Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/locks.c | |
3 | * | |
4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
6 | * | |
7 | * Deadlock detection added. | |
8 | * FIXME: one thing isn't handled yet: | |
9 | * - mandatory locks (requires lots of changes elsewhere) | |
10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
11 | * | |
12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
14 | * | |
15 | * Converted file_lock_table to a linked list from an array, which eliminates | |
16 | * the limits on how many active file locks are open. | |
17 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
18 | * | |
19 | * Removed dependency on file descriptors. dup()'ed file descriptors now | |
20 | * get the same locks as the original file descriptors, and a close() on | |
21 | * any file descriptor removes ALL the locks on the file for the current | |
22 | * process. Since locks still depend on the process id, locks are inherited | |
23 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
24 | * BSD and SVR4 practice. | |
25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
26 | * | |
27 | * Scrapped free list which is redundant now that we allocate locks | |
28 | * dynamically with kmalloc()/kfree(). | |
29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
30 | * | |
31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
32 | * | |
33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
34 | * fcntl() system call. They have the semantics described above. | |
35 | * | |
36 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
37 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
38 | * and will continue to use the old, broken implementation. | |
39 | * | |
40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
41 | * with a file pointer (filp). As a result they can be shared by a parent | |
42 | * process and its children after a fork(). They are removed when the last | |
43 | * file descriptor referring to the file pointer is closed (unless explicitly | |
44 | * unlocked). | |
45 | * | |
46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
47 | * upgrading from shared to exclusive (or vice versa). When this happens | |
48 | * any processes blocked by the current lock are woken up and allowed to | |
49 | * run before the new lock is applied. | |
50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
51 | * | |
52 | * Removed some race conditions in flock_lock_file(), marked other possible | |
53 | * races. Just grep for FIXME to see them. | |
54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. | |
55 | * | |
56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
58 | * once we've checked for blocking and deadlocking. | |
59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
60 | * | |
61 | * Initial implementation of mandatory locks. SunOS turned out to be | |
62 | * a rotten model, so I implemented the "obvious" semantics. | |
395cf969 | 63 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * | |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
69 | * Manual, Section 2. | |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
71 | * | |
72 | * Tidied up block list handling. Added '/proc/locks' interface. | |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
74 | * | |
75 | * Fixed deadlock condition for pathological code that mixes calls to | |
76 | * flock() and fcntl(). | |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
78 | * | |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
81 | * guarantee sensible behaviour in the case where file system modules might | |
82 | * be compiled with different options than the kernel itself. | |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
84 | * | |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
88 | * | |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
91 | * have already been freed. | |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
93 | * | |
94 | * Made the block list a circular list to minimise searching in the list. | |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
96 | * | |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
98 | * locking. | |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
100 | * | |
101 | * Some adaptations for NFS support. | |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
103 | * | |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
106 | * | |
107 | * Use slab allocator instead of kmalloc/kfree. | |
108 | * Use generic list implementation from <linux/list.h>. | |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
111 | * | |
112 | * Leases and LOCK_MAND | |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
115 | */ | |
116 | ||
117 | #include <linux/capability.h> | |
118 | #include <linux/file.h> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
120 | #include <linux/fs.h> |
121 | #include <linux/init.h> | |
122 | #include <linux/module.h> | |
123 | #include <linux/security.h> | |
124 | #include <linux/slab.h> | |
1da177e4 LT |
125 | #include <linux/syscalls.h> |
126 | #include <linux/time.h> | |
4fb3a538 | 127 | #include <linux/rcupdate.h> |
ab1f1611 | 128 | #include <linux/pid_namespace.h> |
48f74186 | 129 | #include <linux/hashtable.h> |
7012b02a JL |
130 | #include <linux/percpu.h> |
131 | #include <linux/lglock.h> | |
1da177e4 | 132 | |
62af4f1f JL |
133 | #define CREATE_TRACE_POINTS |
134 | #include <trace/events/filelock.h> | |
135 | ||
1da177e4 LT |
136 | #include <asm/uaccess.h> |
137 | ||
138 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
139 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
617588d5 | 140 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG)) |
cff2fce5 | 141 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
1da177e4 | 142 | |
ab83fa4b BF |
143 | static bool lease_breaking(struct file_lock *fl) |
144 | { | |
778fc546 BF |
145 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
146 | } | |
147 | ||
148 | static int target_leasetype(struct file_lock *fl) | |
149 | { | |
150 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
151 | return F_UNLCK; | |
152 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
153 | return F_RDLCK; | |
154 | return fl->fl_type; | |
ab83fa4b BF |
155 | } |
156 | ||
1da177e4 LT |
157 | int leases_enable = 1; |
158 | int lease_break_time = 45; | |
159 | ||
160 | #define for_each_lock(inode, lockp) \ | |
161 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
162 | ||
1c8c601a | 163 | /* |
7012b02a JL |
164 | * The global file_lock_list is only used for displaying /proc/locks, so we |
165 | * keep a list on each CPU, with each list protected by its own spinlock via | |
166 | * the file_lock_lglock. Note that alterations to the list also require that | |
167 | * the relevant i_lock is held. | |
1c8c601a | 168 | */ |
7012b02a JL |
169 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
170 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
88974691 | 171 | |
1c8c601a | 172 | /* |
48f74186 | 173 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 174 | * It is protected by blocked_lock_lock. |
48f74186 JL |
175 | * |
176 | * We hash locks by lockowner in order to optimize searching for the lock a | |
177 | * particular lockowner is waiting on. | |
178 | * | |
179 | * FIXME: make this value scale via some heuristic? We generally will want more | |
180 | * buckets when we have more lockowners holding locks, but that's a little | |
181 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 182 | */ |
48f74186 JL |
183 | #define BLOCKED_HASH_BITS 7 |
184 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 185 | |
1c8c601a | 186 | /* |
7b2296af JL |
187 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
188 | * want to be holding this lock. | |
1c8c601a JL |
189 | * |
190 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
191 | * pointer for file_lock structures that are acting as lock requests (in | |
192 | * contrast to those that are acting as records of acquired locks). | |
193 | * | |
194 | * Note that when we acquire this lock in order to change the above fields, | |
195 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
196 | * protected by this lock, we can skip acquiring it iff we already hold the | |
197 | * i_lock. | |
198 | * | |
199 | * In particular, adding an entry to the fl_block list requires that you hold | |
200 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
201 | * an entry from the list however only requires the file_lock_lock. | |
202 | */ | |
7b2296af | 203 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 204 | |
e18b890b | 205 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 206 | |
ee19cc40 | 207 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 208 | { |
139ca04e | 209 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 210 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
211 | INIT_LIST_HEAD(&fl->fl_block); |
212 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
213 | } |
214 | ||
1da177e4 | 215 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 216 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 217 | { |
ee19cc40 | 218 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
219 | |
220 | if (fl) | |
ee19cc40 | 221 | locks_init_lock_heads(fl); |
a51cb91d MS |
222 | |
223 | return fl; | |
1da177e4 | 224 | } |
c5b1f0d9 | 225 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 226 | |
a9e61e25 | 227 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
228 | { |
229 | if (fl->fl_ops) { | |
230 | if (fl->fl_ops->fl_release_private) | |
231 | fl->fl_ops->fl_release_private(fl); | |
232 | fl->fl_ops = NULL; | |
233 | } | |
47831f35 | 234 | |
5c97d7b1 KM |
235 | if (fl->fl_lmops) { |
236 | if (fl->fl_lmops->lm_put_owner) | |
237 | fl->fl_lmops->lm_put_owner(fl); | |
238 | fl->fl_lmops = NULL; | |
239 | } | |
47831f35 | 240 | } |
a9e61e25 | 241 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 242 | |
1da177e4 | 243 | /* Free a lock which is not in use. */ |
05fa3135 | 244 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 245 | { |
5ce29646 | 246 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 247 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 248 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 249 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 250 | |
47831f35 | 251 | locks_release_private(fl); |
1da177e4 LT |
252 | kmem_cache_free(filelock_cache, fl); |
253 | } | |
05fa3135 | 254 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 255 | |
ed9814d8 JL |
256 | static void |
257 | locks_dispose_list(struct list_head *dispose) | |
258 | { | |
259 | struct file_lock *fl; | |
260 | ||
261 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
262 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
263 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
264 | locks_free_lock(fl); |
265 | } | |
266 | } | |
267 | ||
1da177e4 LT |
268 | void locks_init_lock(struct file_lock *fl) |
269 | { | |
ee19cc40 MS |
270 | memset(fl, 0, sizeof(struct file_lock)); |
271 | locks_init_lock_heads(fl); | |
1da177e4 LT |
272 | } |
273 | ||
274 | EXPORT_SYMBOL(locks_init_lock); | |
275 | ||
1da177e4 LT |
276 | /* |
277 | * Initialize a new lock from an existing file_lock structure. | |
278 | */ | |
3fe0fff1 | 279 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
280 | { |
281 | new->fl_owner = fl->fl_owner; | |
282 | new->fl_pid = fl->fl_pid; | |
0996905f | 283 | new->fl_file = NULL; |
1da177e4 LT |
284 | new->fl_flags = fl->fl_flags; |
285 | new->fl_type = fl->fl_type; | |
286 | new->fl_start = fl->fl_start; | |
287 | new->fl_end = fl->fl_end; | |
f328296e | 288 | new->fl_lmops = fl->fl_lmops; |
0996905f | 289 | new->fl_ops = NULL; |
f328296e KM |
290 | |
291 | if (fl->fl_lmops) { | |
292 | if (fl->fl_lmops->lm_get_owner) | |
293 | fl->fl_lmops->lm_get_owner(new, fl); | |
294 | } | |
0996905f | 295 | } |
3fe0fff1 | 296 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
297 | |
298 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
299 | { | |
566709bd JL |
300 | /* "new" must be a freshly-initialized lock */ |
301 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 302 | |
3fe0fff1 | 303 | locks_copy_conflock(new, fl); |
