]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/locks.c | |
3 | * | |
4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
6 | * | |
7 | * Deadlock detection added. | |
8 | * FIXME: one thing isn't handled yet: | |
9 | * - mandatory locks (requires lots of changes elsewhere) | |
10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
11 | * | |
12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
14 | * | |
15 | * Converted file_lock_table to a linked list from an array, which eliminates | |
16 | * the limits on how many active file locks are open. | |
17 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
18 | * | |
19 | * Removed dependency on file descriptors. dup()'ed file descriptors now | |
20 | * get the same locks as the original file descriptors, and a close() on | |
21 | * any file descriptor removes ALL the locks on the file for the current | |
22 | * process. Since locks still depend on the process id, locks are inherited | |
23 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
24 | * BSD and SVR4 practice. | |
25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
26 | * | |
27 | * Scrapped free list which is redundant now that we allocate locks | |
28 | * dynamically with kmalloc()/kfree(). | |
29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
30 | * | |
31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
32 | * | |
33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
34 | * fcntl() system call. They have the semantics described above. | |
35 | * | |
36 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
37 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
38 | * and will continue to use the old, broken implementation. | |
39 | * | |
40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
41 | * with a file pointer (filp). As a result they can be shared by a parent | |
42 | * process and its children after a fork(). They are removed when the last | |
43 | * file descriptor referring to the file pointer is closed (unless explicitly | |
44 | * unlocked). | |
45 | * | |
46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
47 | * upgrading from shared to exclusive (or vice versa). When this happens | |
48 | * any processes blocked by the current lock are woken up and allowed to | |
49 | * run before the new lock is applied. | |
50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
51 | * | |
52 | * Removed some race conditions in flock_lock_file(), marked other possible | |
53 | * races. Just grep for FIXME to see them. | |
54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. | |
55 | * | |
56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
58 | * once we've checked for blocking and deadlocking. | |
59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
60 | * | |
61 | * Initial implementation of mandatory locks. SunOS turned out to be | |
62 | * a rotten model, so I implemented the "obvious" semantics. | |
63 | * See 'Documentation/mandatory.txt' for details. | |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. | |
65 | * | |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
69 | * Manual, Section 2. | |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
71 | * | |
72 | * Tidied up block list handling. Added '/proc/locks' interface. | |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
74 | * | |
75 | * Fixed deadlock condition for pathological code that mixes calls to | |
76 | * flock() and fcntl(). | |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
78 | * | |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
81 | * guarantee sensible behaviour in the case where file system modules might | |
82 | * be compiled with different options than the kernel itself. | |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
84 | * | |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
88 | * | |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
91 | * have already been freed. | |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
93 | * | |
94 | * Made the block list a circular list to minimise searching in the list. | |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
96 | * | |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
98 | * locking. | |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
100 | * | |
101 | * Some adaptations for NFS support. | |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
103 | * | |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
106 | * | |
107 | * Use slab allocator instead of kmalloc/kfree. | |
108 | * Use generic list implementation from <linux/list.h>. | |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
111 | * | |
112 | * Leases and LOCK_MAND | |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
115 | */ | |
116 | ||
117 | #include <linux/capability.h> | |
118 | #include <linux/file.h> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
120 | #include <linux/fs.h> |
121 | #include <linux/init.h> | |
122 | #include <linux/module.h> | |
123 | #include <linux/security.h> | |
124 | #include <linux/slab.h> | |
125 | #include <linux/smp_lock.h> | |
126 | #include <linux/syscalls.h> | |
127 | #include <linux/time.h> | |
4fb3a538 | 128 | #include <linux/rcupdate.h> |
ab1f1611 | 129 | #include <linux/pid_namespace.h> |
1da177e4 | 130 | |
1da177e4 LT |
131 | #include <asm/uaccess.h> |
132 | ||
133 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
134 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
135 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) | |
136 | ||
137 | int leases_enable = 1; | |
138 | int lease_break_time = 45; | |
139 | ||
140 | #define for_each_lock(inode, lockp) \ | |
141 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
142 | ||
26bcbf96 | 143 | static LIST_HEAD(file_lock_list); |
1da177e4 | 144 | static LIST_HEAD(blocked_list); |
72f98e72 | 145 | static DEFINE_SPINLOCK(file_lock_lock); |
1da177e4 | 146 | |
b89f4321 AB |
147 | /* |
148 | * Protects the two list heads above, plus the inode->i_flock list | |
149 | * FIXME: should use a spinlock, once lockd and ceph are ready. | |
150 | */ | |
151 | void lock_flocks(void) | |
152 | { | |
72f98e72 | 153 | spin_lock(&file_lock_lock); |
b89f4321 AB |
154 | } |
155 | EXPORT_SYMBOL_GPL(lock_flocks); | |
156 | ||
157 | void unlock_flocks(void) | |
158 | { | |
72f98e72 | 159 | spin_unlock(&file_lock_lock); |
b89f4321 AB |
160 | } |
161 | EXPORT_SYMBOL_GPL(unlock_flocks); | |
162 | ||
e18b890b | 163 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 LT |
164 | |
165 | /* Allocate an empty lock structure. */ | |
c5b1f0d9 | 166 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 167 | { |
e94b1766 | 168 | return kmem_cache_alloc(filelock_cache, GFP_KERNEL); |
1da177e4 | 169 | } |
c5b1f0d9 | 170 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 171 | |
a9e61e25 | 172 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
173 | { |
174 | if (fl->fl_ops) { | |
175 | if (fl->fl_ops->fl_release_private) | |
176 | fl->fl_ops->fl_release_private(fl); | |
177 | fl->fl_ops = NULL; | |
178 | } | |
179 | if (fl->fl_lmops) { | |
180 | if (fl->fl_lmops->fl_release_private) | |
181 | fl->fl_lmops->fl_release_private(fl); | |
182 | fl->fl_lmops = NULL; | |
183 | } | |
184 | ||
185 | } | |
a9e61e25 | 186 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 187 | |
1da177e4 | 188 | /* Free a lock which is not in use. */ |
05fa3135 | 189 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 190 | { |
5ce29646 MS |
191 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
192 | BUG_ON(!list_empty(&fl->fl_block)); | |
193 | BUG_ON(!list_empty(&fl->fl_link)); | |
1da177e4 | 194 | |
47831f35 | 195 | locks_release_private(fl); |
1da177e4 LT |
196 | kmem_cache_free(filelock_cache, fl); |
197 | } | |
05fa3135 | 198 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
199 | |
200 | void locks_init_lock(struct file_lock *fl) | |
201 | { | |
202 | INIT_LIST_HEAD(&fl->fl_link); | |
203 | INIT_LIST_HEAD(&fl->fl_block); | |
204 | init_waitqueue_head(&fl->fl_wait); | |
205 | fl->fl_next = NULL; | |
206 | fl->fl_fasync = NULL; | |
207 | fl->fl_owner = NULL; | |
208 | fl->fl_pid = 0; | |
ab1f1611 | 209 | fl->fl_nspid = NULL; |
1da177e4 LT |
210 | fl->fl_file = NULL; |
211 | fl->fl_flags = 0; | |
212 | fl->fl_type = 0; | |
213 | fl->fl_start = fl->fl_end = 0; | |
214 | fl->fl_ops = NULL; | |
215 | fl->fl_lmops = NULL; | |
216 | } | |
217 | ||
218 | EXPORT_SYMBOL(locks_init_lock); | |
219 | ||
220 | /* | |
221 | * Initialises the fields of the file lock which are invariant for | |
222 | * free file_locks. | |
223 | */ | |
51cc5068 | 224 | static void init_once(void *foo) |
1da177e4 LT |
225 | { |
226 | struct file_lock *lock = (struct file_lock *) foo; | |
227 | ||
1da177e4 LT |
228 | locks_init_lock(lock); |
229 | } | |
230 | ||
47831f35 TM |
231 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
232 | { | |
233 | if (fl->fl_ops) { | |
234 | if (fl->fl_ops->fl_copy_lock) | |
235 | fl->fl_ops->fl_copy_lock(new, fl); | |
236 | new->fl_ops = fl->fl_ops; | |
237 | } | |
bb8430a2 | 238 | if (fl->fl_lmops) |
47831f35 | 239 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
240 | } |
241 | ||
1da177e4 LT |
242 | /* |
243 | * Initialize a new lock from an existing file_lock structure. | |
244 | */ | |
1a747ee0 | 245 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
246 | { |
247 | new->fl_owner = fl->fl_owner; | |
248 | new->fl_pid = fl->fl_pid; | |
0996905f | 249 | new->fl_file = NULL; |
1da177e4 LT |
250 | new->fl_flags = fl->fl_flags; |
251 | new->fl_type = fl->fl_type; | |
252 | new->fl_start = fl->fl_start; | |
253 | new->fl_end = fl->fl_end; | |
0996905f TM |
254 | new->fl_ops = NULL; |
255 | new->fl_lmops = NULL; | |
256 | } | |
3dd7b71c | 257 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
258 | |
259 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
260 | { | |
261 | locks_release_private(new); | |
262 | ||
263 | __locks_copy_lock(new, fl); | |
264 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
265 | new->fl_ops = fl->fl_ops; |
266 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
267 | |
268 | locks_copy_private(new, fl); | |
1da177e4 LT |
269 | } |
270 | ||
271 | EXPORT_SYMBOL(locks_copy_lock); | |
272 | ||
273 | static inline int flock_translate_cmd(int cmd) { | |
274 | if (cmd & LOCK_MAND) | |
275 | return cmd & (LOCK_MAND | LOCK_RW); | |
