { |
ee19cc40 | 178 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
179 | |
180 | if (fl) | |
ee19cc40 | 181 | locks_init_lock_heads(fl); |
a51cb91d MS |
182 | |
183 | return fl; | |
1da177e4 | 184 | } |
c5b1f0d9 | 185 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 186 | |
a9e61e25 | 187 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
188 | { |
189 | if (fl->fl_ops) { | |
190 | if (fl->fl_ops->fl_release_private) | |
191 | fl->fl_ops->fl_release_private(fl); | |
192 | fl->fl_ops = NULL; | |
193 | } | |
194 | if (fl->fl_lmops) { | |
8fb47a4f BF |
195 | if (fl->fl_lmops->lm_release_private) |
196 | fl->fl_lmops->lm_release_private(fl); | |
47831f35 TM |
197 | fl->fl_lmops = NULL; |
198 | } | |
199 | ||
200 | } | |
a9e61e25 | 201 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 202 | |
1da177e4 | 203 | /* Free a lock which is not in use. */ |
05fa3135 | 204 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 205 | { |
5ce29646 MS |
206 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
207 | BUG_ON(!list_empty(&fl->fl_block)); | |
208 | BUG_ON(!list_empty(&fl->fl_link)); | |
1da177e4 | 209 | |
47831f35 | 210 | locks_release_private(fl); |
1da177e4 LT |
211 | kmem_cache_free(filelock_cache, fl); |
212 | } | |
05fa3135 | 213 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
214 | |
215 | void locks_init_lock(struct file_lock *fl) | |
216 | { | |
ee19cc40 MS |
217 | memset(fl, 0, sizeof(struct file_lock)); |
218 | locks_init_lock_heads(fl); | |
1da177e4 LT |
219 | } |
220 | ||
221 | EXPORT_SYMBOL(locks_init_lock); | |
222 | ||
47831f35 TM |
223 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
224 | { | |
225 | if (fl->fl_ops) { | |
226 | if (fl->fl_ops->fl_copy_lock) | |
227 | fl->fl_ops->fl_copy_lock(new, fl); | |
228 | new->fl_ops = fl->fl_ops; | |
229 | } | |
bb8430a2 | 230 | if (fl->fl_lmops) |
47831f35 | 231 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
232 | } |
233 | ||
1da177e4 LT |
234 | /* |
235 | * Initialize a new lock from an existing file_lock structure. | |
236 | */ | |
1a747ee0 | 237 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
238 | { |
239 | new->fl_owner = fl->fl_owner; | |
240 | new->fl_pid = fl->fl_pid; | |
0996905f | 241 | new->fl_file = NULL; |
1da177e4 LT |
242 | new->fl_flags = fl->fl_flags; |
243 | new->fl_type = fl->fl_type; | |
244 | new->fl_start = fl->fl_start; | |
245 | new->fl_end = fl->fl_end; | |
0996905f TM |
246 | new->fl_ops = NULL; |
247 | new->fl_lmops = NULL; | |
248 | } | |
3dd7b71c | 249 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
250 | |
251 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
252 | { | |
253 | locks_release_private(new); | |
254 | ||
255 | __locks_copy_lock(new, fl); | |
256 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
257 | new->fl_ops = fl->fl_ops; |
258 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
259 | |
260 | locks_copy_private(new, fl); | |
1da177e4 LT |
261 | } |
262 | ||
263 | EXPORT_SYMBOL(locks_copy_lock); | |
264 | ||
265 | static inline int flock_translate_cmd(int cmd) { | |
266 | if (cmd & LOCK_MAND) | |
267 | return cmd & (LOCK_MAND | LOCK_RW); | |
268 | switch (cmd) { | |
269 | case LOCK_SH: | |
270 | return F_RDLCK; | |
271 | case LOCK_EX: | |
272 | return F_WRLCK; | |
273 | case LOCK_UN: | |
274 | return F_UNLCK; | |
275 | } | |
276 | return -EINVAL; | |
277 | } | |
278 | ||
279 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
280 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
281 | unsigned int cmd) | |
282 | { | |
283 | struct file_lock *fl; | |
284 | int type = flock_translate_cmd(cmd); | |
285 | if (type < 0) | |
286 | return type; | |
287 | ||
288 | fl = locks_alloc_lock(); | |
289 | if (fl == NULL) | |
290 | return -ENOMEM; | |
291 | ||
292 | fl->fl_file = filp; | |
293 | fl->fl_pid = current->tgid; | |
294 | fl->fl_flags = FL_FLOCK; | |
295 | fl->fl_type = type; | |
296 | fl->fl_end = OFFSET_MAX; | |
297 | ||
298 | *lock = fl; | |
299 | return 0; | |
300 | } | |
301 | ||
302 | static int assign_type(struct file_lock *fl, int type) | |
303 | { | |
304 | switch (type) { | |
305 | case F_RDLCK: | |
306 | case F_WRLCK: | |
307 | case F_UNLCK: | |
308 | fl->fl_type = type; | |
309 | break; | |
310 | default: | |
311 | return -EINVAL; | |
312 | } | |
313 | return 0; | |
314 | } | |
315 | ||
316 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
317 | * style lock. | |
318 | */ | |
319 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
320 | struct flock *l) | |
321 | { | |
322 | off_t start, end; | |
323 | ||
324 | switch (l->l_whence) { | |
f5579f8c | 325 | case SEEK_SET: |
1da177e4 LT |
326 | start = 0; |
327 | break; | |
f5579f8c | 328 | case SEEK_CUR: |
1da177e4 LT |
329 | start = filp->f_pos; |
330 | break; | |
f5579f8c | 331 | case SEEK_END: |
0f7fc9e4 | 332 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
333 | break; |
334 | default: | |
335 | return -EINVAL; | |
336 | } | |
337 | ||
338 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
339 | POSIX-2001 defines it. */ | |
340 | start += l->l_start; | |
4c780a46 TM |
341 | if (start < 0) |
342 | return -EINVAL; | |
343 | fl->fl_end = OFFSET_MAX; | |
344 | if (l->l_len > 0) { | |
345 | end = start + l->l_len - 1; | |
346 | fl->fl_end = end; | |
347 | } else if (l->l_len < 0) { | |
1da177e4 | 348 | end = start - 1; |
4c780a46 | 349 | fl->fl_end = end; |
1da177e4 | 350 | start += l->l_len; |
4c780a46 TM |
351 | if (start < 0) |
352 | return -EINVAL; | |
1da177e4 | 353 | } |
1da177e4 | 354 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
355 | if (fl->fl_end < fl->fl_start) |
356 | return -EOVERFLOW; | |
1da177e4 LT |
357 | |
358 | fl->fl_owner = current->files; | |
359 | fl->fl_pid = current->tgid; | |
360 | fl->fl_file = filp; | |
361 | fl->fl_flags = FL_POSIX; | |
362 | fl->fl_ops = NULL; | |
363 | fl->fl_lmops = NULL; | |
364 | ||
365 | return assign_type(fl, l->l_type); | |
366 | } | |
367 | ||
368 | #if BITS_PER_LONG == 32 | |
369 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
370 | struct flock64 *l) | |
371 | { | |
372 | loff_t start; | |
373 | ||
374 | switch (l->l_whence) { | |
f5579f8c | 375 | case SEEK_SET: |
1da177e4 LT |
376 | start = 0; |
377 | break; | |
f5579f8c | 378 | case SEEK_CUR: |
1da177e4 LT |
379 | start = filp->f_pos; |
380 | break; | |
f5579f8c | 381 | case SEEK_END: |
0f7fc9e4 | 382 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
383 | break; |
384 | default: | |
385 | return -EINVAL; | |
386 | } | |
387 | ||
4c780a46 TM |
388 | start += l->l_start; |
389 | if (start < 0) | |
1da177e4 | 390 | return -EINVAL; |
4c780a46 TM |
391 | fl->fl_end = OFFSET_MAX; |
392 | if (l->l_len > 0) { | |
393 | fl->fl_end = start + l->l_len - 1; | |
394 | } else if (l->l_len < 0) { | |
395 | fl->fl_end = start - 1; | |
396 | start += l->l_len; | |
397 | if (start < 0) | |
398 | return -EINVAL; | |
399 | } | |
1da177e4 | 400 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
401 | if (fl->fl_end < fl->fl_start) |
402 | return -EOVERFLOW; | |
1da177e4 LT |
403 | |
404 | fl->fl_owner = current->files; | |
405 | fl->fl_pid = current->tgid; | |
406 | fl->fl_file = filp; | |
407 | fl->fl_flags = FL_POSIX; | |
408 | fl->fl_ops = NULL; | |
409 | fl->fl_lmops = NULL; | |
410 | ||
f32cb532 | 411 | return assign_type(fl, l->l_type); |
1da177e4 LT |
412 | } |
413 | #endif | |
414 | ||
415 | /* default lease lock manager operations */ | |
416 | static void lease_break_callback(struct file_lock *fl) | |
417 | { | |
418 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
419 | } | |
420 | ||
421 | static void lease_release_private_callback(struct file_lock *fl) | |
422 | { | |
423 | if (!fl->fl_file) | |
424 | return; | |
425 | ||
426 | f_delown(fl->fl_file); | |
427 | fl->fl_file->f_owner.signum = 0; | |
428 | } | |
429 | ||
7b021967 | 430 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f BF |
431 | .lm_break = lease_break_callback, |
432 | .lm_release_private = lease_release_private_callback, | |
433 | .lm_change = lease_modify, | |
1da177e4 LT |
434 | }; |
435 | ||
436 | /* | |
437 | * Initialize a lease, use the default lock manager operations | |
438 | */ | |
439 | static int lease_init(struct file *filp, int type, struct file_lock *fl) | |
440 | { | |
75dff55a TM |
441 | if (assign_type(fl, type) != 0) |
442 | return -EINVAL; | |
443 | ||
1da177e4 LT |
444 | fl->fl_owner = current->files; |
445 | fl->fl_pid = current->tgid; | |
446 | ||
447 | fl->fl_file = filp; | |
448 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
449 | fl->fl_start = 0; |
450 | fl->fl_end = OFFSET_MAX; | |
451 | fl->fl_ops = NULL; | |
452 | fl->fl_lmops = &lease_manager_ops; | |
453 | return 0; | |
454 | } | |
455 | ||
456 | /* Allocate a file_lock initialised to this type of lease */ | |
e32b8ee2 | 457 | static struct file_lock *lease_alloc(struct file *filp, int type) |
1da177e4 LT |
458 | { |
459 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 460 | int error = -ENOMEM; |
1da177e4 LT |
461 | |
462 | if (fl == NULL) | |
e32b8ee2 | 463 | return ERR_PTR(error); |
1da177e4 LT |
464 | |
465 | error = lease_init(filp, type, fl); | |
75dff55a TM |
466 | if (error) { |
467 | locks_free_lock(fl); | |
e32b8ee2 | 468 | return ERR_PTR(error); |
75dff55a | 469 | } |
e32b8ee2 | 470 | return fl; |
1da177e4 LT |
471 | } |
472 | ||