f328296e | 304 | |
0996905f | 305 | new->fl_file = fl->fl_file; |
1da177e4 | 306 | new->fl_ops = fl->fl_ops; |
47831f35 | 307 | |
f328296e KM |
308 | if (fl->fl_ops) { |
309 | if (fl->fl_ops->fl_copy_lock) | |
310 | fl->fl_ops->fl_copy_lock(new, fl); | |
311 | } | |
1da177e4 LT |
312 | } |
313 | ||
314 | EXPORT_SYMBOL(locks_copy_lock); | |
315 | ||
316 | static inline int flock_translate_cmd(int cmd) { | |
317 | if (cmd & LOCK_MAND) | |
318 | return cmd & (LOCK_MAND | LOCK_RW); | |
319 | switch (cmd) { | |
320 | case LOCK_SH: | |
321 | return F_RDLCK; | |
322 | case LOCK_EX: | |
323 | return F_WRLCK; | |
324 | case LOCK_UN: | |
325 | return F_UNLCK; | |
326 | } | |
327 | return -EINVAL; | |
328 | } | |
329 | ||
330 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
331 | static struct file_lock * |
332 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
333 | { |
334 | struct file_lock *fl; | |
335 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 336 | |
1da177e4 | 337 | if (type < 0) |
6e129d00 | 338 | return ERR_PTR(type); |
1da177e4 LT |
339 | |
340 | fl = locks_alloc_lock(); | |
341 | if (fl == NULL) | |
6e129d00 | 342 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
343 | |
344 | fl->fl_file = filp; | |
73a8f5f7 | 345 | fl->fl_owner = filp; |
1da177e4 LT |
346 | fl->fl_pid = current->tgid; |
347 | fl->fl_flags = FL_FLOCK; | |
348 | fl->fl_type = type; | |
349 | fl->fl_end = OFFSET_MAX; | |
350 | ||
6e129d00 | 351 | return fl; |
1da177e4 LT |
352 | } |
353 | ||
0ec4f431 | 354 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
355 | { |
356 | switch (type) { | |
357 | case F_RDLCK: | |
358 | case F_WRLCK: | |
359 | case F_UNLCK: | |
360 | fl->fl_type = type; | |
361 | break; | |
362 | default: | |
363 | return -EINVAL; | |
364 | } | |
365 | return 0; | |
366 | } | |
367 | ||
ef12e72a BF |
368 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
369 | struct flock64 *l) | |
1da177e4 | 370 | { |
1da177e4 | 371 | switch (l->l_whence) { |
f5579f8c | 372 | case SEEK_SET: |
ef12e72a | 373 | fl->fl_start = 0; |
1da177e4 | 374 | break; |
f5579f8c | 375 | case SEEK_CUR: |
ef12e72a | 376 | fl->fl_start = filp->f_pos; |
1da177e4 | 377 | break; |
f5579f8c | 378 | case SEEK_END: |
ef12e72a | 379 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
380 | break; |
381 | default: | |
382 | return -EINVAL; | |
383 | } | |
ef12e72a BF |
384 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
385 | return -EOVERFLOW; | |
386 | fl->fl_start += l->l_start; | |
387 | if (fl->fl_start < 0) | |
388 | return -EINVAL; | |
1da177e4 LT |
389 | |
390 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
391 | POSIX-2001 defines it. */ | |
4c780a46 | 392 | if (l->l_len > 0) { |
ef12e72a BF |
393 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
394 | return -EOVERFLOW; | |
395 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
396 | ||
4c780a46 | 397 | } else if (l->l_len < 0) { |
ef12e72a | 398 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 399 | return -EINVAL; |
ef12e72a BF |
400 | fl->fl_end = fl->fl_start - 1; |
401 | fl->fl_start += l->l_len; | |
402 | } else | |
403 | fl->fl_end = OFFSET_MAX; | |
404 | ||
1da177e4 LT |
405 | fl->fl_owner = current->files; |
406 | fl->fl_pid = current->tgid; | |
407 | fl->fl_file = filp; | |
408 | fl->fl_flags = FL_POSIX; | |
409 | fl->fl_ops = NULL; | |
410 | fl->fl_lmops = NULL; | |
411 | ||
412 | return assign_type(fl, l->l_type); | |
413 | } | |
414 | ||
ef12e72a BF |
415 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
416 | * style lock. | |
417 | */ | |
418 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
419 | struct flock *l) | |
1da177e4 | 420 | { |
ef12e72a BF |
421 | struct flock64 ll = { |
422 | .l_type = l->l_type, | |
423 | .l_whence = l->l_whence, | |
424 | .l_start = l->l_start, | |
425 | .l_len = l->l_len, | |
426 | }; | |
427 | ||
428 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 429 | } |
1da177e4 LT |
430 | |
431 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
432 | static bool |
433 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
434 | { |
435 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 436 | return false; |
1da177e4 LT |
437 | } |
438 | ||
1c7dd2ff JL |
439 | static void |
440 | lease_setup(struct file_lock *fl, void **priv) | |
441 | { | |
442 | struct file *filp = fl->fl_file; | |
443 | struct fasync_struct *fa = *priv; | |
444 | ||
445 | /* | |
446 | * fasync_insert_entry() returns the old entry if any. If there was no | |
447 | * old entry, then it used "priv" and inserted it into the fasync list. | |
448 | * Clear the pointer to indicate that it shouldn't be freed. | |
449 | */ | |
450 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
451 | *priv = NULL; | |
452 | ||
453 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
454 | } | |
455 | ||
7b021967 | 456 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 457 | .lm_break = lease_break_callback, |
8fb47a4f | 458 | .lm_change = lease_modify, |
1c7dd2ff | 459 | .lm_setup = lease_setup, |
1da177e4 LT |
460 | }; |
461 | ||
462 | /* | |
463 | * Initialize a lease, use the default lock manager operations | |
464 | */ | |
0ec4f431 | 465 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 466 | { |
75dff55a TM |
467 | if (assign_type(fl, type) != 0) |
468 | return -EINVAL; | |
469 | ||
7ca76311 | 470 | fl->fl_owner = filp; |
1da177e4 LT |
471 | fl->fl_pid = current->tgid; |
472 | ||
473 | fl->fl_file = filp; | |
474 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
475 | fl->fl_start = 0; |
476 | fl->fl_end = OFFSET_MAX; | |
477 | fl->fl_ops = NULL; | |
478 | fl->fl_lmops = &lease_manager_ops; | |
479 | return 0; | |
480 | } | |
481 | ||
482 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 483 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
484 | { |
485 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 486 | int error = -ENOMEM; |
1da177e4 LT |
487 | |
488 | if (fl == NULL) | |
e32b8ee2 | 489 | return ERR_PTR(error); |
1da177e4 LT |
490 | |
491 | error = lease_init(filp, type, fl); | |
75dff55a TM |
492 | if (error) { |
493 | locks_free_lock(fl); | |
e32b8ee2 | 494 | return ERR_PTR(error); |
75dff55a | 495 | } |
e32b8ee2 | 496 | return fl; |
1da177e4 LT |
497 | } |
498 | ||
499 | /* Check if two locks overlap each other. | |
500 | */ | |
501 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
502 | { | |
503 | return ((fl1->fl_end >= fl2->fl_start) && | |
504 | (fl2->fl_end >= fl1->fl_start)); | |
505 | } | |
506 | ||
507 | /* | |
508 | * Check whether two locks have the same owner. | |
509 | */ | |
33443c42 | 510 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 511 | { |
8fb47a4f | 512 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 513 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 514 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
515 | return fl1->fl_owner == fl2->fl_owner; |
516 | } | |
517 | ||
7012b02a | 518 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 519 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 520 | { |
7012b02a JL |
521 | lg_local_lock(&file_lock_lglock); |
522 | fl->fl_link_cpu = smp_processor_id(); | |
523 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
524 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
525 | } |
526 | ||
7012b02a | 527 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 528 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 529 | { |
7012b02a JL |
530 | /* |
531 | * Avoid taking lock if already unhashed. This is safe since this check | |
532 | * is done while holding the i_lock, and new insertions into the list | |
533 | * also require that it be held. | |
534 | */ | |
535 | if (hlist_unhashed(&fl->fl_link)) | |
536 | return; | |
537 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 538 | hlist_del_init(&fl->fl_link); |
7012b02a | 539 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
540 | } |
541 | ||
3999e493 JL |
542 | static unsigned long |
543 | posix_owner_key(struct file_lock *fl) | |
544 | { | |
545 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
546 | return fl->fl_lmops->lm_owner_key(fl); | |
547 | return (unsigned long)fl->fl_owner; | |
548 | } | |
549 | ||
6ca10ed8 | 550 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 551 | { |
3999e493 | 552 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
553 | } |
554 | ||
6ca10ed8 | 555 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 556 | { |
48f74186 | 557 | hash_del(&waiter->fl_link); |
88974691 JL |
558 | } |
559 | ||
1da177e4 LT |
560 | /* Remove waiter from blocker's block list. |
561 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 562 | * |
7b2296af | 563 | * Must be called with blocked_lock_lock held. |
1da177e4 | 564 | */ |
33443c42 | 565 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 566 | { |
88974691 | 567 | locks_delete_global_blocked(waiter); |
1da177e4 | 568 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
569 | waiter->fl_next = NULL; |
570 | } | |
571 | ||
1a9e64a7 | 572 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 573 | { |
7b2296af | 574 | spin_lock(&blocked_lock_lock); |
1da177e4 | 575 | __locks_delete_block(waiter); |
7b2296af | 576 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
577 | } |
578 | ||
579 | /* Insert waiter into blocker's block list. | |
580 | * We use a circular list so that processes can be easily woken up in | |
581 | * the order they blocked. The documentation doesn't require this but | |
582 | * it seems like the reasonable thing to do. | |
1c8c601a | 583 | * |
7b2296af | 584 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 585 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 586 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 587 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 588 | */ |
1c8c601a JL |
589 | static void __locks_insert_block(struct file_lock *blocker, |
590 | struct file_lock *waiter) | |
1da177e4 | 591 | { |
6dc0fe8f | 592 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 593 | waiter->fl_next = blocker; |
88974691 | 594 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 595 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
596 | locks_insert_global_blocked(waiter); |
597 | } | |
598 | ||
599 | /* Must be called with i_lock held. */ | |
600 | static void locks_insert_block(struct file_lock *blocker, | |
601 | struct file_lock *waiter) | |
602 | { | |
7b2296af | 603 | spin_lock(&blocked_lock_lock); |
1c8c601a | 604 | __locks_insert_block(blocker, waiter); |
7b2296af | 605 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
606 | } |
607 | ||
1cb36012 JL |
608 | /* |
609 | * Wake up processes blocked waiting for blocker. | |
610 | * | |
1c8c601a | 611 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
612 | */ |
613 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
614 | { | |
4e8c765d JL |
615 | /* |
616 | * Avoid taking global lock if list is empty. This is safe since new | |
617 | * blocked requests are only added to the list under the i_lock, and | |
618 | * the i_lock is always held here. Note that removal from the fl_block | |
619 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 620 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
621 | */ |
622 | if (list_empty(&blocker->fl_block)) | |
623 | return; | |
624 | ||
7b2296af | 625 | spin_lock(&blocked_lock_lock); |
1da177e4 | 626 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
627 | struct file_lock *waiter; |
628 | ||
629 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
630 | struct file_lock, fl_block); |
631 | __locks_delete_block(waiter); | |
8fb47a4f BF |
632 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
633 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
634 | else |
635 | wake_up(&waiter->fl_wait); | |
636 | } | |
7b2296af | 637 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
638 | } |
639 | ||
640 | /* Insert file lock fl into an inode's lock list at the position indicated | |
641 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
642 | * |
643 | * Must be called with the i_lock held! | |
1da177e4 LT |
644 | */ |
645 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