276 | switch (cmd) { | |
277 | case LOCK_SH: | |
278 | return F_RDLCK; | |
279 | case LOCK_EX: | |
280 | return F_WRLCK; | |
281 | case LOCK_UN: | |
282 | return F_UNLCK; | |
283 | } | |
284 | return -EINVAL; | |
285 | } | |
286 | ||
287 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
288 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
289 | unsigned int cmd) | |
290 | { | |
291 | struct file_lock *fl; | |
292 | int type = flock_translate_cmd(cmd); | |
293 | if (type < 0) | |
294 | return type; | |
295 | ||
296 | fl = locks_alloc_lock(); | |
297 | if (fl == NULL) | |
298 | return -ENOMEM; | |
299 | ||
300 | fl->fl_file = filp; | |
301 | fl->fl_pid = current->tgid; | |
302 | fl->fl_flags = FL_FLOCK; | |
303 | fl->fl_type = type; | |
304 | fl->fl_end = OFFSET_MAX; | |
305 | ||
306 | *lock = fl; | |
307 | return 0; | |
308 | } | |
309 | ||
310 | static int assign_type(struct file_lock *fl, int type) | |
311 | { | |
312 | switch (type) { | |
313 | case F_RDLCK: | |
314 | case F_WRLCK: | |
315 | case F_UNLCK: | |
316 | fl->fl_type = type; | |
317 | break; | |
318 | default: | |
319 | return -EINVAL; | |
320 | } | |
321 | return 0; | |
322 | } | |
323 | ||
324 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
325 | * style lock. | |
326 | */ | |
327 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
328 | struct flock *l) | |
329 | { | |
330 | off_t start, end; | |
331 | ||
332 | switch (l->l_whence) { | |
f5579f8c | 333 | case SEEK_SET: |
1da177e4 LT |
334 | start = 0; |
335 | break; | |
f5579f8c | 336 | case SEEK_CUR: |
1da177e4 LT |
337 | start = filp->f_pos; |
338 | break; | |
f5579f8c | 339 | case SEEK_END: |
0f7fc9e4 | 340 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
341 | break; |
342 | default: | |
343 | return -EINVAL; | |
344 | } | |
345 | ||
346 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
347 | POSIX-2001 defines it. */ | |
348 | start += l->l_start; | |
4c780a46 TM |
349 | if (start < 0) |
350 | return -EINVAL; | |
351 | fl->fl_end = OFFSET_MAX; | |
352 | if (l->l_len > 0) { | |
353 | end = start + l->l_len - 1; | |
354 | fl->fl_end = end; | |
355 | } else if (l->l_len < 0) { | |
1da177e4 | 356 | end = start - 1; |
4c780a46 | 357 | fl->fl_end = end; |
1da177e4 | 358 | start += l->l_len; |
4c780a46 TM |
359 | if (start < 0) |
360 | return -EINVAL; | |
1da177e4 | 361 | } |
1da177e4 | 362 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
363 | if (fl->fl_end < fl->fl_start) |
364 | return -EOVERFLOW; | |
1da177e4 LT |
365 | |
366 | fl->fl_owner = current->files; | |
367 | fl->fl_pid = current->tgid; | |
368 | fl->fl_file = filp; | |
369 | fl->fl_flags = FL_POSIX; | |
370 | fl->fl_ops = NULL; | |
371 | fl->fl_lmops = NULL; | |
372 | ||
373 | return assign_type(fl, l->l_type); | |
374 | } | |
375 | ||
376 | #if BITS_PER_LONG == 32 | |
377 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
378 | struct flock64 *l) | |
379 | { | |
380 | loff_t start; | |
381 | ||
382 | switch (l->l_whence) { | |
f5579f8c | 383 | case SEEK_SET: |
1da177e4 LT |
384 | start = 0; |
385 | break; | |
f5579f8c | 386 | case SEEK_CUR: |
1da177e4 LT |
387 | start = filp->f_pos; |
388 | break; | |
f5579f8c | 389 | case SEEK_END: |
0f7fc9e4 | 390 | start = i_size_read(filp->f_path.dentry->d_inode); |
1da177e4 LT |
391 | break; |
392 | default: | |
393 | return -EINVAL; | |
394 | } | |
395 | ||
4c780a46 TM |
396 | start += l->l_start; |
397 | if (start < 0) | |
1da177e4 | 398 | return -EINVAL; |
4c780a46 TM |
399 | fl->fl_end = OFFSET_MAX; |
400 | if (l->l_len > 0) { | |
401 | fl->fl_end = start + l->l_len - 1; | |
402 | } else if (l->l_len < 0) { | |
403 | fl->fl_end = start - 1; | |
404 | start += l->l_len; | |
405 | if (start < 0) | |
406 | return -EINVAL; | |
407 | } | |
1da177e4 | 408 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
409 | if (fl->fl_end < fl->fl_start) |
410 | return -EOVERFLOW; | |
1da177e4 LT |
411 | |
412 | fl->fl_owner = current->files; | |
413 | fl->fl_pid = current->tgid; | |
414 | fl->fl_file = filp; | |
415 | fl->fl_flags = FL_POSIX; | |
416 | fl->fl_ops = NULL; | |
417 | fl->fl_lmops = NULL; | |
418 | ||
419 | switch (l->l_type) { | |
420 | case F_RDLCK: | |
421 | case F_WRLCK: | |
422 | case F_UNLCK: | |
423 | fl->fl_type = l->l_type; | |
424 | break; | |
425 | default: | |
426 | return -EINVAL; | |
427 | } | |
428 | ||
429 | return (0); | |
430 | } | |
431 | #endif | |
432 | ||
433 | /* default lease lock manager operations */ | |
434 | static void lease_break_callback(struct file_lock *fl) | |
435 | { | |
436 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
437 | } | |
438 | ||
439 | static void lease_release_private_callback(struct file_lock *fl) | |
440 | { | |
441 | if (!fl->fl_file) | |
442 | return; | |
443 | ||
444 | f_delown(fl->fl_file); | |
445 | fl->fl_file->f_owner.signum = 0; | |
446 | } | |
447 | ||
75c96f85 | 448 | static int lease_mylease_callback(struct file_lock *fl, struct file_lock *try) |
1da177e4 LT |
449 | { |
450 | return fl->fl_file == try->fl_file; | |
451 | } | |
452 | ||
7b021967 | 453 | static const struct lock_manager_operations lease_manager_ops = { |
1da177e4 LT |
454 | .fl_break = lease_break_callback, |
455 | .fl_release_private = lease_release_private_callback, | |
456 | .fl_mylease = lease_mylease_callback, | |
457 | .fl_change = lease_modify, | |
458 | }; | |
459 | ||
460 | /* | |
461 | * Initialize a lease, use the default lock manager operations | |
462 | */ | |
463 | static int lease_init(struct file *filp, int type, struct file_lock *fl) | |
464 | { | |
75dff55a TM |
465 | if (assign_type(fl, type) != 0) |
466 | return -EINVAL; | |
467 | ||
1da177e4 LT |
468 | fl->fl_owner = current->files; |
469 | fl->fl_pid = current->tgid; | |
470 | ||
471 | fl->fl_file = filp; | |
472 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
473 | fl->fl_start = 0; |
474 | fl->fl_end = OFFSET_MAX; | |
475 | fl->fl_ops = NULL; | |
476 | fl->fl_lmops = &lease_manager_ops; | |
477 | return 0; | |
478 | } | |
479 | ||
480 | /* Allocate a file_lock initialised to this type of lease */ | |
e32b8ee2 | 481 | static struct file_lock *lease_alloc(struct file *filp, int type) |
1da177e4 LT |
482 | { |
483 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 484 | int error = -ENOMEM; |
1da177e4 LT |
485 | |
486 | if (fl == NULL) | |
e32b8ee2 | 487 | return ERR_PTR(error); |
1da177e4 LT |
488 | |
489 | error = lease_init(filp, type, fl); | |
75dff55a TM |
490 | if (error) { |
491 | locks_free_lock(fl); | |
e32b8ee2 | 492 | return ERR_PTR(error); |
75dff55a | 493 | } |
e32b8ee2 | 494 | return fl; |
1da177e4 LT |
495 | } |
496 | ||
497 | /* Check if two locks overlap each other. | |
498 | */ | |
499 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
500 | { | |
501 | return ((fl1->fl_end >= fl2->fl_start) && | |
502 | (fl2->fl_end >= fl1->fl_start)); | |
503 | } | |
504 | ||
505 | /* | |
506 | * Check whether two locks have the same owner. | |
507 | */ | |
33443c42 | 508 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 LT |
509 | { |
510 | if (fl1->fl_lmops && fl1->fl_lmops->fl_compare_owner) | |
511 | return fl2->fl_lmops == fl1->fl_lmops && | |
512 | fl1->fl_lmops->fl_compare_owner(fl1, fl2); | |
513 | return fl1->fl_owner == fl2->fl_owner; | |
514 | } | |
515 | ||
516 | /* Remove waiter from blocker's block list. | |
517 | * When blocker ends up pointing to itself then the list is empty. | |
518 | */ | |
33443c42 | 519 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 LT |
520 | { |
521 | list_del_init(&waiter->fl_block); | |
522 | list_del_init(&waiter->fl_link); | |
523 | waiter->fl_next = NULL; | |
524 | } | |
525 | ||
526 | /* | |
527 | */ | |
528 | static void locks_delete_block(struct file_lock *waiter) | |
529 | { | |
b89f4321 | 530 | lock_flocks(); |
1da177e4 | 531 | __locks_delete_block(waiter); |
b89f4321 | 532 | unlock_flocks(); |
1da177e4 LT |
533 | } |
534 | ||
535 | /* Insert waiter into blocker's block list. | |
536 | * We use a circular list so that processes can be easily woken up in | |
537 | * the order they blocked. The documentation doesn't require this but | |
538 | * it seems like the reasonable thing to do. | |
539 | */ | |
540 | static void locks_insert_block(struct file_lock *blocker, | |
541 | struct file_lock *waiter) | |
542 | { | |
6dc0fe8f | 543 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 LT |
544 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
545 | waiter->fl_next = blocker; | |
546 | if (IS_POSIX(blocker)) | |
547 | list_add(&waiter->fl_link, &blocked_list); | |
548 | } | |
549 | ||
550 | /* Wake up processes blocked waiting for blocker. | |
551 | * If told to wait then schedule the processes until the block list | |
552 | * is empty, otherwise empty the block list ourselves. | |
553 | */ | |
554 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
555 | { | |
556 | while (!list_empty(&blocker->fl_block)) { | |
f0c1cd0e PE |
557 | struct file_lock *waiter; |
558 | ||
559 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
560 | struct file_lock, fl_block); |
561 | __locks_delete_block(waiter); | |
562 | if (waiter->fl_lmops && waiter->fl_lmops->fl_notify) | |
563 | waiter->fl_lmops->fl_notify(waiter); | |
564 | else | |
565 | wake_up(&waiter->fl_wait); | |
566 | } | |
567 | } | |
568 | ||
569 | /* Insert file lock fl into an inode's lock list at the position indicated | |
570 | * by pos. At the same time add the lock to the global file lock list. | |
571 | */ | |
572 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
573 | { | |
574 | list_add(&fl->fl_link, &file_lock_list); | |
575 | ||
ab1f1611 VG |
576 | fl->fl_nspid = get_pid(task_tgid(current)); |
577 | ||
1da177e4 LT |
578 | /* insert into file's list */ |
579 | fl->fl_next = *pos; | |
580 | *pos = fl; | |
1da177e4 LT |
581 | } |
582 | ||
583 | /* | |
584 | * Delete a lock and then free it. | |
585 | * Wake up processes that are blocked waiting for this lock, | |
586 | * notify the FS that the lock has been cleared and | |
587 | * finally free the lock. | |
588 | */ | |
589 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
590 | { | |
591 | struct file_lock *fl = *thisfl_p; | |
592 | ||
593 | *thisfl_p = fl->fl_next; | |
594 | fl->fl_next = NULL; | |
595 | list_del_init(&fl->fl_link); | |
596 | ||
597 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); | |
598 | if (fl->fl_fasync != NULL) { | |
