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> |
1da177e4 | 130 | |
1da177e4 LT |
131 | #include <asm/uaccess.h> |
132 | ||
133 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
134 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
135 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) | |
136 | ||
ab83fa4b BF |
137 | static bool lease_breaking(struct file_lock *fl) |
138 | { | |
778fc546 BF |
139 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
140 | } | |
141 | ||
142 | static int target_leasetype(struct file_lock *fl) | |
143 | { | |
144 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
145 | return F_UNLCK; | |
146 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
147 | return F_RDLCK; | |
148 | return fl->fl_type; | |
ab83fa4b BF |
149 | } |
150 | ||
1da177e4 LT |
151 | int leases_enable = 1; |
152 | int lease_break_time = 45; | |
153 | ||
154 | #define for_each_lock(inode, lockp) \ | |
155 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
156 | ||
1c8c601a JL |
157 | /* |
158 | * The global file_lock_list is only used for displaying /proc/locks. Protected | |
159 | * by the file_lock_lock. | |
160 | */ | |
139ca04e | 161 | static HLIST_HEAD(file_lock_list); |
88974691 | 162 | |
1c8c601a | 163 | /* |
48f74186 JL |
164 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
165 | * It is protected by file_lock_lock. | |
166 | * | |
167 | * We hash locks by lockowner in order to optimize searching for the lock a | |
168 | * particular lockowner is waiting on. | |
169 | * | |
170 | * FIXME: make this value scale via some heuristic? We generally will want more | |
171 | * buckets when we have more lockowners holding locks, but that's a little | |
172 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 173 | */ |
48f74186 JL |
174 | #define BLOCKED_HASH_BITS 7 |
175 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 176 | |
1c8c601a | 177 | /* |
48f74186 | 178 | * This lock protects the blocked_hash and the file_lock_list. Generally, if |
1c8c601a JL |
179 | * you're accessing one of those lists, you want to be holding this lock. |
180 | * | |
181 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
182 | * pointer for file_lock structures that are acting as lock requests (in | |
183 | * contrast to those that are acting as records of acquired locks). | |
184 | * | |
185 | * Note that when we acquire this lock in order to change the above fields, | |
186 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
187 | * protected by this lock, we can skip acquiring it iff we already hold the | |
188 | * i_lock. | |
189 | * | |
190 | * In particular, adding an entry to the fl_block list requires that you hold | |
191 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
192 | * an entry from the list however only requires the file_lock_lock. | |
193 | */ | |
72f98e72 | 194 | static DEFINE_SPINLOCK(file_lock_lock); |
1da177e4 | 195 | |
e18b890b | 196 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 197 | |
ee19cc40 | 198 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 199 | { |
139ca04e | 200 | INIT_HLIST_NODE(&fl->fl_link); |
ee19cc40 MS |
201 | INIT_LIST_HEAD(&fl->fl_block); |
202 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
203 | } |
204 | ||
1da177e4 | 205 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 206 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 207 | { |
ee19cc40 | 208 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
209 | |
210 | if (fl) | |
ee19cc40 | 211 | locks_init_lock_heads(fl); |
a51cb91d MS |
212 | |
213 | return fl; | |
1da177e4 | 214 | } |
c5b1f0d9 | 215 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 216 | |
a9e61e25 | 217 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
218 | { |
219 | if (fl->fl_ops) { | |
220 | if (fl->fl_ops->fl_release_private) | |
221 | fl->fl_ops->fl_release_private(fl); | |
222 | fl->fl_ops = NULL; | |
223 | } | |
068535f1 | 224 | fl->fl_lmops = NULL; |
47831f35 TM |
225 | |
226 | } | |
a9e61e25 | 227 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 228 | |
1da177e4 | 229 | /* Free a lock which is not in use. */ |
05fa3135 | 230 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 231 | { |
5ce29646 MS |
232 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
233 | BUG_ON(!list_empty(&fl->fl_block)); | |
139ca04e | 234 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 235 | |
47831f35 | 236 | locks_release_private(fl); |
1da177e4 LT |
237 | kmem_cache_free(filelock_cache, fl); |
238 | } | |
05fa3135 | 239 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
240 | |
241 | void locks_init_lock(struct file_lock *fl) | |
242 | { | |
ee19cc40 MS |
243 | memset(fl, 0, sizeof(struct file_lock)); |
244 | locks_init_lock_heads(fl); | |
1da177e4 LT |
245 | } |
246 | ||
247 | EXPORT_SYMBOL(locks_init_lock); | |
248 | ||
47831f35 TM |
249 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
250 | { | |
251 | if (fl->fl_ops) { | |
252 | if (fl->fl_ops->fl_copy_lock) | |
253 | fl->fl_ops->fl_copy_lock(new, fl); | |
254 | new->fl_ops = fl->fl_ops; | |
255 | } | |
bb8430a2 | 256 | if (fl->fl_lmops) |
47831f35 | 257 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
258 | } |
259 | ||
1da177e4 LT |
260 | /* |
261 | * Initialize a new lock from an existing file_lock structure. | |
262 | */ | |
1a747ee0 | 263 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
264 | { |
265 | new->fl_owner = fl->fl_owner; | |
266 | new->fl_pid = fl->fl_pid; | |
0996905f | 267 | new->fl_file = NULL; |
1da177e4 LT |
268 | new->fl_flags = fl->fl_flags; |
269 | new->fl_type = fl->fl_type; | |
270 | new->fl_start = fl->fl_start; | |
271 | new->fl_end = fl->fl_end; | |
0996905f TM |
272 | new->fl_ops = NULL; |
273 | new->fl_lmops = NULL; | |
274 | } | |
3dd7b71c | 275 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
276 | |
277 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
278 | { | |
279 | locks_release_private(new); | |
280 | ||
281 | __locks_copy_lock(new, fl); | |
282 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
283 | new->fl_ops = fl->fl_ops; |
284 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
285 | |
286 | locks_copy_private(new, fl); | |
1da177e4 LT |
287 | } |
288 | ||
289 | EXPORT_SYMBOL(locks_copy_lock); | |
290 | ||
291 | static inline int flock_translate_cmd(int cmd) { | |
292 | if (cmd & LOCK_MAND) | |
293 | return cmd & (LOCK_MAND | LOCK_RW); | |
294 | switch (cmd) { | |
295 | case LOCK_SH: | |
296 | return F_RDLCK; | |
297 | case LOCK_EX: | |
298 | return F_WRLCK; | |
299 | case LOCK_UN: | |
300 | return F_UNLCK; | |
301 | } | |
302 | return -EINVAL; | |
303 | } | |
304 | ||
305 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
306 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
307 | unsigned int cmd) | |
308 | { | |
309 | struct file_lock *fl; | |
310 | int type = flock_translate_cmd(cmd); | |
311 | if (type < 0) | |
312 | return type; | |
313 | ||
314 | fl = locks_alloc_lock(); | |
315 | if (fl == NULL) | |
316 | return -ENOMEM; | |
317 | ||
318 | fl->fl_file = filp; | |
319 | fl->fl_pid = current->tgid; | |
320 | fl->fl_flags = FL_FLOCK; | |
321 | fl->fl_type = type; | |
322 | fl->fl_end = OFFSET_MAX; | |
323 | ||
324 | *lock = fl; | |
325 | return 0; | |
326 | } | |
327 | ||
0ec4f431 | 328 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
329 | { |
330 | switch (type) { | |
331 | case F_RDLCK: | |
332 | case F_WRLCK: | |
333 | case F_UNLCK: | |
334 | fl->fl_type = type; | |
335 | break; | |
336 | default: | |
337 | return -EINVAL; | |
338 | } | |
339 | return 0; | |
340 | } | |
341 | ||
342 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
343 | * style lock. | |
344 | */ | |
345 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
346 | struct flock *l) | |
347 | { | |
348 | off_t start, end; | |
349 | ||
350 | switch (l->l_whence) { | |
f5579f8c | 351 | case SEEK_SET: |
1da177e4 LT |
352 | start = 0; |
353 | break; | |
f5579f8c | 354 | case SEEK_CUR: |
1da177e4 LT |
355 | start = filp->f_pos; |
356 | break; | |
f5579f8c | 357 | case SEEK_END: |
496ad9aa | 358 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
359 | break; |
360 | default: | |
361 | return -EINVAL; | |
362 | } | |
363 | ||
364 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
365 | POSIX-2001 defines it. */ | |
366 | start += l->l_start; | |
4c780a46 TM |
367 | if (start < 0) |
368 | return -EINVAL; | |
369 | fl->fl_end = OFFSET_MAX; | |
370 | if (l->l_len > 0) { | |
371 | end = start + l->l_len - 1; | |
372 | fl->fl_end = end; | |
373 | } else if (l->l_len < 0) { | |
1da177e4 | 374 | end = start - 1; |
4c780a46 | 375 | fl->fl_end = end; |
1da177e4 | 376 | start += l->l_len; |
4c780a46 TM |
377 | if (start < 0) |
378 | return -EINVAL; | |
1da177e4 | 379 | } |
1da177e4 | 380 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
381 | if (fl->fl_end < fl->fl_start) |
382 | return -EOVERFLOW; | |
1da177e4 LT |
383 | |
384 | fl->fl_owner = current->files; | |
385 | fl->fl_pid = current->tgid; | |
386 | fl->fl_file = filp; | |
387 | fl->fl_flags = FL_POSIX; | |
388 | fl->fl_ops = NULL; | |
389 | fl->fl_lmops = NULL; | |
390 | ||
391 | return assign_type(fl, l->l_type); | |
392 | } | |
393 | ||
394 | #if BITS_PER_LONG == 32 | |
395 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
396 | struct flock64 *l) | |
397 | { | |
398 | loff_t start; | |
399 | ||
400 | switch (l->l_whence) { | |
f5579f8c | 401 | case SEEK_SET: |
1da177e4 LT |
402 | start = 0; |
403 | break; | |
f5579f8c | 404 | case SEEK_CUR: |
1da177e4 LT |
405 | start = filp->f_pos; |
406 | break; | |
f5579f8c | 407 | case SEEK_END: |
496ad9aa | 408 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
409 | break; |
410 | default: | |
411 | return -EINVAL; | |
412 | } | |
413 | ||
4c780a46 TM |
414 | start += l->l_start; |
415 | if (start < 0) | |
1da177e4 | 416 | return -EINVAL; |
4c780a46 TM |
417 | fl->fl_end = OFFSET_MAX; |
418 | if (l->l_len > 0) { | |
419 | fl->fl_end = start + l->l_len - 1; | |
420 | } else if (l->l_len < 0) { | |
421 | fl->fl_end = start - 1; | |
422 | start += l->l_len; | |
423 | if (start < 0) | |
424 | return -EINVAL; | |
425 | } | |
1da177e4 | 426 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
427 | if (fl->fl_end < fl->fl_start) |
428 | return -EOVERFLOW; | |
1da177e4 LT |
429 | |
430 | fl->fl_owner = current->files; | |
431 | fl->fl_pid = current->tgid; | |
432 | fl->fl_file = filp; | |