646 | { | |
ab1f1611 VG |
647 | fl->fl_nspid = get_pid(task_tgid(current)); |
648 | ||
1da177e4 LT |
649 | /* insert into file's list */ |
650 | fl->fl_next = *pos; | |
651 | *pos = fl; | |
88974691 JL |
652 | |
653 | locks_insert_global_locks(fl); | |
1da177e4 LT |
654 | } |
655 | ||
24cbe784 JL |
656 | /** |
657 | * locks_delete_lock - Delete a lock and then free it. | |
658 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
659 | * inode->i_flock list entry | |
660 | * | |
661 | * Unlink a lock from all lists and free the namespace reference, but don't | |
662 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
663 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
664 | * |
665 | * Must be called with the i_lock held! | |
1da177e4 | 666 | */ |
24cbe784 | 667 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
668 | { |
669 | struct file_lock *fl = *thisfl_p; | |
670 | ||
88974691 JL |
671 | locks_delete_global_locks(fl); |
672 | ||
1da177e4 LT |
673 | *thisfl_p = fl->fl_next; |
674 | fl->fl_next = NULL; | |
1da177e4 | 675 | |
ab1f1611 VG |
676 | if (fl->fl_nspid) { |
677 | put_pid(fl->fl_nspid); | |
678 | fl->fl_nspid = NULL; | |
679 | } | |
680 | ||
1da177e4 | 681 | locks_wake_up_blocks(fl); |
24cbe784 JL |
682 | } |
683 | ||
684 | /* | |
685 | * Unlink a lock from all lists and free it. | |
686 | * | |
687 | * Must be called with i_lock held! | |
688 | */ | |
ed9814d8 JL |
689 | static void locks_delete_lock(struct file_lock **thisfl_p, |
690 | struct list_head *dispose) | |
24cbe784 JL |
691 | { |
692 | struct file_lock *fl = *thisfl_p; | |
693 | ||
694 | locks_unlink_lock(thisfl_p); | |
ed9814d8 | 695 | if (dispose) |
6dee60f6 | 696 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
697 | else |
698 | locks_free_lock(fl); | |
1da177e4 LT |
699 | } |
700 | ||
701 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
702 | * checks for shared/exclusive status of overlapping locks. | |
703 | */ | |
704 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
705 | { | |
706 | if (sys_fl->fl_type == F_WRLCK) | |
707 | return 1; | |
708 | if (caller_fl->fl_type == F_WRLCK) | |
709 | return 1; | |
710 | return 0; | |
711 | } | |
712 | ||
713 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
714 | * checking before calling the locks_conflict(). | |
715 | */ | |
716 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
717 | { | |
718 | /* POSIX locks owned by the same process do not conflict with | |
719 | * each other. | |
720 | */ | |
721 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
722 | return (0); | |
723 | ||
724 | /* Check whether they overlap */ | |
725 | if (!locks_overlap(caller_fl, sys_fl)) | |
726 | return 0; | |
727 | ||
728 | return (locks_conflict(caller_fl, sys_fl)); | |
729 | } | |
730 | ||
731 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
732 | * checking before calling the locks_conflict(). | |
733 | */ | |
734 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
735 | { | |
736 | /* FLOCK locks referring to the same filp do not conflict with | |
737 | * each other. | |
738 | */ | |
739 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
740 | return (0); | |
741 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
742 | return 0; | |
743 | ||
744 | return (locks_conflict(caller_fl, sys_fl)); | |
745 | } | |
746 | ||
6d34ac19 | 747 | void |
9d6a8c5c | 748 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
749 | { |
750 | struct file_lock *cfl; | |
1c8c601a | 751 | struct inode *inode = file_inode(filp); |
1da177e4 | 752 | |
1c8c601a | 753 | spin_lock(&inode->i_lock); |
496ad9aa | 754 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
755 | if (!IS_POSIX(cfl)) |
756 | continue; | |
b842e240 | 757 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
758 | break; |
759 | } | |
ab1f1611 | 760 | if (cfl) { |
3fe0fff1 | 761 | locks_copy_conflock(fl, cfl); |
ab1f1611 | 762 | if (cfl->fl_nspid) |
6c5f3e7b | 763 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 764 | } else |
129a84de | 765 | fl->fl_type = F_UNLCK; |
1c8c601a | 766 | spin_unlock(&inode->i_lock); |
6d34ac19 | 767 | return; |
1da177e4 | 768 | } |
1da177e4 LT |
769 | EXPORT_SYMBOL(posix_test_lock); |
770 | ||
b533184f BF |
771 | /* |
772 | * Deadlock detection: | |
773 | * | |
774 | * We attempt to detect deadlocks that are due purely to posix file | |
775 | * locks. | |
1da177e4 | 776 | * |
b533184f BF |
777 | * We assume that a task can be waiting for at most one lock at a time. |
778 | * So for any acquired lock, the process holding that lock may be | |
779 | * waiting on at most one other lock. That lock in turns may be held by | |
780 | * someone waiting for at most one other lock. Given a requested lock | |
781 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
782 | * follow this chain of waiters to ensure we are not about to create a | |
783 | * cycle. | |
1da177e4 | 784 | * |
b533184f BF |
785 | * Since we do this before we ever put a process to sleep on a lock, we |
786 | * are ensured that there is never a cycle; that is what guarantees that | |
787 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 788 | * |
b533184f BF |
789 | * Note: the above assumption may not be true when handling lock |
790 | * requests from a broken NFS client. It may also fail in the presence | |
791 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 792 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 793 | * |
cff2fce5 | 794 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
795 | * Because the owner is not even nominally tied to a thread of |
796 | * execution, the deadlock detection below can't reasonably work well. Just | |
797 | * skip it for those. | |
798 | * | |
cff2fce5 | 799 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
800 | * locks that just checks for the case where two tasks are attempting to |
801 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 802 | */ |
97855b49 BF |
803 | |
804 | #define MAX_DEADLK_ITERATIONS 10 | |
805 | ||
b533184f BF |
806 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
807 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
808 | { | |
809 | struct file_lock *fl; | |
810 | ||
3999e493 | 811 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
812 | if (posix_same_owner(fl, block_fl)) |
813 | return fl->fl_next; | |
814 | } | |
815 | return NULL; | |
816 | } | |
817 | ||
7b2296af | 818 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 819 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
820 | struct file_lock *block_fl) |
821 | { | |
97855b49 | 822 | int i = 0; |
1da177e4 | 823 | |
57b65325 JL |
824 | /* |
825 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 826 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 827 | */ |
cff2fce5 | 828 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
829 | return 0; |
830 | ||
b533184f BF |
831 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
832 | if (i++ > MAX_DEADLK_ITERATIONS) | |
833 | return 0; | |
834 | if (posix_same_owner(caller_fl, block_fl)) | |
835 | return 1; | |
1da177e4 LT |
836 | } |
837 | return 0; | |
838 | } | |
839 | ||
1da177e4 | 840 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 841 | * after any leases, but before any posix locks. |
f475ae95 TM |
842 | * |
843 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
844 | * whether or not a lock was successfully freed by testing the return | |
845 | * value for -ENOENT. | |
1da177e4 | 846 | */ |
993dfa87 | 847 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 848 | { |
993dfa87 | 849 | struct file_lock *new_fl = NULL; |
1da177e4 | 850 | struct file_lock **before; |
496ad9aa | 851 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
852 | int error = 0; |
853 | int found = 0; | |
ed9814d8 | 854 | LIST_HEAD(dispose); |
1da177e4 | 855 | |
b89f4321 | 856 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 857 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
858 | if (!new_fl) |
859 | return -ENOMEM; | |
84d535ad PE |
860 | } |
861 | ||
1c8c601a | 862 | spin_lock(&inode->i_lock); |
b89f4321 AB |
863 | if (request->fl_flags & FL_ACCESS) |
864 | goto find_conflict; | |
865 | ||
1da177e4 LT |
866 | for_each_lock(inode, before) { |
867 | struct file_lock *fl = *before; | |
868 | if (IS_POSIX(fl)) | |
869 | break; | |
870 | if (IS_LEASE(fl)) | |
871 | continue; | |
872 | if (filp != fl->fl_file) | |
873 | continue; | |
993dfa87 | 874 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
875 | goto out; |
876 | found = 1; | |
ed9814d8 | 877 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
878 | break; |
879 | } | |
1da177e4 | 880 | |
f475ae95 TM |
881 | if (request->fl_type == F_UNLCK) { |
882 | if ((request->fl_flags & FL_EXISTS) && !found) | |
883 | error = -ENOENT; | |
993dfa87 | 884 | goto out; |
f475ae95 | 885 | } |
1da177e4 LT |
886 | |
887 | /* | |
888 | * If a higher-priority process was blocked on the old file lock, | |
889 | * give it the opportunity to lock the file. | |
890 | */ | |
b89f4321 | 891 | if (found) { |
1c8c601a | 892 | spin_unlock(&inode->i_lock); |
def01bc5 | 893 | cond_resched(); |
1c8c601a | 894 | spin_lock(&inode->i_lock); |
b89f4321 | 895 | } |
1da177e4 | 896 | |
f07f18dd | 897 | find_conflict: |
1da177e4 LT |
898 | for_each_lock(inode, before) { |
899 | struct file_lock *fl = *before; | |
900 | if (IS_POSIX(fl)) | |
901 | break; | |
902 | if (IS_LEASE(fl)) | |
903 | continue; | |
993dfa87 | 904 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
905 | continue; |
906 | error = -EAGAIN; | |
bde74e4b MS |
907 | if (!(request->fl_flags & FL_SLEEP)) |
908 | goto out; | |
909 | error = FILE_LOCK_DEFERRED; | |
910 | locks_insert_block(fl, request); | |
1da177e4 LT |
911 | goto out; |
912 | } | |
f07f18dd TM |
913 | if (request->fl_flags & FL_ACCESS) |
914 | goto out; | |
993dfa87 | 915 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 916 | locks_insert_lock(before, new_fl); |
993dfa87 | 917 | new_fl = NULL; |
9cedc194 | 918 | error = 0; |
1da177e4 LT |
919 | |
920 | out: | |
1c8c601a | 921 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
922 | if (new_fl) |
923 | locks_free_lock(new_fl); | |
ed9814d8 | 924 | locks_dispose_list(&dispose); |
1da177e4 LT |
925 | return error; |
926 | } | |
927 | ||
150b3934 | 928 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
929 | { |
930 | struct file_lock *fl; | |
39005d02 MS |
931 | struct file_lock *new_fl = NULL; |
932 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
933 | struct file_lock *left = NULL; |
934 | struct file_lock *right = NULL; | |
935 | struct file_lock **before; | |
b9746ef8 JL |
936 | int error; |
937 | bool added = false; | |
ed9814d8 | 938 | LIST_HEAD(dispose); |
1da177e4 LT |
939 | |
940 | /* | |
941 | * We may need two file_lock structures for this operation, | |
942 | * so we get them in advance to avoid races. | |
39005d02 MS |
943 | * |
944 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 945 | */ |
39005d02 MS |
946 | if (!(request->fl_flags & FL_ACCESS) && |
947 | (request->fl_type != F_UNLCK || | |
948 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
949 | new_fl = locks_alloc_lock(); | |
950 | new_fl2 = locks_alloc_lock(); | |
951 | } | |
1da177e4 | 952 | |
1c8c601a | 953 | spin_lock(&inode->i_lock); |
1cb36012 JL |
954 | /* |
955 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
956 | * there are any, either return error or put the request on the | |
48f74186 | 957 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 958 | */ |
1da177e4 LT |
959 | if (request->fl_type != F_UNLCK) { |
960 | for_each_lock(inode, before) { | |
526985b9 | 961 | fl = *before; |
1da177e4 LT |
962 | if (!IS_POSIX(fl)) |
963 | continue; | |
964 | if (!posix_locks_conflict(request, fl)) | |
965 | continue; | |
5842add2 | 966 | if (conflock) |
3fe0fff1 | 967 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
968 | error = -EAGAIN; |
969 | if (!(request->fl_flags & FL_SLEEP)) | |
970 | goto out; | |
1c8c601a JL |
971 | /* |
972 | * Deadlock detection and insertion into the blocked | |
973 | * locks list must be done while holding the same lock! | |
974 | */ | |
1da177e4 | 975 | error = -EDEADLK; |
7b2296af | 976 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