599 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
600 | fl->fl_fasync = NULL; | |
601 | } | |
602 | ||
ab1f1611 VG |
603 | if (fl->fl_nspid) { |
604 | put_pid(fl->fl_nspid); | |
605 | fl->fl_nspid = NULL; | |
606 | } | |
607 | ||
1da177e4 LT |
608 | locks_wake_up_blocks(fl); |
609 | locks_free_lock(fl); | |
610 | } | |
611 | ||
612 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
613 | * checks for shared/exclusive status of overlapping locks. | |
614 | */ | |
615 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
616 | { | |
617 | if (sys_fl->fl_type == F_WRLCK) | |
618 | return 1; | |
619 | if (caller_fl->fl_type == F_WRLCK) | |
620 | return 1; | |
621 | return 0; | |
622 | } | |
623 | ||
624 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
625 | * checking before calling the locks_conflict(). | |
626 | */ | |
627 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
628 | { | |
629 | /* POSIX locks owned by the same process do not conflict with | |
630 | * each other. | |
631 | */ | |
632 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
633 | return (0); | |
634 | ||
635 | /* Check whether they overlap */ | |
636 | if (!locks_overlap(caller_fl, sys_fl)) | |
637 | return 0; | |
638 | ||
639 | return (locks_conflict(caller_fl, sys_fl)); | |
640 | } | |
641 | ||
642 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
643 | * checking before calling the locks_conflict(). | |
644 | */ | |
645 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
646 | { | |
647 | /* FLOCK locks referring to the same filp do not conflict with | |
648 | * each other. | |
649 | */ | |
650 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
651 | return (0); | |
652 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
653 | return 0; | |
654 | ||
655 | return (locks_conflict(caller_fl, sys_fl)); | |
656 | } | |
657 | ||
6d34ac19 | 658 | void |
9d6a8c5c | 659 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
660 | { |
661 | struct file_lock *cfl; | |
662 | ||
b89f4321 | 663 | lock_flocks(); |
0f7fc9e4 | 664 | for (cfl = filp->f_path.dentry->d_inode->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
665 | if (!IS_POSIX(cfl)) |
666 | continue; | |
b842e240 | 667 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
668 | break; |
669 | } | |
ab1f1611 | 670 | if (cfl) { |
9d6a8c5c | 671 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 672 | if (cfl->fl_nspid) |
6c5f3e7b | 673 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 674 | } else |
129a84de | 675 | fl->fl_type = F_UNLCK; |
b89f4321 | 676 | unlock_flocks(); |
6d34ac19 | 677 | return; |
1da177e4 | 678 | } |
1da177e4 LT |
679 | EXPORT_SYMBOL(posix_test_lock); |
680 | ||
b533184f BF |
681 | /* |
682 | * Deadlock detection: | |
683 | * | |
684 | * We attempt to detect deadlocks that are due purely to posix file | |
685 | * locks. | |
1da177e4 | 686 | * |
b533184f BF |
687 | * We assume that a task can be waiting for at most one lock at a time. |
688 | * So for any acquired lock, the process holding that lock may be | |
689 | * waiting on at most one other lock. That lock in turns may be held by | |
690 | * someone waiting for at most one other lock. Given a requested lock | |
691 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
692 | * follow this chain of waiters to ensure we are not about to create a | |
693 | * cycle. | |
1da177e4 | 694 | * |
b533184f BF |
695 | * Since we do this before we ever put a process to sleep on a lock, we |
696 | * are ensured that there is never a cycle; that is what guarantees that | |
697 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 698 | * |
b533184f BF |
699 | * Note: the above assumption may not be true when handling lock |
700 | * requests from a broken NFS client. It may also fail in the presence | |
701 | * of tasks (such as posix threads) sharing the same open file table. | |
702 | * | |
703 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 704 | */ |
97855b49 BF |
705 | |
706 | #define MAX_DEADLK_ITERATIONS 10 | |
707 | ||
b533184f BF |
708 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
709 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
710 | { | |
711 | struct file_lock *fl; | |
712 | ||
713 | list_for_each_entry(fl, &blocked_list, fl_link) { | |
714 | if (posix_same_owner(fl, block_fl)) | |
715 | return fl->fl_next; | |
716 | } | |
717 | return NULL; | |
718 | } | |
719 | ||
b0904e14 | 720 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
721 | struct file_lock *block_fl) |
722 | { | |
97855b49 | 723 | int i = 0; |
1da177e4 | 724 | |
b533184f BF |
725 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
726 | if (i++ > MAX_DEADLK_ITERATIONS) | |
727 | return 0; | |
728 | if (posix_same_owner(caller_fl, block_fl)) | |
729 | return 1; | |
1da177e4 LT |
730 | } |
731 | return 0; | |
732 | } | |
733 | ||
1da177e4 | 734 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 735 | * after any leases, but before any posix locks. |
f475ae95 TM |
736 | * |
737 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
738 | * whether or not a lock was successfully freed by testing the return | |
739 | * value for -ENOENT. | |
1da177e4 | 740 | */ |
993dfa87 | 741 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 742 | { |
993dfa87 | 743 | struct file_lock *new_fl = NULL; |
1da177e4 | 744 | struct file_lock **before; |
0f7fc9e4 | 745 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
746 | int error = 0; |
747 | int found = 0; | |
748 | ||
b89f4321 | 749 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 750 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
751 | if (!new_fl) |
752 | return -ENOMEM; | |
84d535ad PE |
753 | } |
754 | ||
b89f4321 AB |
755 | lock_flocks(); |
756 | if (request->fl_flags & FL_ACCESS) | |
757 | goto find_conflict; | |
758 | ||
1da177e4 LT |
759 | for_each_lock(inode, before) { |
760 | struct file_lock *fl = *before; | |
761 | if (IS_POSIX(fl)) | |
762 | break; | |
763 | if (IS_LEASE(fl)) | |
764 | continue; | |
765 | if (filp != fl->fl_file) | |
766 | continue; | |
993dfa87 | 767 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
768 | goto out; |
769 | found = 1; | |
770 | locks_delete_lock(before); | |
771 | break; | |
772 | } | |
1da177e4 | 773 | |
f475ae95 TM |
774 | if (request->fl_type == F_UNLCK) { |
775 | if ((request->fl_flags & FL_EXISTS) && !found) | |
776 | error = -ENOENT; | |
993dfa87 | 777 | goto out; |
f475ae95 | 778 | } |
1da177e4 LT |
779 | |
780 | /* | |
781 | * If a higher-priority process was blocked on the old file lock, | |
782 | * give it the opportunity to lock the file. | |
783 | */ | |
b89f4321 AB |
784 | if (found) { |
785 | unlock_flocks(); | |
def01bc5 | 786 | cond_resched(); |
b89f4321 AB |
787 | lock_flocks(); |
788 | } | |
1da177e4 | 789 | |
f07f18dd | 790 | find_conflict: |
1da177e4 LT |
791 | for_each_lock(inode, before) { |
792 | struct file_lock *fl = *before; | |
793 | if (IS_POSIX(fl)) | |
794 | break; | |
795 | if (IS_LEASE(fl)) | |
796 | continue; | |
993dfa87 | 797 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
798 | continue; |
799 | error = -EAGAIN; | |
bde74e4b MS |
800 | if (!(request->fl_flags & FL_SLEEP)) |
801 | goto out; | |
802 | error = FILE_LOCK_DEFERRED; | |
803 | locks_insert_block(fl, request); | |
1da177e4 LT |
804 | goto out; |
805 | } | |
f07f18dd TM |
806 | if (request->fl_flags & FL_ACCESS) |
807 | goto out; | |
993dfa87 | 808 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 809 | locks_insert_lock(before, new_fl); |
993dfa87 | 810 | new_fl = NULL; |
9cedc194 | 811 | error = 0; |
1da177e4 LT |
812 | |
813 | out: | |
b89f4321 | 814 | unlock_flocks(); |
993dfa87 TM |
815 | if (new_fl) |
816 | locks_free_lock(new_fl); | |
1da177e4 LT |
817 | return error; |
818 | } | |
819 | ||
150b3934 | 820 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
821 | { |
822 | struct file_lock *fl; | |
39005d02 MS |
823 | struct file_lock *new_fl = NULL; |
824 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
825 | struct file_lock *left = NULL; |
826 | struct file_lock *right = NULL; | |
827 | struct file_lock **before; | |
828 | int error, added = 0; | |
829 | ||
830 | /* | |
831 | * We may need two file_lock structures for this operation, | |
832 | * so we get them in advance to avoid races. | |
39005d02 MS |
833 | * |
834 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 835 | */ |
39005d02 MS |
836 | if (!(request->fl_flags & FL_ACCESS) && |
837 | (request->fl_type != F_UNLCK || | |
838 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
839 | new_fl = locks_alloc_lock(); | |
840 | new_fl2 = locks_alloc_lock(); | |
841 | } | |
1da177e4 | 842 | |
b89f4321 | 843 | lock_flocks(); |
1da177e4 LT |
844 | if (request->fl_type != F_UNLCK) { |
845 | for_each_lock(inode, before) { | |
526985b9 | 846 | fl = *before; |
1da177e4 LT |
847 | if (!IS_POSIX(fl)) |
848 | continue; | |
849 | if (!posix_locks_conflict(request, fl)) | |
850 | continue; | |
5842add2 | 851 | if (conflock) |
1a747ee0 | 852 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
853 | error = -EAGAIN; |
854 | if (!(request->fl_flags & FL_SLEEP)) | |
855 | goto out; | |
856 | error = -EDEADLK; | |
857 | if (posix_locks_deadlock(request, fl)) | |
858 | goto out; | |
bde74e4b | 859 | error = FILE_LOCK_DEFERRED; |
1da177e4 LT |
860 | locks_insert_block(fl, request); |
861 | goto out; | |
862 | } | |
863 | } | |
864 | ||
865 | /* If we're just looking for a conflict, we're done. */ | |
866 | error = 0; | |
867 | if (request->fl_flags & FL_ACCESS) | |
868 | goto out; | |
869 | ||
1da177e4 | 870 | /* |
1da177e4 LT |
871 | * Find the first old lock with the same owner as the new lock. |
872 | */ | |
873 | ||
874 | before = &inode->i_flock; | |
875 | ||
876 | /* First skip locks owned by other processes. */ | |
877 | while ((fl = *before) && (!IS_POSIX(fl) || | |
878 | !posix_same_owner(request, fl))) { | |
879 | before = &fl->fl_next; | |
880 | } | |
881 | ||
882 | /* Process locks with this owner. */ | |
883 | while ((fl = *before) && posix_same_owner(request, fl)) { | |
884 | /* Detect adjacent or overlapping regions (if same lock type) | |
885 | */ | |
886 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
887 | /* In all comparisons of start vs end, use |
888 | * "start - 1" rather than "end + 1". If end | |
889 | * is OFFSET_MAX, end + 1 will become negative. | |