433 | fl->fl_flags = FL_POSIX; | |
434 | fl->fl_ops = NULL; | |
435 | fl->fl_lmops = NULL; | |
436 | ||
f32cb532 | 437 | return assign_type(fl, l->l_type); |
1da177e4 LT |
438 | } |
439 | #endif | |
440 | ||
441 | /* default lease lock manager operations */ | |
442 | static void lease_break_callback(struct file_lock *fl) | |
443 | { | |
444 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
445 | } | |
446 | ||
7b021967 | 447 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 448 | .lm_break = lease_break_callback, |
8fb47a4f | 449 | .lm_change = lease_modify, |
1da177e4 LT |
450 | }; |
451 | ||
452 | /* | |
453 | * Initialize a lease, use the default lock manager operations | |
454 | */ | |
0ec4f431 | 455 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 456 | { |
75dff55a TM |
457 | if (assign_type(fl, type) != 0) |
458 | return -EINVAL; | |
459 | ||
1da177e4 LT |
460 | fl->fl_owner = current->files; |
461 | fl->fl_pid = current->tgid; | |
462 | ||
463 | fl->fl_file = filp; | |
464 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
465 | fl->fl_start = 0; |
466 | fl->fl_end = OFFSET_MAX; | |
467 | fl->fl_ops = NULL; | |
468 | fl->fl_lmops = &lease_manager_ops; | |
469 | return 0; | |
470 | } | |
471 | ||
472 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 473 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
474 | { |
475 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 476 | int error = -ENOMEM; |
1da177e4 LT |
477 | |
478 | if (fl == NULL) | |
e32b8ee2 | 479 | return ERR_PTR(error); |
1da177e4 LT |
480 | |
481 | error = lease_init(filp, type, fl); | |
75dff55a TM |
482 | if (error) { |
483 | locks_free_lock(fl); | |
e32b8ee2 | 484 | return ERR_PTR(error); |
75dff55a | 485 | } |
e32b8ee2 | 486 | return fl; |
1da177e4 LT |
487 | } |
488 | ||
489 | /* Check if two locks overlap each other. | |
490 | */ | |
491 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
492 | { | |
493 | return ((fl1->fl_end >= fl2->fl_start) && | |
494 | (fl2->fl_end >= fl1->fl_start)); | |
495 | } | |
496 | ||
497 | /* | |
498 | * Check whether two locks have the same owner. | |
499 | */ | |
33443c42 | 500 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 501 | { |
8fb47a4f | 502 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 503 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 504 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
505 | return fl1->fl_owner == fl2->fl_owner; |
506 | } | |
507 | ||
88974691 JL |
508 | static inline void |
509 | locks_insert_global_locks(struct file_lock *fl) | |
510 | { | |
1c8c601a | 511 | spin_lock(&file_lock_lock); |
139ca04e | 512 | hlist_add_head(&fl->fl_link, &file_lock_list); |
1c8c601a | 513 | spin_unlock(&file_lock_lock); |
88974691 JL |
514 | } |
515 | ||
516 | static inline void | |
517 | locks_delete_global_locks(struct file_lock *fl) | |
518 | { | |
1c8c601a | 519 | spin_lock(&file_lock_lock); |
139ca04e | 520 | hlist_del_init(&fl->fl_link); |
1c8c601a | 521 | spin_unlock(&file_lock_lock); |
88974691 JL |
522 | } |
523 | ||
3999e493 JL |
524 | static unsigned long |
525 | posix_owner_key(struct file_lock *fl) | |
526 | { | |
527 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
528 | return fl->fl_lmops->lm_owner_key(fl); | |
529 | return (unsigned long)fl->fl_owner; | |
530 | } | |
531 | ||
88974691 JL |
532 | static inline void |
533 | locks_insert_global_blocked(struct file_lock *waiter) | |
534 | { | |
3999e493 | 535 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
536 | } |
537 | ||
538 | static inline void | |
539 | locks_delete_global_blocked(struct file_lock *waiter) | |
540 | { | |
48f74186 | 541 | hash_del(&waiter->fl_link); |
88974691 JL |
542 | } |
543 | ||
1da177e4 LT |
544 | /* Remove waiter from blocker's block list. |
545 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a JL |
546 | * |
547 | * Must be called with file_lock_lock held. | |
1da177e4 | 548 | */ |
33443c42 | 549 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 550 | { |
88974691 | 551 | locks_delete_global_blocked(waiter); |
1da177e4 | 552 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
553 | waiter->fl_next = NULL; |
554 | } | |
555 | ||
1a9e64a7 | 556 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 557 | { |
1c8c601a | 558 | spin_lock(&file_lock_lock); |
1da177e4 | 559 | __locks_delete_block(waiter); |
1c8c601a | 560 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
561 | } |
562 | ||
563 | /* Insert waiter into blocker's block list. | |
564 | * We use a circular list so that processes can be easily woken up in | |
565 | * the order they blocked. The documentation doesn't require this but | |
566 | * it seems like the reasonable thing to do. | |
1c8c601a | 567 | * |
4e8c765d JL |
568 | * Must be called with both the i_lock and file_lock_lock held. The fl_block |
569 | * list itself is protected by the file_lock_list, but by ensuring that the | |
570 | * i_lock is also held on insertions we can avoid taking the file_lock_lock | |
571 | * in some cases when we see that the fl_block list is empty. | |
1da177e4 | 572 | */ |
1c8c601a JL |
573 | static void __locks_insert_block(struct file_lock *blocker, |
574 | struct file_lock *waiter) | |
1da177e4 | 575 | { |
6dc0fe8f | 576 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 577 | waiter->fl_next = blocker; |
88974691 | 578 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
1da177e4 | 579 | if (IS_POSIX(blocker)) |
1c8c601a JL |
580 | locks_insert_global_blocked(waiter); |
581 | } | |
582 | ||
583 | /* Must be called with i_lock held. */ | |
584 | static void locks_insert_block(struct file_lock *blocker, | |
585 | struct file_lock *waiter) | |
586 | { | |
587 | spin_lock(&file_lock_lock); | |
588 | __locks_insert_block(blocker, waiter); | |
589 | spin_unlock(&file_lock_lock); | |
1da177e4 LT |
590 | } |
591 | ||
1cb36012 JL |
592 | /* |
593 | * Wake up processes blocked waiting for blocker. | |
594 | * | |
1c8c601a | 595 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
596 | */ |
597 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
598 | { | |
4e8c765d JL |
599 | /* |
600 | * Avoid taking global lock if list is empty. This is safe since new | |
601 | * blocked requests are only added to the list under the i_lock, and | |
602 | * the i_lock is always held here. Note that removal from the fl_block | |
603 | * list does not require the i_lock, so we must recheck list_empty() | |
604 | * after acquiring the file_lock_lock. | |
605 | */ | |
606 | if (list_empty(&blocker->fl_block)) | |
607 | return; | |
608 | ||
1c8c601a | 609 | spin_lock(&file_lock_lock); |
1da177e4 | 610 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
611 | struct file_lock *waiter; |
612 | ||
613 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
614 | struct file_lock, fl_block); |
615 | __locks_delete_block(waiter); | |
8fb47a4f BF |
616 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
617 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
618 | else |
619 | wake_up(&waiter->fl_wait); | |
620 | } | |
1c8c601a | 621 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
622 | } |
623 | ||
624 | /* Insert file lock fl into an inode's lock list at the position indicated | |
625 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
626 | * |
627 | * Must be called with the i_lock held! | |
1da177e4 LT |
628 | */ |
629 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
630 | { | |
ab1f1611 VG |
631 | fl->fl_nspid = get_pid(task_tgid(current)); |
632 | ||
1da177e4 LT |
633 | /* insert into file's list */ |
634 | fl->fl_next = *pos; | |
635 | *pos = fl; | |
88974691 JL |
636 | |
637 | locks_insert_global_locks(fl); | |
1da177e4 LT |
638 | } |
639 | ||
640 | /* | |
641 | * Delete a lock and then free it. | |
642 | * Wake up processes that are blocked waiting for this lock, | |
643 | * notify the FS that the lock has been cleared and | |
644 | * finally free the lock. | |
1c8c601a JL |
645 | * |
646 | * Must be called with the i_lock held! | |
1da177e4 LT |
647 | */ |
648 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
649 | { | |
650 | struct file_lock *fl = *thisfl_p; | |
651 | ||
88974691 JL |
652 | locks_delete_global_locks(fl); |
653 | ||
1da177e4 LT |
654 | *thisfl_p = fl->fl_next; |
655 | fl->fl_next = NULL; | |
1da177e4 | 656 | |
ab1f1611 VG |
657 | if (fl->fl_nspid) { |
658 | put_pid(fl->fl_nspid); | |
659 | fl->fl_nspid = NULL; | |
660 | } | |
661 | ||
1da177e4 LT |
662 | locks_wake_up_blocks(fl); |
663 | locks_free_lock(fl); | |
664 | } | |
665 | ||
666 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
667 | * checks for shared/exclusive status of overlapping locks. | |
668 | */ | |
669 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
670 | { | |
671 | if (sys_fl->fl_type == F_WRLCK) | |
672 | return 1; | |
673 | if (caller_fl->fl_type == F_WRLCK) | |
674 | return 1; | |
675 | return 0; | |
676 | } | |
677 | ||
678 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
679 | * checking before calling the locks_conflict(). | |
680 | */ | |
681 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
682 | { | |
683 | /* POSIX locks owned by the same process do not conflict with | |
684 | * each other. | |
685 | */ | |
686 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
687 | return (0); | |
688 | ||
689 | /* Check whether they overlap */ | |
690 | if (!locks_overlap(caller_fl, sys_fl)) | |
691 | return 0; | |
692 | ||
693 | return (locks_conflict(caller_fl, sys_fl)); | |
694 | } | |
695 | ||
696 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
697 | * checking before calling the locks_conflict(). | |
698 | */ | |
699 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
700 | { | |
701 | /* FLOCK locks referring to the same filp do not conflict with | |
702 | * each other. | |
703 | */ | |
704 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
705 | return (0); | |
706 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
707 | return 0; | |
708 | ||
709 | return (locks_conflict(caller_fl, sys_fl)); | |
710 | } | |
711 | ||
6d34ac19 | 712 | void |
9d6a8c5c | 713 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
714 | { |
715 | struct file_lock *cfl; | |
1c8c601a | 716 | struct inode *inode = file_inode(filp); |
1da177e4 | 717 | |
1c8c601a | 718 | spin_lock(&inode->i_lock); |
496ad9aa | 719 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
720 | if (!IS_POSIX(cfl)) |