977 | if (likely(!posix_locks_deadlock(request, fl))) { |
978 | error = FILE_LOCK_DEFERRED; | |
979 | __locks_insert_block(fl, request); | |
980 | } | |
7b2296af | 981 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
982 | goto out; |
983 | } | |
984 | } | |
985 | ||
986 | /* If we're just looking for a conflict, we're done. */ | |
987 | error = 0; | |
988 | if (request->fl_flags & FL_ACCESS) | |
989 | goto out; | |
990 | ||
1da177e4 | 991 | /* |
1da177e4 LT |
992 | * Find the first old lock with the same owner as the new lock. |
993 | */ | |
994 | ||
995 | before = &inode->i_flock; | |
996 | ||
997 | /* First skip locks owned by other processes. */ | |
998 | while ((fl = *before) && (!IS_POSIX(fl) || | |
999 | !posix_same_owner(request, fl))) { | |
1000 | before = &fl->fl_next; | |
1001 | } | |
1002 | ||
1cb36012 | 1003 | /* Process locks with this owner. */ |
1da177e4 LT |
1004 | while ((fl = *before) && posix_same_owner(request, fl)) { |
1005 | /* Detect adjacent or overlapping regions (if same lock type) | |
1006 | */ | |
1007 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
1008 | /* In all comparisons of start vs end, use |
1009 | * "start - 1" rather than "end + 1". If end | |
1010 | * is OFFSET_MAX, end + 1 will become negative. | |
1011 | */ | |
1da177e4 LT |
1012 | if (fl->fl_end < request->fl_start - 1) |
1013 | goto next_lock; | |
1014 | /* If the next lock in the list has entirely bigger | |
1015 | * addresses than the new one, insert the lock here. | |
1016 | */ | |
449231d6 | 1017 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1018 | break; |
1019 | ||
1020 | /* If we come here, the new and old lock are of the | |
1021 | * same type and adjacent or overlapping. Make one | |
1022 | * lock yielding from the lower start address of both | |
1023 | * locks to the higher end address. | |
1024 | */ | |
1025 | if (fl->fl_start > request->fl_start) | |
1026 | fl->fl_start = request->fl_start; | |
1027 | else | |
1028 | request->fl_start = fl->fl_start; | |
1029 | if (fl->fl_end < request->fl_end) | |
1030 | fl->fl_end = request->fl_end; | |
1031 | else | |
1032 | request->fl_end = fl->fl_end; | |
1033 | if (added) { | |
ed9814d8 | 1034 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1035 | continue; |
1036 | } | |
1037 | request = fl; | |
b9746ef8 | 1038 | added = true; |
1da177e4 LT |
1039 | } |
1040 | else { | |
1041 | /* Processing for different lock types is a bit | |
1042 | * more complex. | |
1043 | */ | |
1044 | if (fl->fl_end < request->fl_start) | |
1045 | goto next_lock; | |
1046 | if (fl->fl_start > request->fl_end) | |
1047 | break; | |
1048 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1049 | added = true; |
1da177e4 LT |
1050 | if (fl->fl_start < request->fl_start) |
1051 | left = fl; | |
1052 | /* If the next lock in the list has a higher end | |
1053 | * address than the new one, insert the new one here. | |
1054 | */ | |
1055 | if (fl->fl_end > request->fl_end) { | |
1056 | right = fl; | |
1057 | break; | |
1058 | } | |
1059 | if (fl->fl_start >= request->fl_start) { | |
1060 | /* The new lock completely replaces an old | |
1061 | * one (This may happen several times). | |
1062 | */ | |
1063 | if (added) { | |
ed9814d8 | 1064 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1065 | continue; |
1066 | } | |
b84d49f9 JL |
1067 | /* |
1068 | * Replace the old lock with new_fl, and | |
1069 | * remove the old one. It's safe to do the | |
1070 | * insert here since we know that we won't be | |
1071 | * using new_fl later, and that the lock is | |
1072 | * just replacing an existing lock. | |
1da177e4 | 1073 | */ |
b84d49f9 JL |
1074 | error = -ENOLCK; |
1075 | if (!new_fl) | |
1076 | goto out; | |
1077 | locks_copy_lock(new_fl, request); | |
1078 | request = new_fl; | |
1079 | new_fl = NULL; | |
ed9814d8 | 1080 | locks_delete_lock(before, &dispose); |
b84d49f9 | 1081 | locks_insert_lock(before, request); |
b9746ef8 | 1082 | added = true; |
1da177e4 LT |
1083 | } |
1084 | } | |
1085 | /* Go on to next lock. | |
1086 | */ | |
1087 | next_lock: | |
1088 | before = &fl->fl_next; | |
1089 | } | |
1090 | ||
0d9a490a | 1091 | /* |
1cb36012 JL |
1092 | * The above code only modifies existing locks in case of merging or |
1093 | * replacing. If new lock(s) need to be inserted all modifications are | |
1094 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1095 | */ |
1096 | error = -ENOLCK; /* "no luck" */ | |
1097 | if (right && left == right && !new_fl2) | |
1098 | goto out; | |
1099 | ||
1da177e4 LT |
1100 | error = 0; |
1101 | if (!added) { | |
f475ae95 TM |
1102 | if (request->fl_type == F_UNLCK) { |
1103 | if (request->fl_flags & FL_EXISTS) | |
1104 | error = -ENOENT; | |
1da177e4 | 1105 | goto out; |
f475ae95 | 1106 | } |
0d9a490a MS |
1107 | |
1108 | if (!new_fl) { | |
1109 | error = -ENOLCK; | |
1110 | goto out; | |
1111 | } | |
1da177e4 LT |
1112 | locks_copy_lock(new_fl, request); |
1113 | locks_insert_lock(before, new_fl); | |
1114 | new_fl = NULL; | |
1115 | } | |
1116 | if (right) { | |
1117 | if (left == right) { | |
1118 | /* The new lock breaks the old one in two pieces, | |
1119 | * so we have to use the second new lock. | |
1120 | */ | |
1121 | left = new_fl2; | |
1122 | new_fl2 = NULL; | |
1123 | locks_copy_lock(left, right); | |
1124 | locks_insert_lock(before, left); | |
1125 | } | |
1126 | right->fl_start = request->fl_end + 1; | |
1127 | locks_wake_up_blocks(right); | |
1128 | } | |
1129 | if (left) { | |
1130 | left->fl_end = request->fl_start - 1; | |
1131 | locks_wake_up_blocks(left); | |
1132 | } | |
1133 | out: | |
1c8c601a | 1134 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1135 | /* |
1136 | * Free any unused locks. | |
1137 | */ | |
1138 | if (new_fl) | |
1139 | locks_free_lock(new_fl); | |
1140 | if (new_fl2) | |
1141 | locks_free_lock(new_fl2); | |
ed9814d8 | 1142 | locks_dispose_list(&dispose); |
1da177e4 LT |
1143 | return error; |
1144 | } | |
1145 | ||
1146 | /** | |
1147 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1148 | * @filp: The file to apply the lock to | |
1149 | * @fl: The lock to be applied | |
150b3934 | 1150 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1151 | * |
1152 | * Add a POSIX style lock to a file. | |
1153 | * We merge adjacent & overlapping locks whenever possible. | |
1154 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1155 | * |
1156 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1157 | * whether or not a lock was successfully freed by testing the return | |
1158 | * value for -ENOENT. | |
1da177e4 | 1159 | */ |
150b3934 | 1160 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1161 | struct file_lock *conflock) |
1162 | { | |
496ad9aa | 1163 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1164 | } |
150b3934 | 1165 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1166 | |
1167 | /** | |
1168 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1169 | * @filp: The file to apply the lock to | |
1170 | * @fl: The lock to be applied | |
1171 | * | |
1172 | * Add a POSIX style lock to a file. | |
1173 | * We merge adjacent & overlapping locks whenever possible. | |
1174 | * POSIX locks are sorted by owner task, then by starting address | |
1175 | */ | |
1176 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1177 | { | |
1178 | int error; | |
1179 | might_sleep (); | |
1180 | for (;;) { | |
150b3934 | 1181 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1182 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1183 | break; |
1184 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1185 | if (!error) | |
1186 | continue; | |
1187 | ||
1188 | locks_delete_block(fl); | |
1189 | break; | |
1190 | } | |
1191 | return error; | |
1192 | } | |
1193 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1194 | ||
1195 | /** | |
1196 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1197 | * @file: the file to check |
1da177e4 LT |
1198 | * |
1199 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1200 | * This function is called from locks_verify_locked() only. | |
1201 | */ | |
d7a06983 | 1202 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1203 | { |
d7a06983 | 1204 | struct inode *inode = file_inode(file); |
1da177e4 LT |
1205 | struct file_lock *fl; |
1206 | ||
1207 | /* | |
1208 | * Search the lock list for this inode for any POSIX locks. | |
1209 | */ | |
1c8c601a | 1210 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1211 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1212 | if (!IS_POSIX(fl)) | |
1213 | continue; | |
73a8f5f7 CH |
1214 | if (fl->fl_owner != current->files && |
1215 | fl->fl_owner != file) | |
1da177e4 LT |
1216 | break; |
1217 | } | |
1c8c601a | 1218 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1219 | return fl ? -EAGAIN : 0; |
1220 | } | |
1221 | ||
1222 | /** | |
1223 | * locks_mandatory_area - Check for a conflicting lock | |
1224 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1225 | * for shared | |
1226 | * @inode: the file to check | |
1227 | * @filp: how the file was opened (if it was) | |
1228 | * @offset: start of area to check | |
1229 | * @count: length of area to check | |
1230 | * | |
1231 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1232 | * This function is called from rw_verify_area() and | |
1233 | * locks_verify_truncate(). | |
1234 | */ | |
1235 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1236 | struct file *filp, loff_t offset, | |
1237 | size_t count) | |
1238 | { | |
1239 | struct file_lock fl; | |
1240 | int error; | |
29723ade | 1241 | bool sleep = false; |
1da177e4 LT |
1242 | |
1243 | locks_init_lock(&fl); | |
1da177e4 LT |
1244 | fl.fl_pid = current->tgid; |
1245 | fl.fl_file = filp; | |
1246 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1247 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1248 | sleep = true; |
1da177e4 LT |
1249 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1250 | fl.fl_start = offset; | |
1251 | fl.fl_end = offset + count - 1; | |
1252 | ||
1253 | for (;;) { | |
29723ade | 1254 | if (filp) { |
73a8f5f7 | 1255 | fl.fl_owner = filp; |
29723ade JL |
1256 | fl.fl_flags &= ~FL_SLEEP; |
1257 | error = __posix_lock_file(inode, &fl, NULL); | |
1258 | if (!error) | |
1259 | break; | |
1260 | } | |
1261 | ||
1262 | if (sleep) | |
1263 | fl.fl_flags |= FL_SLEEP; | |
1264 | fl.fl_owner = current->files; | |
150b3934 | 1265 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1266 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1267 | break; |
1268 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1269 | if (!error) { | |
1270 | /* | |
1271 | * If we've been sleeping someone might have | |
1272 | * changed the permissions behind our back. | |
1273 | */ | |
a16877ca | 1274 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1275 | continue; |
1276 | } | |
1277 | ||
1278 | locks_delete_block(&fl); | |
1279 | break; | |
1280 | } | |
1281 | ||
1282 | return error; | |
1283 | } | |
1284 | ||
1285 | EXPORT_SYMBOL(locks_mandatory_area); | |
1286 | ||
778fc546 BF |
1287 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1288 | { | |
1289 | switch (arg) { | |
1290 | case F_UNLCK: | |
1291 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1292 | /* fall through: */ | |
1293 | case F_RDLCK: | |
1294 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1295 | } | |
1296 | } | |
1297 | ||
1da177e4 | 1298 | /* We already had a lease on this file; just change its type */ |
c45198ed | 1299 | int lease_modify(struct file_lock **before, int arg, struct list_head *dispose) |
1da177e4 LT |
1300 | { |
1301 | struct file_lock *fl = *before; | |
1302 | int error = assign_type(fl, arg); | |
1303 | ||
1304 | if (error) | |
1305 | return error; | |
778fc546 | 1306 | lease_clear_pending(fl, arg); |
1da177e4 | 1307 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1308 | if (arg == F_UNLCK) { |
1309 | struct file *filp = fl->fl_file; | |
1310 | ||
1311 | f_delown(filp); | |
1312 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1313 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1314 | if (fl->fl_fasync != NULL) { | |
1315 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1316 | fl->fl_fasync = NULL; | |
