Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/locks.c | |
3 | * | |
4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
6 | * | |
7 | * Deadlock detection added. | |
8 | * FIXME: one thing isn't handled yet: | |
9 | * - mandatory locks (requires lots of changes elsewhere) | |
10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
11 | * | |
12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
14 | * | |
15 | * Converted file_lock_table to a linked list from an array, which eliminates | |
16 | * the limits on how many active file locks are open. | |
17 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
18 | * | |
19 | * Removed dependency on file descriptors. dup()'ed file descriptors now | |
20 | * get the same locks as the original file descriptors, and a close() on | |
21 | * any file descriptor removes ALL the locks on the file for the current | |
22 | * process. Since locks still depend on the process id, locks are inherited | |
23 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
24 | * BSD and SVR4 practice. | |
25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
26 | * | |
27 | * Scrapped free list which is redundant now that we allocate locks | |
28 | * dynamically with kmalloc()/kfree(). | |
29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
30 | * | |
31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
32 | * | |
33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
34 | * fcntl() system call. They have the semantics described above. | |
35 | * | |
36 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
37 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
38 | * and will continue to use the old, broken implementation. | |
39 | * | |
40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
41 | * with a file pointer (filp). As a result they can be shared by a parent | |
42 | * process and its children after a fork(). They are removed when the last | |
43 | * file descriptor referring to the file pointer is closed (unless explicitly | |
44 | * unlocked). | |
45 | * | |
46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
47 | * upgrading from shared to exclusive (or vice versa). When this happens | |
48 | * any processes blocked by the current lock are woken up and allowed to | |
49 | * run before the new lock is applied. | |
50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
51 | * | |
52 | * Removed some race conditions in flock_lock_file(), marked other possible | |
53 | * races. Just grep for FIXME to see them. | |
54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. | |
55 | * | |
56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
58 | * once we've checked for blocking and deadlocking. | |
59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
60 | * | |
61 | * Initial implementation of mandatory locks. SunOS turned out to be | |
62 | * a rotten model, so I implemented the "obvious" semantics. | |
395cf969 | 63 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * | |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
69 | * Manual, Section 2. | |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
71 | * | |
72 | * Tidied up block list handling. Added '/proc/locks' interface. | |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
74 | * | |
75 | * Fixed deadlock condition for pathological code that mixes calls to | |
76 | * flock() and fcntl(). | |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
78 | * | |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
81 | * guarantee sensible behaviour in the case where file system modules might | |
82 | * be compiled with different options than the kernel itself. | |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
84 | * | |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
88 | * | |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
91 | * have already been freed. | |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
93 | * | |
94 | * Made the block list a circular list to minimise searching in the list. | |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
96 | * | |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