473 | /* Check if two locks overlap each other. | |
474 | */ | |
475 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
476 | { | |
477 | return ((fl1->fl_end >= fl2->fl_start) && | |
478 | (fl2->fl_end >= fl1->fl_start)); | |
479 | } | |
480 | ||
481 | /* | |
482 | * Check whether two locks have the same owner. | |
483 | */ | |
33443c42 | 484 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 485 | { |
8fb47a4f | 486 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 487 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 488 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
489 | return fl1->fl_owner == fl2->fl_owner; |
490 | } | |
491 | ||
492 | /* Remove waiter from blocker's block list. | |
493 | * When blocker ends up pointing to itself then the list is empty. | |
494 | */ | |
33443c42 | 495 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 LT |
496 | { |
497 | list_del_init(&waiter->fl_block); | |
498 | list_del_init(&waiter->fl_link); | |
499 | waiter->fl_next = NULL; | |
500 | } | |
501 | ||
502 | /* | |
503 | */ | |
504 | static void locks_delete_block(struct file_lock *waiter) | |
505 | { | |
b89f4321 | 506 | lock_flocks(); |
1da177e4 | 507 | __locks_delete_block(waiter); |
b89f4321 | 508 | unlock_flocks(); |
1da177e4 LT |
509 | } |
510 | ||
511 | /* Insert waiter into blocker's block list. | |
512 | * We use a circular list so that processes can be easily woken up in | |
513 | * the order they blocked. The documentation doesn't require this but | |
514 | * it seems like the reasonable thing to do. | |
515 | */ | |
516 | static void locks_insert_block(struct file_lock *blocker, | |
517 | struct file_lock *waiter) | |
518 | { | |
6dc0fe8f | 519 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 LT |
520 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
521 | waiter->fl_next = blocker; | |
522 | if (IS_POSIX(blocker)) | |
523 | list_add(&waiter->fl_link, &blocked_list); | |
524 | } | |
525 | ||
526 | /* Wake up processes blocked waiting for blocker. | |
527 | * If told to wait then schedule the processes until the block list | |
528 | * is empty, otherwise empty the block list ourselves. | |
529 | */ | |
530 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
531 | { | |
532 | while (!list_empty(&blocker->fl_block)) { | |
f0c1cd0e PE |
533 | struct file_lock *waiter; |
534 | ||
535 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
536 | struct file_lock, fl_block); |
537 | __locks_delete_block(waiter); | |
8fb47a4f BF |
538 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
539 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
540 | else |
541 | wake_up(&waiter->fl_wait); | |
542 | } | |
543 | } | |
544 | ||
545 | /* Insert file lock fl into an inode's lock list at the position indicated | |
546 | * by pos. At the same time add the lock to the global file lock list. | |
547 | */ | |
548 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
549 | { | |
550 | list_add(&fl->fl_link, &file_lock_list); | |
551 | ||
ab1f1611 VG |
552 | fl->fl_nspid = get_pid(task_tgid(current)); |
553 | ||
1da177e4 LT |
554 | /* insert into file's list */ |
555 | fl->fl_next = *pos; | |
556 | *pos = fl; | |
1da177e4 LT |
557 | } |
558 | ||
559 | /* | |
560 | * Delete a lock and then free it. | |
561 | * Wake up processes that are blocked waiting for this lock, | |
562 | * notify the FS that the lock has been cleared and | |
563 | * finally free the lock. | |
564 | */ | |
565 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
566 | { | |
567 | struct file_lock *fl = *thisfl_p; | |
568 | ||
569 | *thisfl_p = fl->fl_next; | |
570 | fl->fl_next = NULL; | |
571 | list_del_init(&fl->fl_link); | |
572 | ||
573 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); | |
574 | if (fl->fl_fasync != NULL) { | |
575 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
576 | fl->fl_fasync = NULL; | |
577 | } | |
578 | ||
ab1f1611 VG |
579 | if (fl->fl_nspid) { |
580 | put_pid(fl->fl_nspid); | |
581 | fl->fl_nspid = NULL; | |
582 | } | |
583 | ||
1da177e4 LT |
584 | locks_wake_up_blocks(fl); |
585 | locks_free_lock(fl); | |
586 | } | |
587 | ||
588 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
589 | * checks for shared/exclusive status of overlapping locks. | |
590 | */ | |
591 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
592 | { | |
593 | if (sys_fl->fl_type == F_WRLCK) | |
594 | return 1; | |
595 | if (caller_fl->fl_type == F_WRLCK) | |
596 | return 1; | |
597 | return 0; | |
598 | } | |
599 | ||
600 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
601 | * checking before calling the locks_conflict(). | |
602 | */ | |
603 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
604 | { | |
605 | /* POSIX locks owned by the same process do not conflict with | |
606 | * each other. | |
607 | */ | |
608 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
609 | return (0); | |
610 | ||
611 | /* Check whether they overlap */ | |
612 | if (!locks_overlap(caller_fl, sys_fl)) | |
613 | return 0; | |
614 | ||
615 | return (locks_conflict(caller_fl, sys_fl)); | |
616 | } | |
617 | ||
618 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
619 | * checking before calling the locks_conflict(). | |
620 | */ | |
621 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
622 | { | |
623 | /* FLOCK locks referring to the same filp do not conflict with | |
624 | * each other. | |
625 | */ | |
626 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
627 | return (0); | |
628 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
629 | return 0; | |
630 | ||
631 | return (locks_conflict(caller_fl, sys_fl)); | |
632 | } | |
633 | ||
6d34ac19 | 634 | void |
9d6a8c5c | 635 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
636 | { |
637 | struct file_lock *cfl; | |
638 | ||
b89f4321 | 639 | lock_flocks(); |
0f7fc9e4 | 640 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
641 | if (!IS_POSIX(cfl)) |
642 | continue; | |
b842e240 | 643 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
644 | break; |
645 | } | |
ab1f1611 | 646 | if (cfl) { |
9d6a8c5c | 647 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 648 | if (cfl->fl_nspid) |
6c5f3e7b | 649 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 650 | } else |
129a84de | 651 | fl->fl_type = F_UNLCK; |
b89f4321 | 652 | unlock_flocks(); |
6d34ac19 | 653 | return; |
1da177e4 | 654 | } |
1da177e4 LT |
655 | EXPORT_SYMBOL(posix_test_lock); |
656 | ||
b533184f BF |
657 | /* |
658 | * Deadlock detection: | |
659 | * | |
660 | * We attempt to detect deadlocks that are due purely to posix file | |
661 | * locks. | |
1da177e4 | 662 | * |
b533184f BF |
663 | * We assume that a task can be waiting for at most one lock at a time. |
664 | * So for any acquired lock, the process holding that lock may be | |
665 | * waiting on at most one other lock. That lock in turns may be held by | |
666 | * someone waiting for at most one other lock. Given a requested lock | |
667 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
668 | * follow this chain of waiters to ensure we are not about to create a | |
669 | * cycle. | |
1da177e4 | 670 | * |
b533184f BF |
671 | * Since we do this before we ever put a process to sleep on a lock, we |
672 | * are ensured that there is never a cycle; that is what guarantees that | |
673 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 674 | * |
b533184f BF |
675 | * Note: the above assumption may not be true when handling lock |
676 | * requests from a broken NFS client. It may also fail in the presence | |
677 | * of tasks (such as posix threads) sharing the same open file table. | |
678 | * | |
679 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 680 | */ |
97855b49 BF |
681 | |
682 | #define MAX_DEADLK_ITERATIONS 10 | |
683 | ||
b533184f BF |
684 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
685 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
686 | { | |
687 | struct file_lock *fl; | |
688 | ||
689 | list_for_each_entry(fl, &blocked_list, fl_link) { | |
690 | if (posix_same_owner(fl, block_fl)) | |
691 | return fl->fl_next; | |
692 | } | |
693 | return NULL; | |
694 | } | |
695 | ||
b0904e14 | 696 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
697 | struct file_lock *block_fl) |
698 | { | |
97855b49 | 699 | int i = 0; |
1da177e4 | 700 | |
b533184f BF |
701 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
702 | if (i++ > MAX_DEADLK_ITERATIONS) | |
703 | return 0; | |
704 | if (posix_same_owner(caller_fl, block_fl)) | |
705 | return 1; | |
1da177e4 LT |
706 | } |
707 | return 0; | |
708 | } | |
709 | ||
1da177e4 | 710 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 711 | * after any leases, but before any posix locks. |
f475ae95 TM |
712 | * |
713 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
714 | * whether or not a lock was successfully freed by testing the return | |
715 | * value for -ENOENT. | |
1da177e4 | 716 | */ |
993dfa87 | 717 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 718 | { |
993dfa87 | 719 | struct file_lock *new_fl = NULL; |
1da177e4 | 720 | struct file_lock **before; |
0f7fc9e4 | 721 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
722 | int error = 0; |
723 | int found = 0; | |
724 | ||
b89f4321 | 725 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 726 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