1317 | } | |
c45198ed | 1318 | locks_delete_lock(before, dispose); |
3b6e2723 | 1319 | } |
1da177e4 LT |
1320 | return 0; |
1321 | } | |
1da177e4 LT |
1322 | EXPORT_SYMBOL(lease_modify); |
1323 | ||
778fc546 BF |
1324 | static bool past_time(unsigned long then) |
1325 | { | |
1326 | if (!then) | |
1327 | /* 0 is a special value meaning "this never expires": */ | |
1328 | return false; | |
1329 | return time_after(jiffies, then); | |
1330 | } | |
1331 | ||
c45198ed | 1332 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 LT |
1333 | { |
1334 | struct file_lock **before; | |
1335 | struct file_lock *fl; | |
1336 | ||
f82b4b67 JL |
1337 | lockdep_assert_held(&inode->i_lock); |
1338 | ||
1da177e4 | 1339 | before = &inode->i_flock; |
ab83fa4b | 1340 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1341 | trace_time_out_leases(inode, fl); |
778fc546 | 1342 | if (past_time(fl->fl_downgrade_time)) |
c45198ed | 1343 | lease_modify(before, F_RDLCK, dispose); |
778fc546 | 1344 | if (past_time(fl->fl_break_time)) |
c45198ed | 1345 | lease_modify(before, F_UNLCK, dispose); |
1da177e4 LT |
1346 | if (fl == *before) /* lease_modify may have freed fl */ |
1347 | before = &fl->fl_next; | |
1348 | } | |
1349 | } | |
1350 | ||
df4e8d2c BF |
1351 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1352 | { | |
1353 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1354 | return false; | |
1355 | return locks_conflict(breaker, lease); | |
1356 | } | |
1357 | ||
03d12ddf JL |
1358 | static bool |
1359 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1360 | { | |
1361 | struct file_lock *fl; | |
1362 | ||
1363 | lockdep_assert_held(&inode->i_lock); | |
1364 | ||
1365 | for (fl = inode->i_flock ; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1366 | if (leases_conflict(fl, breaker)) | |
1367 | return true; | |
1368 | } | |
1369 | return false; | |
1370 | } | |
1371 | ||
1da177e4 LT |
1372 | /** |
1373 | * __break_lease - revoke all outstanding leases on file | |
1374 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1375 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1376 | * break all leases | |
1377 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1378 | * only delegations | |
1da177e4 | 1379 | * |
87250dd2 | 1380 | * break_lease (inlined for speed) has checked there already is at least |
1381 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1382 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1383 | * specified %O_NONBLOCK to your open(). |
1384 | */ | |
df4e8d2c | 1385 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1386 | { |
778fc546 | 1387 | int error = 0; |
03d12ddf | 1388 | struct file_lock *new_fl; |
4d01b7f5 | 1389 | struct file_lock *fl, **before; |
1da177e4 | 1390 | unsigned long break_time; |
8737c930 | 1391 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1392 | LIST_HEAD(dispose); |
1da177e4 | 1393 | |
8737c930 | 1394 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1395 | if (IS_ERR(new_fl)) |
1396 | return PTR_ERR(new_fl); | |
df4e8d2c | 1397 | new_fl->fl_flags = type; |
1da177e4 | 1398 | |
1c8c601a | 1399 | spin_lock(&inode->i_lock); |
1da177e4 | 1400 | |
c45198ed | 1401 | time_out_leases(inode, &dispose); |
1da177e4 | 1402 | |
03d12ddf | 1403 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1404 | goto out; |
1405 | ||
1da177e4 LT |
1406 | break_time = 0; |
1407 | if (lease_break_time > 0) { | |
1408 | break_time = jiffies + lease_break_time * HZ; | |
1409 | if (break_time == 0) | |
1410 | break_time++; /* so that 0 means no break time */ | |
1411 | } | |
1412 | ||
4d01b7f5 JL |
1413 | for (before = &inode->i_flock; |
1414 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1415 | before = &fl->fl_next) { | |
df4e8d2c BF |
1416 | if (!leases_conflict(fl, new_fl)) |
1417 | continue; | |
778fc546 BF |
1418 | if (want_write) { |
1419 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1420 | continue; | |
1421 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1422 | fl->fl_break_time = break_time; |
778fc546 | 1423 | } else { |
03d12ddf | 1424 | if (lease_breaking(inode->i_flock)) |
778fc546 BF |
1425 | continue; |
1426 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1427 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1428 | } |
4d01b7f5 JL |
1429 | if (fl->fl_lmops->lm_break(fl)) |
1430 | locks_delete_lock(before, &dispose); | |
1da177e4 LT |
1431 | } |
1432 | ||
4d01b7f5 JL |
1433 | fl = inode->i_flock; |
1434 | if (!fl || !IS_LEASE(fl)) | |
1435 | goto out; | |
1436 | ||
843c6b2f | 1437 | if (mode & O_NONBLOCK) { |
62af4f1f | 1438 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1439 | error = -EWOULDBLOCK; |
1440 | goto out; | |
1441 | } | |
1442 | ||
1443 | restart: | |
03d12ddf | 1444 | break_time = inode->i_flock->fl_break_time; |
f1c6bb2c | 1445 | if (break_time != 0) |
1da177e4 | 1446 | break_time -= jiffies; |
f1c6bb2c JL |
1447 | if (break_time == 0) |
1448 | break_time++; | |
03d12ddf | 1449 | locks_insert_block(inode->i_flock, new_fl); |
62af4f1f | 1450 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1451 | spin_unlock(&inode->i_lock); |
c45198ed | 1452 | locks_dispose_list(&dispose); |
4321e01e MW |
1453 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1454 | !new_fl->fl_next, break_time); | |
1c8c601a | 1455 | spin_lock(&inode->i_lock); |
62af4f1f | 1456 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1457 | locks_delete_block(new_fl); |
1da177e4 | 1458 | if (error >= 0) { |
778fc546 BF |
1459 | /* |
1460 | * Wait for the next conflicting lease that has not been | |
1461 | * broken yet | |
1462 | */ | |
03d12ddf JL |
1463 | if (error == 0) |
1464 | time_out_leases(inode, &dispose); | |
1465 | if (any_leases_conflict(inode, new_fl)) | |
1466 | goto restart; | |
1467 | ||
1da177e4 LT |
1468 | error = 0; |
1469 | } | |
1470 | ||
1471 | out: | |
1c8c601a | 1472 | spin_unlock(&inode->i_lock); |
c45198ed | 1473 | locks_dispose_list(&dispose); |
6d4b9e38 | 1474 | locks_free_lock(new_fl); |
1da177e4 LT |
1475 | return error; |
1476 | } | |
1477 | ||
1478 | EXPORT_SYMBOL(__break_lease); | |
1479 | ||
1480 | /** | |
a6b91919 | 1481 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1482 | * @inode: the inode |
1483 | * @time: pointer to a timespec which will contain the last modified time | |
1484 | * | |
1485 | * This is to force NFS clients to flush their caches for files with | |
1486 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1487 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1488 | */ |
1489 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1490 | { | |
bfe86024 JL |
1491 | bool has_lease = false; |
1492 | struct file_lock *flock; | |
1493 | ||
1494 | if (inode->i_flock) { | |
1495 | spin_lock(&inode->i_lock); | |
1496 | flock = inode->i_flock; | |
1497 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) | |
1498 | has_lease = true; | |
1499 | spin_unlock(&inode->i_lock); | |
1500 | } | |
1501 | ||
1502 | if (has_lease) | |
1da177e4 LT |
1503 | *time = current_fs_time(inode->i_sb); |
1504 | else | |
1505 | *time = inode->i_mtime; | |
1506 | } | |
1507 | ||
1508 | EXPORT_SYMBOL(lease_get_mtime); | |
1509 | ||
1510 | /** | |
1511 | * fcntl_getlease - Enquire what lease is currently active | |
1512 | * @filp: the file | |
1513 | * | |
1514 | * The value returned by this function will be one of | |
1515 | * (if no lease break is pending): | |
1516 | * | |
1517 | * %F_RDLCK to indicate a shared lease is held. | |
1518 | * | |
1519 | * %F_WRLCK to indicate an exclusive lease is held. | |
1520 | * | |
1521 | * %F_UNLCK to indicate no lease is held. | |
1522 | * | |
1523 | * (if a lease break is pending): | |
1524 | * | |
1525 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1526 | * changed to a shared lease (or removed). | |
1527 | * | |
1528 | * %F_UNLCK to indicate the lease needs to be removed. | |
1529 | * | |
1530 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1531 | * should be returned to userspace. | |
1532 | */ | |
1533 | int fcntl_getlease(struct file *filp) | |
1534 | { | |
1535 | struct file_lock *fl; | |
1c8c601a | 1536 | struct inode *inode = file_inode(filp); |
1da177e4 | 1537 | int type = F_UNLCK; |
c45198ed | 1538 | LIST_HEAD(dispose); |
1da177e4 | 1539 | |
1c8c601a | 1540 | spin_lock(&inode->i_lock); |
c45198ed | 1541 | time_out_leases(file_inode(filp), &dispose); |
496ad9aa | 1542 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); |
1da177e4 LT |
1543 | fl = fl->fl_next) { |
1544 | if (fl->fl_file == filp) { | |
778fc546 | 1545 | type = target_leasetype(fl); |
1da177e4 LT |
1546 | break; |
1547 | } | |
1548 | } | |
1c8c601a | 1549 | spin_unlock(&inode->i_lock); |
c45198ed | 1550 | locks_dispose_list(&dispose); |
1da177e4 LT |
1551 | return type; |
1552 | } | |
1553 | ||
24cbe784 JL |
1554 | /** |
1555 | * check_conflicting_open - see if the given dentry points to a file that has | |
1556 | * an existing open that would conflict with the | |
1557 | * desired lease. | |
1558 | * @dentry: dentry to check | |
1559 | * @arg: type of lease that we're trying to acquire | |
1560 | * | |
1561 | * Check to see if there's an existing open fd on this file that would | |
1562 | * conflict with the lease we're trying to set. | |
1563 | */ | |
1564 | static int | |
1565 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1566 | { | |
1567 | int ret = 0; | |
1568 | struct inode *inode = dentry->d_inode; | |
1569 | ||
1570 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1571 | return -EAGAIN; | |
1572 | ||
1573 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1574 | (atomic_read(&inode->i_count) > 1))) | |
1575 | ret = -EAGAIN; | |
1576 | ||
1577 | return ret; | |
1578 | } | |
1579 | ||
e6f5c789 JL |
1580 | static int |
1581 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1582 | { |
7eaae282 | 1583 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1584 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1585 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1586 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1587 | int error; |
c45198ed | 1588 | LIST_HEAD(dispose); |
1da177e4 | 1589 | |
096657b6 | 1590 | lease = *flp; |
62af4f1f JL |
1591 | trace_generic_add_lease(inode, lease); |
1592 | ||
df4e8d2c BF |
1593 | /* |
1594 | * In the delegation case we need mutual exclusion with | |
1595 | * a number of operations that take the i_mutex. We trylock | |
1596 | * because delegations are an optional optimization, and if | |
1597 | * there's some chance of a conflict--we'd rather not | |
1598 | * bother, maybe that's a sign this just isn't a good file to | |
1599 | * hand out a delegation on. | |
1600 | */ | |
1601 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1602 | return -EAGAIN; | |
1603 | ||
1604 | if (is_deleg && arg == F_WRLCK) { | |
1605 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1606 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1607 | WARN_ON_ONCE(1); |
1608 | return -EINVAL; | |
1609 | } | |
096657b6 | 1610 | |
f82b4b67 | 1611 | spin_lock(&inode->i_lock); |
c45198ed | 1612 | time_out_leases(inode, &dispose); |
24cbe784 JL |
1613 | error = check_conflicting_open(dentry, arg); |
1614 | if (error) | |
096657b6 | 1615 | goto out; |
6d5e8b05 | 1616 | |
1da177e4 LT |
1617 | /* |
1618 | * At this point, we know that if there is an exclusive | |
1619 | * lease on this file, then we hold it on this filp | |
1620 | * (otherwise our open of this file would have blocked). | |
1621 | * And if we are trying to acquire an exclusive lease, | |
1622 | * then the file is not open by anyone (including us) | |
1623 | * except for this filp. | |
1624 | */ | |
c1f24ef4 | 1625 | error = -EAGAIN; |
1da177e4 LT |
1626 | for (before = &inode->i_flock; |
1627 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1628 | before = &fl->fl_next) { | |
c1f24ef4 | 1629 | if (fl->fl_file == filp) { |
1da177e4 | 1630 | my_before = before; |
c1f24ef4 BF |
1631 | continue; |
1632 | } | |
1633 | /* | |
1634 | * No exclusive leases if someone else has a lease on | |
1635 | * this file: | |
1636 | */ | |
1637 | if (arg == F_WRLCK) | |
1638 | goto out; | |
1639 | /* | |