98 | * locking. | |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
100 | * | |
101 | * Some adaptations for NFS support. | |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
103 | * | |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
106 | * | |
107 | * Use slab allocator instead of kmalloc/kfree. | |
108 | * Use generic list implementation from <linux/list.h>. | |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
111 | * | |
112 | * Leases and LOCK_MAND | |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
115 | */ | |
116 | ||
117 | #include <linux/capability.h> | |
118 | #include <linux/file.h> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
120 | #include <linux/fs.h> |
121 | #include <linux/init.h> | |
122 | #include <linux/module.h> | |
123 | #include <linux/security.h> | |
124 | #include <linux/slab.h> | |
1da177e4 LT |
125 | #include <linux/syscalls.h> |
126 | #include <linux/time.h> | |
4fb3a538 | 127 | #include <linux/rcupdate.h> |
ab1f1611 | 128 | #include <linux/pid_namespace.h> |
48f74186 | 129 | #include <linux/hashtable.h> |
7012b02a JL |
130 | #include <linux/percpu.h> |
131 | #include <linux/lglock.h> | |
1da177e4 | 132 | |
1da177e4 LT |
133 | #include <asm/uaccess.h> |
134 | ||
135 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
136 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
617588d5 | 137 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG)) |
1da177e4 | 138 | |
ab83fa4b BF |
139 | static bool lease_breaking(struct file_lock *fl) |
140 | { | |
778fc546 BF |
141 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
142 | } | |
143 | ||
144 | static int target_leasetype(struct file_lock *fl) | |
145 | { | |
146 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
147 | return F_UNLCK; | |
148 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
149 | return F_RDLCK; | |
150 | return fl->fl_type; | |
ab83fa4b BF |
151 | } |
152 | ||
1da177e4 LT |
153 | int leases_enable = 1; |
154 | int lease_break_time = 45; | |
155 | ||
156 | #define for_each_lock(inode, lockp) \ | |
157 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
158 | ||
1c8c601a | 159 | /* |
7012b02a JL |
160 | * The global file_lock_list is only used for displaying /proc/locks, so we |
161 | * keep a list on each CPU, with each list protected by its own spinlock via | |
162 | * the file_lock_lglock. Note that alterations to the list also require that | |
163 | * the relevant i_lock is held. | |
1c8c601a | 164 | */ |
7012b02a JL |
165 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
166 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
88974691 | 167 | |
1c8c601a | 168 | /* |
48f74186 | 169 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 170 | * It is protected by blocked_lock_lock. |
48f74186 JL |
171 | * |
172 | * We hash locks by lockowner in order to optimize searching for the lock a | |
173 | * particular lockowner is waiting on. | |
174 | * | |
175 | * FIXME: make this value scale via some heuristic? We generally will want more | |
176 | * buckets when we have more lockowners holding locks, but that's a little | |
177 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 178 | */ |
48f74186 JL |
179 | #define BLOCKED_HASH_BITS 7 |
180 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 181 | |
1c8c601a | 182 | /* |
7b2296af JL |
183 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
184 | * want to be holding this lock. | |
1c8c601a JL |
185 | * |
186 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
187 | * pointer for file_lock structures that are acting as lock requests (in | |
188 | * contrast to those that are acting as records of acquired locks). | |
189 | * | |
190 | * Note that when we acquire this lock in order to change the above fields, | |
191 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
192 | * protected by this lock, we can skip acquiring it iff we already hold the | |
193 | * i_lock. | |
194 | * | |
195 | * In particular, adding an entry to the fl_block list requires that you hold | |
196 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