721 | continue; | |
b842e240 | 722 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
723 | break; |
724 | } | |
ab1f1611 | 725 | if (cfl) { |
9d6a8c5c | 726 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 727 | if (cfl->fl_nspid) |
6c5f3e7b | 728 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 729 | } else |
129a84de | 730 | fl->fl_type = F_UNLCK; |
1c8c601a | 731 | spin_unlock(&inode->i_lock); |
6d34ac19 | 732 | return; |
1da177e4 | 733 | } |
1da177e4 LT |
734 | EXPORT_SYMBOL(posix_test_lock); |
735 | ||
b533184f BF |
736 | /* |
737 | * Deadlock detection: | |
738 | * | |
739 | * We attempt to detect deadlocks that are due purely to posix file | |
740 | * locks. | |
1da177e4 | 741 | * |
b533184f BF |
742 | * We assume that a task can be waiting for at most one lock at a time. |
743 | * So for any acquired lock, the process holding that lock may be | |
744 | * waiting on at most one other lock. That lock in turns may be held by | |
745 | * someone waiting for at most one other lock. Given a requested lock | |
746 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
747 | * follow this chain of waiters to ensure we are not about to create a | |
748 | * cycle. | |
1da177e4 | 749 | * |
b533184f BF |
750 | * Since we do this before we ever put a process to sleep on a lock, we |
751 | * are ensured that there is never a cycle; that is what guarantees that | |
752 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 753 | * |
b533184f BF |
754 | * Note: the above assumption may not be true when handling lock |
755 | * requests from a broken NFS client. It may also fail in the presence | |
756 | * of tasks (such as posix threads) sharing the same open file table. | |
757 | * | |
758 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 759 | */ |
97855b49 BF |
760 | |
761 | #define MAX_DEADLK_ITERATIONS 10 | |
762 | ||
b533184f BF |
763 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
764 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
765 | { | |
766 | struct file_lock *fl; | |
767 | ||
3999e493 | 768 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
769 | if (posix_same_owner(fl, block_fl)) |
770 | return fl->fl_next; | |
771 | } | |
772 | return NULL; | |
773 | } | |
774 | ||
1c8c601a | 775 | /* Must be called with the file_lock_lock held! */ |
b0904e14 | 776 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
777 | struct file_lock *block_fl) |
778 | { | |
97855b49 | 779 | int i = 0; |
1da177e4 | 780 | |
b533184f BF |
781 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
782 | if (i++ > MAX_DEADLK_ITERATIONS) | |
783 | return 0; | |
784 | if (posix_same_owner(caller_fl, block_fl)) | |
785 | return 1; | |
1da177e4 LT |
786 | } |
787 | return 0; | |
788 | } | |
789 | ||
1da177e4 | 790 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 791 | * after any leases, but before any posix locks. |
f475ae95 TM |
792 | * |
793 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
794 | * whether or not a lock was successfully freed by testing the return | |
795 | * value for -ENOENT. | |
1da177e4 | 796 | */ |
993dfa87 | 797 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 798 | { |
993dfa87 | 799 | struct file_lock *new_fl = NULL; |
1da177e4 | 800 | struct file_lock **before; |
496ad9aa | 801 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
802 | int error = 0; |
803 | int found = 0; | |
804 | ||
b89f4321 | 805 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 806 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
807 | if (!new_fl) |
808 | return -ENOMEM; | |
84d535ad PE |
809 | } |
810 | ||
1c8c601a | 811 | spin_lock(&inode->i_lock); |
b89f4321 AB |
812 | if (request->fl_flags & FL_ACCESS) |
813 | goto find_conflict; | |
814 | ||
1da177e4 LT |
815 | for_each_lock(inode, before) { |
816 | struct file_lock *fl = *before; | |
817 | if (IS_POSIX(fl)) | |
818 | break; | |
819 | if (IS_LEASE(fl)) | |
820 | continue; | |
821 | if (filp != fl->fl_file) | |
822 | continue; | |
993dfa87 | 823 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
824 | goto out; |
825 | found = 1; | |
826 | locks_delete_lock(before); | |
827 | break; | |
828 | } | |
1da177e4 | 829 | |
f475ae95 TM |
830 | if (request->fl_type == F_UNLCK) { |
831 | if ((request->fl_flags & FL_EXISTS) && !found) | |
832 | error = -ENOENT; | |
993dfa87 | 833 | goto out; |
f475ae95 | 834 | } |
1da177e4 LT |
835 | |
836 | /* | |
837 | * If a higher-priority process was blocked on the old file lock, | |
838 | * give it the opportunity to lock the file. | |
839 | */ | |
b89f4321 | 840 | if (found) { |
1c8c601a | 841 | spin_unlock(&inode->i_lock); |
def01bc5 | 842 | cond_resched(); |
1c8c601a | 843 | spin_lock(&inode->i_lock); |
b89f4321 | 844 | } |
1da177e4 | 845 | |
f07f18dd | 846 | find_conflict: |
1da177e4 LT |
847 | for_each_lock(inode, before) { |
848 | struct file_lock *fl = *before; | |
849 | if (IS_POSIX(fl)) | |
850 | break; | |
851 | if (IS_LEASE(fl)) | |
852 | continue; | |
993dfa87 | 853 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
854 | continue; |
855 | error = -EAGAIN; | |
bde74e4b MS |
856 | if (!(request->fl_flags & FL_SLEEP)) |
857 | goto out; | |
858 | error = FILE_LOCK_DEFERRED; | |
859 | locks_insert_block(fl, request); | |
1da177e4 LT |
860 | goto out; |
861 | } | |
f07f18dd TM |
862 | if (request->fl_flags & FL_ACCESS) |
863 | goto out; | |
993dfa87 | 864 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 865 | locks_insert_lock(before, new_fl); |
993dfa87 | 866 | new_fl = NULL; |
9cedc194 | 867 | error = 0; |
1da177e4 LT |
868 | |
869 | out: | |
1c8c601a | 870 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
871 | if (new_fl) |
872 | locks_free_lock(new_fl); | |
1da177e4 LT |
873 | return error; |
874 | } | |
875 | ||
150b3934 | 876 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
877 | { |
878 | struct file_lock *fl; | |
39005d02 MS |
879 | struct file_lock *new_fl = NULL; |
880 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
881 | struct file_lock *left = NULL; |
882 | struct file_lock *right = NULL; | |
883 | struct file_lock **before; | |
b9746ef8 JL |
884 | int error; |
885 | bool added = false; | |
1da177e4 LT |
886 | |
887 | /* | |
888 | * We may need two file_lock structures for this operation, | |
889 | * so we get them in advance to avoid races. | |
39005d02 MS |
890 | * |
891 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 892 | */ |
39005d02 MS |
893 | if (!(request->fl_flags & FL_ACCESS) && |
894 | (request->fl_type != F_UNLCK || | |
895 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
896 | new_fl = locks_alloc_lock(); | |
897 | new_fl2 = locks_alloc_lock(); | |
898 | } | |
1da177e4 | 899 | |
1c8c601a | 900 | spin_lock(&inode->i_lock); |
1cb36012 JL |
901 | /* |
902 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
903 | * there are any, either return error or put the request on the | |
48f74186 | 904 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 905 | */ |
1da177e4 LT |
906 | if (request->fl_type != F_UNLCK) { |
907 | for_each_lock(inode, before) { | |
526985b9 | 908 | fl = *before; |
1da177e4 LT |
909 | if (!IS_POSIX(fl)) |
910 | continue; | |
911 | if (!posix_locks_conflict(request, fl)) | |
912 | continue; | |
5842add2 | 913 | if (conflock) |
1a747ee0 | 914 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
915 | error = -EAGAIN; |
916 | if (!(request->fl_flags & FL_SLEEP)) | |
917 | goto out; | |
1c8c601a JL |
918 | /* |
919 | * Deadlock detection and insertion into the blocked | |
920 | * locks list must be done while holding the same lock! | |
921 | */ | |
1da177e4 | 922 | error = -EDEADLK; |
1c8c601a JL |
923 | spin_lock(&file_lock_lock); |
924 | if (likely(!posix_locks_deadlock(request, fl))) { | |
925 | error = FILE_LOCK_DEFERRED; | |
926 | __locks_insert_block(fl, request); | |
927 | } | |
928 | spin_unlock(&file_lock_lock); | |
1da177e4 LT |
929 | goto out; |
930 | } | |
931 | } | |
932 | ||
933 | /* If we're just looking for a conflict, we're done. */ | |
934 | error = 0; | |
935 | if (request->fl_flags & FL_ACCESS) | |
936 | goto out; | |
937 | ||
1da177e4 | 938 | /* |
1da177e4 LT |
939 | * Find the first old lock with the same owner as the new lock. |
940 | */ | |
941 | ||
942 | before = &inode->i_flock; | |
943 | ||
944 | /* First skip locks owned by other processes. */ | |
945 | while ((fl = *before) && (!IS_POSIX(fl) || | |
946 | !posix_same_owner(request, fl))) { | |
947 | before = &fl->fl_next; | |
948 | } | |
949 | ||
1cb36012 | 950 | /* Process locks with this owner. */ |
1da177e4 LT |
951 | while ((fl = *before) && posix_same_owner(request, fl)) { |
952 | /* Detect adjacent or overlapping regions (if same lock type) | |
953 | */ | |
954 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
955 | /* In all comparisons of start vs end, use |
956 | * "start - 1" rather than "end + 1". If end | |
957 | * is OFFSET_MAX, end + 1 will become negative. | |
958 | */ | |
1da177e4 LT |
959 | if (fl->fl_end < request->fl_start - 1) |
960 | goto next_lock; | |
961 | /* If the next lock in the list has entirely bigger | |
962 | * addresses than the new one, insert the lock here. | |
963 | */ | |
449231d6 | 964 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
965 | break; |
966 | ||
967 | /* If we come here, the new and old lock are of the | |
968 | * same type and adjacent or overlapping. Make one | |
969 | * lock yielding from the lower start address of both | |
970 | * locks to the higher end address. | |
971 | */ | |
972 | if (fl->fl_start > request->fl_start) | |
973 | fl->fl_start = request->fl_start; | |
974 | else | |
975 | request->fl_start = fl->fl_start; | |
976 | if (fl->fl_end < request->fl_end) | |
977 | fl->fl_end = request->fl_end; | |
978 | else | |
979 | request->fl_end = fl->fl_end; | |
980 | if (added) { | |
981 | locks_delete_lock(before); | |
982 | continue; | |
983 | } | |
984 | request = fl; | |
b9746ef8 | 985 | added = true; |
1da177e4 LT |
986 | } |
987 | else { | |
988 | /* Processing for different lock types is a bit | |
989 | * more complex. | |
990 | */ | |
991 | if (fl->fl_end < request->fl_start) | |
992 | goto next_lock; | |
993 | if (fl->fl_start > request->fl_end) | |
994 | break; | |
995 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 996 | added = true; |
1da177e4 LT |
997 | if (fl->fl_start < request->fl_start) |
998 | left = fl; | |
999 | /* If the next lock in the list has a higher end | |