1640 | * Modifying our existing lease is OK, but no getting a | |
1641 | * new lease if someone else is opening for write: | |
1642 | */ | |
1643 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1644 | goto out; | |
1da177e4 LT |
1645 | } |
1646 | ||
1da177e4 | 1647 | if (my_before != NULL) { |
1c7dd2ff | 1648 | lease = *my_before; |
c45198ed | 1649 | error = lease->fl_lmops->lm_change(my_before, arg, &dispose); |
1c7dd2ff JL |
1650 | if (error) |
1651 | goto out; | |
1652 | goto out_setup; | |
1da177e4 LT |
1653 | } |
1654 | ||
1da177e4 LT |
1655 | error = -EINVAL; |
1656 | if (!leases_enable) | |
1657 | goto out; | |
1658 | ||
c5b1f0d9 | 1659 | locks_insert_lock(before, lease); |
24cbe784 JL |
1660 | /* |
1661 | * The check in break_lease() is lockless. It's possible for another | |
1662 | * open to race in after we did the earlier check for a conflicting | |
1663 | * open but before the lease was inserted. Check again for a | |
1664 | * conflicting open and cancel the lease if there is one. | |
1665 | * | |
1666 | * We also add a barrier here to ensure that the insertion of the lock | |
1667 | * precedes these checks. | |
1668 | */ | |
1669 | smp_mb(); | |
1670 | error = check_conflicting_open(dentry, arg); | |
1671 | if (error) | |
e6f5c789 | 1672 | goto out_unlink; |
1c7dd2ff JL |
1673 | |
1674 | out_setup: | |
1675 | if (lease->fl_lmops->lm_setup) | |
1676 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1677 | out: |
f82b4b67 | 1678 | spin_unlock(&inode->i_lock); |
c45198ed | 1679 | locks_dispose_list(&dispose); |
df4e8d2c BF |
1680 | if (is_deleg) |
1681 | mutex_unlock(&inode->i_mutex); | |
1c7dd2ff JL |
1682 | if (!error && !my_before) |
1683 | *flp = NULL; | |
1da177e4 | 1684 | return error; |
e6f5c789 JL |
1685 | out_unlink: |
1686 | locks_unlink_lock(before); | |
1687 | goto out; | |
1da177e4 | 1688 | } |
8335ebd9 | 1689 | |
0efaa7e8 | 1690 | static int generic_delete_lease(struct file *filp) |
8335ebd9 | 1691 | { |
0efaa7e8 | 1692 | int error = -EAGAIN; |
8335ebd9 BF |
1693 | struct file_lock *fl, **before; |
1694 | struct dentry *dentry = filp->f_path.dentry; | |
1695 | struct inode *inode = dentry->d_inode; | |
c45198ed | 1696 | LIST_HEAD(dispose); |
8335ebd9 | 1697 | |
f82b4b67 | 1698 | spin_lock(&inode->i_lock); |
c45198ed | 1699 | time_out_leases(inode, &dispose); |
8335ebd9 BF |
1700 | for (before = &inode->i_flock; |
1701 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1702 | before = &fl->fl_next) { | |
0efaa7e8 JL |
1703 | if (fl->fl_file == filp) |
1704 | break; | |
8335ebd9 | 1705 | } |
0efaa7e8 | 1706 | trace_generic_delete_lease(inode, fl); |
52d304eb | 1707 | if (fl && IS_LEASE(fl)) |
c45198ed | 1708 | error = fl->fl_lmops->lm_change(before, F_UNLCK, &dispose); |
f82b4b67 | 1709 | spin_unlock(&inode->i_lock); |
c45198ed | 1710 | locks_dispose_list(&dispose); |
0efaa7e8 | 1711 | return error; |
8335ebd9 BF |
1712 | } |
1713 | ||
1714 | /** | |
1715 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1716 | * @filp: file pointer |
1717 | * @arg: type of lease to obtain | |
1718 | * @flp: input - file_lock to use, output - file_lock inserted | |
1719 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1720 | * doesn't require it) | |
8335ebd9 BF |
1721 | * |
1722 | * The (input) flp->fl_lmops->lm_break function is required | |
1723 | * by break_lease(). | |
8335ebd9 | 1724 | */ |
e6f5c789 JL |
1725 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1726 | void **priv) | |
8335ebd9 BF |
1727 | { |
1728 | struct dentry *dentry = filp->f_path.dentry; | |
1729 | struct inode *inode = dentry->d_inode; | |
1730 | int error; | |
1731 | ||
8e96e3b7 | 1732 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1733 | return -EACCES; |
1734 | if (!S_ISREG(inode->i_mode)) | |
1735 | return -EINVAL; | |
1736 | error = security_file_lock(filp, arg); | |
1737 | if (error) | |
1738 | return error; | |
1739 | ||
8335ebd9 BF |
1740 | switch (arg) { |
1741 | case F_UNLCK: | |
0efaa7e8 | 1742 | return generic_delete_lease(filp); |
8335ebd9 BF |
1743 | case F_RDLCK: |
1744 | case F_WRLCK: | |
0efaa7e8 JL |
1745 | if (!(*flp)->fl_lmops->lm_break) { |
1746 | WARN_ON_ONCE(1); | |
1747 | return -ENOLCK; | |
1748 | } | |
e6f5c789 | 1749 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1750 | default: |
8d657eb3 | 1751 | return -EINVAL; |
8335ebd9 BF |
1752 | } |
1753 | } | |
0af1a450 | 1754 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1755 | |
b89f4321 | 1756 | /** |
e51673aa | 1757 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1758 | * @filp: file pointer |
1759 | * @arg: type of lease to obtain | |
1760 | * @lease: file_lock to use when adding a lease | |
1761 | * @priv: private info for lm_setup when adding a lease (may be | |
1762 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1763 | * |
1764 | * Call this to establish a lease on the file. The "lease" argument is not | |
1765 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1766 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1767 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1768 | * stack trace). | |
1c7dd2ff JL |
1769 | * |
1770 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1771 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1772 | */ |
e6f5c789 JL |
1773 | int |
1774 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1775 | { |
1c7dd2ff | 1776 | if (filp->f_op->setlease) |
f82b4b67 | 1777 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1778 | else |
f82b4b67 | 1779 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1780 | } |
a9933cea | 1781 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1782 | |
0ceaf6c7 | 1783 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1784 | { |
1c7dd2ff | 1785 | struct file_lock *fl; |
f7347ce4 | 1786 | struct fasync_struct *new; |
1da177e4 LT |
1787 | int error; |
1788 | ||
c5b1f0d9 AB |
1789 | fl = lease_alloc(filp, arg); |
1790 | if (IS_ERR(fl)) | |
1791 | return PTR_ERR(fl); | |
1da177e4 | 1792 | |
f7347ce4 LT |
1793 | new = fasync_alloc(); |
1794 | if (!new) { | |
1795 | locks_free_lock(fl); | |
1796 | return -ENOMEM; | |
1797 | } | |
1c7dd2ff | 1798 | new->fa_fd = fd; |
f7347ce4 | 1799 | |
1c7dd2ff | 1800 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1801 | if (fl) |
1802 | locks_free_lock(fl); | |
f7347ce4 LT |
1803 | if (new) |
1804 | fasync_free(new); | |
1da177e4 LT |
1805 | return error; |
1806 | } | |
1807 | ||
0ceaf6c7 BF |
1808 | /** |
1809 | * fcntl_setlease - sets a lease on an open file | |
1810 | * @fd: open file descriptor | |
1811 | * @filp: file pointer | |
1812 | * @arg: type of lease to obtain | |
1813 | * | |
1814 | * Call this fcntl to establish a lease on the file. | |
1815 | * Note that you also need to call %F_SETSIG to | |
1816 | * receive a signal when the lease is broken. | |
1817 | */ | |
1818 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1819 | { | |
1820 | if (arg == F_UNLCK) | |
e6f5c789 | 1821 | return vfs_setlease(filp, F_UNLCK, NULL, NULL); |
0ceaf6c7 BF |
1822 | return do_fcntl_add_lease(fd, filp, arg); |
1823 | } | |
1824 | ||
1da177e4 LT |
1825 | /** |
1826 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1827 | * @filp: The file to apply the lock to | |
1828 | * @fl: The lock to be applied | |
1829 | * | |
1830 | * Add a FLOCK style lock to a file. | |
1831 | */ | |
1832 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1833 | { | |
1834 | int error; | |
1835 | might_sleep(); | |
1836 | for (;;) { | |
1837 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1838 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1839 | break; |
1840 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1841 | if (!error) | |
1842 | continue; | |
1843 | ||
1844 | locks_delete_block(fl); | |
1845 | break; | |
1846 | } | |
1847 | return error; | |
1848 | } | |
1849 | ||
1850 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1851 | ||
1852 | /** | |
1853 | * sys_flock: - flock() system call. | |
1854 | * @fd: the file descriptor to lock. | |
1855 | * @cmd: the type of lock to apply. | |
1856 | * | |
1857 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1858 | * The @cmd can be one of | |
1859 | * | |
1860 | * %LOCK_SH -- a shared lock. | |
1861 | * | |
1862 | * %LOCK_EX -- an exclusive lock. | |
1863 | * | |
1864 | * %LOCK_UN -- remove an existing lock. | |
1865 | * | |
1866 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1867 | * | |
1868 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1869 | * processes read and write access respectively. | |
1870 | */ | |
002c8976 | 1871 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1872 | { |
2903ff01 | 1873 | struct fd f = fdget(fd); |
1da177e4 LT |
1874 | struct file_lock *lock; |
1875 | int can_sleep, unlock; | |
1876 | int error; | |
1877 | ||
1878 | error = -EBADF; | |
2903ff01 | 1879 | if (!f.file) |
1da177e4 LT |
1880 | goto out; |
1881 | ||
1882 | can_sleep = !(cmd & LOCK_NB); | |
1883 | cmd &= ~LOCK_NB; | |
1884 | unlock = (cmd == LOCK_UN); | |
1885 | ||
aeb5d727 | 1886 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1887 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1888 | goto out_putf; |
1889 | ||
6e129d00 JL |
1890 | lock = flock_make_lock(f.file, cmd); |
1891 | if (IS_ERR(lock)) { | |
1892 | error = PTR_ERR(lock); | |
1da177e4 | 1893 | goto out_putf; |
6e129d00 JL |
1894 | } |
1895 | ||
1da177e4 LT |
1896 | if (can_sleep) |
1897 | lock->fl_flags |= FL_SLEEP; | |
1898 | ||
2903ff01 | 1899 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1900 | if (error) |
1901 | goto out_free; | |
1902 | ||
72c2d531 | 1903 | if (f.file->f_op->flock) |
2903ff01 | 1904 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1905 | (can_sleep) ? F_SETLKW : F_SETLK, |
1906 | lock); | |
1907 | else | |
2903ff01 | 1908 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1909 | |
1910 | out_free: | |
993dfa87 | 1911 | locks_free_lock(lock); |
1da177e4 LT |
1912 | |
1913 | out_putf: | |
2903ff01 | 1914 | fdput(f); |
1da177e4 LT |
1915 | out: |
1916 | return error; | |
1917 | } | |
1918 | ||
3ee17abd BF |
1919 | /** |
1920 | * vfs_test_lock - test file byte range lock | |
1921 | * @filp: The file to test lock for | |
6924c554 | 1922 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1923 | * |
1924 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1925 | * setting conf->fl_type to something other than F_UNLCK. | |
1926 | */ | |
1927 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1928 | { | |
72c2d531 | 1929 | if (filp->f_op->lock) |
3ee17abd BF |
1930 | return filp->f_op->lock(filp, F_GETLK, fl); |
1931 | posix_test_lock(filp, fl); | |
1932 | return 0; | |
1933 | } | |
1934 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1935 | ||
c2fa1b8a BF |
1936 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1937 | { | |
cff2fce5 | 1938 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1939 | #if BITS_PER_LONG == 32 |
1940 | /* | |
1941 | * Make sure we can represent the posix lock via | |
1942 | * legacy 32bit flock. | |
1943 | */ | |
1944 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1945 | return -EOVERFLOW; | |
1946 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1947 | return -EOVERFLOW; | |
1948 | #endif | |
1949 | flock->l_start = fl->fl_start; | |
1950 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1951 | fl->fl_end - fl->fl_start + 1; | |
1952 | flock->l_whence = 0; | |
129a84de | 1953 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1954 | return 0; |
1955 | } | |
1956 | ||
1957 | #if BITS_PER_LONG == 32 | |
1958 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1959 | { | |
cff2fce5 | 1960 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1961 | flock->l_start = fl->fl_start; |
1962 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1963 | fl->fl_end - fl->fl_start + 1; | |
1964 | flock->l_whence = 0; | |
1965 | flock->l_type = fl->fl_type; | |
1966 | } | |
1967 | #endif | |
1968 | ||
1da177e4 LT |
1969 | /* Report the first existing lock that would conflict with l. |
1970 | * This implements the F_GETLK command of fcntl(). | |
1971 | */ | |
c1e62b8f | 1972 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 1973 | { |
9d6a8c5c | 1974 | struct file_lock file_lock; |