197 | * an entry from the list however only requires the file_lock_lock. | |
198 | */ | |
7b2296af | 199 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 200 | |
e18b890b | 201 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 202 | |
ee19cc40 | 203 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 204 | { |
139ca04e | 205 | INIT_HLIST_NODE(&fl->fl_link); |
ee19cc40 MS |
206 | INIT_LIST_HEAD(&fl->fl_block); |
207 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
208 | } |
209 | ||
1da177e4 | 210 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 211 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 212 | { |
ee19cc40 | 213 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
214 | |
215 | if (fl) | |
ee19cc40 | 216 | locks_init_lock_heads(fl); |
a51cb91d MS |
217 | |
218 | return fl; | |
1da177e4 | 219 | } |
c5b1f0d9 | 220 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 221 | |
a9e61e25 | 222 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
223 | { |
224 | if (fl->fl_ops) { | |
225 | if (fl->fl_ops->fl_release_private) | |
226 | fl->fl_ops->fl_release_private(fl); | |
227 | fl->fl_ops = NULL; | |
228 | } | |
068535f1 | 229 | fl->fl_lmops = NULL; |
47831f35 TM |
230 | |
231 | } | |
a9e61e25 | 232 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 233 | |
1da177e4 | 234 | /* Free a lock which is not in use. */ |
05fa3135 | 235 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 236 | { |
5ce29646 MS |
237 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
238 | BUG_ON(!list_empty(&fl->fl_block)); | |
139ca04e | 239 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 240 | |
47831f35 | 241 | locks_release_private(fl); |
1da177e4 LT |
242 | kmem_cache_free(filelock_cache, fl); |
243 | } | |
05fa3135 | 244 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
245 | |
246 | void locks_init_lock(struct file_lock *fl) | |
247 | { | |
ee19cc40 MS |
248 | memset(fl, 0, sizeof(struct file_lock)); |
249 | locks_init_lock_heads(fl); | |
1da177e4 LT |
250 | } |
251 | ||
252 | EXPORT_SYMBOL(locks_init_lock); | |
253 | ||
47831f35 TM |
254 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
255 | { | |
256 | if (fl->fl_ops) { | |
257 | if (fl->fl_ops->fl_copy_lock) | |
258 | fl->fl_ops->fl_copy_lock(new, fl); | |
259 | new->fl_ops = fl->fl_ops; | |
260 | } | |
bb8430a2 | 261 | if (fl->fl_lmops) |
47831f35 | 262 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
263 | } |
264 | ||
1da177e4 LT |
265 | /* |
266 | * Initialize a new lock from an existing file_lock structure. | |
267 | */ | |
1a747ee0 | 268 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
269 | { |
270 | new->fl_owner = fl->fl_owner; | |
271 | new->fl_pid = fl->fl_pid; | |
0996905f | 272 | new->fl_file = NULL; |
1da177e4 LT |
273 | new->fl_flags = fl->fl_flags; |
274 | new->fl_type = fl->fl_type; | |
275 | new->fl_start = fl->fl_start; | |
276 | new->fl_end = fl->fl_end; | |
0996905f TM |
277 | new->fl_ops = NULL; |
278 | new->fl_lmops = NULL; | |
279 | } | |
3dd7b71c | 280 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
281 | |
282 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
283 | { | |
284 | locks_release_private(new); | |
285 | ||
286 | __locks_copy_lock(new, fl); | |
287 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
288 | new->fl_ops = fl->fl_ops; |
289 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
290 | |
291 | locks_copy_private(new, fl); | |
1da177e4 LT |
292 | } |
293 | ||
294 | EXPORT_SYMBOL(locks_copy_lock); | |
295 | ||
296 | static inline int flock_translate_cmd(int cmd) { | |
297 | if (cmd & LOCK_MAND) | |
298 | return cmd & (LOCK_MAND | LOCK_RW); | |
299 | switch (cmd) { | |
300 | case LOCK_SH: | |
301 | return F_RDLCK; | |
302 | case LOCK_EX: | |
303 | return F_WRLCK; | |
304 | case LOCK_UN: | |
305 | return F_UNLCK; | |
306 | } | |
307 | return -EINVAL; | |
308 | } | |
309 | ||
310 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
311 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
312 | unsigned int cmd) | |
313 | { | |
314 | struct file_lock *fl; | |