727 | if (!new_fl) |
728 | return -ENOMEM; | |
84d535ad PE |
729 | } |
730 | ||
b89f4321 AB |
731 | lock_flocks(); |
732 | if (request->fl_flags & FL_ACCESS) | |
733 | goto find_conflict; | |
734 | ||
1da177e4 LT |
735 | for_each_lock(inode, before) { |
736 | struct file_lock *fl = *before; | |
737 | if (IS_POSIX(fl)) | |
738 | break; | |
739 | if (IS_LEASE(fl)) | |
740 | continue; | |
741 | if (filp != fl->fl_file) | |
742 | continue; | |
993dfa87 | 743 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
744 | goto out; |
745 | found = 1; | |
746 | locks_delete_lock(before); | |
747 | break; | |
748 | } | |
1da177e4 | 749 | |
f475ae95 TM |
750 | if (request->fl_type == F_UNLCK) { |
751 | if ((request->fl_flags & FL_EXISTS) && !found) | |
752 | error = -ENOENT; | |
993dfa87 | 753 | goto out; |
f475ae95 | 754 | } |
1da177e4 LT |
755 | |
756 | /* | |
757 | * If a higher-priority process was blocked on the old file lock, | |
758 | * give it the opportunity to lock the file. | |
759 | */ | |
b89f4321 AB |
760 | if (found) { |
761 | unlock_flocks(); | |
def01bc5 | 762 | cond_resched(); |
b89f4321 AB |
763 | lock_flocks(); |
764 | } | |
1da177e4 | 765 | |
f07f18dd | 766 | find_conflict: |
1da177e4 LT |
767 | for_each_lock(inode, before) { |
768 | struct file_lock *fl = *before; | |
769 | if (IS_POSIX(fl)) | |
770 | break; | |
771 | if (IS_LEASE(fl)) | |
772 | continue; | |
993dfa87 | 773 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
774 | continue; |
775 | error = -EAGAIN; | |
bde74e4b MS |
776 | if (!(request->fl_flags & FL_SLEEP)) |
777 | goto out; | |
778 | error = FILE_LOCK_DEFERRED; | |
779 | locks_insert_block(fl, request); | |
1da177e4 LT |
780 | goto out; |
781 | } | |
f07f18dd TM |
782 | if (request->fl_flags & FL_ACCESS) |
783 | goto out; | |
993dfa87 | 784 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 785 | locks_insert_lock(before, new_fl); |
993dfa87 | 786 | new_fl = NULL; |
9cedc194 | 787 | error = 0; |
1da177e4 LT |
788 | |
789 | out: | |
b89f4321 | 790 | unlock_flocks(); |
993dfa87 TM |
791 | if (new_fl) |
792 | locks_free_lock(new_fl); | |
1da177e4 LT |
793 | return error; |
794 | } | |
795 | ||
150b3934 | 796 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
797 | { |
798 | struct file_lock *fl; | |
39005d02 MS |
799 | struct file_lock *new_fl = NULL; |
800 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
801 | struct file_lock *left = NULL; |
802 | struct file_lock *right = NULL; | |
803 | struct file_lock **before; | |
804 | int error, added = 0; | |
805 | ||
806 | /* | |
807 | * We may need two file_lock structures for this operation, | |
808 | * so we get them in advance to avoid races. | |
39005d02 MS |
809 | * |
810 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 811 | */ |
39005d02 MS |
812 | if (!(request->fl_flags & FL_ACCESS) && |
813 | (request->fl_type != F_UNLCK || | |
814 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
815 | new_fl = locks_alloc_lock(); | |
816 | new_fl2 = locks_alloc_lock(); | |
817 | } | |
1da177e4 | 818 | |
b89f4321 | 819 | lock_flocks(); |
1da177e4 LT |
820 | if (request->fl_type != F_UNLCK) { |
821 | for_each_lock(inode, before) { | |
526985b9 | 822 | fl = *before; |
1da177e4 LT |
823 | if (!IS_POSIX(fl)) |
824 | continue; | |
825 | if (!posix_locks_conflict(request, fl)) | |
826 | continue; | |
5842add2 | 827 | if (conflock) |
1a747ee0 | 828 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
829 | error = -EAGAIN; |
830 | if (!(request->fl_flags & FL_SLEEP)) | |
831 | goto out; | |
832 | error = -EDEADLK; | |
833 | if (posix_locks_deadlock(request, fl)) | |
834 | goto out; | |
bde74e4b | 835 | error = FILE_LOCK_DEFERRED; |
1da177e4 LT |
836 | locks_insert_block(fl, request); |
837 | goto out; | |
838 | } | |
839 | } | |
840 | ||
841 | /* If we're just looking for a conflict, we're done. */ | |
842 | error = 0; | |
843 | if (request->fl_flags & FL_ACCESS) | |
844 | goto out; | |
845 | ||
1da177e4 | 846 | /* |
1da177e4 LT |
847 | * Find the first old lock with the same owner as the new lock. |
848 | */ | |
849 | ||
850 | before = &inode->i_flock; | |
851 | ||
852 | /* First skip locks owned by other processes. */ | |
853 | while ((fl = *before) && (!IS_POSIX(fl) || | |
854 | !posix_same_owner(request, fl))) { | |
855 | before = &fl->fl_next; | |
856 | } | |
857 | ||
858 | /* Process locks with this owner. */ | |
859 | while ((fl = *before) && posix_same_owner(request, fl)) { | |
860 | /* Detect adjacent or overlapping regions (if same lock type) | |
861 | */ | |
862 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
863 | /* In all comparisons of start vs end, use |
864 | * "start - 1" rather than "end + 1". If end | |
865 | * is OFFSET_MAX, end + 1 will become negative. | |
866 | */ | |
1da177e4 LT |
867 | if (fl->fl_end < request->fl_start - 1) |
868 | goto next_lock; | |
869 | /* If the next lock in the list has entirely bigger | |
870 | * addresses than the new one, insert the lock here. | |
871 | */ | |
449231d6 | 872 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
873 | break; |
874 | ||
875 | /* If we come here, the new and old lock are of the | |
876 | * same type and adjacent or overlapping. Make one | |
877 | * lock yielding from the lower start address of both | |
878 | * locks to the higher end address. | |
879 | */ | |
880 | if (fl->fl_start > request->fl_start) | |
881 | fl->fl_start = request->fl_start; | |
882 | else | |
883 | request->fl_start = fl->fl_start; | |
884 | if (fl->fl_end < request->fl_end) | |
885 | fl->fl_end = request->fl_end; | |
886 | else | |
887 | request->fl_end = fl->fl_end; | |
888 | if (added) { | |
889 | locks_delete_lock(before); | |
890 | continue; | |
891 | } | |
892 | request = fl; | |
893 | added = 1; | |
894 | } | |
895 | else { | |
896 | /* Processing for different lock types is a bit | |
897 | * more complex. | |
898 | */ | |
899 | if (fl->fl_end < request->fl_start) | |
900 | goto next_lock; | |
901 | if (fl->fl_start > request->fl_end) | |
902 | break; | |
903 | if (request->fl_type == F_UNLCK) | |
904 | added = 1; | |
905 | if (fl->fl_start < request->fl_start) | |
906 | left = fl; | |
907 | /* If the next lock in the list has a higher end | |
908 | * address than the new one, insert the new one here. | |
909 | */ | |
910 | if (fl->fl_end > request->fl_end) { | |
911 | right = fl; | |
912 | break; | |
913 | } | |
914 | if (fl->fl_start >= request->fl_start) { | |
915 | /* The new lock completely replaces an old | |
916 | * one (This may happen several times). | |
917 | */ | |
918 | if (added) { | |
919 | locks_delete_lock(before); | |
920 | continue; | |
921 | } | |
922 | /* Replace the old lock with the new one. | |
923 | * Wake up anybody waiting for the old one, | |
924 | * as the change in lock type might satisfy | |
925 | * their needs. | |
926 | */ | |
927 | locks_wake_up_blocks(fl); | |
928 | fl->fl_start = request->fl_start; | |
929 | fl->fl_end = request->fl_end; | |
930 | fl->fl_type = request->fl_type; | |
47831f35 TM |
931 | locks_release_private(fl); |
932 | locks_copy_private(fl, request); | |
1da177e4 LT |
933 | request = fl; |
934 | added = 1; | |
935 | } | |
936 | } | |
937 | /* Go on to next lock. | |
938 | */ | |
939 | next_lock: | |
940 | before = &fl->fl_next; | |
941 | } | |
942 | ||
0d9a490a MS |
943 | /* |
944 | * The above code only modifies existing locks in case of | |
945 | * merging or replacing. If new lock(s) need to be inserted | |
946 | * all modifications are done bellow this, so it's safe yet to | |
947 | * bail out. | |
948 | */ | |
949 | error = -ENOLCK; /* "no luck" */ | |
950 | if (right && left == right && !new_fl2) | |
951 | goto out; | |
952 | ||
1da177e4 LT |
953 | error = 0; |
954 | if (!added) { | |
f475ae95 TM |
955 | if (request->fl_type == F_UNLCK) { |
956 | if (request->fl_flags & FL_EXISTS) | |
957 | error = -ENOENT; | |
1da177e4 | 958 | goto out; |
f475ae95 | 959 | } |
0d9a490a MS |
960 | |
961 | if (!new_fl) { | |
962 | error = -ENOLCK; | |
963 | goto out; | |
964 | } | |
1da177e4 LT |
965 | locks_copy_lock(new_fl, request); |
966 | locks_insert_lock(before, new_fl); | |
967 | new_fl = NULL; | |
968 | } | |
969 | if (right) { | |
970 | if (left == right) { | |
971 | /* The new lock breaks the old one in two pieces, | |
972 | * so we have to use the second new lock. | |
973 | */ | |
974 | left = new_fl2; | |
975 | new_fl2 = NULL; | |
976 | locks_copy_lock(left, right); | |
977 | locks_insert_lock(before, left); | |
978 | } | |
979 | right->fl_start = request->fl_end + 1; | |
980 | locks_wake_up_blocks(right); | |
981 | } | |
982 | if (left) { | |
983 | left->fl_end = request->fl_start - 1; | |
984 | locks_wake_up_blocks(left); | |
985 | } | |
986 | out: | |
b89f4321 | 987 | unlock_flocks(); |
1da177e4 LT |
988 | /* |
989 | * Free any unused locks. | |
990 | */ | |
991 | if (new_fl) | |
992 | locks_free_lock(new_fl); | |
993 | if (new_fl2) | |
994 | locks_free_lock(new_fl2); | |
995 | return error; | |
996 | } | |
997 | ||
998 | /** | |
999 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1000 | * @filp: The file to apply the lock to | |
1001 | * @fl: The lock to be applied | |
150b3934 | 1002 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1003 | * |
1004 | * Add a POSIX style lock to a file. | |
1005 | * We merge adjacent & overlapping locks whenever possible. | |