1da177e4 LT |
1975 | struct flock flock; |
1976 | int error; | |
1977 | ||
1978 | error = -EFAULT; | |
1979 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1980 | goto out; | |
1981 | error = -EINVAL; | |
1982 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1983 | goto out; | |
1984 | ||
1985 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1986 | if (error) | |
1987 | goto out; | |
1988 | ||
0d3f7a2d | 1989 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
1990 | error = -EINVAL; |
1991 | if (flock.l_pid != 0) | |
1992 | goto out; | |
1993 | ||
5d50ffd7 | 1994 | cmd = F_GETLK; |
cff2fce5 | 1995 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 1996 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
1997 | } |
1998 | ||
3ee17abd BF |
1999 | error = vfs_test_lock(filp, &file_lock); |
2000 | if (error) | |
2001 | goto out; | |
1da177e4 | 2002 | |
9d6a8c5c ME |
2003 | flock.l_type = file_lock.fl_type; |
2004 | if (file_lock.fl_type != F_UNLCK) { | |
2005 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2006 | if (error) |
f328296e | 2007 | goto rel_priv; |
1da177e4 LT |
2008 | } |
2009 | error = -EFAULT; | |
2010 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2011 | error = 0; | |
f328296e KM |
2012 | rel_priv: |
2013 | locks_release_private(&file_lock); | |
1da177e4 LT |
2014 | out: |
2015 | return error; | |
2016 | } | |
2017 | ||
7723ec97 ME |
2018 | /** |
2019 | * vfs_lock_file - file byte range lock | |
2020 | * @filp: The file to apply the lock to | |
2021 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2022 | * @fl: The lock to be applied | |
150b3934 ME |
2023 | * @conf: Place to return a copy of the conflicting lock, if found. |
2024 | * | |
2025 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2026 | * as the final argument. | |
2027 | * | |
2028 | * If the filesystem defines a private ->lock() method, then @conf will | |
2029 | * be left unchanged; so a caller that cares should initialize it to | |
2030 | * some acceptable default. | |
2beb6614 ME |
2031 | * |
2032 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2033 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2034 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2035 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2036 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2037 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2038 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2039 | * request completes. |
2040 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2041 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2042 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2043 | * nonzero return code and the file system should release the lock. The file |
2044 | * system is also responsible to keep a corresponding posix lock when it | |
2045 | * grants a lock so the VFS can find out which locks are locally held and do | |
2046 | * the correct lock cleanup when required. | |
2047 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2048 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2049 | * return code. |
7723ec97 | 2050 | */ |
150b3934 | 2051 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2052 | { |
72c2d531 | 2053 | if (filp->f_op->lock) |
7723ec97 ME |
2054 | return filp->f_op->lock(filp, cmd, fl); |
2055 | else | |
150b3934 | 2056 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2057 | } |
2058 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2059 | ||
b648a6de MS |
2060 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2061 | struct file_lock *fl) | |
2062 | { | |
2063 | int error; | |
2064 | ||
2065 | error = security_file_lock(filp, fl->fl_type); | |
2066 | if (error) | |
2067 | return error; | |
2068 | ||
764c76b3 MS |
2069 | for (;;) { |
2070 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2071 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2072 | break; |
764c76b3 MS |
2073 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2074 | if (!error) | |
2075 | continue; | |
2076 | ||
2077 | locks_delete_block(fl); | |
2078 | break; | |
b648a6de MS |
2079 | } |
2080 | ||
2081 | return error; | |
2082 | } | |
2083 | ||
cf01f4ee JL |
2084 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2085 | static int | |
2086 | check_fmode_for_setlk(struct file_lock *fl) | |
2087 | { | |
2088 | switch (fl->fl_type) { | |
2089 | case F_RDLCK: | |
2090 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2091 | return -EBADF; | |
2092 | break; | |
2093 | case F_WRLCK: | |
2094 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2095 | return -EBADF; | |
2096 | } | |
2097 | return 0; | |
2098 | } | |
2099 | ||
1da177e4 LT |
2100 | /* Apply the lock described by l to an open file descriptor. |
2101 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2102 | */ | |
c293621b PS |
2103 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2104 | struct flock __user *l) | |
1da177e4 LT |
2105 | { |
2106 | struct file_lock *file_lock = locks_alloc_lock(); | |
2107 | struct flock flock; | |
2108 | struct inode *inode; | |
0b2bac2f | 2109 | struct file *f; |
1da177e4 LT |
2110 | int error; |
2111 | ||
2112 | if (file_lock == NULL) | |
2113 | return -ENOLCK; | |
2114 | ||
2115 | /* | |
2116 | * This might block, so we do it before checking the inode. | |
2117 | */ | |
2118 | error = -EFAULT; | |
2119 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2120 | goto out; | |
2121 | ||
496ad9aa | 2122 | inode = file_inode(filp); |
1da177e4 LT |
2123 | |
2124 | /* Don't allow mandatory locks on files that may be memory mapped | |
2125 | * and shared. | |
2126 | */ | |
a16877ca | 2127 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2128 | error = -EAGAIN; |
2129 | goto out; | |
2130 | } | |
2131 | ||
c293621b | 2132 | again: |
1da177e4 LT |
2133 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2134 | if (error) | |
2135 | goto out; | |
5d50ffd7 | 2136 | |
cf01f4ee JL |
2137 | error = check_fmode_for_setlk(file_lock); |
2138 | if (error) | |
2139 | goto out; | |
2140 | ||
5d50ffd7 JL |
2141 | /* |
2142 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2143 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2144 | */ |
2145 | switch (cmd) { | |
0d3f7a2d | 2146 | case F_OFD_SETLK: |
90478939 JL |
2147 | error = -EINVAL; |
2148 | if (flock.l_pid != 0) | |
2149 | goto out; | |
2150 | ||
5d50ffd7 | 2151 | cmd = F_SETLK; |
cff2fce5 | 2152 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2153 | file_lock->fl_owner = filp; |
5d50ffd7 | 2154 | break; |
0d3f7a2d | 2155 | case F_OFD_SETLKW: |
90478939 JL |
2156 | error = -EINVAL; |
2157 | if (flock.l_pid != 0) | |
2158 | goto out; | |
2159 | ||
5d50ffd7 | 2160 | cmd = F_SETLKW; |
cff2fce5 | 2161 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2162 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2163 | /* Fallthrough */ |
2164 | case F_SETLKW: | |
1da177e4 LT |
2165 | file_lock->fl_flags |= FL_SLEEP; |
2166 | } | |
5d50ffd7 | 2167 | |
b648a6de | 2168 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2169 | |
c293621b PS |
2170 | /* |
2171 | * Attempt to detect a close/fcntl race and recover by | |
2172 | * releasing the lock that was just acquired. | |
2173 | */ | |
0b2bac2f AV |
2174 | /* |
2175 | * we need that spin_lock here - it prevents reordering between | |
2176 | * update of inode->i_flock and check for it done in close(). | |
2177 | * rcu_read_lock() wouldn't do. | |
2178 | */ | |
2179 | spin_lock(¤t->files->file_lock); | |
2180 | f = fcheck(fd); | |
2181 | spin_unlock(¤t->files->file_lock); | |
2182 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2183 | flock.l_type = F_UNLCK; |
2184 | goto again; | |
1da177e4 LT |
2185 | } |
2186 | ||
c293621b | 2187 | out: |
1da177e4 LT |
2188 | locks_free_lock(file_lock); |
2189 | return error; | |
2190 | } | |
2191 | ||
2192 | #if BITS_PER_LONG == 32 | |
2193 | /* Report the first existing lock that would conflict with l. | |
2194 | * This implements the F_GETLK command of fcntl(). | |
2195 | */ | |
c1e62b8f | 2196 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2197 | { |
9d6a8c5c | 2198 | struct file_lock file_lock; |
1da177e4 LT |
2199 | struct flock64 flock; |
2200 | int error; | |
2201 | ||
2202 | error = -EFAULT; | |
2203 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2204 | goto out; | |
2205 | error = -EINVAL; | |
2206 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2207 | goto out; | |
2208 | ||
2209 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2210 | if (error) | |
2211 | goto out; | |
2212 | ||
0d3f7a2d | 2213 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2214 | error = -EINVAL; |
2215 | if (flock.l_pid != 0) | |
2216 | goto out; | |
2217 | ||
5d50ffd7 | 2218 | cmd = F_GETLK64; |
cff2fce5 | 2219 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2220 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2221 | } |
2222 | ||
3ee17abd BF |
2223 | error = vfs_test_lock(filp, &file_lock); |
2224 | if (error) | |
2225 | goto out; | |
2226 | ||
9d6a8c5c ME |
2227 | flock.l_type = file_lock.fl_type; |
2228 | if (file_lock.fl_type != F_UNLCK) | |
2229 | posix_lock_to_flock64(&flock, &file_lock); | |
2230 | ||
1da177e4 LT |
2231 | error = -EFAULT; |
2232 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2233 | error = 0; | |
f328296e KM |
2234 | |
2235 | locks_release_private(&file_lock); | |
1da177e4 LT |
2236 | out: |
2237 | return error; | |
2238 | } | |
2239 | ||
2240 | /* Apply the lock described by l to an open file descriptor. | |
2241 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2242 | */ | |
c293621b PS |
2243 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2244 | struct flock64 __user *l) | |
1da177e4 LT |
2245 | { |
2246 | struct file_lock *file_lock = locks_alloc_lock(); | |
2247 | struct flock64 flock; | |
2248 | struct inode *inode; | |
0b2bac2f | 2249 | struct file *f; |
1da177e4 LT |
2250 | int error; |
2251 | ||
2252 | if (file_lock == NULL) | |
2253 | return -ENOLCK; | |
2254 | ||
2255 | /* | |
2256 | * This might block, so we do it before checking the inode. | |
2257 | */ | |
2258 | error = -EFAULT; | |
2259 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2260 | goto out; | |
2261 | ||
496ad9aa | 2262 | inode = file_inode(filp); |
1da177e4 LT |
2263 | |
2264 | /* Don't allow mandatory locks on files that may be memory mapped | |
2265 | * and shared. | |
2266 | */ | |
a16877ca | 2267 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2268 | error = -EAGAIN; |
2269 | goto out; | |
2270 | } | |
2271 | ||
c293621b | 2272 | again: |
1da177e4 LT |
2273 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2274 | if (error) | |
2275 | goto out; | |
5d50ffd7 | 2276 | |
cf01f4ee JL |
2277 | error = check_fmode_for_setlk(file_lock); |
2278 | if (error) | |
2279 | goto out; | |
2280 | ||
5d50ffd7 JL |
2281 | /* |
2282 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2283 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2284 | */ |
2285 | switch (cmd) { | |
0d3f7a2d | 2286 | case F_OFD_SETLK: |
90478939 JL |
2287 | error = -EINVAL; |
2288 | if (flock.l_pid != 0) | |
2289 | goto out; | |
2290 | ||
5d50ffd7 | 2291 | cmd = F_SETLK64; |
cff2fce5 | 2292 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2293 | file_lock->fl_owner = filp; |
5d50ffd7 | 2294 | break; |
0d3f7a2d | 2295 | case F_OFD_SETLKW: |
90478939 JL |
2296 | error = -EINVAL; |
2297 | if (flock.l_pid != 0) | |
2298 | goto out; | |
2299 | ||
5d50ffd7 | 2300 | cmd = F_SETLKW64; |
cff2fce5 | 2301 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2302 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2303 | /* Fallthrough */ |
2304 | case F_SETLKW64: | |
1da177e4 LT |
2305 | file_lock->fl_flags |= FL_SLEEP; |
2306 | } | |
5d50ffd7 | 2307 | |
b648a6de | 2308 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2309 | |
c293621b PS |
2310 | /* |
2311 | * Attempt to detect a close/fcntl race and recover by | |
2312 | * releasing the lock that was just acquired. | |
2313 | */ | |
0b2bac2f AV |
2314 | spin_lock(¤t->files->file_lock); |
2315 | f = fcheck(fd); | |
2316 | spin_unlock(¤t->files->file_lock); | |
2317 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2318 | flock.l_type = F_UNLCK; |