315 | int type = flock_translate_cmd(cmd); | |
316 | if (type < 0) | |
317 | return type; | |
318 | ||
319 | fl = locks_alloc_lock(); | |
320 | if (fl == NULL) | |
321 | return -ENOMEM; | |
322 | ||
323 | fl->fl_file = filp; | |
324 | fl->fl_pid = current->tgid; | |
325 | fl->fl_flags = FL_FLOCK; | |
326 | fl->fl_type = type; | |
327 | fl->fl_end = OFFSET_MAX; | |
328 | ||
329 | *lock = fl; | |
330 | return 0; | |
331 | } | |
332 | ||
0ec4f431 | 333 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
334 | { |
335 | switch (type) { | |
336 | case F_RDLCK: | |
337 | case F_WRLCK: | |
338 | case F_UNLCK: | |
339 | fl->fl_type = type; | |
340 | break; | |
341 | default: | |
342 | return -EINVAL; | |
343 | } | |
344 | return 0; | |
345 | } | |
346 | ||
347 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
348 | * style lock. | |
349 | */ | |
350 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
351 | struct flock *l) | |
352 | { | |
353 | off_t start, end; | |
354 | ||
355 | switch (l->l_whence) { | |
f5579f8c | 356 | case SEEK_SET: |
1da177e4 LT |
357 | start = 0; |
358 | break; | |
f5579f8c | 359 | case SEEK_CUR: |
1da177e4 LT |
360 | start = filp->f_pos; |
361 | break; | |
f5579f8c | 362 | case SEEK_END: |
496ad9aa | 363 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
364 | break; |
365 | default: | |
366 | return -EINVAL; | |
367 | } | |
368 | ||
369 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
370 | POSIX-2001 defines it. */ | |
371 | start += l->l_start; | |
4c780a46 TM |
372 | if (start < 0) |
373 | return -EINVAL; | |
374 | fl->fl_end = OFFSET_MAX; | |
375 | if (l->l_len > 0) { | |
376 | end = start + l->l_len - 1; | |
377 | fl->fl_end = end; | |
378 | } else if (l->l_len < 0) { | |
1da177e4 | 379 | end = start - 1; |
4c780a46 | 380 | fl->fl_end = end; |
1da177e4 | 381 | start += l->l_len; |
4c780a46 TM |
382 | if (start < 0) |
383 | return -EINVAL; | |
1da177e4 | 384 | } |
1da177e4 | 385 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
386 | if (fl->fl_end < fl->fl_start) |
387 | return -EOVERFLOW; | |
1da177e4 LT |
388 | |
389 | fl->fl_owner = current->files; | |
390 | fl->fl_pid = current->tgid; | |
391 | fl->fl_file = filp; | |
392 | fl->fl_flags = FL_POSIX; | |
393 | fl->fl_ops = NULL; | |
394 | fl->fl_lmops = NULL; | |
395 | ||
396 | return assign_type(fl, l->l_type); | |
397 | } | |
398 | ||
399 | #if BITS_PER_LONG == 32 | |
400 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
401 | struct flock64 *l) | |
402 | { | |
403 | loff_t start; | |
404 | ||
405 | switch (l->l_whence) { | |
f5579f8c | 406 | case SEEK_SET: |
1da177e4 LT |
407 | start = 0; |
408 | break; | |
f5579f8c | 409 | case SEEK_CUR: |
1da177e4 LT |
410 | start = filp->f_pos; |
411 | break; | |
f5579f8c | 412 | case SEEK_END: |
496ad9aa | 413 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
414 | break; |
415 | default: | |
416 | return -EINVAL; | |
417 | } | |
418 | ||
4c780a46 TM |
419 | start += l->l_start; |
420 | if (start < 0) | |
1da177e4 | 421 | return -EINVAL; |
4c780a46 TM |
422 | fl->fl_end = OFFSET_MAX; |
423 | if (l->l_len > 0) { | |
424 | fl->fl_end = start + l->l_len - 1; | |
425 | } else if (l->l_len < 0) { | |
426 | fl->fl_end = start - 1; | |
427 | start += l->l_len; | |
428 | if (start < 0) | |
429 | return -EINVAL; | |
430 | } | |
1da177e4 | 431 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
432 | if (fl->fl_end < fl->fl_start) |
433 | return -EOVERFLOW; | |
1da177e4 LT |
434 | |
435 | fl->fl_owner = current->files; | |
436 | fl->fl_pid = current->tgid; | |
437 | fl->fl_file = filp; | |
438 | fl->fl_flags = FL_POSIX; | |
439 | fl->fl_ops = NULL; | |
440 | fl->fl_lmops = NULL; | |
441 | ||
f32cb532 | 442 | return assign_type(fl, l->l_type); |
1da177e4 LT |
443 | } |
444 | #endif | |
445 | ||
446 | /* default lease lock manager operations */ | |
447 | static void lease_break_callback(struct file_lock *fl) | |
448 | { | |
449 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
450 | } | |
451 | ||
7b021967 | 452 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 453 | .lm_break = lease_break_callback, |