1006 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1007 | * |
1008 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1009 | * whether or not a lock was successfully freed by testing the return | |
1010 | * value for -ENOENT. | |
1da177e4 | 1011 | */ |
150b3934 | 1012 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1013 | struct file_lock *conflock) |
1014 | { | |
150b3934 | 1015 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
1da177e4 | 1016 | } |
150b3934 | 1017 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1018 | |
1019 | /** | |
1020 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1021 | * @filp: The file to apply the lock to | |
1022 | * @fl: The lock to be applied | |
1023 | * | |
1024 | * Add a POSIX style lock to a file. | |
1025 | * We merge adjacent & overlapping locks whenever possible. | |
1026 | * POSIX locks are sorted by owner task, then by starting address | |
1027 | */ | |
1028 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1029 | { | |
1030 | int error; | |
1031 | might_sleep (); | |
1032 | for (;;) { | |
150b3934 | 1033 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1034 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1035 | break; |
1036 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1037 | if (!error) | |
1038 | continue; | |
1039 | ||
1040 | locks_delete_block(fl); | |
1041 | break; | |
1042 | } | |
1043 | return error; | |
1044 | } | |
1045 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1046 | ||
1047 | /** | |
1048 | * locks_mandatory_locked - Check for an active lock | |
1049 | * @inode: the file to check | |
1050 | * | |
1051 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1052 | * This function is called from locks_verify_locked() only. | |
1053 | */ | |
1054 | int locks_mandatory_locked(struct inode *inode) | |
1055 | { | |
1056 | fl_owner_t owner = current->files; | |
1057 | struct file_lock *fl; | |
1058 | ||
1059 | /* | |
1060 | * Search the lock list for this inode for any POSIX locks. | |
1061 | */ | |
b89f4321 | 1062 | lock_flocks(); |
1da177e4 LT |
1063 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1064 | if (!IS_POSIX(fl)) | |
1065 | continue; | |
1066 | if (fl->fl_owner != owner) | |
1067 | break; | |
1068 | } | |
b89f4321 | 1069 | unlock_flocks(); |
1da177e4 LT |
1070 | return fl ? -EAGAIN : 0; |
1071 | } | |
1072 | ||
1073 | /** | |
1074 | * locks_mandatory_area - Check for a conflicting lock | |
1075 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1076 | * for shared | |
1077 | * @inode: the file to check | |
1078 | * @filp: how the file was opened (if it was) | |
1079 | * @offset: start of area to check | |
1080 | * @count: length of area to check | |
1081 | * | |
1082 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1083 | * This function is called from rw_verify_area() and | |
1084 | * locks_verify_truncate(). | |
1085 | */ | |
1086 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1087 | struct file *filp, loff_t offset, | |
1088 | size_t count) | |
1089 | { | |
1090 | struct file_lock fl; | |
1091 | int error; | |
1092 | ||
1093 | locks_init_lock(&fl); | |
1094 | fl.fl_owner = current->files; | |
1095 | fl.fl_pid = current->tgid; | |
1096 | fl.fl_file = filp; | |
1097 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1098 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1099 | fl.fl_flags |= FL_SLEEP; | |
1100 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1101 | fl.fl_start = offset; | |
1102 | fl.fl_end = offset + count - 1; | |
1103 | ||
1104 | for (;;) { | |
150b3934 | 1105 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1106 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1107 | break; |
1108 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1109 | if (!error) { | |
1110 | /* | |
1111 | * If we've been sleeping someone might have | |
1112 | * changed the permissions behind our back. | |
1113 | */ | |
a16877ca | 1114 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1115 | continue; |
1116 | } | |
1117 | ||
1118 | locks_delete_block(&fl); | |
1119 | break; | |
1120 | } | |
1121 | ||
1122 | return error; | |
1123 | } | |
1124 | ||
1125 | EXPORT_SYMBOL(locks_mandatory_area); | |
1126 | ||
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; | |
710b7216 | 1135 | fl->fl_flags &= ~FL_INPROGRESS; |
1da177e4 LT |
1136 | locks_wake_up_blocks(fl); |
1137 | if (arg == F_UNLCK) | |
1138 | locks_delete_lock(before); | |
1139 | return 0; | |
1140 | } | |
1141 | ||
1142 | EXPORT_SYMBOL(lease_modify); | |
1143 | ||
1144 | static void time_out_leases(struct inode *inode) | |
1145 | { | |
1146 | struct file_lock **before; | |
1147 | struct file_lock *fl; | |
1148 | ||
1149 | before = &inode->i_flock; | |
ab83fa4b | 1150 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
1da177e4 LT |
1151 | if ((fl->fl_break_time == 0) |
1152 | || time_before(jiffies, fl->fl_break_time)) { | |
1153 | before = &fl->fl_next; | |
1154 | continue; | |
1155 | } | |
710b7216 | 1156 | lease_modify(before, fl->fl_type); |
1da177e4 LT |
1157 | if (fl == *before) /* lease_modify may have freed fl */ |
1158 | before = &fl->fl_next; | |
1159 | } | |
1160 | } | |
1161 | ||
1162 | /** | |
1163 | * __break_lease - revoke all outstanding leases on file | |
1164 | * @inode: the inode of the file to return | |
1165 | * @mode: the open mode (read or write) | |
1166 | * | |
87250dd2 | 1167 | * break_lease (inlined for speed) has checked there already is at least |
1168 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1169 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1170 | * specified %O_NONBLOCK to your open(). |
1171 | */ | |
1172 | int __break_lease(struct inode *inode, unsigned int mode) | |
1173 | { | |
1174 | int error = 0, future; | |
1175 | struct file_lock *new_fl, *flock; | |
1176 | struct file_lock *fl; | |
1da177e4 LT |
1177 | unsigned long break_time; |
1178 | int i_have_this_lease = 0; | |
8737c930 | 1179 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1180 | |
8737c930 | 1181 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
1da177e4 | 1182 | |
b89f4321 | 1183 | lock_flocks(); |
1da177e4 LT |
1184 | |
1185 | time_out_leases(inode); | |
1186 | ||
1187 | flock = inode->i_flock; | |
1188 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1189 | goto out; | |
1190 | ||
1191 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) | |
1192 | if (fl->fl_owner == current->files) | |
1193 | i_have_this_lease = 1; | |
1194 | ||
8737c930 | 1195 | if (want_write) { |
1da177e4 | 1196 | /* If we want write access, we have to revoke any lease. */ |
710b7216 | 1197 | future = F_UNLCK; |
ab83fa4b | 1198 | } else if (lease_breaking(flock)) { |
1da177e4 LT |
1199 | /* If the lease is already being broken, we just leave it */ |
1200 | future = flock->fl_type; | |
1201 | } else if (flock->fl_type & F_WRLCK) { | |
1202 | /* Downgrade the exclusive lease to a read-only lease. */ | |
710b7216 | 1203 | future = F_RDLCK; |
1da177e4 LT |
1204 | } else { |
1205 | /* the existing lease was read-only, so we can read too. */ | |
1206 | goto out; | |
1207 | } | |
1208 | ||
e32b8ee2 BF |
1209 | if (IS_ERR(new_fl) && !i_have_this_lease |
1210 | && ((mode & O_NONBLOCK) == 0)) { | |
1211 | error = PTR_ERR(new_fl); | |
1da177e4 LT |
1212 | goto out; |
1213 | } | |
1214 | ||
1215 | break_time = 0; | |
1216 | if (lease_break_time > 0) { | |
1217 | break_time = jiffies + lease_break_time * HZ; | |
1218 | if (break_time == 0) | |
1219 | break_time++; /* so that 0 means no break time */ | |
1220 | } | |
1221 | ||
1222 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1223 | if (fl->fl_type != future) { | |
1224 | fl->fl_type = future; | |
710b7216 | 1225 | fl->fl_flags |= FL_INPROGRESS; |
1da177e4 LT |
1226 | fl->fl_break_time = break_time; |
1227 | /* lease must have lmops break callback */ | |
8fb47a4f | 1228 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1229 | } |
1230 | } | |
1231 | ||
1232 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1233 | error = -EWOULDBLOCK; | |
1234 | goto out; | |
1235 | } | |
1236 | ||
1237 | restart: | |
1238 | break_time = flock->fl_break_time; | |
1239 | if (break_time != 0) { | |
1240 | break_time -= jiffies; | |
1241 | if (break_time == 0) | |
1242 | break_time++; | |
1243 | } | |
4321e01e | 1244 | locks_insert_block(flock, new_fl); |
b89f4321 | 1245 | unlock_flocks(); |
4321e01e MW |
1246 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1247 | !new_fl->fl_next, break_time); | |
b89f4321 | 1248 | lock_flocks(); |
4321e01e | 1249 | __locks_delete_block(new_fl); |
1da177e4 LT |
1250 | if (error >= 0) { |
1251 | if (error == 0) | |
1252 | time_out_leases(inode); | |
1253 | /* Wait for the next lease that has not been broken yet */ | |
1254 | for (flock = inode->i_flock; flock && IS_LEASE(flock); | |
1255 | flock = flock->fl_next) { | |
ab83fa4b | 1256 | if (lease_breaking(flock)) |
1da177e4 LT |
1257 | goto restart; |
1258 | } | |
1259 | error = 0; | |
1260 | } | |
1261 | ||
1262 | out: | |
b89f4321 | 1263 | unlock_flocks(); |
e32b8ee2 | 1264 | if (!IS_ERR(new_fl)) |
1da177e4 LT |
1265 | locks_free_lock(new_fl); |
1266 | return error; | |
1267 | } | |
1268 | ||
1269 | EXPORT_SYMBOL(__break_lease); | |
1270 | ||
1271 | /** | |
a6b91919 | 1272 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1273 | * @inode: the inode |