2319 | goto again; | |
1da177e4 LT |
2320 | } |
2321 | ||
2322 | out: | |
2323 | locks_free_lock(file_lock); | |
2324 | return error; | |
2325 | } | |
2326 | #endif /* BITS_PER_LONG == 32 */ | |
2327 | ||
2328 | /* | |
2329 | * This function is called when the file is being removed | |
2330 | * from the task's fd array. POSIX locks belonging to this task | |
2331 | * are deleted at this time. | |
2332 | */ | |
2333 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2334 | { | |
ff7b86b8 | 2335 | struct file_lock lock; |
1da177e4 LT |
2336 | |
2337 | /* | |
2338 | * If there are no locks held on this file, we don't need to call | |
2339 | * posix_lock_file(). Another process could be setting a lock on this | |
2340 | * file at the same time, but we wouldn't remove that lock anyway. | |
2341 | */ | |
496ad9aa | 2342 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2343 | return; |
2344 | ||
2345 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2346 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2347 | lock.fl_start = 0; |
2348 | lock.fl_end = OFFSET_MAX; | |
2349 | lock.fl_owner = owner; | |
2350 | lock.fl_pid = current->tgid; | |
2351 | lock.fl_file = filp; | |
2352 | lock.fl_ops = NULL; | |
2353 | lock.fl_lmops = NULL; | |
2354 | ||
150b3934 | 2355 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2356 | |
1da177e4 LT |
2357 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2358 | lock.fl_ops->fl_release_private(&lock); | |
2359 | } | |
2360 | ||
2361 | EXPORT_SYMBOL(locks_remove_posix); | |
2362 | ||
dd459bb1 JL |
2363 | static void |
2364 | locks_remove_flock(struct file *filp) | |
2365 | { | |
2366 | struct file_lock fl = { | |
2367 | .fl_owner = filp, | |
2368 | .fl_pid = current->tgid, | |
2369 | .fl_file = filp, | |
2370 | .fl_flags = FL_FLOCK, | |
2371 | .fl_type = F_UNLCK, | |
2372 | .fl_end = OFFSET_MAX, | |
2373 | }; | |
2374 | ||
2375 | if (!file_inode(filp)->i_flock) | |
2376 | return; | |
2377 | ||
2378 | if (filp->f_op->flock) | |
2379 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2380 | else | |
2381 | flock_lock_file(filp, &fl); | |
2382 | ||
2383 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2384 | fl.fl_ops->fl_release_private(&fl); | |
2385 | } | |
2386 | ||
1da177e4 LT |
2387 | /* |
2388 | * This function is called on the last close of an open file. | |
2389 | */ | |
78ed8a13 | 2390 | void locks_remove_file(struct file *filp) |
1da177e4 | 2391 | { |
496ad9aa | 2392 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2393 | struct file_lock *fl; |
2394 | struct file_lock **before; | |
ed9814d8 | 2395 | LIST_HEAD(dispose); |
1da177e4 | 2396 | |
dd459bb1 | 2397 | /* remove any OFD locks */ |
73a8f5f7 | 2398 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2399 | |
dd459bb1 JL |
2400 | /* remove flock locks */ |
2401 | locks_remove_flock(filp); | |
2402 | ||
2403 | if (!inode->i_flock) | |
2404 | return; | |
1da177e4 | 2405 | |
1c8c601a | 2406 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2407 | before = &inode->i_flock; |
2408 | ||
2409 | while ((fl = *before) != NULL) { | |
2410 | if (fl->fl_file == filp) { | |
1da177e4 | 2411 | if (IS_LEASE(fl)) { |
c45198ed | 2412 | lease_modify(before, F_UNLCK, &dispose); |
1da177e4 LT |
2413 | continue; |
2414 | } | |
8c3cac5e JL |
2415 | |
2416 | /* | |
2417 | * There's a leftover lock on the list of a type that | |
2418 | * we didn't expect to see. Most likely a classic | |
2419 | * POSIX lock that ended up not getting released | |
2420 | * properly, or that raced onto the list somehow. Log | |
2421 | * some info about it and then just remove it from | |
2422 | * the list. | |
2423 | */ | |
dd459bb1 | 2424 | WARN(1, "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", |
8c3cac5e JL |
2425 | MAJOR(inode->i_sb->s_dev), |
2426 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2427 | fl->fl_type, fl->fl_flags, | |
2428 | fl->fl_start, fl->fl_end); | |
2429 | ||
ed9814d8 | 2430 | locks_delete_lock(before, &dispose); |
8c3cac5e | 2431 | continue; |
1da177e4 LT |
2432 | } |
2433 | before = &fl->fl_next; | |
2434 | } | |
1c8c601a | 2435 | spin_unlock(&inode->i_lock); |
ed9814d8 | 2436 | locks_dispose_list(&dispose); |
1da177e4 LT |
2437 | } |
2438 | ||
1da177e4 LT |
2439 | /** |
2440 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2441 | * @waiter: the lock which was waiting |
2442 | * | |
2443 | * lockd needs to block waiting for locks. | |
2444 | */ | |
64a318ee | 2445 | int |
f891a29f | 2446 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2447 | { |
64a318ee BF |
2448 | int status = 0; |
2449 | ||
7b2296af | 2450 | spin_lock(&blocked_lock_lock); |
5996a298 | 2451 | if (waiter->fl_next) |
1da177e4 | 2452 | __locks_delete_block(waiter); |
64a318ee BF |
2453 | else |
2454 | status = -ENOENT; | |
7b2296af | 2455 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2456 | return status; |
1da177e4 | 2457 | } |
1da177e4 LT |
2458 | EXPORT_SYMBOL(posix_unblock_lock); |
2459 | ||
9b9d2ab4 ME |
2460 | /** |
2461 | * vfs_cancel_lock - file byte range unblock lock | |
2462 | * @filp: The file to apply the unblock to | |
2463 | * @fl: The lock to be unblocked | |
2464 | * | |
2465 | * Used by lock managers to cancel blocked requests | |
2466 | */ | |
2467 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2468 | { | |
72c2d531 | 2469 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2470 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2471 | return 0; | |
2472 | } | |
2473 | ||
2474 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2475 | ||
7f8ada98 | 2476 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2477 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2478 | #include <linux/seq_file.h> |
2479 | ||
7012b02a JL |
2480 | struct locks_iterator { |
2481 | int li_cpu; | |
2482 | loff_t li_pos; | |
2483 | }; | |
2484 | ||
7f8ada98 | 2485 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2486 | loff_t id, char *pfx) |
1da177e4 LT |
2487 | { |
2488 | struct inode *inode = NULL; | |
ab1f1611 VG |
2489 | unsigned int fl_pid; |
2490 | ||
2491 | if (fl->fl_nspid) | |
6c5f3e7b | 2492 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2493 | else |
2494 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2495 | |
2496 | if (fl->fl_file != NULL) | |
496ad9aa | 2497 | inode = file_inode(fl->fl_file); |
1da177e4 | 2498 | |
99dc8292 | 2499 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2500 | if (IS_POSIX(fl)) { |
c918d42a | 2501 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2502 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2503 | else if (IS_OFDLCK(fl)) |
5315c26a | 2504 | seq_puts(f, "OFDLCK"); |
c918d42a | 2505 | else |
5315c26a | 2506 | seq_puts(f, "POSIX "); |
c918d42a JL |
2507 | |
2508 | seq_printf(f, " %s ", | |
1da177e4 | 2509 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2510 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2511 | } else if (IS_FLOCK(fl)) { |
2512 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2513 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2514 | } else { |
5315c26a | 2515 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2516 | } |
2517 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2518 | if (fl->fl_flags & FL_DELEG) |
2519 | seq_puts(f, "DELEG "); | |
2520 | else | |
2521 | seq_puts(f, "LEASE "); | |
2522 | ||
ab83fa4b | 2523 | if (lease_breaking(fl)) |
5315c26a | 2524 | seq_puts(f, "BREAKING "); |
1da177e4 | 2525 | else if (fl->fl_file) |
5315c26a | 2526 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2527 | else |
5315c26a | 2528 | seq_puts(f, "BREAKER "); |
1da177e4 | 2529 | } else { |
5315c26a | 2530 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2531 | } |
2532 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2533 | seq_printf(f, "%s ", |
1da177e4 LT |
2534 | (fl->fl_type & LOCK_READ) |
2535 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2536 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2537 | } else { | |
7f8ada98 | 2538 | seq_printf(f, "%s ", |
ab83fa4b | 2539 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2540 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2541 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2542 | } |
2543 | if (inode) { | |
2544 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2545 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2546 | inode->i_sb->s_id, inode->i_ino); |
2547 | #else | |
2548 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2549 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2550 | MAJOR(inode->i_sb->s_dev), |
2551 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2552 | #endif | |
2553 | } else { | |
ab1f1611 | 2554 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2555 | } |
2556 | if (IS_POSIX(fl)) { | |
2557 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2558 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2559 | else |
7f8ada98 | 2560 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2561 | } else { |
5315c26a | 2562 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2563 | } |
2564 | } | |
2565 | ||
7f8ada98 | 2566 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2567 | { |
7012b02a | 2568 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2569 | struct file_lock *fl, *bfl; |
1da177e4 | 2570 | |
139ca04e | 2571 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2572 | |
7012b02a | 2573 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2574 | |
7f8ada98 | 2575 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2576 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2577 | |
7f8ada98 PE |
2578 | return 0; |
2579 | } | |
1da177e4 | 2580 | |
7f8ada98 | 2581 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2582 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2583 | { |
7012b02a | 2584 | struct locks_iterator *iter = f->private; |
99dc8292 | 2585 | |
7012b02a JL |
2586 | iter->li_pos = *pos + 1; |
2587 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2588 | spin_lock(&blocked_lock_lock); |
7012b02a | 2589 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2590 | } |
1da177e4 | 2591 | |
7f8ada98 PE |
2592 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2593 | { | |
7012b02a JL |
2594 | struct locks_iterator *iter = f->private; |
2595 | ||
2596 | ++iter->li_pos; | |
2597 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2598 | } |
1da177e4 | 2599 | |
7f8ada98 | 2600 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2601 | __releases(&blocked_lock_lock) |
7f8ada98 | 2602 | { |
7b2296af | 2603 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2604 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2605 | } |
2606 | ||
d8ba7a36 | 2607 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2608 | .start = locks_start, |
2609 | .next = locks_next, | |
2610 | .stop = locks_stop, | |
2611 | .show = locks_show, | |
2612 | }; | |
d8ba7a36 AD |
2613 | |
2614 | static int locks_open(struct inode *inode, struct file *filp) | |
2615 | { | |
7012b02a JL |
2616 | return seq_open_private(filp, &locks_seq_operations, |
2617 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2618 | } |
2619 | ||
2620 | static const struct file_operations proc_locks_operations = { | |
2621 | .open = locks_open, | |
2622 | .read = seq_read, | |
2623 | .llseek = seq_lseek, | |
99dc8292 | 2624 | .release = seq_release_private, |
d8ba7a36 AD |
2625 | }; |
2626 | ||
2627 | static int __init proc_locks_init(void) | |
2628 | { | |
2629 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2630 | return 0; | |
2631 | } | |
2632 | module_init(proc_locks_init); | |
7f8ada98 PE |
2633 | #endif |
2634 | ||
1da177e4 LT |
2635 | static int __init filelock_init(void) |
2636 | { | |
7012b02a JL |
2637 | int i; |
2638 | ||
1da177e4 | 2639 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2640 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2641 | ||
7012b02a JL |
2642 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2643 | ||
2644 | for_each_possible_cpu(i) | |
2645 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2646 | ||
1da177e4 LT |
2647 | return 0; |
2648 | } | |
2649 | ||
2650 | core_initcall(filelock_init); |