890 | */ | |
1da177e4 LT |
891 | if (fl->fl_end < request->fl_start - 1) |
892 | goto next_lock; | |
893 | /* If the next lock in the list has entirely bigger | |
894 | * addresses than the new one, insert the lock here. | |
895 | */ | |
449231d6 | 896 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
897 | break; |
898 | ||
899 | /* If we come here, the new and old lock are of the | |
900 | * same type and adjacent or overlapping. Make one | |
901 | * lock yielding from the lower start address of both | |
902 | * locks to the higher end address. | |
903 | */ | |
904 | if (fl->fl_start > request->fl_start) | |
905 | fl->fl_start = request->fl_start; | |
906 | else | |
907 | request->fl_start = fl->fl_start; | |
908 | if (fl->fl_end < request->fl_end) | |
909 | fl->fl_end = request->fl_end; | |
910 | else | |
911 | request->fl_end = fl->fl_end; | |
912 | if (added) { | |
913 | locks_delete_lock(before); | |
914 | continue; | |
915 | } | |
916 | request = fl; | |
917 | added = 1; | |
918 | } | |
919 | else { | |
920 | /* Processing for different lock types is a bit | |
921 | * more complex. | |
922 | */ | |
923 | if (fl->fl_end < request->fl_start) | |
924 | goto next_lock; | |
925 | if (fl->fl_start > request->fl_end) | |
926 | break; | |
927 | if (request->fl_type == F_UNLCK) | |
928 | added = 1; | |
929 | if (fl->fl_start < request->fl_start) | |
930 | left = fl; | |
931 | /* If the next lock in the list has a higher end | |
932 | * address than the new one, insert the new one here. | |
933 | */ | |
934 | if (fl->fl_end > request->fl_end) { | |
935 | right = fl; | |
936 | break; | |
937 | } | |
938 | if (fl->fl_start >= request->fl_start) { | |
939 | /* The new lock completely replaces an old | |
940 | * one (This may happen several times). | |
941 | */ | |
942 | if (added) { | |
943 | locks_delete_lock(before); | |
944 | continue; | |
945 | } | |
946 | /* Replace the old lock with the new one. | |
947 | * Wake up anybody waiting for the old one, | |
948 | * as the change in lock type might satisfy | |
949 | * their needs. | |
950 | */ | |
951 | locks_wake_up_blocks(fl); | |
952 | fl->fl_start = request->fl_start; | |
953 | fl->fl_end = request->fl_end; | |
954 | fl->fl_type = request->fl_type; | |
47831f35 TM |
955 | locks_release_private(fl); |
956 | locks_copy_private(fl, request); | |
1da177e4 LT |
957 | request = fl; |
958 | added = 1; | |
959 | } | |
960 | } | |
961 | /* Go on to next lock. | |
962 | */ | |
963 | next_lock: | |
964 | before = &fl->fl_next; | |
965 | } | |
966 | ||
0d9a490a MS |
967 | /* |
968 | * The above code only modifies existing locks in case of | |
969 | * merging or replacing. If new lock(s) need to be inserted | |
970 | * all modifications are done bellow this, so it's safe yet to | |
971 | * bail out. | |
972 | */ | |
973 | error = -ENOLCK; /* "no luck" */ | |
974 | if (right && left == right && !new_fl2) | |
975 | goto out; | |
976 | ||
1da177e4 LT |
977 | error = 0; |
978 | if (!added) { | |
f475ae95 TM |
979 | if (request->fl_type == F_UNLCK) { |
980 | if (request->fl_flags & FL_EXISTS) | |
981 | error = -ENOENT; | |
1da177e4 | 982 | goto out; |
f475ae95 | 983 | } |
0d9a490a MS |
984 | |
985 | if (!new_fl) { | |
986 | error = -ENOLCK; | |
987 | goto out; | |
988 | } | |
1da177e4 LT |
989 | locks_copy_lock(new_fl, request); |
990 | locks_insert_lock(before, new_fl); | |
991 | new_fl = NULL; | |
992 | } | |
993 | if (right) { | |
994 | if (left == right) { | |
995 | /* The new lock breaks the old one in two pieces, | |
996 | * so we have to use the second new lock. | |
997 | */ | |
998 | left = new_fl2; | |
999 | new_fl2 = NULL; | |
1000 | locks_copy_lock(left, right); | |
1001 | locks_insert_lock(before, left); | |
1002 | } | |
1003 | right->fl_start = request->fl_end + 1; | |
1004 | locks_wake_up_blocks(right); | |
1005 | } | |
1006 | if (left) { | |
1007 | left->fl_end = request->fl_start - 1; | |
1008 | locks_wake_up_blocks(left); | |
1009 | } | |
1010 | out: | |
b89f4321 | 1011 | unlock_flocks(); |
1da177e4 LT |
1012 | /* |
1013 | * Free any unused locks. | |
1014 | */ | |
1015 | if (new_fl) | |
1016 | locks_free_lock(new_fl); | |
1017 | if (new_fl2) | |
1018 | locks_free_lock(new_fl2); | |
1019 | return error; | |
1020 | } | |
1021 | ||
1022 | /** | |
1023 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1024 | * @filp: The file to apply the lock to | |
1025 | * @fl: The lock to be applied | |
150b3934 | 1026 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1027 | * |
1028 | * Add a POSIX style lock to a file. | |
1029 | * We merge adjacent & overlapping locks whenever possible. | |
1030 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1031 | * |
1032 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1033 | * whether or not a lock was successfully freed by testing the return | |
1034 | * value for -ENOENT. | |
1da177e4 | 1035 | */ |
150b3934 | 1036 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1037 | struct file_lock *conflock) |
1038 | { | |
150b3934 | 1039 | return __posix_lock_file(filp->f_path.dentry->d_inode, fl, conflock); |
1da177e4 | 1040 | } |
150b3934 | 1041 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1042 | |
1043 | /** | |
1044 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1045 | * @filp: The file to apply the lock to | |
1046 | * @fl: The lock to be applied | |
1047 | * | |
1048 | * Add a POSIX style lock to a file. | |
1049 | * We merge adjacent & overlapping locks whenever possible. | |
1050 | * POSIX locks are sorted by owner task, then by starting address | |
1051 | */ | |
1052 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1053 | { | |
1054 | int error; | |
1055 | might_sleep (); | |
1056 | for (;;) { | |
150b3934 | 1057 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1058 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1059 | break; |
1060 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1061 | if (!error) | |
1062 | continue; | |
1063 | ||
1064 | locks_delete_block(fl); | |
1065 | break; | |
1066 | } | |
1067 | return error; | |
1068 | } | |
1069 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1070 | ||
1071 | /** | |
1072 | * locks_mandatory_locked - Check for an active lock | |
1073 | * @inode: the file to check | |
1074 | * | |
1075 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1076 | * This function is called from locks_verify_locked() only. | |
1077 | */ | |
1078 | int locks_mandatory_locked(struct inode *inode) | |
1079 | { | |
1080 | fl_owner_t owner = current->files; | |
1081 | struct file_lock *fl; | |
1082 | ||
1083 | /* | |
1084 | * Search the lock list for this inode for any POSIX locks. | |
1085 | */ | |
b89f4321 | 1086 | lock_flocks(); |
1da177e4 LT |
1087 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1088 | if (!IS_POSIX(fl)) | |
1089 | continue; | |
1090 | if (fl->fl_owner != owner) | |
1091 | break; | |
1092 | } | |
b89f4321 | 1093 | unlock_flocks(); |
1da177e4 LT |
1094 | return fl ? -EAGAIN : 0; |
1095 | } | |
1096 | ||
1097 | /** | |
1098 | * locks_mandatory_area - Check for a conflicting lock | |
1099 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1100 | * for shared | |
1101 | * @inode: the file to check | |
1102 | * @filp: how the file was opened (if it was) | |
1103 | * @offset: start of area to check | |
1104 | * @count: length of area to check | |
1105 | * | |
1106 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1107 | * This function is called from rw_verify_area() and | |
1108 | * locks_verify_truncate(). | |
1109 | */ | |
1110 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1111 | struct file *filp, loff_t offset, | |
1112 | size_t count) | |
1113 | { | |
1114 | struct file_lock fl; | |
1115 | int error; | |
1116 | ||
1117 | locks_init_lock(&fl); | |
1118 | fl.fl_owner = current->files; | |
1119 | fl.fl_pid = current->tgid; | |
1120 | fl.fl_file = filp; | |
1121 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1122 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1123 | fl.fl_flags |= FL_SLEEP; | |
1124 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1125 | fl.fl_start = offset; | |
1126 | fl.fl_end = offset + count - 1; | |
1127 | ||
1128 | for (;;) { | |
150b3934 | 1129 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1130 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1131 | break; |
1132 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1133 | if (!error) { | |
1134 | /* | |
1135 | * If we've been sleeping someone might have | |
1136 | * changed the permissions behind our back. | |
1137 | */ | |
a16877ca | 1138 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1139 | continue; |
1140 | } | |
1141 | ||
1142 | locks_delete_block(&fl); | |
1143 | break; | |
1144 | } | |
1145 | ||
1146 | return error; | |
1147 | } | |
1148 | ||
1149 | EXPORT_SYMBOL(locks_mandatory_area); | |
1150 | ||
1151 | /* We already had a lease on this file; just change its type */ | |
1152 | int lease_modify(struct file_lock **before, int arg) | |
1153 | { | |
1154 | struct file_lock *fl = *before; | |
1155 | int error = assign_type(fl, arg); | |
1156 | ||
1157 | if (error) | |
1158 | return error; | |
1159 | locks_wake_up_blocks(fl); | |
1160 | if (arg == F_UNLCK) | |
1161 | locks_delete_lock(before); | |
1162 | return 0; | |
1163 | } | |
1164 | ||
1165 | EXPORT_SYMBOL(lease_modify); | |
1166 | ||
1167 | static void time_out_leases(struct inode *inode) | |
1168 | { | |
1169 | struct file_lock **before; | |
1170 | struct file_lock *fl; | |
1171 | ||
1172 | before = &inode->i_flock; | |
1173 | while ((fl = *before) && IS_LEASE(fl) && (fl->fl_type & F_INPROGRESS)) { | |
1174 | if ((fl->fl_break_time == 0) | |
1175 | || time_before(jiffies, fl->fl_break_time)) { | |
1176 | before = &fl->fl_next; | |
1177 | continue; | |
1178 | } | |
1da177e4 LT |
1179 | lease_modify(before, fl->fl_type & ~F_INPROGRESS); |
1180 | if (fl == *before) /* lease_modify may have freed fl */ | |
1181 | before = &fl->fl_next; | |
1182 | } | |
1183 | } | |
1184 | ||
1185 | /** | |
1186 | * __break_lease - revoke all outstanding leases on file | |
1187 | * @inode: the inode of the file to return | |
1188 | * @mode: the open mode (read or write) | |
1189 | * | |
87250dd2 | 1190 | * break_lease (inlined for speed) has checked there already is at least |