1274 | * @time: pointer to a timespec which will contain the last modified time | |
1275 | * | |
1276 | * This is to force NFS clients to flush their caches for files with | |
1277 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1278 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1279 | */ |
1280 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1281 | { | |
1282 | struct file_lock *flock = inode->i_flock; | |
1283 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) | |
1284 | *time = current_fs_time(inode->i_sb); | |
1285 | else | |
1286 | *time = inode->i_mtime; | |
1287 | } | |
1288 | ||
1289 | EXPORT_SYMBOL(lease_get_mtime); | |
1290 | ||
1291 | /** | |
1292 | * fcntl_getlease - Enquire what lease is currently active | |
1293 | * @filp: the file | |
1294 | * | |
1295 | * The value returned by this function will be one of | |
1296 | * (if no lease break is pending): | |
1297 | * | |
1298 | * %F_RDLCK to indicate a shared lease is held. | |
1299 | * | |
1300 | * %F_WRLCK to indicate an exclusive lease is held. | |
1301 | * | |
1302 | * %F_UNLCK to indicate no lease is held. | |
1303 | * | |
1304 | * (if a lease break is pending): | |
1305 | * | |
1306 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1307 | * changed to a shared lease (or removed). | |
1308 | * | |
1309 | * %F_UNLCK to indicate the lease needs to be removed. | |
1310 | * | |
1311 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1312 | * should be returned to userspace. | |
1313 | */ | |
1314 | int fcntl_getlease(struct file *filp) | |
1315 | { | |
1316 | struct file_lock *fl; | |
1317 | int type = F_UNLCK; | |
1318 | ||
b89f4321 | 1319 | lock_flocks(); |
0f7fc9e4 JJS |
1320 | time_out_leases(filp->f_path.dentry->d_inode); |
1321 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1322 | fl = fl->fl_next) { |
1323 | if (fl->fl_file == filp) { | |
710b7216 | 1324 | type = fl->fl_type; |
1da177e4 LT |
1325 | break; |
1326 | } | |
1327 | } | |
b89f4321 | 1328 | unlock_flocks(); |
1da177e4 LT |
1329 | return type; |
1330 | } | |
1331 | ||
1332 | /** | |
0af1a450 | 1333 | * generic_setlease - sets a lease on an open file |
1da177e4 LT |
1334 | * @filp: file pointer |
1335 | * @arg: type of lease to obtain | |
1336 | * @flp: input - file_lock to use, output - file_lock inserted | |
1337 | * | |
8fb47a4f | 1338 | * The (input) flp->fl_lmops->lm_break function is required |
1da177e4 LT |
1339 | * by break_lease(). |
1340 | * | |
b89f4321 | 1341 | * Called with file_lock_lock held. |
1da177e4 | 1342 | */ |
0af1a450 | 1343 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1344 | { |
7eaae282 | 1345 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1346 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 LT |
1347 | struct inode *inode = dentry->d_inode; |
1348 | int error, rdlease_count = 0, wrlease_count = 0; | |
1349 | ||
096657b6 BF |
1350 | lease = *flp; |
1351 | ||
1352 | error = -EACCES; | |
da9592ed | 1353 | if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE)) |
096657b6 BF |
1354 | goto out; |
1355 | error = -EINVAL; | |
6d5e8b05 | 1356 | if (!S_ISREG(inode->i_mode)) |
096657b6 | 1357 | goto out; |
6d5e8b05 BF |
1358 | error = security_file_lock(filp, arg); |
1359 | if (error) | |
096657b6 | 1360 | goto out; |
6d5e8b05 | 1361 | |
1da177e4 LT |
1362 | time_out_leases(inode); |
1363 | ||
8fb47a4f | 1364 | BUG_ON(!(*flp)->fl_lmops->lm_break); |
1da177e4 | 1365 | |
288b2fd8 DR |
1366 | if (arg != F_UNLCK) { |
1367 | error = -EAGAIN; | |
1368 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1369 | goto out; | |
1370 | if ((arg == F_WRLCK) | |
b7ab39f6 | 1371 | && ((dentry->d_count > 1) |
288b2fd8 DR |
1372 | || (atomic_read(&inode->i_count) > 1))) |
1373 | goto out; | |
288b2fd8 | 1374 | } |
85c59580 | 1375 | |
1da177e4 LT |
1376 | /* |
1377 | * At this point, we know that if there is an exclusive | |
1378 | * lease on this file, then we hold it on this filp | |
1379 | * (otherwise our open of this file would have blocked). | |
1380 | * And if we are trying to acquire an exclusive lease, | |
1381 | * then the file is not open by anyone (including us) | |
1382 | * except for this filp. | |
1383 | */ | |
1384 | for (before = &inode->i_flock; | |
1385 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1386 | before = &fl->fl_next) { | |
255c7cf8 | 1387 | if (fl->fl_file == filp) |
1da177e4 | 1388 | my_before = before; |
710b7216 | 1389 | else if ((fl->fl_type == F_UNLCK) && lease_breaking(fl)) |
1da177e4 LT |
1390 | /* |
1391 | * Someone is in the process of opening this | |
1392 | * file for writing so we may not take an | |
1393 | * exclusive lease on it. | |
1394 | */ | |
1395 | wrlease_count++; | |
1396 | else | |
1397 | rdlease_count++; | |
1398 | } | |
1399 | ||
5fcc60c3 | 1400 | error = -EAGAIN; |
1da177e4 LT |
1401 | if ((arg == F_RDLCK && (wrlease_count > 0)) || |
1402 | (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0))) | |
1403 | goto out; | |
1404 | ||
1405 | if (my_before != NULL) { | |
8fb47a4f | 1406 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1407 | if (!error) |
1408 | *flp = *my_before; | |
1da177e4 LT |
1409 | goto out; |
1410 | } | |
1411 | ||
1da177e4 LT |
1412 | if (arg == F_UNLCK) |
1413 | goto out; | |
1414 | ||
1415 | error = -EINVAL; | |
1416 | if (!leases_enable) | |
1417 | goto out; | |
1418 | ||
c5b1f0d9 | 1419 | locks_insert_lock(before, lease); |
85c59580 | 1420 | return 0; |
1da177e4 | 1421 | |
1da177e4 LT |
1422 | out: |
1423 | return error; | |
1424 | } | |
0af1a450 | 1425 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1426 | |
b89f4321 AB |
1427 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1428 | { | |
1429 | if (filp->f_op && filp->f_op->setlease) | |
1430 | return filp->f_op->setlease(filp, arg, lease); | |
1431 | else | |
1432 | return generic_setlease(filp, arg, lease); | |
1433 | } | |
1434 | ||
1435 | /** | |
a9933cea | 1436 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1437 | * @filp: file pointer |
1438 | * @arg: type of lease to obtain | |
1439 | * @lease: file_lock to use | |
1440 | * | |
1441 | * Call this to establish a lease on the file. | |
8fb47a4f | 1442 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1443 | * break_lease will oops! |
1444 | * | |
1445 | * This will call the filesystem's setlease file method, if | |
1446 | * defined. Note that there is no getlease method; instead, the | |
1447 | * filesystem setlease method should call back to setlease() to | |
1448 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1449 | * find it. Since fcntl_getlease() only reports whether the current | |
1450 | * task holds a lease, a cluster filesystem need only do this for | |
1451 | * leases held by processes on this node. | |
1452 | * | |
1453 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1454 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1455 | * filesystem's open, create, and (on truncate) setattr methods. |
1456 | * | |
1457 | * Warning: the only current setlease methods exist only to disable | |
1458 | * leases in certain cases. More vfs changes may be required to | |
1459 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1460 | */ |
1461 | ||
a9933cea | 1462 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1463 | { |
1da177e4 LT |
1464 | int error; |
1465 | ||
b89f4321 AB |
1466 | lock_flocks(); |
1467 | error = __vfs_setlease(filp, arg, lease); | |
1468 | unlock_flocks(); | |
1da177e4 LT |
1469 | |
1470 | return error; | |
1471 | } | |
a9933cea | 1472 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1473 | |
0ceaf6c7 BF |
1474 | static int do_fcntl_delete_lease(struct file *filp) |
1475 | { | |
1476 | struct file_lock fl, *flp = &fl; | |
1477 | ||
1478 | lease_init(filp, F_UNLCK, flp); | |
1479 | ||
1480 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1481 | } | |
1482 | ||
1483 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1484 | { |
3df057ac | 1485 | struct file_lock *fl, *ret; |
f7347ce4 | 1486 | struct fasync_struct *new; |
1da177e4 LT |
1487 | int error; |
1488 | ||
c5b1f0d9 AB |
1489 | fl = lease_alloc(filp, arg); |
1490 | if (IS_ERR(fl)) | |
1491 | return PTR_ERR(fl); | |
1da177e4 | 1492 | |
f7347ce4 LT |
1493 | new = fasync_alloc(); |
1494 | if (!new) { | |
1495 | locks_free_lock(fl); | |
1496 | return -ENOMEM; | |
1497 | } | |
3df057ac | 1498 | ret = fl; |
b89f4321 | 1499 | lock_flocks(); |
8896b93f | 1500 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 CH |
1501 | if (error) { |
1502 | unlock_flocks(); | |
1503 | locks_free_lock(fl); | |
1504 | goto out_free_fasync; | |
1505 | } | |
3df057ac BF |
1506 | if (ret != fl) |
1507 | locks_free_lock(fl); | |
1da177e4 | 1508 | |
f7347ce4 LT |
1509 | /* |
1510 | * fasync_insert_entry() returns the old entry if any. | |
1511 | * If there was no old entry, then it used 'new' and | |
1512 | * inserted it into the fasync list. Clear new so that | |
1513 | * we don't release it here. | |
1514 | */ | |
3df057ac | 1515 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1516 | new = NULL; |
1517 | ||
8896b93f | 1518 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
b89f4321 | 1519 | unlock_flocks(); |
51ee4b84 CH |
1520 | |
1521 | out_free_fasync: | |
f7347ce4 LT |
1522 | if (new) |
1523 | fasync_free(new); | |
1da177e4 LT |