1000 | * address than the new one, insert the new one here. | |
1001 | */ | |
1002 | if (fl->fl_end > request->fl_end) { | |
1003 | right = fl; | |
1004 | break; | |
1005 | } | |
1006 | if (fl->fl_start >= request->fl_start) { | |
1007 | /* The new lock completely replaces an old | |
1008 | * one (This may happen several times). | |
1009 | */ | |
1010 | if (added) { | |
1011 | locks_delete_lock(before); | |
1012 | continue; | |
1013 | } | |
1014 | /* Replace the old lock with the new one. | |
1015 | * Wake up anybody waiting for the old one, | |
1016 | * as the change in lock type might satisfy | |
1017 | * their needs. | |
1018 | */ | |
1019 | locks_wake_up_blocks(fl); | |
1020 | fl->fl_start = request->fl_start; | |
1021 | fl->fl_end = request->fl_end; | |
1022 | fl->fl_type = request->fl_type; | |
47831f35 TM |
1023 | locks_release_private(fl); |
1024 | locks_copy_private(fl, request); | |
1da177e4 | 1025 | request = fl; |
b9746ef8 | 1026 | added = true; |
1da177e4 LT |
1027 | } |
1028 | } | |
1029 | /* Go on to next lock. | |
1030 | */ | |
1031 | next_lock: | |
1032 | before = &fl->fl_next; | |
1033 | } | |
1034 | ||
0d9a490a | 1035 | /* |
1cb36012 JL |
1036 | * The above code only modifies existing locks in case of merging or |
1037 | * replacing. If new lock(s) need to be inserted all modifications are | |
1038 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1039 | */ |
1040 | error = -ENOLCK; /* "no luck" */ | |
1041 | if (right && left == right && !new_fl2) | |
1042 | goto out; | |
1043 | ||
1da177e4 LT |
1044 | error = 0; |
1045 | if (!added) { | |
f475ae95 TM |
1046 | if (request->fl_type == F_UNLCK) { |
1047 | if (request->fl_flags & FL_EXISTS) | |
1048 | error = -ENOENT; | |
1da177e4 | 1049 | goto out; |
f475ae95 | 1050 | } |
0d9a490a MS |
1051 | |
1052 | if (!new_fl) { | |
1053 | error = -ENOLCK; | |
1054 | goto out; | |
1055 | } | |
1da177e4 LT |
1056 | locks_copy_lock(new_fl, request); |
1057 | locks_insert_lock(before, new_fl); | |
1058 | new_fl = NULL; | |
1059 | } | |
1060 | if (right) { | |
1061 | if (left == right) { | |
1062 | /* The new lock breaks the old one in two pieces, | |
1063 | * so we have to use the second new lock. | |
1064 | */ | |
1065 | left = new_fl2; | |
1066 | new_fl2 = NULL; | |
1067 | locks_copy_lock(left, right); | |
1068 | locks_insert_lock(before, left); | |
1069 | } | |
1070 | right->fl_start = request->fl_end + 1; | |
1071 | locks_wake_up_blocks(right); | |
1072 | } | |
1073 | if (left) { | |
1074 | left->fl_end = request->fl_start - 1; | |
1075 | locks_wake_up_blocks(left); | |
1076 | } | |
1077 | out: | |
1c8c601a | 1078 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1079 | /* |
1080 | * Free any unused locks. | |
1081 | */ | |
1082 | if (new_fl) | |
1083 | locks_free_lock(new_fl); | |
1084 | if (new_fl2) | |
1085 | locks_free_lock(new_fl2); | |
1086 | return error; | |
1087 | } | |
1088 | ||
1089 | /** | |
1090 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1091 | * @filp: The file to apply the lock to | |
1092 | * @fl: The lock to be applied | |
150b3934 | 1093 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1094 | * |
1095 | * Add a POSIX style lock to a file. | |
1096 | * We merge adjacent & overlapping locks whenever possible. | |
1097 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1098 | * |
1099 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1100 | * whether or not a lock was successfully freed by testing the return | |
1101 | * value for -ENOENT. | |
1da177e4 | 1102 | */ |
150b3934 | 1103 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1104 | struct file_lock *conflock) |
1105 | { | |
496ad9aa | 1106 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1107 | } |
150b3934 | 1108 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1109 | |
1110 | /** | |
1111 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1112 | * @filp: The file to apply the lock to | |
1113 | * @fl: The lock to be applied | |
1114 | * | |
1115 | * Add a POSIX style lock to a file. | |
1116 | * We merge adjacent & overlapping locks whenever possible. | |
1117 | * POSIX locks are sorted by owner task, then by starting address | |
1118 | */ | |
1119 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1120 | { | |
1121 | int error; | |
1122 | might_sleep (); | |
1123 | for (;;) { | |
150b3934 | 1124 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1125 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1126 | break; |
1127 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1128 | if (!error) | |
1129 | continue; | |
1130 | ||
1131 | locks_delete_block(fl); | |
1132 | break; | |
1133 | } | |
1134 | return error; | |
1135 | } | |
1136 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1137 | ||
1138 | /** | |
1139 | * locks_mandatory_locked - Check for an active lock | |
1140 | * @inode: the file to check | |
1141 | * | |
1142 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1143 | * This function is called from locks_verify_locked() only. | |
1144 | */ | |
1145 | int locks_mandatory_locked(struct inode *inode) | |
1146 | { | |
1147 | fl_owner_t owner = current->files; | |
1148 | struct file_lock *fl; | |
1149 | ||
1150 | /* | |
1151 | * Search the lock list for this inode for any POSIX locks. | |
1152 | */ | |
1c8c601a | 1153 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1154 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1155 | if (!IS_POSIX(fl)) | |
1156 | continue; | |
1157 | if (fl->fl_owner != owner) | |
1158 | break; | |
1159 | } | |
1c8c601a | 1160 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1161 | return fl ? -EAGAIN : 0; |
1162 | } | |
1163 | ||
1164 | /** | |
1165 | * locks_mandatory_area - Check for a conflicting lock | |
1166 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1167 | * for shared | |
1168 | * @inode: the file to check | |
1169 | * @filp: how the file was opened (if it was) | |
1170 | * @offset: start of area to check | |
1171 | * @count: length of area to check | |
1172 | * | |
1173 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1174 | * This function is called from rw_verify_area() and | |
1175 | * locks_verify_truncate(). | |
1176 | */ | |
1177 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1178 | struct file *filp, loff_t offset, | |
1179 | size_t count) | |
1180 | { | |
1181 | struct file_lock fl; | |
1182 | int error; | |
1183 | ||
1184 | locks_init_lock(&fl); | |
1185 | fl.fl_owner = current->files; | |
1186 | fl.fl_pid = current->tgid; | |
1187 | fl.fl_file = filp; | |
1188 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1189 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1190 | fl.fl_flags |= FL_SLEEP; | |
1191 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1192 | fl.fl_start = offset; | |
1193 | fl.fl_end = offset + count - 1; | |
1194 | ||
1195 | for (;;) { | |
150b3934 | 1196 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1197 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1198 | break; |
1199 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1200 | if (!error) { | |
1201 | /* | |
1202 | * If we've been sleeping someone might have | |
1203 | * changed the permissions behind our back. | |
1204 | */ | |
a16877ca | 1205 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1206 | continue; |
1207 | } | |
1208 | ||
1209 | locks_delete_block(&fl); | |
1210 | break; | |
1211 | } | |
1212 | ||
1213 | return error; | |
1214 | } | |
1215 | ||
1216 | EXPORT_SYMBOL(locks_mandatory_area); | |
1217 | ||
778fc546 BF |
1218 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1219 | { | |
1220 | switch (arg) { | |
1221 | case F_UNLCK: | |
1222 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1223 | /* fall through: */ | |
1224 | case F_RDLCK: | |
1225 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1226 | } | |
1227 | } | |
1228 | ||
1da177e4 LT |
1229 | /* We already had a lease on this file; just change its type */ |
1230 | int lease_modify(struct file_lock **before, int arg) | |
1231 | { | |
1232 | struct file_lock *fl = *before; | |
1233 | int error = assign_type(fl, arg); | |
1234 | ||
1235 | if (error) | |
1236 | return error; | |
778fc546 | 1237 | lease_clear_pending(fl, arg); |
1da177e4 | 1238 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1239 | if (arg == F_UNLCK) { |
1240 | struct file *filp = fl->fl_file; | |
1241 | ||
1242 | f_delown(filp); | |
1243 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1244 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1245 | if (fl->fl_fasync != NULL) { | |
1246 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1247 | fl->fl_fasync = NULL; | |
1248 | } | |
1da177e4 | 1249 | locks_delete_lock(before); |
3b6e2723 | 1250 | } |
1da177e4 LT |
1251 | return 0; |
1252 | } | |
1253 | ||
1254 | EXPORT_SYMBOL(lease_modify); | |
1255 | ||
778fc546 BF |
1256 | static bool past_time(unsigned long then) |
1257 | { | |
1258 | if (!then) | |
1259 | /* 0 is a special value meaning "this never expires": */ | |
1260 | return false; | |
1261 | return time_after(jiffies, then); | |
1262 | } | |
1263 | ||
1da177e4 LT |
1264 | static void time_out_leases(struct inode *inode) |
1265 | { | |
1266 | struct file_lock **before; | |
1267 | struct file_lock *fl; | |
1268 | ||
1269 | before = &inode->i_flock; | |
ab83fa4b | 1270 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
778fc546 BF |
1271 | if (past_time(fl->fl_downgrade_time)) |
1272 | lease_modify(before, F_RDLCK); | |
1273 | if (past_time(fl->fl_break_time)) | |
1274 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1275 | if (fl == *before) /* lease_modify may have freed fl */ |
1276 | before = &fl->fl_next; | |
1277 | } | |
1278 | } | |
1279 | ||
1280 | /** | |
1281 | * __break_lease - revoke all outstanding leases on file | |
1282 | * @inode: the inode of the file to return | |
1283 | * @mode: the open mode (read or write) | |
1284 | * | |
87250dd2 | 1285 | * break_lease (inlined for speed) has checked there already is at least |
1286 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1287 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1288 | * specified %O_NONBLOCK to your open(). |
1289 | */ | |
1290 | int __break_lease(struct inode *inode, unsigned int mode) | |
1291 | { | |
778fc546 | 1292 | int error = 0; |
1da177e4 LT |
1293 | struct file_lock *new_fl, *flock; |
1294 | struct file_lock *fl; | |
1da177e4 LT |