1191 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1192 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1193 | * specified %O_NONBLOCK to your open(). |
1194 | */ | |
1195 | int __break_lease(struct inode *inode, unsigned int mode) | |
1196 | { | |
1197 | int error = 0, future; | |
1198 | struct file_lock *new_fl, *flock; | |
1199 | struct file_lock *fl; | |
1da177e4 LT |
1200 | unsigned long break_time; |
1201 | int i_have_this_lease = 0; | |
8737c930 | 1202 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1203 | |
8737c930 | 1204 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
1da177e4 | 1205 | |
b89f4321 | 1206 | lock_flocks(); |
1da177e4 LT |
1207 | |
1208 | time_out_leases(inode); | |
1209 | ||
1210 | flock = inode->i_flock; | |
1211 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1212 | goto out; | |
1213 | ||
1214 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) | |
1215 | if (fl->fl_owner == current->files) | |
1216 | i_have_this_lease = 1; | |
1217 | ||
8737c930 | 1218 | if (want_write) { |
1da177e4 LT |
1219 | /* If we want write access, we have to revoke any lease. */ |
1220 | future = F_UNLCK | F_INPROGRESS; | |
1221 | } else if (flock->fl_type & F_INPROGRESS) { | |
1222 | /* If the lease is already being broken, we just leave it */ | |
1223 | future = flock->fl_type; | |
1224 | } else if (flock->fl_type & F_WRLCK) { | |
1225 | /* Downgrade the exclusive lease to a read-only lease. */ | |
1226 | future = F_RDLCK | F_INPROGRESS; | |
1227 | } else { | |
1228 | /* the existing lease was read-only, so we can read too. */ | |
1229 | goto out; | |
1230 | } | |
1231 | ||
e32b8ee2 BF |
1232 | if (IS_ERR(new_fl) && !i_have_this_lease |
1233 | && ((mode & O_NONBLOCK) == 0)) { | |
1234 | error = PTR_ERR(new_fl); | |
1da177e4 LT |
1235 | goto out; |
1236 | } | |
1237 | ||
1238 | break_time = 0; | |
1239 | if (lease_break_time > 0) { | |
1240 | break_time = jiffies + lease_break_time * HZ; | |
1241 | if (break_time == 0) | |
1242 | break_time++; /* so that 0 means no break time */ | |
1243 | } | |
1244 | ||
1245 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1246 | if (fl->fl_type != future) { | |
1247 | fl->fl_type = future; | |
1248 | fl->fl_break_time = break_time; | |
1249 | /* lease must have lmops break callback */ | |
1250 | fl->fl_lmops->fl_break(fl); | |
1251 | } | |
1252 | } | |
1253 | ||
1254 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1255 | error = -EWOULDBLOCK; | |
1256 | goto out; | |
1257 | } | |
1258 | ||
1259 | restart: | |
1260 | break_time = flock->fl_break_time; | |
1261 | if (break_time != 0) { | |
1262 | break_time -= jiffies; | |
1263 | if (break_time == 0) | |
1264 | break_time++; | |
1265 | } | |
4321e01e | 1266 | locks_insert_block(flock, new_fl); |
b89f4321 | 1267 | unlock_flocks(); |
4321e01e MW |
1268 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1269 | !new_fl->fl_next, break_time); | |
b89f4321 | 1270 | lock_flocks(); |
4321e01e | 1271 | __locks_delete_block(new_fl); |
1da177e4 LT |
1272 | if (error >= 0) { |
1273 | if (error == 0) | |
1274 | time_out_leases(inode); | |
1275 | /* Wait for the next lease that has not been broken yet */ | |
1276 | for (flock = inode->i_flock; flock && IS_LEASE(flock); | |
1277 | flock = flock->fl_next) { | |
1278 | if (flock->fl_type & F_INPROGRESS) | |
1279 | goto restart; | |
1280 | } | |
1281 | error = 0; | |
1282 | } | |
1283 | ||
1284 | out: | |
b89f4321 | 1285 | unlock_flocks(); |
e32b8ee2 | 1286 | if (!IS_ERR(new_fl)) |
1da177e4 LT |
1287 | locks_free_lock(new_fl); |
1288 | return error; | |
1289 | } | |
1290 | ||
1291 | EXPORT_SYMBOL(__break_lease); | |
1292 | ||
1293 | /** | |
a6b91919 | 1294 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1295 | * @inode: the inode |
1296 | * @time: pointer to a timespec which will contain the last modified time | |
1297 | * | |
1298 | * This is to force NFS clients to flush their caches for files with | |
1299 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1300 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1301 | */ |
1302 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1303 | { | |
1304 | struct file_lock *flock = inode->i_flock; | |
1305 | if (flock && IS_LEASE(flock) && (flock->fl_type & F_WRLCK)) | |
1306 | *time = current_fs_time(inode->i_sb); | |
1307 | else | |
1308 | *time = inode->i_mtime; | |
1309 | } | |
1310 | ||
1311 | EXPORT_SYMBOL(lease_get_mtime); | |
1312 | ||
1313 | /** | |
1314 | * fcntl_getlease - Enquire what lease is currently active | |
1315 | * @filp: the file | |
1316 | * | |
1317 | * The value returned by this function will be one of | |
1318 | * (if no lease break is pending): | |
1319 | * | |
1320 | * %F_RDLCK to indicate a shared lease is held. | |
1321 | * | |
1322 | * %F_WRLCK to indicate an exclusive lease is held. | |
1323 | * | |
1324 | * %F_UNLCK to indicate no lease is held. | |
1325 | * | |
1326 | * (if a lease break is pending): | |
1327 | * | |
1328 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1329 | * changed to a shared lease (or removed). | |
1330 | * | |
1331 | * %F_UNLCK to indicate the lease needs to be removed. | |
1332 | * | |
1333 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1334 | * should be returned to userspace. | |
1335 | */ | |
1336 | int fcntl_getlease(struct file *filp) | |
1337 | { | |
1338 | struct file_lock *fl; | |
1339 | int type = F_UNLCK; | |
1340 | ||
b89f4321 | 1341 | lock_flocks(); |
0f7fc9e4 JJS |
1342 | time_out_leases(filp->f_path.dentry->d_inode); |
1343 | for (fl = filp->f_path.dentry->d_inode->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1344 | fl = fl->fl_next) { |
1345 | if (fl->fl_file == filp) { | |
1346 | type = fl->fl_type & ~F_INPROGRESS; | |
1347 | break; | |
1348 | } | |
1349 | } | |
b89f4321 | 1350 | unlock_flocks(); |
1da177e4 LT |
1351 | return type; |
1352 | } | |
1353 | ||
1354 | /** | |
0af1a450 | 1355 | * generic_setlease - sets a lease on an open file |
1da177e4 LT |
1356 | * @filp: file pointer |
1357 | * @arg: type of lease to obtain | |
1358 | * @flp: input - file_lock to use, output - file_lock inserted | |
1359 | * | |
1360 | * The (input) flp->fl_lmops->fl_break function is required | |
1361 | * by break_lease(). | |
1362 | * | |
b89f4321 | 1363 | * Called with file_lock_lock held. |
1da177e4 | 1364 | */ |
0af1a450 | 1365 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1366 | { |
7eaae282 | 1367 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1368 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 LT |
1369 | struct inode *inode = dentry->d_inode; |
1370 | int error, rdlease_count = 0, wrlease_count = 0; | |
1371 | ||
096657b6 BF |
1372 | lease = *flp; |
1373 | ||
1374 | error = -EACCES; | |
da9592ed | 1375 | if ((current_fsuid() != inode->i_uid) && !capable(CAP_LEASE)) |
096657b6 BF |
1376 | goto out; |
1377 | error = -EINVAL; | |
6d5e8b05 | 1378 | if (!S_ISREG(inode->i_mode)) |
096657b6 | 1379 | goto out; |
6d5e8b05 BF |
1380 | error = security_file_lock(filp, arg); |
1381 | if (error) | |
096657b6 | 1382 | goto out; |
6d5e8b05 | 1383 | |
1da177e4 LT |
1384 | time_out_leases(inode); |
1385 | ||
d2ab0b0c | 1386 | BUG_ON(!(*flp)->fl_lmops->fl_break); |
1da177e4 | 1387 | |
288b2fd8 DR |
1388 | if (arg != F_UNLCK) { |
1389 | error = -EAGAIN; | |
1390 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1391 | goto out; | |
1392 | if ((arg == F_WRLCK) | |
1393 | && ((atomic_read(&dentry->d_count) > 1) | |
1394 | || (atomic_read(&inode->i_count) > 1))) | |
1395 | goto out; | |
288b2fd8 | 1396 | } |
85c59580 | 1397 | |
1da177e4 LT |
1398 | /* |
1399 | * At this point, we know that if there is an exclusive | |
1400 | * lease on this file, then we hold it on this filp | |
1401 | * (otherwise our open of this file would have blocked). | |
1402 | * And if we are trying to acquire an exclusive lease, | |
1403 | * then the file is not open by anyone (including us) | |
1404 | * except for this filp. | |
1405 | */ | |
1406 | for (before = &inode->i_flock; | |
1407 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1408 | before = &fl->fl_next) { | |
1409 | if (lease->fl_lmops->fl_mylease(fl, lease)) | |
1410 | my_before = before; | |
1411 | else if (fl->fl_type == (F_INPROGRESS | F_UNLCK)) | |
1412 | /* | |
1413 | * Someone is in the process of opening this | |
1414 | * file for writing so we may not take an | |
1415 | * exclusive lease on it. | |
1416 | */ | |
1417 | wrlease_count++; | |
1418 | else | |
1419 | rdlease_count++; | |
1420 | } | |
1421 | ||
5fcc60c3 | 1422 | error = -EAGAIN; |
1da177e4 LT |
1423 | if ((arg == F_RDLCK && (wrlease_count > 0)) || |
1424 | (arg == F_WRLCK && ((rdlease_count + wrlease_count) > 0))) | |
1425 | goto out; | |
1426 | ||
1427 | if (my_before != NULL) { | |
1428 | error = lease->fl_lmops->fl_change(my_before, arg); | |
51ee4b84 CH |
1429 | if (!error) |
1430 | *flp = *my_before; | |
1da177e4 LT |
1431 | goto out; |
1432 | } | |
1433 | ||
1da177e4 LT |
1434 | if (arg == F_UNLCK) |
1435 | goto out; | |
1436 | ||
1437 | error = -EINVAL; | |
1438 | if (!leases_enable) | |
1439 | goto out; | |
1440 | ||
c5b1f0d9 | 1441 | locks_insert_lock(before, lease); |
85c59580 | 1442 | return 0; |
1da177e4 | 1443 | |
1da177e4 LT |
1444 | out: |
1445 | return error; | |
1446 | } | |
0af1a450 | 1447 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1448 | |
b89f4321 AB |
1449 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1450 | { | |
1451 | if (filp->f_op && filp->f_op->setlease) | |
1452 | return filp->f_op->setlease(filp, arg, lease); | |
1453 | else | |
1454 | return generic_setlease(filp, arg, lease); | |
1455 | } | |
1456 | ||
1457 | /** | |
a9933cea | 1458 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1459 | * @filp: file pointer |
1460 | * @arg: type of lease to obtain | |
1461 | * @lease: file_lock to use | |
1462 | * | |
1463 | * Call this to establish a lease on the file. | |
f9ffed26 BF |
1464 | * The (*lease)->fl_lmops->fl_break operation must be set; if not, |
1465 | * break_lease will oops! | |
1466 | * | |
1467 | * This will call the filesystem's setlease file method, if | |
1468 | * defined. Note that there is no getlease method; instead, the | |
1469 | * filesystem setlease method should call back to setlease() to | |
1470 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1471 | * find it. Since fcntl_getlease() only reports whether the current | |