1524 | return error; |
1525 | } | |
1526 | ||
0ceaf6c7 BF |
1527 | /** |
1528 | * fcntl_setlease - sets a lease on an open file | |
1529 | * @fd: open file descriptor | |
1530 | * @filp: file pointer | |
1531 | * @arg: type of lease to obtain | |
1532 | * | |
1533 | * Call this fcntl to establish a lease on the file. | |
1534 | * Note that you also need to call %F_SETSIG to | |
1535 | * receive a signal when the lease is broken. | |
1536 | */ | |
1537 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1538 | { | |
1539 | if (arg == F_UNLCK) | |
1540 | return do_fcntl_delete_lease(filp); | |
1541 | return do_fcntl_add_lease(fd, filp, arg); | |
1542 | } | |
1543 | ||
1da177e4 LT |
1544 | /** |
1545 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1546 | * @filp: The file to apply the lock to | |
1547 | * @fl: The lock to be applied | |
1548 | * | |
1549 | * Add a FLOCK style lock to a file. | |
1550 | */ | |
1551 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1552 | { | |
1553 | int error; | |
1554 | might_sleep(); | |
1555 | for (;;) { | |
1556 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1557 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1558 | break; |
1559 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1560 | if (!error) | |
1561 | continue; | |
1562 | ||
1563 | locks_delete_block(fl); | |
1564 | break; | |
1565 | } | |
1566 | return error; | |
1567 | } | |
1568 | ||
1569 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1570 | ||
1571 | /** | |
1572 | * sys_flock: - flock() system call. | |
1573 | * @fd: the file descriptor to lock. | |
1574 | * @cmd: the type of lock to apply. | |
1575 | * | |
1576 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1577 | * The @cmd can be one of | |
1578 | * | |
1579 | * %LOCK_SH -- a shared lock. | |
1580 | * | |
1581 | * %LOCK_EX -- an exclusive lock. | |
1582 | * | |
1583 | * %LOCK_UN -- remove an existing lock. | |
1584 | * | |
1585 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1586 | * | |
1587 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1588 | * processes read and write access respectively. | |
1589 | */ | |
002c8976 | 1590 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 LT |
1591 | { |
1592 | struct file *filp; | |
1593 | struct file_lock *lock; | |
1594 | int can_sleep, unlock; | |
1595 | int error; | |
1596 | ||
1597 | error = -EBADF; | |
1598 | filp = fget(fd); | |
1599 | if (!filp) | |
1600 | goto out; | |
1601 | ||
1602 | can_sleep = !(cmd & LOCK_NB); | |
1603 | cmd &= ~LOCK_NB; | |
1604 | unlock = (cmd == LOCK_UN); | |
1605 | ||
aeb5d727 AV |
1606 | if (!unlock && !(cmd & LOCK_MAND) && |
1607 | !(filp->f_mode & (FMODE_READ|FMODE_WRITE))) | |
1da177e4 LT |
1608 | goto out_putf; |
1609 | ||
1610 | error = flock_make_lock(filp, &lock, cmd); | |
1611 | if (error) | |
1612 | goto out_putf; | |
1613 | if (can_sleep) | |
1614 | lock->fl_flags |= FL_SLEEP; | |
1615 | ||
713c0ecd | 1616 | error = security_file_lock(filp, lock->fl_type); |
1da177e4 LT |
1617 | if (error) |
1618 | goto out_free; | |
1619 | ||
1620 | if (filp->f_op && filp->f_op->flock) | |
1621 | error = filp->f_op->flock(filp, | |
1622 | (can_sleep) ? F_SETLKW : F_SETLK, | |
1623 | lock); | |
1624 | else | |
1625 | error = flock_lock_file_wait(filp, lock); | |
1626 | ||
1627 | out_free: | |
993dfa87 | 1628 | locks_free_lock(lock); |
1da177e4 LT |
1629 | |
1630 | out_putf: | |
1631 | fput(filp); | |
1632 | out: | |
1633 | return error; | |
1634 | } | |
1635 | ||
3ee17abd BF |
1636 | /** |
1637 | * vfs_test_lock - test file byte range lock | |
1638 | * @filp: The file to test lock for | |
6924c554 | 1639 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1640 | * |
1641 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1642 | * setting conf->fl_type to something other than F_UNLCK. | |
1643 | */ | |
1644 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1645 | { | |
1646 | if (filp->f_op && filp->f_op->lock) | |
1647 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1648 | posix_test_lock(filp, fl); | |
1649 | return 0; | |
1650 | } | |
1651 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1652 | ||
c2fa1b8a BF |
1653 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1654 | { | |
1655 | flock->l_pid = fl->fl_pid; | |
1656 | #if BITS_PER_LONG == 32 | |
1657 | /* | |
1658 | * Make sure we can represent the posix lock via | |
1659 | * legacy 32bit flock. | |
1660 | */ | |
1661 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1662 | return -EOVERFLOW; | |
1663 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1664 | return -EOVERFLOW; | |
1665 | #endif | |
1666 | flock->l_start = fl->fl_start; | |
1667 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1668 | fl->fl_end - fl->fl_start + 1; | |
1669 | flock->l_whence = 0; | |
129a84de | 1670 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1671 | return 0; |
1672 | } | |
1673 | ||
1674 | #if BITS_PER_LONG == 32 | |
1675 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1676 | { | |
1677 | flock->l_pid = fl->fl_pid; | |
1678 | flock->l_start = fl->fl_start; | |
1679 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1680 | fl->fl_end - fl->fl_start + 1; | |
1681 | flock->l_whence = 0; | |
1682 | flock->l_type = fl->fl_type; | |
1683 | } | |
1684 | #endif | |
1685 | ||
1da177e4 LT |
1686 | /* Report the first existing lock that would conflict with l. |
1687 | * This implements the F_GETLK command of fcntl(). | |
1688 | */ | |
1689 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1690 | { | |
9d6a8c5c | 1691 | struct file_lock file_lock; |
1da177e4 LT |
1692 | struct flock flock; |
1693 | int error; | |
1694 | ||
1695 | error = -EFAULT; | |
1696 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1697 | goto out; | |
1698 | error = -EINVAL; | |
1699 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1700 | goto out; | |
1701 | ||
1702 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1703 | if (error) | |
1704 | goto out; | |
1705 | ||
3ee17abd BF |
1706 | error = vfs_test_lock(filp, &file_lock); |
1707 | if (error) | |
1708 | goto out; | |
1da177e4 | 1709 | |
9d6a8c5c ME |
1710 | flock.l_type = file_lock.fl_type; |
1711 | if (file_lock.fl_type != F_UNLCK) { | |
1712 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1713 | if (error) |
1da177e4 | 1714 | goto out; |
1da177e4 LT |
1715 | } |
1716 | error = -EFAULT; | |
1717 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1718 | error = 0; | |
1719 | out: | |
1720 | return error; | |
1721 | } | |
1722 | ||
7723ec97 ME |
1723 | /** |
1724 | * vfs_lock_file - file byte range lock | |
1725 | * @filp: The file to apply the lock to | |
1726 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1727 | * @fl: The lock to be applied | |
150b3934 ME |
1728 | * @conf: Place to return a copy of the conflicting lock, if found. |
1729 | * | |
1730 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1731 | * as the final argument. | |
1732 | * | |
1733 | * If the filesystem defines a private ->lock() method, then @conf will | |
1734 | * be left unchanged; so a caller that cares should initialize it to | |
1735 | * some acceptable default. | |
2beb6614 ME |
1736 | * |
1737 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1738 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1739 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1740 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1741 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1742 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1743 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1744 | * request completes. |
1745 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1746 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1747 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1748 | * nonzero return code and the file system should release the lock. The file |
1749 | * system is also responsible to keep a corresponding posix lock when it | |
1750 | * grants a lock so the VFS can find out which locks are locally held and do | |
1751 | * the correct lock cleanup when required. | |
1752 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1753 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1754 | * return code. |
7723ec97 | 1755 | */ |
150b3934 | 1756 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1757 | { |
1758 | if (filp->f_op && filp->f_op->lock) | |
1759 | return filp->f_op->lock(filp, cmd, fl); | |
1760 | else | |
150b3934 | 1761 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1762 | } |
1763 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1764 | ||
b648a6de MS |
1765 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1766 | struct file_lock *fl) | |
1767 | { | |
1768 | int error; | |
1769 | ||
1770 | error = security_file_lock(filp, fl->fl_type); | |
1771 | if (error) | |
1772 | return error; | |
1773 | ||
764c76b3 MS |
1774 | for (;;) { |
1775 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1776 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1777 | break; |
764c76b3 MS |
1778 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1779 | if (!error) | |
1780 | continue; | |
1781 | ||
1782 | locks_delete_block(fl); | |
1783 | break; | |
b648a6de MS |
1784 | } |
1785 | ||
1786 | return error; | |
1787 | } | |
1788 | ||
1da177e4 LT |