1295 | unsigned long break_time; |
1296 | int i_have_this_lease = 0; | |
8737c930 | 1297 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1298 | |
8737c930 | 1299 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1300 | if (IS_ERR(new_fl)) |
1301 | return PTR_ERR(new_fl); | |
1da177e4 | 1302 | |
1c8c601a | 1303 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1304 | |
1305 | time_out_leases(inode); | |
1306 | ||
1307 | flock = inode->i_flock; | |
1308 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1309 | goto out; | |
1310 | ||
778fc546 BF |
1311 | if (!locks_conflict(flock, new_fl)) |
1312 | goto out; | |
1313 | ||
1da177e4 LT |
1314 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
1315 | if (fl->fl_owner == current->files) | |
1316 | i_have_this_lease = 1; | |
1317 | ||
1da177e4 LT |
1318 | break_time = 0; |
1319 | if (lease_break_time > 0) { | |
1320 | break_time = jiffies + lease_break_time * HZ; | |
1321 | if (break_time == 0) | |
1322 | break_time++; /* so that 0 means no break time */ | |
1323 | } | |
1324 | ||
1325 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
778fc546 BF |
1326 | if (want_write) { |
1327 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1328 | continue; | |
1329 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1330 | fl->fl_break_time = break_time; |
778fc546 BF |
1331 | } else { |
1332 | if (lease_breaking(flock)) | |
1333 | continue; | |
1334 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1335 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1336 | } |
778fc546 | 1337 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1338 | } |
1339 | ||
1340 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1341 | error = -EWOULDBLOCK; | |
1342 | goto out; | |
1343 | } | |
1344 | ||
1345 | restart: | |
1346 | break_time = flock->fl_break_time; | |
1347 | if (break_time != 0) { | |
1348 | break_time -= jiffies; | |
1349 | if (break_time == 0) | |
1350 | break_time++; | |
1351 | } | |
4321e01e | 1352 | locks_insert_block(flock, new_fl); |
1c8c601a | 1353 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1354 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1355 | !new_fl->fl_next, break_time); | |
1c8c601a JL |
1356 | spin_lock(&inode->i_lock); |
1357 | locks_delete_block(new_fl); | |
1da177e4 LT |
1358 | if (error >= 0) { |
1359 | if (error == 0) | |
1360 | time_out_leases(inode); | |
778fc546 BF |
1361 | /* |
1362 | * Wait for the next conflicting lease that has not been | |
1363 | * broken yet | |
1364 | */ | |
1da177e4 LT |
1365 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1366 | flock = flock->fl_next) { | |
778fc546 | 1367 | if (locks_conflict(new_fl, flock)) |
1da177e4 LT |
1368 | goto restart; |
1369 | } | |
1370 | error = 0; | |
1371 | } | |
1372 | ||
1373 | out: | |
1c8c601a | 1374 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1375 | locks_free_lock(new_fl); |
1da177e4 LT |
1376 | return error; |
1377 | } | |
1378 | ||
1379 | EXPORT_SYMBOL(__break_lease); | |
1380 | ||
1381 | /** | |
a6b91919 | 1382 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1383 | * @inode: the inode |
1384 | * @time: pointer to a timespec which will contain the last modified time | |
1385 | * | |
1386 | * This is to force NFS clients to flush their caches for files with | |
1387 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1388 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1389 | */ |
1390 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1391 | { | |
1392 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1393 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1394 | *time = current_fs_time(inode->i_sb); |
1395 | else | |
1396 | *time = inode->i_mtime; | |
1397 | } | |
1398 | ||
1399 | EXPORT_SYMBOL(lease_get_mtime); | |
1400 | ||
1401 | /** | |
1402 | * fcntl_getlease - Enquire what lease is currently active | |
1403 | * @filp: the file | |
1404 | * | |
1405 | * The value returned by this function will be one of | |
1406 | * (if no lease break is pending): | |
1407 | * | |
1408 | * %F_RDLCK to indicate a shared lease is held. | |
1409 | * | |
1410 | * %F_WRLCK to indicate an exclusive lease is held. | |
1411 | * | |
1412 | * %F_UNLCK to indicate no lease is held. | |
1413 | * | |
1414 | * (if a lease break is pending): | |
1415 | * | |
1416 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1417 | * changed to a shared lease (or removed). | |
1418 | * | |
1419 | * %F_UNLCK to indicate the lease needs to be removed. | |
1420 | * | |
1421 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1422 | * should be returned to userspace. | |
1423 | */ | |
1424 | int fcntl_getlease(struct file *filp) | |
1425 | { | |
1426 | struct file_lock *fl; | |
1c8c601a | 1427 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1428 | int type = F_UNLCK; |
1429 | ||
1c8c601a | 1430 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1431 | time_out_leases(file_inode(filp)); |
1432 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1433 | fl = fl->fl_next) { |
1434 | if (fl->fl_file == filp) { | |
778fc546 | 1435 | type = target_leasetype(fl); |
1da177e4 LT |
1436 | break; |
1437 | } | |
1438 | } | |
1c8c601a | 1439 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1440 | return type; |
1441 | } | |
1442 | ||
d4f22d19 | 1443 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1444 | { |
7eaae282 | 1445 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1446 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1447 | struct inode *inode = dentry->d_inode; |
c1f24ef4 | 1448 | int error; |
1da177e4 | 1449 | |
096657b6 BF |
1450 | lease = *flp; |
1451 | ||
8335ebd9 BF |
1452 | error = -EAGAIN; |
1453 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
096657b6 | 1454 | goto out; |
8335ebd9 BF |
1455 | if ((arg == F_WRLCK) |
1456 | && ((dentry->d_count > 1) | |
1457 | || (atomic_read(&inode->i_count) > 1))) | |
096657b6 | 1458 | goto out; |
6d5e8b05 | 1459 | |
1da177e4 LT |
1460 | /* |
1461 | * At this point, we know that if there is an exclusive | |
1462 | * lease on this file, then we hold it on this filp | |
1463 | * (otherwise our open of this file would have blocked). | |
1464 | * And if we are trying to acquire an exclusive lease, | |
1465 | * then the file is not open by anyone (including us) | |
1466 | * except for this filp. | |
1467 | */ | |
c1f24ef4 | 1468 | error = -EAGAIN; |
1da177e4 LT |
1469 | for (before = &inode->i_flock; |
1470 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1471 | before = &fl->fl_next) { | |
c1f24ef4 | 1472 | if (fl->fl_file == filp) { |
1da177e4 | 1473 | my_before = before; |
c1f24ef4 BF |
1474 | continue; |
1475 | } | |
1476 | /* | |
1477 | * No exclusive leases if someone else has a lease on | |
1478 | * this file: | |
1479 | */ | |
1480 | if (arg == F_WRLCK) | |
1481 | goto out; | |
1482 | /* | |
1483 | * Modifying our existing lease is OK, but no getting a | |
1484 | * new lease if someone else is opening for write: | |
1485 | */ | |
1486 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1487 | goto out; | |
1da177e4 LT |
1488 | } |
1489 | ||
1da177e4 | 1490 | if (my_before != NULL) { |
8fb47a4f | 1491 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1492 | if (!error) |
1493 | *flp = *my_before; | |
1da177e4 LT |
1494 | goto out; |
1495 | } | |
1496 | ||
1da177e4 LT |
1497 | error = -EINVAL; |
1498 | if (!leases_enable) | |
1499 | goto out; | |
1500 | ||
c5b1f0d9 | 1501 | locks_insert_lock(before, lease); |
85c59580 | 1502 | return 0; |
1da177e4 | 1503 | |
1da177e4 LT |
1504 | out: |
1505 | return error; | |
1506 | } | |
8335ebd9 | 1507 | |
d4f22d19 | 1508 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1509 | { |
1510 | struct file_lock *fl, **before; | |
1511 | struct dentry *dentry = filp->f_path.dentry; | |
1512 | struct inode *inode = dentry->d_inode; | |
1513 | ||
1514 | for (before = &inode->i_flock; | |
1515 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1516 | before = &fl->fl_next) { | |
1517 | if (fl->fl_file != filp) | |
1518 | continue; | |
1519 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1520 | } | |
1521 | return -EAGAIN; | |
1522 | } | |
1523 | ||
1524 | /** | |
1525 | * generic_setlease - sets a lease on an open file | |
1526 | * @filp: file pointer | |
1527 | * @arg: type of lease to obtain | |
1528 | * @flp: input - file_lock to use, output - file_lock inserted | |
1529 | * | |
1530 | * The (input) flp->fl_lmops->lm_break function is required | |
1531 | * by break_lease(). | |
1532 | * | |
1c8c601a | 1533 | * Called with inode->i_lock held. |
8335ebd9 BF |
1534 | */ |
1535 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1536 | { | |
1537 | struct dentry *dentry = filp->f_path.dentry; | |
1538 | struct inode *inode = dentry->d_inode; | |
1539 | int error; | |
1540 | ||
8e96e3b7 | 1541 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1542 | return -EACCES; |
1543 | if (!S_ISREG(inode->i_mode)) | |
1544 | return -EINVAL; | |
1545 | error = security_file_lock(filp, arg); | |
1546 | if (error) | |
1547 | return error; | |
1548 | ||
1549 | time_out_leases(inode); | |
1550 | ||
1551 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1552 | ||
1553 | switch (arg) { | |
1554 | case F_UNLCK: | |
1555 | return generic_delete_lease(filp, flp); | |
1556 | case F_RDLCK: | |
1557 | case F_WRLCK: | |
1558 | return generic_add_lease(filp, arg, flp); | |
1559 | default: | |
8d657eb3 | 1560 | return -EINVAL; |
8335ebd9 BF |
1561 | } |
1562 | } | |
0af1a450 | 1563 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1564 | |
b89f4321 AB |
1565 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1566 | { | |
1567 | if (filp->f_op && filp->f_op->setlease) | |
1568 | return filp->f_op->setlease(filp, arg, lease); | |
1569 | else | |
1570 | return generic_setlease(filp, arg, lease); | |
1571 | } | |
1572 | ||
1573 | /** | |
a9933cea | 1574 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1575 | * @filp: file pointer |
1576 | * @arg: type of lease to obtain | |
1577 | * @lease: file_lock to use | |
1578 | * | |
1579 | * Call this to establish a lease on the file. | |
8fb47a4f | 1580 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1581 | * break_lease will oops! |
1582 | * | |
1583 | * This will call the filesystem's setlease file method, if | |
1584 | * defined. Note that there is no getlease method; instead, the | |
1585 | * filesystem setlease method should call back to setlease() to | |