1472 | * task holds a lease, a cluster filesystem need only do this for | |
1473 | * leases held by processes on this node. | |
1474 | * | |
1475 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1476 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1477 | * filesystem's open, create, and (on truncate) setattr methods. |
1478 | * | |
1479 | * Warning: the only current setlease methods exist only to disable | |
1480 | * leases in certain cases. More vfs changes may be required to | |
1481 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1482 | */ |
1483 | ||
a9933cea | 1484 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1485 | { |
1da177e4 LT |
1486 | int error; |
1487 | ||
b89f4321 AB |
1488 | lock_flocks(); |
1489 | error = __vfs_setlease(filp, arg, lease); | |
1490 | unlock_flocks(); | |
1da177e4 LT |
1491 | |
1492 | return error; | |
1493 | } | |
a9933cea | 1494 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1495 | |
0ceaf6c7 BF |
1496 | static int do_fcntl_delete_lease(struct file *filp) |
1497 | { | |
1498 | struct file_lock fl, *flp = &fl; | |
1499 | ||
1500 | lease_init(filp, F_UNLCK, flp); | |
1501 | ||
1502 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1503 | } | |
1504 | ||
1505 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1506 | { |
c5b1f0d9 | 1507 | struct file_lock *fl; |
f7347ce4 | 1508 | struct fasync_struct *new; |
9d91cdcc | 1509 | struct inode *inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1510 | int error; |
1511 | ||
c5b1f0d9 AB |
1512 | fl = lease_alloc(filp, arg); |
1513 | if (IS_ERR(fl)) | |
1514 | return PTR_ERR(fl); | |
1da177e4 | 1515 | |
f7347ce4 LT |
1516 | new = fasync_alloc(); |
1517 | if (!new) { | |
1518 | locks_free_lock(fl); | |
1519 | return -ENOMEM; | |
1520 | } | |
b89f4321 | 1521 | lock_flocks(); |
c5b1f0d9 | 1522 | error = __vfs_setlease(filp, arg, &fl); |
51ee4b84 CH |
1523 | if (error) { |
1524 | unlock_flocks(); | |
1525 | locks_free_lock(fl); | |
1526 | goto out_free_fasync; | |
1527 | } | |
1da177e4 | 1528 | |
f7347ce4 LT |
1529 | /* |
1530 | * fasync_insert_entry() returns the old entry if any. | |
1531 | * If there was no old entry, then it used 'new' and | |
1532 | * inserted it into the fasync list. Clear new so that | |
1533 | * we don't release it here. | |
1534 | */ | |
1535 | if (!fasync_insert_entry(fd, filp, &fl->fl_fasync, new)) | |
1536 | new = NULL; | |
1537 | ||
1da177e4 | 1538 | if (error < 0) { |
6d5e8b05 | 1539 | /* remove lease just inserted by setlease */ |
c5b1f0d9 AB |
1540 | fl->fl_type = F_UNLCK | F_INPROGRESS; |
1541 | fl->fl_break_time = jiffies - 10; | |
1da177e4 | 1542 | time_out_leases(inode); |
51ee4b84 CH |
1543 | } else { |
1544 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
1da177e4 | 1545 | } |
b89f4321 | 1546 | unlock_flocks(); |
51ee4b84 CH |
1547 | |
1548 | out_free_fasync: | |
f7347ce4 LT |
1549 | if (new) |
1550 | fasync_free(new); | |
1da177e4 LT |
1551 | return error; |
1552 | } | |
1553 | ||
0ceaf6c7 BF |
1554 | /** |
1555 | * fcntl_setlease - sets a lease on an open file | |
1556 | * @fd: open file descriptor | |
1557 | * @filp: file pointer | |
1558 | * @arg: type of lease to obtain | |
1559 | * | |
1560 | * Call this fcntl to establish a lease on the file. | |
1561 | * Note that you also need to call %F_SETSIG to | |
1562 | * receive a signal when the lease is broken. | |
1563 | */ | |
1564 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1565 | { | |
1566 | if (arg == F_UNLCK) | |
1567 | return do_fcntl_delete_lease(filp); | |
1568 | return do_fcntl_add_lease(fd, filp, arg); | |
1569 | } | |
1570 | ||
1da177e4 LT |
1571 | /** |
1572 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1573 | * @filp: The file to apply the lock to | |
1574 | * @fl: The lock to be applied | |
1575 | * | |
1576 | * Add a FLOCK style lock to a file. | |
1577 | */ | |
1578 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1579 | { | |
1580 | int error; | |
1581 | might_sleep(); | |
1582 | for (;;) { | |
1583 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1584 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1585 | break; |
1586 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1587 | if (!error) | |
1588 | continue; | |
1589 | ||
1590 | locks_delete_block(fl); | |
1591 | break; | |
1592 | } | |
1593 | return error; | |
1594 | } | |
1595 | ||
1596 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1597 | ||
1598 | /** | |
1599 | * sys_flock: - flock() system call. | |
1600 | * @fd: the file descriptor to lock. | |
1601 | * @cmd: the type of lock to apply. | |
1602 | * | |
1603 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1604 | * The @cmd can be one of | |
1605 | * | |
1606 | * %LOCK_SH -- a shared lock. | |
1607 | * | |
1608 | * %LOCK_EX -- an exclusive lock. | |
1609 | * | |
1610 | * %LOCK_UN -- remove an existing lock. | |
1611 | * | |
1612 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1613 | * | |
1614 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1615 | * processes read and write access respectively. | |
1616 | */ | |
002c8976 | 1617 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 LT |
1618 | { |
1619 | struct file *filp; | |
1620 | struct file_lock *lock; | |
1621 | int can_sleep, unlock; | |
1622 | int error; | |
1623 | ||
1624 | error = -EBADF; | |
1625 | filp = fget(fd); | |
1626 | if (!filp) | |
1627 | goto out; | |
1628 | ||
1629 | can_sleep = !(cmd & LOCK_NB); | |
1630 | cmd &= ~LOCK_NB; | |
1631 | unlock = (cmd == LOCK_UN); | |
1632 | ||
aeb5d727 AV |
1633 | if (!unlock && !(cmd & LOCK_MAND) && |
1634 | !(filp->f_mode & (FMODE_READ|FMODE_WRITE))) | |
1da177e4 LT |
1635 | goto out_putf; |
1636 | ||
1637 | error = flock_make_lock(filp, &lock, cmd); | |
1638 | if (error) | |
1639 | goto out_putf; | |
1640 | if (can_sleep) | |
1641 | lock->fl_flags |= FL_SLEEP; | |
1642 | ||
713c0ecd | 1643 | error = security_file_lock(filp, lock->fl_type); |
1da177e4 LT |
1644 | if (error) |
1645 | goto out_free; | |
1646 | ||
1647 | if (filp->f_op && filp->f_op->flock) | |
1648 | error = filp->f_op->flock(filp, | |
1649 | (can_sleep) ? F_SETLKW : F_SETLK, | |
1650 | lock); | |
1651 | else | |
1652 | error = flock_lock_file_wait(filp, lock); | |
1653 | ||
1654 | out_free: | |
993dfa87 | 1655 | locks_free_lock(lock); |
1da177e4 LT |
1656 | |
1657 | out_putf: | |
1658 | fput(filp); | |
1659 | out: | |
1660 | return error; | |
1661 | } | |
1662 | ||
3ee17abd BF |
1663 | /** |
1664 | * vfs_test_lock - test file byte range lock | |
1665 | * @filp: The file to test lock for | |
6924c554 | 1666 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1667 | * |
1668 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1669 | * setting conf->fl_type to something other than F_UNLCK. | |
1670 | */ | |
1671 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1672 | { | |
1673 | if (filp->f_op && filp->f_op->lock) | |
1674 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1675 | posix_test_lock(filp, fl); | |
1676 | return 0; | |
1677 | } | |
1678 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1679 | ||
c2fa1b8a BF |
1680 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1681 | { | |
1682 | flock->l_pid = fl->fl_pid; | |
1683 | #if BITS_PER_LONG == 32 | |
1684 | /* | |
1685 | * Make sure we can represent the posix lock via | |
1686 | * legacy 32bit flock. | |
1687 | */ | |
1688 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1689 | return -EOVERFLOW; | |
1690 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1691 | return -EOVERFLOW; | |
1692 | #endif | |
1693 | flock->l_start = fl->fl_start; | |
1694 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1695 | fl->fl_end - fl->fl_start + 1; | |
1696 | flock->l_whence = 0; | |
129a84de | 1697 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1698 | return 0; |
1699 | } | |
1700 | ||
1701 | #if BITS_PER_LONG == 32 | |
1702 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1703 | { | |
1704 | flock->l_pid = fl->fl_pid; | |
1705 | flock->l_start = fl->fl_start; | |
1706 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1707 | fl->fl_end - fl->fl_start + 1; | |
1708 | flock->l_whence = 0; | |
1709 | flock->l_type = fl->fl_type; | |
1710 | } | |
1711 | #endif | |
1712 | ||
1da177e4 LT |
1713 | /* Report the first existing lock that would conflict with l. |
1714 | * This implements the F_GETLK command of fcntl(). | |
1715 | */ | |
1716 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1717 | { | |
9d6a8c5c | 1718 | struct file_lock file_lock; |
1da177e4 LT |
1719 | struct flock flock; |
1720 | int error; | |
1721 | ||
1722 | error = -EFAULT; | |
1723 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1724 | goto out; | |
1725 | error = -EINVAL; | |
1726 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1727 | goto out; | |
1728 | ||
1729 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1730 | if (error) | |
1731 | goto out; | |
1732 | ||
3ee17abd BF |
1733 | error = vfs_test_lock(filp, &file_lock); |
1734 | if (error) | |
1735 | goto out; | |
1da177e4 | 1736 | |
9d6a8c5c ME |
1737 | flock.l_type = file_lock.fl_type; |
1738 | if (file_lock.fl_type != F_UNLCK) { | |
1739 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1740 | if (error) |
1da177e4 | 1741 | goto out; |
1da177e4 LT |
1742 | } |
1743 | error = -EFAULT; | |
1744 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1745 | error = 0; | |
1746 | out: | |
1747 | return error; | |
1748 | } | |
1749 | ||
7723ec97 ME |
1750 | /** |
1751 | * vfs_lock_file - file byte range lock | |
1752 | * @filp: The file to apply the lock to | |
1753 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1754 | * @fl: The lock to be applied | |
150b3934 ME |
1755 | * @conf: Place to return a copy of the conflicting lock, if found. |
1756 | * | |
1757 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1758 | * as the final argument. | |
1759 | * | |
1760 | * If the filesystem defines a private ->lock() method, then @conf will | |
1761 | * be left unchanged; so a caller that cares should initialize it to | |
1762 | * some acceptable default. | |
2beb6614 ME |
1763 | * |
1764 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1765 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1766 | * been granted or denied by the underlying filesystem, if (and only if) | |
1767 | * fl_grant is set. Callers expecting ->lock() to return asynchronously | |