1789 | /* Apply the lock described by l to an open file descriptor. |
1790 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1791 | */ | |
c293621b PS |
1792 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1793 | struct flock __user *l) | |
1da177e4 LT |
1794 | { |
1795 | struct file_lock *file_lock = locks_alloc_lock(); | |
1796 | struct flock flock; | |
1797 | struct inode *inode; | |
0b2bac2f | 1798 | struct file *f; |
1da177e4 LT |
1799 | int error; |
1800 | ||
1801 | if (file_lock == NULL) | |
1802 | return -ENOLCK; | |
1803 | ||
1804 | /* | |
1805 | * This might block, so we do it before checking the inode. | |
1806 | */ | |
1807 | error = -EFAULT; | |
1808 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1809 | goto out; | |
1810 | ||
0f7fc9e4 | 1811 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1812 | |
1813 | /* Don't allow mandatory locks on files that may be memory mapped | |
1814 | * and shared. | |
1815 | */ | |
a16877ca | 1816 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1817 | error = -EAGAIN; |
1818 | goto out; | |
1819 | } | |
1820 | ||
c293621b | 1821 | again: |
1da177e4 LT |
1822 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1823 | if (error) | |
1824 | goto out; | |
1825 | if (cmd == F_SETLKW) { | |
1826 | file_lock->fl_flags |= FL_SLEEP; | |
1827 | } | |
1828 | ||
1829 | error = -EBADF; | |
1830 | switch (flock.l_type) { | |
1831 | case F_RDLCK: | |
1832 | if (!(filp->f_mode & FMODE_READ)) | |
1833 | goto out; | |
1834 | break; | |
1835 | case F_WRLCK: | |
1836 | if (!(filp->f_mode & FMODE_WRITE)) | |
1837 | goto out; | |
1838 | break; | |
1839 | case F_UNLCK: | |
1840 | break; | |
1841 | default: | |
1842 | error = -EINVAL; | |
1843 | goto out; | |
1844 | } | |
1845 | ||
b648a6de | 1846 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1847 | |
c293621b PS |
1848 | /* |
1849 | * Attempt to detect a close/fcntl race and recover by | |
1850 | * releasing the lock that was just acquired. | |
1851 | */ | |
0b2bac2f AV |
1852 | /* |
1853 | * we need that spin_lock here - it prevents reordering between | |
1854 | * update of inode->i_flock and check for it done in close(). | |
1855 | * rcu_read_lock() wouldn't do. | |
1856 | */ | |
1857 | spin_lock(¤t->files->file_lock); | |
1858 | f = fcheck(fd); | |
1859 | spin_unlock(¤t->files->file_lock); | |
1860 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1861 | flock.l_type = F_UNLCK; |
1862 | goto again; | |
1da177e4 LT |
1863 | } |
1864 | ||
c293621b | 1865 | out: |
1da177e4 LT |
1866 | locks_free_lock(file_lock); |
1867 | return error; | |
1868 | } | |
1869 | ||
1870 | #if BITS_PER_LONG == 32 | |
1871 | /* Report the first existing lock that would conflict with l. | |
1872 | * This implements the F_GETLK command of fcntl(). | |
1873 | */ | |
1874 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
1875 | { | |
9d6a8c5c | 1876 | struct file_lock file_lock; |
1da177e4 LT |
1877 | struct flock64 flock; |
1878 | int error; | |
1879 | ||
1880 | error = -EFAULT; | |
1881 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1882 | goto out; | |
1883 | error = -EINVAL; | |
1884 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1885 | goto out; | |
1886 | ||
1887 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
1888 | if (error) | |
1889 | goto out; | |
1890 | ||
3ee17abd BF |
1891 | error = vfs_test_lock(filp, &file_lock); |
1892 | if (error) | |
1893 | goto out; | |
1894 | ||
9d6a8c5c ME |
1895 | flock.l_type = file_lock.fl_type; |
1896 | if (file_lock.fl_type != F_UNLCK) | |
1897 | posix_lock_to_flock64(&flock, &file_lock); | |
1898 | ||
1da177e4 LT |
1899 | error = -EFAULT; |
1900 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1901 | error = 0; | |
1902 | ||
1903 | out: | |
1904 | return error; | |
1905 | } | |
1906 | ||
1907 | /* Apply the lock described by l to an open file descriptor. | |
1908 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1909 | */ | |
c293621b PS |
1910 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1911 | struct flock64 __user *l) | |
1da177e4 LT |
1912 | { |
1913 | struct file_lock *file_lock = locks_alloc_lock(); | |
1914 | struct flock64 flock; | |
1915 | struct inode *inode; | |
0b2bac2f | 1916 | struct file *f; |
1da177e4 LT |
1917 | int error; |
1918 | ||
1919 | if (file_lock == NULL) | |
1920 | return -ENOLCK; | |
1921 | ||
1922 | /* | |
1923 | * This might block, so we do it before checking the inode. | |
1924 | */ | |
1925 | error = -EFAULT; | |
1926 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1927 | goto out; | |
1928 | ||
0f7fc9e4 | 1929 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1930 | |
1931 | /* Don't allow mandatory locks on files that may be memory mapped | |
1932 | * and shared. | |
1933 | */ | |
a16877ca | 1934 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1935 | error = -EAGAIN; |
1936 | goto out; | |
1937 | } | |
1938 | ||
c293621b | 1939 | again: |
1da177e4 LT |
1940 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1941 | if (error) | |
1942 | goto out; | |
1943 | if (cmd == F_SETLKW64) { | |
1944 | file_lock->fl_flags |= FL_SLEEP; | |
1945 | } | |
1946 | ||
1947 | error = -EBADF; | |
1948 | switch (flock.l_type) { | |
1949 | case F_RDLCK: | |
1950 | if (!(filp->f_mode & FMODE_READ)) | |
1951 | goto out; | |
1952 | break; | |
1953 | case F_WRLCK: | |
1954 | if (!(filp->f_mode & FMODE_WRITE)) | |
1955 | goto out; | |
1956 | break; | |
1957 | case F_UNLCK: | |
1958 | break; | |
1959 | default: | |
1960 | error = -EINVAL; | |
1961 | goto out; | |
1962 | } | |
1963 | ||
b648a6de | 1964 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1965 | |
c293621b PS |
1966 | /* |
1967 | * Attempt to detect a close/fcntl race and recover by | |
1968 | * releasing the lock that was just acquired. | |
1969 | */ | |
0b2bac2f AV |
1970 | spin_lock(¤t->files->file_lock); |
1971 | f = fcheck(fd); | |
1972 | spin_unlock(¤t->files->file_lock); | |
1973 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1974 | flock.l_type = F_UNLCK; |
1975 | goto again; | |
1da177e4 LT |
1976 | } |
1977 | ||
1978 | out: | |
1979 | locks_free_lock(file_lock); | |
1980 | return error; | |
1981 | } | |
1982 | #endif /* BITS_PER_LONG == 32 */ | |
1983 | ||
1984 | /* | |
1985 | * This function is called when the file is being removed | |
1986 | * from the task's fd array. POSIX locks belonging to this task | |
1987 | * are deleted at this time. | |
1988 | */ | |
1989 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
1990 | { | |
ff7b86b8 | 1991 | struct file_lock lock; |
1da177e4 LT |
1992 | |
1993 | /* | |
1994 | * If there are no locks held on this file, we don't need to call | |
1995 | * posix_lock_file(). Another process could be setting a lock on this | |
1996 | * file at the same time, but we wouldn't remove that lock anyway. | |
1997 | */ | |
0f7fc9e4 | 1998 | if (!filp->f_path.dentry->d_inode->i_flock) |
1da177e4 LT |
1999 | return; |
2000 | ||
2001 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2002 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2003 | lock.fl_start = 0; |
2004 | lock.fl_end = OFFSET_MAX; | |
2005 | lock.fl_owner = owner; | |
2006 | lock.fl_pid = current->tgid; | |
2007 | lock.fl_file = filp; | |
2008 | lock.fl_ops = NULL; | |
2009 | lock.fl_lmops = NULL; | |
2010 | ||
150b3934 | 2011 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2012 | |
1da177e4 LT |
2013 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2014 | lock.fl_ops->fl_release_private(&lock); | |
2015 | } | |
2016 | ||
2017 | EXPORT_SYMBOL(locks_remove_posix); | |
2018 | ||
2019 | /* | |
2020 | * This function is called on the last close of an open file. | |
2021 | */ | |
2022 | void locks_remove_flock(struct file *filp) | |
2023 | { | |
0f7fc9e4 | 2024 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
2025 | struct file_lock *fl; |
2026 | struct file_lock **before; | |
2027 | ||
2028 | if (!inode->i_flock) | |
2029 | return; | |
2030 | ||
2031 | if (filp->f_op && filp->f_op->flock) { | |
2032 | struct file_lock fl = { | |
2033 | .fl_pid = current->tgid, | |
2034 | .fl_file = filp, | |
2035 | .fl_flags = FL_FLOCK, | |
2036 | .fl_type = F_UNLCK, | |
2037 | .fl_end = OFFSET_MAX, | |
2038 | }; | |
2039 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2040 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2041 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2042 | } |
2043 | ||
b89f4321 | 2044 | lock_flocks(); |
1da177e4 LT |
2045 | before = &inode->i_flock; |
2046 | ||
2047 | while ((fl = *before) != NULL) { | |
2048 | if (fl->fl_file == filp) { | |
c293621b | 2049 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2050 | locks_delete_lock(before); |
2051 | continue; | |
2052 | } | |
2053 | if (IS_LEASE(fl)) { | |
2054 | lease_modify(before, F_UNLCK); | |
2055 | continue; | |
2056 | } | |
2057 | /* What? */ | |
2058 | BUG(); | |
2059 | } | |
2060 | before = &fl->fl_next; | |
2061 | } | |
b89f4321 | 2062 | unlock_flocks(); |
1da177e4 LT |
2063 | } |
2064 | ||
1da177e4 LT |
2065 | /** |
2066 | * posix_unblock_lock - stop waiting for a file lock | |
2067 | * @filp: how the file was opened | |
2068 | * @waiter: the lock which was waiting | |
2069 | * | |
2070 | * lockd needs to block waiting for locks. | |
2071 | */ | |
64a318ee | 2072 | int |
1da177e4 LT |
2073 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