1586 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1587 | * find it. Since fcntl_getlease() only reports whether the current | |
1588 | * task holds a lease, a cluster filesystem need only do this for | |
1589 | * leases held by processes on this node. | |
1590 | * | |
1591 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1592 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1593 | * filesystem's open, create, and (on truncate) setattr methods. |
1594 | * | |
1595 | * Warning: the only current setlease methods exist only to disable | |
1596 | * leases in certain cases. More vfs changes may be required to | |
1597 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1598 | */ |
1599 | ||
a9933cea | 1600 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1601 | { |
1c8c601a | 1602 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1603 | int error; |
1604 | ||
1c8c601a | 1605 | spin_lock(&inode->i_lock); |
b89f4321 | 1606 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1607 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1608 | |
1609 | return error; | |
1610 | } | |
a9933cea | 1611 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1612 | |
0ceaf6c7 BF |
1613 | static int do_fcntl_delete_lease(struct file *filp) |
1614 | { | |
1615 | struct file_lock fl, *flp = &fl; | |
1616 | ||
1617 | lease_init(filp, F_UNLCK, flp); | |
1618 | ||
1619 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1620 | } | |
1621 | ||
1622 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1623 | { |
3df057ac | 1624 | struct file_lock *fl, *ret; |
1c8c601a | 1625 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1626 | struct fasync_struct *new; |
1da177e4 LT |
1627 | int error; |
1628 | ||
c5b1f0d9 AB |
1629 | fl = lease_alloc(filp, arg); |
1630 | if (IS_ERR(fl)) | |
1631 | return PTR_ERR(fl); | |
1da177e4 | 1632 | |
f7347ce4 LT |
1633 | new = fasync_alloc(); |
1634 | if (!new) { | |
1635 | locks_free_lock(fl); | |
1636 | return -ENOMEM; | |
1637 | } | |
3df057ac | 1638 | ret = fl; |
1c8c601a | 1639 | spin_lock(&inode->i_lock); |
8896b93f | 1640 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 | 1641 | if (error) { |
1c8c601a | 1642 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1643 | locks_free_lock(fl); |
1644 | goto out_free_fasync; | |
1645 | } | |
3df057ac BF |
1646 | if (ret != fl) |
1647 | locks_free_lock(fl); | |
1da177e4 | 1648 | |
f7347ce4 LT |
1649 | /* |
1650 | * fasync_insert_entry() returns the old entry if any. | |
1651 | * If there was no old entry, then it used 'new' and | |
1652 | * inserted it into the fasync list. Clear new so that | |
1653 | * we don't release it here. | |
1654 | */ | |
3df057ac | 1655 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1656 | new = NULL; |
1657 | ||
8896b93f | 1658 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1c8c601a | 1659 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1660 | |
1661 | out_free_fasync: | |
f7347ce4 LT |
1662 | if (new) |
1663 | fasync_free(new); | |
1da177e4 LT |
1664 | return error; |
1665 | } | |
1666 | ||
0ceaf6c7 BF |
1667 | /** |
1668 | * fcntl_setlease - sets a lease on an open file | |
1669 | * @fd: open file descriptor | |
1670 | * @filp: file pointer | |
1671 | * @arg: type of lease to obtain | |
1672 | * | |
1673 | * Call this fcntl to establish a lease on the file. | |
1674 | * Note that you also need to call %F_SETSIG to | |
1675 | * receive a signal when the lease is broken. | |
1676 | */ | |
1677 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1678 | { | |
1679 | if (arg == F_UNLCK) | |
1680 | return do_fcntl_delete_lease(filp); | |
1681 | return do_fcntl_add_lease(fd, filp, arg); | |
1682 | } | |
1683 | ||
1da177e4 LT |
1684 | /** |
1685 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1686 | * @filp: The file to apply the lock to | |
1687 | * @fl: The lock to be applied | |
1688 | * | |
1689 | * Add a FLOCK style lock to a file. | |
1690 | */ | |
1691 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1692 | { | |
1693 | int error; | |
1694 | might_sleep(); | |
1695 | for (;;) { | |
1696 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1697 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1698 | break; |
1699 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1700 | if (!error) | |
1701 | continue; | |
1702 | ||
1703 | locks_delete_block(fl); | |
1704 | break; | |
1705 | } | |
1706 | return error; | |
1707 | } | |
1708 | ||
1709 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1710 | ||
1711 | /** | |
1712 | * sys_flock: - flock() system call. | |
1713 | * @fd: the file descriptor to lock. | |
1714 | * @cmd: the type of lock to apply. | |
1715 | * | |
1716 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1717 | * The @cmd can be one of | |
1718 | * | |
1719 | * %LOCK_SH -- a shared lock. | |
1720 | * | |
1721 | * %LOCK_EX -- an exclusive lock. | |
1722 | * | |
1723 | * %LOCK_UN -- remove an existing lock. | |
1724 | * | |
1725 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1726 | * | |
1727 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1728 | * processes read and write access respectively. | |
1729 | */ | |
002c8976 | 1730 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1731 | { |
2903ff01 | 1732 | struct fd f = fdget(fd); |
1da177e4 LT |
1733 | struct file_lock *lock; |
1734 | int can_sleep, unlock; | |
1735 | int error; | |
1736 | ||
1737 | error = -EBADF; | |
2903ff01 | 1738 | if (!f.file) |
1da177e4 LT |
1739 | goto out; |
1740 | ||
1741 | can_sleep = !(cmd & LOCK_NB); | |
1742 | cmd &= ~LOCK_NB; | |
1743 | unlock = (cmd == LOCK_UN); | |
1744 | ||
aeb5d727 | 1745 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1746 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1747 | goto out_putf; |
1748 | ||
2903ff01 | 1749 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1750 | if (error) |
1751 | goto out_putf; | |
1752 | if (can_sleep) | |
1753 | lock->fl_flags |= FL_SLEEP; | |
1754 | ||
2903ff01 | 1755 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1756 | if (error) |
1757 | goto out_free; | |
1758 | ||
2903ff01 AV |
1759 | if (f.file->f_op && f.file->f_op->flock) |
1760 | error = f.file->f_op->flock(f.file, | |
1da177e4 LT |
1761 | (can_sleep) ? F_SETLKW : F_SETLK, |
1762 | lock); | |
1763 | else | |
2903ff01 | 1764 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1765 | |
1766 | out_free: | |
993dfa87 | 1767 | locks_free_lock(lock); |
1da177e4 LT |
1768 | |
1769 | out_putf: | |
2903ff01 | 1770 | fdput(f); |
1da177e4 LT |
1771 | out: |
1772 | return error; | |
1773 | } | |
1774 | ||
3ee17abd BF |
1775 | /** |
1776 | * vfs_test_lock - test file byte range lock | |
1777 | * @filp: The file to test lock for | |
6924c554 | 1778 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1779 | * |
1780 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1781 | * setting conf->fl_type to something other than F_UNLCK. | |
1782 | */ | |
1783 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1784 | { | |
1785 | if (filp->f_op && filp->f_op->lock) | |
1786 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1787 | posix_test_lock(filp, fl); | |
1788 | return 0; | |
1789 | } | |
1790 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1791 | ||
c2fa1b8a BF |
1792 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1793 | { | |
1794 | flock->l_pid = fl->fl_pid; | |
1795 | #if BITS_PER_LONG == 32 | |
1796 | /* | |
1797 | * Make sure we can represent the posix lock via | |
1798 | * legacy 32bit flock. | |
1799 | */ | |
1800 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1801 | return -EOVERFLOW; | |
1802 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1803 | return -EOVERFLOW; | |
1804 | #endif | |
1805 | flock->l_start = fl->fl_start; | |
1806 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1807 | fl->fl_end - fl->fl_start + 1; | |
1808 | flock->l_whence = 0; | |
129a84de | 1809 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1810 | return 0; |
1811 | } | |
1812 | ||
1813 | #if BITS_PER_LONG == 32 | |
1814 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1815 | { | |
1816 | flock->l_pid = fl->fl_pid; | |
1817 | flock->l_start = fl->fl_start; | |
1818 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1819 | fl->fl_end - fl->fl_start + 1; | |
1820 | flock->l_whence = 0; | |
1821 | flock->l_type = fl->fl_type; | |
1822 | } | |
1823 | #endif | |
1824 | ||
1da177e4 LT |
1825 | /* Report the first existing lock that would conflict with l. |
1826 | * This implements the F_GETLK command of fcntl(). | |
1827 | */ | |
1828 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1829 | { | |
9d6a8c5c | 1830 | struct file_lock file_lock; |
1da177e4 LT |
1831 | struct flock flock; |
1832 | int error; | |
1833 | ||
1834 | error = -EFAULT; | |
1835 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1836 | goto out; | |
1837 | error = -EINVAL; | |
1838 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1839 | goto out; | |
1840 | ||
1841 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1842 | if (error) | |
1843 | goto out; | |
1844 | ||
3ee17abd BF |
1845 | error = vfs_test_lock(filp, &file_lock); |
1846 | if (error) | |
1847 | goto out; | |
1da177e4 | 1848 | |
9d6a8c5c ME |
1849 | flock.l_type = file_lock.fl_type; |
1850 | if (file_lock.fl_type != F_UNLCK) { | |
1851 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1852 | if (error) |
1da177e4 | 1853 | goto out; |
1da177e4 LT |
1854 | } |
1855 | error = -EFAULT; | |
1856 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1857 | error = 0; | |
1858 | out: | |
1859 | return error; | |
1860 | } | |
1861 | ||
7723ec97 ME |
1862 | /** |
1863 | * vfs_lock_file - file byte range lock | |
1864 | * @filp: The file to apply the lock to | |
1865 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1866 | * @fl: The lock to be applied | |
150b3934 ME |
1867 | * @conf: Place to return a copy of the conflicting lock, if found. |
1868 | * | |
1869 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1870 | * as the final argument. | |
1871 | * | |
1872 | * If the filesystem defines a private ->lock() method, then @conf will | |
1873 | * be left unchanged; so a caller that cares should initialize it to | |
1874 | * some acceptable default. | |
2beb6614 ME |
1875 | * |