1768 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) | |
1769 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
bde74e4b | 1770 | * it must return FILE_LOCK_DEFERRED, and call ->fl_grant() when the lock |
2beb6614 ME |
1771 | * request completes. |
1772 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1773 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1774 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1775 | * nonzero return code and the file system should release the lock. The file |
1776 | * system is also responsible to keep a corresponding posix lock when it | |
1777 | * grants a lock so the VFS can find out which locks are locally held and do | |
1778 | * the correct lock cleanup when required. | |
1779 | * The underlying filesystem must not drop the kernel lock or call | |
bde74e4b | 1780 | * ->fl_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1781 | * return code. |
7723ec97 | 1782 | */ |
150b3934 | 1783 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1784 | { |
1785 | if (filp->f_op && filp->f_op->lock) | |
1786 | return filp->f_op->lock(filp, cmd, fl); | |
1787 | else | |
150b3934 | 1788 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1789 | } |
1790 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1791 | ||
b648a6de MS |
1792 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1793 | struct file_lock *fl) | |
1794 | { | |
1795 | int error; | |
1796 | ||
1797 | error = security_file_lock(filp, fl->fl_type); | |
1798 | if (error) | |
1799 | return error; | |
1800 | ||
764c76b3 MS |
1801 | for (;;) { |
1802 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1803 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1804 | break; |
764c76b3 MS |
1805 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1806 | if (!error) | |
1807 | continue; | |
1808 | ||
1809 | locks_delete_block(fl); | |
1810 | break; | |
b648a6de MS |
1811 | } |
1812 | ||
1813 | return error; | |
1814 | } | |
1815 | ||
1da177e4 LT |
1816 | /* Apply the lock described by l to an open file descriptor. |
1817 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1818 | */ | |
c293621b PS |
1819 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1820 | struct flock __user *l) | |
1da177e4 LT |
1821 | { |
1822 | struct file_lock *file_lock = locks_alloc_lock(); | |
1823 | struct flock flock; | |
1824 | struct inode *inode; | |
0b2bac2f | 1825 | struct file *f; |
1da177e4 LT |
1826 | int error; |
1827 | ||
1828 | if (file_lock == NULL) | |
1829 | return -ENOLCK; | |
1830 | ||
1831 | /* | |
1832 | * This might block, so we do it before checking the inode. | |
1833 | */ | |
1834 | error = -EFAULT; | |
1835 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1836 | goto out; | |
1837 | ||
0f7fc9e4 | 1838 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1839 | |
1840 | /* Don't allow mandatory locks on files that may be memory mapped | |
1841 | * and shared. | |
1842 | */ | |
a16877ca | 1843 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1844 | error = -EAGAIN; |
1845 | goto out; | |
1846 | } | |
1847 | ||
c293621b | 1848 | again: |
1da177e4 LT |
1849 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1850 | if (error) | |
1851 | goto out; | |
1852 | if (cmd == F_SETLKW) { | |
1853 | file_lock->fl_flags |= FL_SLEEP; | |
1854 | } | |
1855 | ||
1856 | error = -EBADF; | |
1857 | switch (flock.l_type) { | |
1858 | case F_RDLCK: | |
1859 | if (!(filp->f_mode & FMODE_READ)) | |
1860 | goto out; | |
1861 | break; | |
1862 | case F_WRLCK: | |
1863 | if (!(filp->f_mode & FMODE_WRITE)) | |
1864 | goto out; | |
1865 | break; | |
1866 | case F_UNLCK: | |
1867 | break; | |
1868 | default: | |
1869 | error = -EINVAL; | |
1870 | goto out; | |
1871 | } | |
1872 | ||
b648a6de | 1873 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1874 | |
c293621b PS |
1875 | /* |
1876 | * Attempt to detect a close/fcntl race and recover by | |
1877 | * releasing the lock that was just acquired. | |
1878 | */ | |
0b2bac2f AV |
1879 | /* |
1880 | * we need that spin_lock here - it prevents reordering between | |
1881 | * update of inode->i_flock and check for it done in close(). | |
1882 | * rcu_read_lock() wouldn't do. | |
1883 | */ | |
1884 | spin_lock(¤t->files->file_lock); | |
1885 | f = fcheck(fd); | |
1886 | spin_unlock(¤t->files->file_lock); | |
1887 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1888 | flock.l_type = F_UNLCK; |
1889 | goto again; | |
1da177e4 LT |
1890 | } |
1891 | ||
c293621b | 1892 | out: |
1da177e4 LT |
1893 | locks_free_lock(file_lock); |
1894 | return error; | |
1895 | } | |
1896 | ||
1897 | #if BITS_PER_LONG == 32 | |
1898 | /* Report the first existing lock that would conflict with l. | |
1899 | * This implements the F_GETLK command of fcntl(). | |
1900 | */ | |
1901 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
1902 | { | |
9d6a8c5c | 1903 | struct file_lock file_lock; |
1da177e4 LT |
1904 | struct flock64 flock; |
1905 | int error; | |
1906 | ||
1907 | error = -EFAULT; | |
1908 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1909 | goto out; | |
1910 | error = -EINVAL; | |
1911 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1912 | goto out; | |
1913 | ||
1914 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
1915 | if (error) | |
1916 | goto out; | |
1917 | ||
3ee17abd BF |
1918 | error = vfs_test_lock(filp, &file_lock); |
1919 | if (error) | |
1920 | goto out; | |
1921 | ||
9d6a8c5c ME |
1922 | flock.l_type = file_lock.fl_type; |
1923 | if (file_lock.fl_type != F_UNLCK) | |
1924 | posix_lock_to_flock64(&flock, &file_lock); | |
1925 | ||
1da177e4 LT |
1926 | error = -EFAULT; |
1927 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1928 | error = 0; | |
1929 | ||
1930 | out: | |
1931 | return error; | |
1932 | } | |
1933 | ||
1934 | /* Apply the lock described by l to an open file descriptor. | |
1935 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1936 | */ | |
c293621b PS |
1937 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
1938 | struct flock64 __user *l) | |
1da177e4 LT |
1939 | { |
1940 | struct file_lock *file_lock = locks_alloc_lock(); | |
1941 | struct flock64 flock; | |
1942 | struct inode *inode; | |
0b2bac2f | 1943 | struct file *f; |
1da177e4 LT |
1944 | int error; |
1945 | ||
1946 | if (file_lock == NULL) | |
1947 | return -ENOLCK; | |
1948 | ||
1949 | /* | |
1950 | * This might block, so we do it before checking the inode. | |
1951 | */ | |
1952 | error = -EFAULT; | |
1953 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1954 | goto out; | |
1955 | ||
0f7fc9e4 | 1956 | inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
1957 | |
1958 | /* Don't allow mandatory locks on files that may be memory mapped | |
1959 | * and shared. | |
1960 | */ | |
a16877ca | 1961 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1962 | error = -EAGAIN; |
1963 | goto out; | |
1964 | } | |
1965 | ||
c293621b | 1966 | again: |
1da177e4 LT |
1967 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
1968 | if (error) | |
1969 | goto out; | |
1970 | if (cmd == F_SETLKW64) { | |
1971 | file_lock->fl_flags |= FL_SLEEP; | |
1972 | } | |
1973 | ||
1974 | error = -EBADF; | |
1975 | switch (flock.l_type) { | |
1976 | case F_RDLCK: | |
1977 | if (!(filp->f_mode & FMODE_READ)) | |
1978 | goto out; | |
1979 | break; | |
1980 | case F_WRLCK: | |
1981 | if (!(filp->f_mode & FMODE_WRITE)) | |
1982 | goto out; | |
1983 | break; | |
1984 | case F_UNLCK: | |
1985 | break; | |
1986 | default: | |
1987 | error = -EINVAL; | |
1988 | goto out; | |
1989 | } | |
1990 | ||
b648a6de | 1991 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1992 | |
c293621b PS |
1993 | /* |
1994 | * Attempt to detect a close/fcntl race and recover by | |
1995 | * releasing the lock that was just acquired. | |
1996 | */ | |
0b2bac2f AV |
1997 | spin_lock(¤t->files->file_lock); |
1998 | f = fcheck(fd); | |
1999 | spin_unlock(¤t->files->file_lock); | |
2000 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2001 | flock.l_type = F_UNLCK; |
2002 | goto again; | |
1da177e4 LT |
2003 | } |
2004 | ||
2005 | out: | |
2006 | locks_free_lock(file_lock); | |
2007 | return error; | |
2008 | } | |
2009 | #endif /* BITS_PER_LONG == 32 */ | |
2010 | ||
2011 | /* | |
2012 | * This function is called when the file is being removed | |
2013 | * from the task's fd array. POSIX locks belonging to this task | |
2014 | * are deleted at this time. | |
2015 | */ | |
2016 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2017 | { | |
ff7b86b8 | 2018 | struct file_lock lock; |
1da177e4 LT |
2019 | |
2020 | /* | |
2021 | * If there are no locks held on this file, we don't need to call | |
2022 | * posix_lock_file(). Another process could be setting a lock on this | |
2023 | * file at the same time, but we wouldn't remove that lock anyway. | |
2024 | */ | |
0f7fc9e4 | 2025 | if (!filp->f_path.dentry->d_inode->i_flock) |
1da177e4 LT |
2026 | return; |
2027 | ||
2028 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2029 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2030 | lock.fl_start = 0; |
2031 | lock.fl_end = OFFSET_MAX; | |
2032 | lock.fl_owner = owner; | |
2033 | lock.fl_pid = current->tgid; | |
2034 | lock.fl_file = filp; | |
2035 | lock.fl_ops = NULL; | |
2036 | lock.fl_lmops = NULL; | |
2037 | ||
150b3934 | 2038 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2039 | |
1da177e4 LT |
2040 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2041 | lock.fl_ops->fl_release_private(&lock); | |
2042 | } | |
2043 | ||
2044 | EXPORT_SYMBOL(locks_remove_posix); | |
2045 | ||
2046 | /* | |
2047 | * This function is called on the last close of an open file. | |
2048 | */ | |
2049 | void locks_remove_flock(struct file *filp) | |
2050 | { | |
0f7fc9e4 | 2051 | struct inode * inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
2052 | struct file_lock *fl; |
2053 | struct file_lock **before; | |
2054 | ||
2055 | if (!inode->i_flock) | |
2056 | return; | |
2057 | ||
2058 | if (filp->f_op && filp->f_op->flock) { | |
2059 | struct file_lock fl = { | |
2060 | .fl_pid = current->tgid, | |
2061 | .fl_file = filp, | |
2062 | .fl_flags = FL_FLOCK, | |
2063 | .fl_type = F_UNLCK, | |
2064 | .fl_end = OFFSET_MAX, | |
2065 | }; | |
2066 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2067 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2068 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2069 | } |