2074 | { | |
64a318ee BF |
2075 | int status = 0; |
2076 | ||
b89f4321 | 2077 | lock_flocks(); |
5996a298 | 2078 | if (waiter->fl_next) |
1da177e4 | 2079 | __locks_delete_block(waiter); |
64a318ee BF |
2080 | else |
2081 | status = -ENOENT; | |
b89f4321 | 2082 | unlock_flocks(); |
64a318ee | 2083 | return status; |
1da177e4 LT |
2084 | } |
2085 | ||
2086 | EXPORT_SYMBOL(posix_unblock_lock); | |
2087 | ||
9b9d2ab4 ME |
2088 | /** |
2089 | * vfs_cancel_lock - file byte range unblock lock | |
2090 | * @filp: The file to apply the unblock to | |
2091 | * @fl: The lock to be unblocked | |
2092 | * | |
2093 | * Used by lock managers to cancel blocked requests | |
2094 | */ | |
2095 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2096 | { | |
2097 | if (filp->f_op && filp->f_op->lock) | |
2098 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2099 | return 0; | |
2100 | } | |
2101 | ||
2102 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2103 | ||
7f8ada98 | 2104 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2105 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2106 | #include <linux/seq_file.h> |
2107 | ||
2108 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
99dc8292 | 2109 | loff_t id, char *pfx) |
1da177e4 LT |
2110 | { |
2111 | struct inode *inode = NULL; | |
ab1f1611 VG |
2112 | unsigned int fl_pid; |
2113 | ||
2114 | if (fl->fl_nspid) | |
6c5f3e7b | 2115 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2116 | else |
2117 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2118 | |
2119 | if (fl->fl_file != NULL) | |
0f7fc9e4 | 2120 | inode = fl->fl_file->f_path.dentry->d_inode; |
1da177e4 | 2121 | |
99dc8292 | 2122 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2123 | if (IS_POSIX(fl)) { |
7f8ada98 | 2124 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2125 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2126 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2127 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2128 | } else if (IS_FLOCK(fl)) { |
2129 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2130 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2131 | } else { |
7f8ada98 | 2132 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2133 | } |
2134 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2135 | seq_printf(f, "LEASE "); |
ab83fa4b | 2136 | if (lease_breaking(fl)) |
7f8ada98 | 2137 | seq_printf(f, "BREAKING "); |
1da177e4 | 2138 | else if (fl->fl_file) |
7f8ada98 | 2139 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2140 | else |
7f8ada98 | 2141 | seq_printf(f, "BREAKER "); |
1da177e4 | 2142 | } else { |
7f8ada98 | 2143 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2144 | } |
2145 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2146 | seq_printf(f, "%s ", |
1da177e4 LT |
2147 | (fl->fl_type & LOCK_READ) |
2148 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2149 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2150 | } else { | |
7f8ada98 | 2151 | seq_printf(f, "%s ", |
ab83fa4b | 2152 | (lease_breaking(fl)) |
1da177e4 LT |
2153 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " |
2154 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); | |
2155 | } | |
2156 | if (inode) { | |
2157 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2158 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2159 | inode->i_sb->s_id, inode->i_ino); |
2160 | #else | |
2161 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2162 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2163 | MAJOR(inode->i_sb->s_dev), |
2164 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2165 | #endif | |
2166 | } else { | |
ab1f1611 | 2167 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2168 | } |
2169 | if (IS_POSIX(fl)) { | |
2170 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2171 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2172 | else |
7f8ada98 | 2173 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2174 | } else { |
7f8ada98 | 2175 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2176 | } |
2177 | } | |
2178 | ||
7f8ada98 | 2179 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2180 | { |
7f8ada98 | 2181 | struct file_lock *fl, *bfl; |
1da177e4 | 2182 | |
7f8ada98 | 2183 | fl = list_entry(v, struct file_lock, fl_link); |
1da177e4 | 2184 | |
99dc8292 | 2185 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
1da177e4 | 2186 | |
7f8ada98 | 2187 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
99dc8292 | 2188 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
094f2825 | 2189 | |
7f8ada98 PE |
2190 | return 0; |
2191 | } | |
1da177e4 | 2192 | |
7f8ada98 PE |
2193 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2194 | { | |
99dc8292 JM |
2195 | loff_t *p = f->private; |
2196 | ||
b89f4321 | 2197 | lock_flocks(); |
99dc8292 | 2198 | *p = (*pos + 1); |
7f8ada98 PE |
2199 | return seq_list_start(&file_lock_list, *pos); |
2200 | } | |
1da177e4 | 2201 | |
7f8ada98 PE |
2202 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2203 | { | |
99dc8292 JM |
2204 | loff_t *p = f->private; |
2205 | ++*p; | |
7f8ada98 PE |
2206 | return seq_list_next(v, &file_lock_list, pos); |
2207 | } | |
1da177e4 | 2208 | |
7f8ada98 PE |
2209 | static void locks_stop(struct seq_file *f, void *v) |
2210 | { | |
b89f4321 | 2211 | unlock_flocks(); |
1da177e4 LT |
2212 | } |
2213 | ||
d8ba7a36 | 2214 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2215 | .start = locks_start, |
2216 | .next = locks_next, | |
2217 | .stop = locks_stop, | |
2218 | .show = locks_show, | |
2219 | }; | |
d8ba7a36 AD |
2220 | |
2221 | static int locks_open(struct inode *inode, struct file *filp) | |
2222 | { | |
99dc8292 | 2223 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
d8ba7a36 AD |
2224 | } |
2225 | ||
2226 | static const struct file_operations proc_locks_operations = { | |
2227 | .open = locks_open, | |
2228 | .read = seq_read, | |
2229 | .llseek = seq_lseek, | |
99dc8292 | 2230 | .release = seq_release_private, |
d8ba7a36 AD |
2231 | }; |
2232 | ||
2233 | static int __init proc_locks_init(void) | |
2234 | { | |
2235 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2236 | return 0; | |
2237 | } | |
2238 | module_init(proc_locks_init); | |
7f8ada98 PE |
2239 | #endif |
2240 | ||
1da177e4 LT |
2241 | /** |
2242 | * lock_may_read - checks that the region is free of locks | |
2243 | * @inode: the inode that is being read | |
2244 | * @start: the first byte to read | |
2245 | * @len: the number of bytes to read | |
2246 | * | |
2247 | * Emulates Windows locking requirements. Whole-file | |
2248 | * mandatory locks (share modes) can prohibit a read and | |
2249 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2250 | * | |
2251 | * N.B. this function is only ever called | |
2252 | * from knfsd and ownership of locks is never checked. | |
2253 | */ | |
2254 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2255 | { | |
2256 | struct file_lock *fl; | |
2257 | int result = 1; | |
b89f4321 | 2258 | lock_flocks(); |
1da177e4 LT |
2259 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2260 | if (IS_POSIX(fl)) { | |
2261 | if (fl->fl_type == F_RDLCK) | |
2262 | continue; | |
2263 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2264 | continue; | |
2265 | } else if (IS_FLOCK(fl)) { | |
2266 | if (!(fl->fl_type & LOCK_MAND)) | |
2267 | continue; | |
2268 | if (fl->fl_type & LOCK_READ) | |
2269 | continue; | |
2270 | } else | |
2271 | continue; | |
2272 | result = 0; | |
2273 | break; | |
2274 | } | |
b89f4321 | 2275 | unlock_flocks(); |
1da177e4 LT |
2276 | return result; |
2277 | } | |
2278 | ||
2279 | EXPORT_SYMBOL(lock_may_read); | |
2280 | ||
2281 | /** | |
2282 | * lock_may_write - checks that the region is free of locks | |
2283 | * @inode: the inode that is being written | |
2284 | * @start: the first byte to write | |
2285 | * @len: the number of bytes to write | |
2286 | * | |
2287 | * Emulates Windows locking requirements. Whole-file | |
2288 | * mandatory locks (share modes) can prohibit a write and | |
2289 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2290 | * | |
2291 | * N.B. this function is only ever called | |
2292 | * from knfsd and ownership of locks is never checked. | |
2293 | */ | |
2294 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2295 | { | |
2296 | struct file_lock *fl; | |
2297 | int result = 1; | |
b89f4321 | 2298 | lock_flocks(); |
1da177e4 LT |
2299 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2300 | if (IS_POSIX(fl)) { | |
2301 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2302 | continue; | |
2303 | } else if (IS_FLOCK(fl)) { | |
2304 | if (!(fl->fl_type & LOCK_MAND)) | |
2305 | continue; | |
2306 | if (fl->fl_type & LOCK_WRITE) | |
2307 | continue; | |
2308 | } else | |
2309 | continue; | |
2310 | result = 0; | |
2311 | break; | |
2312 | } | |
b89f4321 | 2313 | unlock_flocks(); |
1da177e4 LT |
2314 | return result; |
2315 | } | |
2316 | ||
2317 | EXPORT_SYMBOL(lock_may_write); | |
2318 | ||
1da177e4 LT |
2319 | static int __init filelock_init(void) |
2320 | { | |
2321 | filelock_cache = kmem_cache_create("file_lock_cache", | |
ee19cc40 MS |
2322 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2323 | ||
1da177e4 LT |
2324 | return 0; |
2325 | } | |
2326 | ||
2327 | core_initcall(filelock_init); |