1876 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1877 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1878 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1879 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1880 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1881 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1882 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1883 | * request completes. |
1884 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1885 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1886 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1887 | * nonzero return code and the file system should release the lock. The file |
1888 | * system is also responsible to keep a corresponding posix lock when it | |
1889 | * grants a lock so the VFS can find out which locks are locally held and do | |
1890 | * the correct lock cleanup when required. | |
1891 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1892 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1893 | * return code. |
7723ec97 | 1894 | */ |
150b3934 | 1895 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1896 | { |
1897 | if (filp->f_op && filp->f_op->lock) | |
1898 | return filp->f_op->lock(filp, cmd, fl); | |
1899 | else | |
150b3934 | 1900 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1901 | } |
1902 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1903 | ||
b648a6de MS |
1904 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1905 | struct file_lock *fl) | |
1906 | { | |
1907 | int error; | |
1908 | ||
1909 | error = security_file_lock(filp, fl->fl_type); | |
1910 | if (error) | |
1911 | return error; | |
1912 | ||
764c76b3 MS |
1913 | for (;;) { |
1914 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1915 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1916 | break; |
764c76b3 MS |
1917 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1918 | if (!error) | |
1919 | continue; | |
1920 | ||
1921 | locks_delete_block(fl); | |
1922 | break; | |
b648a6de MS |
1923 | } |
1924 | ||
1925 | return error; | |
1926 | } | |
1927 | ||
1da177e4 LT |
1928 | /* Apply the lock described by l to an open file descriptor. |
1929 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1930 | */ | |
c293621b PS |
1931 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1932 | struct flock __user *l) | |
1da177e4 LT |
1933 | { |
1934 | struct file_lock *file_lock = locks_alloc_lock(); | |
1935 | struct flock flock; | |
1936 | struct inode *inode; | |
0b2bac2f | 1937 | struct file *f; |
1da177e4 LT |
1938 | int error; |
1939 | ||
1940 | if (file_lock == NULL) | |
1941 | return -ENOLCK; | |
1942 | ||
1943 | /* | |
1944 | * This might block, so we do it before checking the inode. | |
1945 | */ | |
1946 | error = -EFAULT; | |
1947 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1948 | goto out; | |
1949 | ||
496ad9aa | 1950 | inode = file_inode(filp); |
1da177e4 LT |
1951 | |
1952 | /* Don't allow mandatory locks on files that may be memory mapped | |
1953 | * and shared. | |
1954 | */ | |
a16877ca | 1955 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1956 | error = -EAGAIN; |
1957 | goto out; | |
1958 | } | |
1959 | ||
c293621b | 1960 | again: |
1da177e4 LT |
1961 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1962 | if (error) | |
1963 | goto out; | |
1964 | if (cmd == F_SETLKW) { | |
1965 | file_lock->fl_flags |= FL_SLEEP; | |
1966 | } | |
1967 | ||
1968 | error = -EBADF; | |
1969 | switch (flock.l_type) { | |
1970 | case F_RDLCK: | |
1971 | if (!(filp->f_mode & FMODE_READ)) | |
1972 | goto out; | |
1973 | break; | |
1974 | case F_WRLCK: | |
1975 | if (!(filp->f_mode & FMODE_WRITE)) | |
1976 | goto out; | |
1977 | break; | |
1978 | case F_UNLCK: | |
1979 | break; | |
1980 | default: | |
1981 | error = -EINVAL; | |
1982 | goto out; | |
1983 | } | |
1984 | ||
b648a6de | 1985 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1986 | |
c293621b PS |
1987 | /* |
1988 | * Attempt to detect a close/fcntl race and recover by | |
1989 | * releasing the lock that was just acquired. | |
1990 | */ | |
0b2bac2f AV |
1991 | /* |
1992 | * we need that spin_lock here - it prevents reordering between | |
1993 | * update of inode->i_flock and check for it done in close(). | |
1994 | * rcu_read_lock() wouldn't do. | |
1995 | */ | |
1996 | spin_lock(¤t->files->file_lock); | |
1997 | f = fcheck(fd); | |
1998 | spin_unlock(¤t->files->file_lock); | |
1999 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2000 | flock.l_type = F_UNLCK; |
2001 | goto again; | |
1da177e4 LT |
2002 | } |
2003 | ||
c293621b | 2004 | out: |
1da177e4 LT |
2005 | locks_free_lock(file_lock); |
2006 | return error; | |
2007 | } | |
2008 | ||
2009 | #if BITS_PER_LONG == 32 | |
2010 | /* Report the first existing lock that would conflict with l. | |
2011 | * This implements the F_GETLK command of fcntl(). | |
2012 | */ | |
2013 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
2014 | { | |
9d6a8c5c | 2015 | struct file_lock file_lock; |
1da177e4 LT |
2016 | struct flock64 flock; |
2017 | int error; | |
2018 | ||
2019 | error = -EFAULT; | |
2020 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2021 | goto out; | |
2022 | error = -EINVAL; | |
2023 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2024 | goto out; | |
2025 | ||
2026 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2027 | if (error) | |
2028 | goto out; | |
2029 | ||
3ee17abd BF |
2030 | error = vfs_test_lock(filp, &file_lock); |
2031 | if (error) | |
2032 | goto out; | |
2033 | ||
9d6a8c5c ME |
2034 | flock.l_type = file_lock.fl_type; |
2035 | if (file_lock.fl_type != F_UNLCK) | |
2036 | posix_lock_to_flock64(&flock, &file_lock); | |
2037 | ||
1da177e4 LT |
2038 | error = -EFAULT; |
2039 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2040 | error = 0; | |
2041 | ||
2042 | out: | |
2043 | return error; | |
2044 | } | |
2045 | ||
2046 | /* Apply the lock described by l to an open file descriptor. | |
2047 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2048 | */ | |
c293621b PS |
2049 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2050 | struct flock64 __user *l) | |
1da177e4 LT |
2051 | { |
2052 | struct file_lock *file_lock = locks_alloc_lock(); | |
2053 | struct flock64 flock; | |
2054 | struct inode *inode; | |
0b2bac2f | 2055 | struct file *f; |
1da177e4 LT |
2056 | int error; |
2057 | ||
2058 | if (file_lock == NULL) | |
2059 | return -ENOLCK; | |
2060 | ||
2061 | /* | |
2062 | * This might block, so we do it before checking the inode. | |
2063 | */ | |
2064 | error = -EFAULT; | |
2065 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2066 | goto out; | |
2067 | ||
496ad9aa | 2068 | inode = file_inode(filp); |
1da177e4 LT |
2069 | |
2070 | /* Don't allow mandatory locks on files that may be memory mapped | |
2071 | * and shared. | |
2072 | */ | |
a16877ca | 2073 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2074 | error = -EAGAIN; |
2075 | goto out; | |
2076 | } | |
2077 | ||
c293621b | 2078 | again: |
1da177e4 LT |
2079 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2080 | if (error) | |
2081 | goto out; | |
2082 | if (cmd == F_SETLKW64) { | |
2083 | file_lock->fl_flags |= FL_SLEEP; | |
2084 | } | |
2085 | ||
2086 | error = -EBADF; | |
2087 | switch (flock.l_type) { | |
2088 | case F_RDLCK: | |
2089 | if (!(filp->f_mode & FMODE_READ)) | |
2090 | goto out; | |
2091 | break; | |
2092 | case F_WRLCK: | |
2093 | if (!(filp->f_mode & FMODE_WRITE)) | |
2094 | goto out; | |
2095 | break; | |
2096 | case F_UNLCK: | |
2097 | break; | |
2098 | default: | |
2099 | error = -EINVAL; | |
2100 | goto out; | |
2101 | } | |
2102 | ||
b648a6de | 2103 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2104 | |
c293621b PS |
2105 | /* |
2106 | * Attempt to detect a close/fcntl race and recover by | |
2107 | * releasing the lock that was just acquired. | |
2108 | */ | |
0b2bac2f AV |
2109 | spin_lock(¤t->files->file_lock); |
2110 | f = fcheck(fd); | |
2111 | spin_unlock(¤t->files->file_lock); | |
2112 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2113 | flock.l_type = F_UNLCK; |
2114 | goto again; | |
1da177e4 LT |
2115 | } |
2116 | ||
2117 | out: | |
2118 | locks_free_lock(file_lock); | |
2119 | return error; | |
2120 | } | |
2121 | #endif /* BITS_PER_LONG == 32 */ | |
2122 | ||
2123 | /* | |
2124 | * This function is called when the file is being removed | |
2125 | * from the task's fd array. POSIX locks belonging to this task | |
2126 | * are deleted at this time. | |
2127 | */ | |
2128 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2129 | { | |
ff7b86b8 | 2130 | struct file_lock lock; |
1da177e4 LT |
2131 | |
2132 | /* | |
2133 | * If there are no locks held on this file, we don't need to call | |
2134 | * posix_lock_file(). Another process could be setting a lock on this | |
2135 | * file at the same time, but we wouldn't remove that lock anyway. | |
2136 | */ | |
496ad9aa | 2137 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2138 | return; |
2139 | ||
2140 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2141 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2142 | lock.fl_start = 0; |
2143 | lock.fl_end = OFFSET_MAX; | |
2144 | lock.fl_owner = owner; | |
2145 | lock.fl_pid = current->tgid; | |
2146 | lock.fl_file = filp; | |
2147 | lock.fl_ops = NULL; | |
2148 | lock.fl_lmops = NULL; | |
2149 | ||
150b3934 | 2150 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2151 | |
1da177e4 LT |
2152 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2153 | lock.fl_ops->fl_release_private(&lock); | |
2154 | } | |
2155 | ||
2156 | EXPORT_SYMBOL(locks_remove_posix); | |
2157 | ||
2158 | /* | |
2159 | * This function is called on the last close of an open file. | |
2160 | */ | |
2161 | void locks_remove_flock(struct file *filp) | |
2162 | { | |
496ad9aa | 2163 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2164 | struct file_lock *fl; |
2165 | struct file_lock **before; | |
2166 | ||
2167 | if (!inode->i_flock) | |
2168 | return; | |
2169 | ||
2170 | if (filp->f_op && filp->f_op->flock) { | |
2171 | struct file_lock fl = { | |
2172 | .fl_pid = current->tgid, | |
2173 | .fl_file = filp, | |
2174 | .fl_flags = FL_FLOCK, | |
2175 | .fl_type = F_UNLCK, | |
2176 | .fl_end = OFFSET_MAX, | |
2177 | }; | |
2178 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2179 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2180 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2181 | } |
2182 | ||