2070 | ||
b89f4321 | 2071 | lock_flocks(); |
1da177e4 LT |
2072 | before = &inode->i_flock; |
2073 | ||
2074 | while ((fl = *before) != NULL) { | |
2075 | if (fl->fl_file == filp) { | |
c293621b | 2076 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2077 | locks_delete_lock(before); |
2078 | continue; | |
2079 | } | |
2080 | if (IS_LEASE(fl)) { | |
2081 | lease_modify(before, F_UNLCK); | |
2082 | continue; | |
2083 | } | |
2084 | /* What? */ | |
2085 | BUG(); | |
2086 | } | |
2087 | before = &fl->fl_next; | |
2088 | } | |
b89f4321 | 2089 | unlock_flocks(); |
1da177e4 LT |
2090 | } |
2091 | ||
1da177e4 LT |
2092 | /** |
2093 | * posix_unblock_lock - stop waiting for a file lock | |
2094 | * @filp: how the file was opened | |
2095 | * @waiter: the lock which was waiting | |
2096 | * | |
2097 | * lockd needs to block waiting for locks. | |
2098 | */ | |
64a318ee | 2099 | int |
1da177e4 LT |
2100 | posix_unblock_lock(struct file *filp, struct file_lock *waiter) |
2101 | { | |
64a318ee BF |
2102 | int status = 0; |
2103 | ||
b89f4321 | 2104 | lock_flocks(); |
5996a298 | 2105 | if (waiter->fl_next) |
1da177e4 | 2106 | __locks_delete_block(waiter); |
64a318ee BF |
2107 | else |
2108 | status = -ENOENT; | |
b89f4321 | 2109 | unlock_flocks(); |
64a318ee | 2110 | return status; |
1da177e4 LT |
2111 | } |
2112 | ||
2113 | EXPORT_SYMBOL(posix_unblock_lock); | |
2114 | ||
9b9d2ab4 ME |
2115 | /** |
2116 | * vfs_cancel_lock - file byte range unblock lock | |
2117 | * @filp: The file to apply the unblock to | |
2118 | * @fl: The lock to be unblocked | |
2119 | * | |
2120 | * Used by lock managers to cancel blocked requests | |
2121 | */ | |
2122 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2123 | { | |
2124 | if (filp->f_op && filp->f_op->lock) | |
2125 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2126 | return 0; | |
2127 | } | |
2128 | ||
2129 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2130 | ||
7f8ada98 | 2131 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2132 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2133 | #include <linux/seq_file.h> |
2134 | ||
2135 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
99dc8292 | 2136 | loff_t id, char *pfx) |
1da177e4 LT |
2137 | { |
2138 | struct inode *inode = NULL; | |
ab1f1611 VG |
2139 | unsigned int fl_pid; |
2140 | ||
2141 | if (fl->fl_nspid) | |
6c5f3e7b | 2142 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2143 | else |
2144 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2145 | |
2146 | if (fl->fl_file != NULL) | |
0f7fc9e4 | 2147 | inode = fl->fl_file->f_path.dentry->d_inode; |
1da177e4 | 2148 | |
99dc8292 | 2149 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2150 | if (IS_POSIX(fl)) { |
7f8ada98 | 2151 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2152 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2153 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2154 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2155 | } else if (IS_FLOCK(fl)) { |
2156 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2157 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2158 | } else { |
7f8ada98 | 2159 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2160 | } |
2161 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2162 | seq_printf(f, "LEASE "); |
1da177e4 | 2163 | if (fl->fl_type & F_INPROGRESS) |
7f8ada98 | 2164 | seq_printf(f, "BREAKING "); |
1da177e4 | 2165 | else if (fl->fl_file) |
7f8ada98 | 2166 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2167 | else |
7f8ada98 | 2168 | seq_printf(f, "BREAKER "); |
1da177e4 | 2169 | } else { |
7f8ada98 | 2170 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2171 | } |
2172 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2173 | seq_printf(f, "%s ", |
1da177e4 LT |
2174 | (fl->fl_type & LOCK_READ) |
2175 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2176 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2177 | } else { | |
7f8ada98 | 2178 | seq_printf(f, "%s ", |
1da177e4 LT |
2179 | (fl->fl_type & F_INPROGRESS) |
2180 | ? (fl->fl_type & F_UNLCK) ? "UNLCK" : "READ " | |
2181 | : (fl->fl_type & F_WRLCK) ? "WRITE" : "READ "); | |
2182 | } | |
2183 | if (inode) { | |
2184 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2185 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2186 | inode->i_sb->s_id, inode->i_ino); |
2187 | #else | |
2188 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2189 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2190 | MAJOR(inode->i_sb->s_dev), |
2191 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2192 | #endif | |
2193 | } else { | |
ab1f1611 | 2194 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2195 | } |
2196 | if (IS_POSIX(fl)) { | |
2197 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2198 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2199 | else |
7f8ada98 | 2200 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2201 | } else { |
7f8ada98 | 2202 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2203 | } |
2204 | } | |
2205 | ||
7f8ada98 | 2206 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2207 | { |
7f8ada98 | 2208 | struct file_lock *fl, *bfl; |
1da177e4 | 2209 | |
7f8ada98 | 2210 | fl = list_entry(v, struct file_lock, fl_link); |
1da177e4 | 2211 | |
99dc8292 | 2212 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
1da177e4 | 2213 | |
7f8ada98 | 2214 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
99dc8292 | 2215 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
094f2825 | 2216 | |
7f8ada98 PE |
2217 | return 0; |
2218 | } | |
1da177e4 | 2219 | |
7f8ada98 PE |
2220 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2221 | { | |
99dc8292 JM |
2222 | loff_t *p = f->private; |
2223 | ||
b89f4321 | 2224 | lock_flocks(); |
99dc8292 | 2225 | *p = (*pos + 1); |
7f8ada98 PE |
2226 | return seq_list_start(&file_lock_list, *pos); |
2227 | } | |
1da177e4 | 2228 | |
7f8ada98 PE |
2229 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2230 | { | |
99dc8292 JM |
2231 | loff_t *p = f->private; |
2232 | ++*p; | |
7f8ada98 PE |
2233 | return seq_list_next(v, &file_lock_list, pos); |
2234 | } | |
1da177e4 | 2235 | |
7f8ada98 PE |
2236 | static void locks_stop(struct seq_file *f, void *v) |
2237 | { | |
b89f4321 | 2238 | unlock_flocks(); |
1da177e4 LT |
2239 | } |
2240 | ||
d8ba7a36 | 2241 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2242 | .start = locks_start, |
2243 | .next = locks_next, | |
2244 | .stop = locks_stop, | |
2245 | .show = locks_show, | |
2246 | }; | |
d8ba7a36 AD |
2247 | |
2248 | static int locks_open(struct inode *inode, struct file *filp) | |
2249 | { | |
99dc8292 | 2250 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
d8ba7a36 AD |
2251 | } |
2252 | ||
2253 | static const struct file_operations proc_locks_operations = { | |
2254 | .open = locks_open, | |
2255 | .read = seq_read, | |
2256 | .llseek = seq_lseek, | |
99dc8292 | 2257 | .release = seq_release_private, |
d8ba7a36 AD |
2258 | }; |
2259 | ||
2260 | static int __init proc_locks_init(void) | |
2261 | { | |
2262 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2263 | return 0; | |
2264 | } | |
2265 | module_init(proc_locks_init); | |
7f8ada98 PE |
2266 | #endif |
2267 | ||
1da177e4 LT |
2268 | /** |
2269 | * lock_may_read - checks that the region is free of locks | |
2270 | * @inode: the inode that is being read | |
2271 | * @start: the first byte to read | |
2272 | * @len: the number of bytes to read | |
2273 | * | |
2274 | * Emulates Windows locking requirements. Whole-file | |
2275 | * mandatory locks (share modes) can prohibit a read and | |
2276 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2277 | * | |
2278 | * N.B. this function is only ever called | |
2279 | * from knfsd and ownership of locks is never checked. | |
2280 | */ | |
2281 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2282 | { | |
2283 | struct file_lock *fl; | |
2284 | int result = 1; | |
b89f4321 | 2285 | lock_flocks(); |
1da177e4 LT |
2286 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2287 | if (IS_POSIX(fl)) { | |
2288 | if (fl->fl_type == F_RDLCK) | |
2289 | continue; | |
2290 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2291 | continue; | |
2292 | } else if (IS_FLOCK(fl)) { | |
2293 | if (!(fl->fl_type & LOCK_MAND)) | |
2294 | continue; | |
2295 | if (fl->fl_type & LOCK_READ) | |
2296 | continue; | |
2297 | } else | |
2298 | continue; | |
2299 | result = 0; | |
2300 | break; | |
2301 | } | |
b89f4321 | 2302 | unlock_flocks(); |
1da177e4 LT |
2303 | return result; |
2304 | } | |
2305 | ||
2306 | EXPORT_SYMBOL(lock_may_read); | |
2307 | ||
2308 | /** | |
2309 | * lock_may_write - checks that the region is free of locks | |
2310 | * @inode: the inode that is being written | |
2311 | * @start: the first byte to write | |
2312 | * @len: the number of bytes to write | |
2313 | * | |
2314 | * Emulates Windows locking requirements. Whole-file | |
2315 | * mandatory locks (share modes) can prohibit a write and | |
2316 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2317 | * | |
2318 | * N.B. this function is only ever called | |
2319 | * from knfsd and ownership of locks is never checked. | |
2320 | */ | |
2321 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2322 | { | |
2323 | struct file_lock *fl; | |
2324 | int result = 1; | |
b89f4321 | 2325 | lock_flocks(); |
1da177e4 LT |
2326 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2327 | if (IS_POSIX(fl)) { | |
2328 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2329 | continue; | |
2330 | } else if (IS_FLOCK(fl)) { | |
2331 | if (!(fl->fl_type & LOCK_MAND)) | |
2332 | continue; | |
2333 | if (fl->fl_type & LOCK_WRITE) | |
2334 | continue; | |
2335 | } else | |
2336 | continue; | |
2337 | result = 0; | |
2338 | break; | |
2339 | } | |
b89f4321 | 2340 | unlock_flocks(); |
1da177e4 LT |
2341 | return result; |
2342 | } | |
2343 | ||
2344 | EXPORT_SYMBOL(lock_may_write); | |
2345 | ||
1da177e4 LT |
2346 | static int __init filelock_init(void) |
2347 | { | |
2348 | filelock_cache = kmem_cache_create("file_lock_cache", | |
2349 | sizeof(struct file_lock), 0, SLAB_PANIC, | |
20c2df83 | 2350 | init_once); |
1da177e4 LT |
2351 | return 0; |
2352 | } | |
2353 | ||
2354 | core_initcall(filelock_init); |