8fb47a4f | 454 | .lm_change = lease_modify, |
1da177e4 LT |
455 | }; |
456 | ||
457 | /* | |
458 | * Initialize a lease, use the default lock manager operations | |
459 | */ | |
0ec4f431 | 460 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 461 | { |
75dff55a TM |
462 | if (assign_type(fl, type) != 0) |
463 | return -EINVAL; | |
464 | ||
1da177e4 LT |
465 | fl->fl_owner = current->files; |
466 | fl->fl_pid = current->tgid; | |
467 | ||
468 | fl->fl_file = filp; | |
469 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
470 | fl->fl_start = 0; |
471 | fl->fl_end = OFFSET_MAX; | |
472 | fl->fl_ops = NULL; | |
473 | fl->fl_lmops = &lease_manager_ops; | |
474 | return 0; | |
475 | } | |
476 | ||
477 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 478 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
479 | { |
480 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 481 | int error = -ENOMEM; |
1da177e4 LT |
482 | |
483 | if (fl == NULL) | |
e32b8ee2 | 484 | return ERR_PTR(error); |
1da177e4 LT |
485 | |
486 | error = lease_init(filp, type, fl); | |
75dff55a TM |
487 | if (error) { |
488 | locks_free_lock(fl); | |
e32b8ee2 | 489 | return ERR_PTR(error); |
75dff55a | 490 | } |
e32b8ee2 | 491 | return fl; |
1da177e4 LT |
492 | } |
493 | ||
494 | /* Check if two locks overlap each other. | |
495 | */ | |
496 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
497 | { | |
498 | return ((fl1->fl_end >= fl2->fl_start) && | |
499 | (fl2->fl_end >= fl1->fl_start)); | |
500 | } | |
501 | ||
502 | /* | |
503 | * Check whether two locks have the same owner. | |
504 | */ | |
33443c42 | 505 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 506 | { |
8fb47a4f | 507 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 508 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 509 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
510 | return fl1->fl_owner == fl2->fl_owner; |
511 | } | |
512 | ||
7012b02a | 513 | /* Must be called with the i_lock held! */ |
88974691 JL |
514 | static inline void |
515 | locks_insert_global_locks(struct file_lock *fl) | |
516 | { | |
7012b02a JL |
517 | lg_local_lock(&file_lock_lglock); |
518 | fl->fl_link_cpu = smp_processor_id(); | |
519 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
520 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
521 | } |
522 | ||
7012b02a | 523 | /* Must be called with the i_lock held! */ |
88974691 JL |
524 | static inline void |
525 | locks_delete_global_locks(struct file_lock *fl) | |
526 | { | |
7012b02a JL |
527 | /* |
528 | * Avoid taking lock if already unhashed. This is safe since this check | |
529 | * is done while holding the i_lock, and new insertions into the list | |
530 | * also require that it be held. | |
531 | */ | |
532 | if (hlist_unhashed(&fl->fl_link)) | |
533 | return; | |
534 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 535 | hlist_del_init(&fl->fl_link); |
7012b02a | 536 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
537 | } |
538 | ||
3999e493 JL |
539 | static unsigned long |
540 | posix_owner_key(struct file_lock *fl) | |
541 | { | |
542 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
543 | return fl->fl_lmops->lm_owner_key(fl); | |
544 | return (unsigned long)fl->fl_owner; | |
545 | } | |
546 | ||
88974691 JL |
547 | static inline void |
548 | locks_insert_global_blocked(struct file_lock *waiter) | |
549 | { | |
3999e493 | 550 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
551 | } |
552 | ||
553 | static inline void | |
554 | locks_delete_global_blocked(struct file_lock *waiter) | |
555 | { | |
48f74186 | 556 | hash_del(&waiter->fl_link); |
88974691 JL |
557 | } |
558 | ||
1da177e4 LT |
559 | /* Remove waiter from blocker's block list. |
560 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 561 | * |
7b2296af | 562 | * Must be called with blocked_lock_lock held. |
1da177e4 | 563 | */ |
33443c42 | 564 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 565 | { |
88974691 | 566 | locks_delete_global_blocked(waiter); |
1da177e4 | 567 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