1c8c601a | 2183 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2184 | before = &inode->i_flock; |
2185 | ||
2186 | while ((fl = *before) != NULL) { | |
2187 | if (fl->fl_file == filp) { | |
c293621b | 2188 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2189 | locks_delete_lock(before); |
2190 | continue; | |
2191 | } | |
2192 | if (IS_LEASE(fl)) { | |
2193 | lease_modify(before, F_UNLCK); | |
2194 | continue; | |
2195 | } | |
2196 | /* What? */ | |
2197 | BUG(); | |
2198 | } | |
2199 | before = &fl->fl_next; | |
2200 | } | |
1c8c601a | 2201 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2202 | } |
2203 | ||
1da177e4 LT |
2204 | /** |
2205 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2206 | * @waiter: the lock which was waiting |
2207 | * | |
2208 | * lockd needs to block waiting for locks. | |
2209 | */ | |
64a318ee | 2210 | int |
f891a29f | 2211 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2212 | { |
64a318ee BF |
2213 | int status = 0; |
2214 | ||
1c8c601a | 2215 | spin_lock(&file_lock_lock); |
5996a298 | 2216 | if (waiter->fl_next) |
1da177e4 | 2217 | __locks_delete_block(waiter); |
64a318ee BF |
2218 | else |
2219 | status = -ENOENT; | |
1c8c601a | 2220 | spin_unlock(&file_lock_lock); |
64a318ee | 2221 | return status; |
1da177e4 | 2222 | } |
1da177e4 LT |
2223 | EXPORT_SYMBOL(posix_unblock_lock); |
2224 | ||
9b9d2ab4 ME |
2225 | /** |
2226 | * vfs_cancel_lock - file byte range unblock lock | |
2227 | * @filp: The file to apply the unblock to | |
2228 | * @fl: The lock to be unblocked | |
2229 | * | |
2230 | * Used by lock managers to cancel blocked requests | |
2231 | */ | |
2232 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2233 | { | |
2234 | if (filp->f_op && filp->f_op->lock) | |
2235 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2236 | return 0; | |
2237 | } | |
2238 | ||
2239 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2240 | ||
7f8ada98 | 2241 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2242 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2243 | #include <linux/seq_file.h> |
2244 | ||
2245 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
99dc8292 | 2246 | loff_t id, char *pfx) |
1da177e4 LT |
2247 | { |
2248 | struct inode *inode = NULL; | |
ab1f1611 VG |
2249 | unsigned int fl_pid; |
2250 | ||
2251 | if (fl->fl_nspid) | |
6c5f3e7b | 2252 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2253 | else |
2254 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2255 | |
2256 | if (fl->fl_file != NULL) | |
496ad9aa | 2257 | inode = file_inode(fl->fl_file); |
1da177e4 | 2258 | |
99dc8292 | 2259 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2260 | if (IS_POSIX(fl)) { |
7f8ada98 | 2261 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2262 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2263 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2264 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2265 | } else if (IS_FLOCK(fl)) { |
2266 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2267 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2268 | } else { |
7f8ada98 | 2269 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2270 | } |
2271 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2272 | seq_printf(f, "LEASE "); |
ab83fa4b | 2273 | if (lease_breaking(fl)) |
7f8ada98 | 2274 | seq_printf(f, "BREAKING "); |
1da177e4 | 2275 | else if (fl->fl_file) |
7f8ada98 | 2276 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2277 | else |
7f8ada98 | 2278 | seq_printf(f, "BREAKER "); |
1da177e4 | 2279 | } else { |
7f8ada98 | 2280 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2281 | } |
2282 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2283 | seq_printf(f, "%s ", |
1da177e4 LT |
2284 | (fl->fl_type & LOCK_READ) |
2285 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2286 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2287 | } else { | |
7f8ada98 | 2288 | seq_printf(f, "%s ", |
ab83fa4b | 2289 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2290 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2291 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2292 | } |
2293 | if (inode) { | |
2294 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2295 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2296 | inode->i_sb->s_id, inode->i_ino); |
2297 | #else | |
2298 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2299 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2300 | MAJOR(inode->i_sb->s_dev), |
2301 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2302 | #endif | |
2303 | } else { | |
ab1f1611 | 2304 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2305 | } |
2306 | if (IS_POSIX(fl)) { | |
2307 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2308 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2309 | else |
7f8ada98 | 2310 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2311 | } else { |
7f8ada98 | 2312 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2313 | } |
2314 | } | |
2315 | ||
7f8ada98 | 2316 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2317 | { |
7f8ada98 | 2318 | struct file_lock *fl, *bfl; |
1da177e4 | 2319 | |
139ca04e | 2320 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2321 | |
99dc8292 | 2322 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
1da177e4 | 2323 | |
7f8ada98 | 2324 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
99dc8292 | 2325 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
094f2825 | 2326 | |
7f8ada98 PE |
2327 | return 0; |
2328 | } | |
1da177e4 | 2329 | |
7f8ada98 PE |
2330 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2331 | { | |
99dc8292 JM |
2332 | loff_t *p = f->private; |
2333 | ||
1c8c601a | 2334 | spin_lock(&file_lock_lock); |
99dc8292 | 2335 | *p = (*pos + 1); |
139ca04e | 2336 | return seq_hlist_start(&file_lock_list, *pos); |
7f8ada98 | 2337 | } |
1da177e4 | 2338 | |
7f8ada98 PE |
2339 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2340 | { | |
99dc8292 JM |
2341 | loff_t *p = f->private; |
2342 | ++*p; | |
139ca04e | 2343 | return seq_hlist_next(v, &file_lock_list, pos); |
7f8ada98 | 2344 | } |
1da177e4 | 2345 | |
7f8ada98 PE |
2346 | static void locks_stop(struct seq_file *f, void *v) |
2347 | { | |
1c8c601a | 2348 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
2349 | } |
2350 | ||
d8ba7a36 | 2351 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2352 | .start = locks_start, |
2353 | .next = locks_next, | |
2354 | .stop = locks_stop, | |
2355 | .show = locks_show, | |
2356 | }; | |
d8ba7a36 AD |
2357 | |
2358 | static int locks_open(struct inode *inode, struct file *filp) | |
2359 | { | |
99dc8292 | 2360 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
d8ba7a36 AD |
2361 | } |
2362 | ||
2363 | static const struct file_operations proc_locks_operations = { | |
2364 | .open = locks_open, | |
2365 | .read = seq_read, | |
2366 | .llseek = seq_lseek, | |
99dc8292 | 2367 | .release = seq_release_private, |
d8ba7a36 AD |
2368 | }; |
2369 | ||
2370 | static int __init proc_locks_init(void) | |
2371 | { | |
2372 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2373 | return 0; | |
2374 | } | |
2375 | module_init(proc_locks_init); | |
7f8ada98 PE |
2376 | #endif |
2377 | ||
1da177e4 LT |
2378 | /** |
2379 | * lock_may_read - checks that the region is free of locks | |
2380 | * @inode: the inode that is being read | |
2381 | * @start: the first byte to read | |
2382 | * @len: the number of bytes to read | |
2383 | * | |
2384 | * Emulates Windows locking requirements. Whole-file | |
2385 | * mandatory locks (share modes) can prohibit a read and | |
2386 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2387 | * | |
2388 | * N.B. this function is only ever called | |
2389 | * from knfsd and ownership of locks is never checked. | |
2390 | */ | |
2391 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2392 | { | |
2393 | struct file_lock *fl; | |
2394 | int result = 1; | |
1c8c601a JL |
2395 | |
2396 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2397 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2398 | if (IS_POSIX(fl)) { | |
2399 | if (fl->fl_type == F_RDLCK) | |
2400 | continue; | |
2401 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2402 | continue; | |
2403 | } else if (IS_FLOCK(fl)) { | |
2404 | if (!(fl->fl_type & LOCK_MAND)) | |
2405 | continue; | |
2406 | if (fl->fl_type & LOCK_READ) | |
2407 | continue; | |
2408 | } else | |
2409 | continue; | |
2410 | result = 0; | |
2411 | break; | |
2412 | } | |
1c8c601a | 2413 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2414 | return result; |
2415 | } | |
2416 | ||
2417 | EXPORT_SYMBOL(lock_may_read); | |
2418 | ||
2419 | /** | |
2420 | * lock_may_write - checks that the region is free of locks | |
2421 | * @inode: the inode that is being written | |
2422 | * @start: the first byte to write | |
2423 | * @len: the number of bytes to write | |
2424 | * | |
2425 | * Emulates Windows locking requirements. Whole-file | |
2426 | * mandatory locks (share modes) can prohibit a write and | |
2427 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2428 | * | |
2429 | * N.B. this function is only ever called | |
2430 | * from knfsd and ownership of locks is never checked. | |
2431 | */ | |
2432 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2433 | { | |
2434 | struct file_lock *fl; | |
2435 | int result = 1; | |
1c8c601a JL |
2436 | |
2437 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2438 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2439 | if (IS_POSIX(fl)) { | |
2440 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2441 | continue; | |
2442 | } else if (IS_FLOCK(fl)) { | |
2443 | if (!(fl->fl_type & LOCK_MAND)) | |
2444 | continue; | |
2445 | if (fl->fl_type & LOCK_WRITE) | |
2446 | continue; | |
2447 | } else | |
2448 | continue; | |
2449 | result = 0; | |
2450 | break; | |
2451 | } | |
1c8c601a | 2452 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2453 | return result; |
2454 | } | |
2455 | ||
2456 | EXPORT_SYMBOL(lock_may_write); | |
2457 | ||
1da177e4 LT |
2458 | static int __init filelock_init(void) |
2459 | { | |
2460 | filelock_cache = kmem_cache_create("file_lock_cache", | |
ee19cc40 MS |
2461 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2462 | ||
1da177e4 LT |
2463 | return 0; |
2464 | } | |
2465 | ||
2466 | core_initcall(filelock_init); |