568 | waiter->fl_next = NULL; |
569 | } | |
570 | ||
1a9e64a7 | 571 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 572 | { |
7b2296af | 573 | spin_lock(&blocked_lock_lock); |
1da177e4 | 574 | __locks_delete_block(waiter); |
7b2296af | 575 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
576 | } |
577 | ||
578 | /* Insert waiter into blocker's block list. | |
579 | * We use a circular list so that processes can be easily woken up in | |
580 | * the order they blocked. The documentation doesn't require this but | |
581 | * it seems like the reasonable thing to do. | |
1c8c601a | 582 | * |
7b2296af | 583 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
4e8c765d | 584 | * list itself is protected by the file_lock_list, but by ensuring that the |
7b2296af | 585 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 586 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 587 | */ |
1c8c601a JL |
588 | static void __locks_insert_block(struct file_lock *blocker, |
589 | struct file_lock *waiter) | |
1da177e4 | 590 | { |
6dc0fe8f | 591 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 592 | waiter->fl_next = blocker; |
88974691 | 593 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
1da177e4 | 594 | if (IS_POSIX(blocker)) |
1c8c601a JL |
595 | locks_insert_global_blocked(waiter); |
596 | } | |
597 | ||
598 | /* Must be called with i_lock held. */ | |
599 | static void locks_insert_block(struct file_lock *blocker, | |
600 | struct file_lock *waiter) | |
601 | { | |
7b2296af | 602 | spin_lock(&blocked_lock_lock); |
1c8c601a | 603 | __locks_insert_block(blocker, waiter); |
7b2296af | 604 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
605 | } |
606 | ||
1cb36012 JL |
607 | /* |
608 | * Wake up processes blocked waiting for blocker. | |
609 | * | |
1c8c601a | 610 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
611 | */ |
612 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
613 | { | |
4e8c765d JL |
614 | /* |
615 | * Avoid taking global lock if list is empty. This is safe since new | |
616 | * blocked requests are only added to the list under the i_lock, and | |
617 | * the i_lock is always held here. Note that removal from the fl_block | |
618 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 619 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
620 | */ |
621 | if (list_empty(&blocker->fl_block)) | |
622 | return; | |
623 | ||
7b2296af | 624 | spin_lock(&blocked_lock_lock); |
1da177e4 | 625 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
626 | struct file_lock *waiter; |
627 | ||
628 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
629 | struct file_lock, fl_block); |
630 | __locks_delete_block(waiter); | |
8fb47a4f BF |
631 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
632 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
633 | else |
634 | wake_up(&waiter->fl_wait); | |
635 | } | |
7b2296af | 636 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
637 | } |
638 | ||
639 | /* Insert file lock fl into an inode's lock list at the position indicated | |
640 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
641 | * |
642 | * Must be called with the i_lock held! | |
1da177e4 LT |
643 | */ |
644 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
645 | { | |
ab1f1611 VG |
646 | fl->fl_nspid = get_pid(task_tgid(current)); |
647 | ||
1da177e4 LT |
648 | /* insert into file's list */ |
649 | fl->fl_next = *pos; | |
650 | *pos = fl; | |
88974691 JL |
651 | |
652 | locks_insert_global_locks(fl); | |
1da177e4 LT |
653 | } |
654 | ||
24cbe784 JL |
655 | /** |
656 | * locks_delete_lock - Delete a lock and then free it. | |
657 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
658 | * inode->i_flock list entry | |
659 | * | |
660 | * Unlink a lock from all lists and free the namespace reference, but don't | |
661 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
662 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
663 | * |
664 | * Must be called with the i_lock held! | |
1da177e4 | 665 | */ |
24cbe784 | 666 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
667 | { |
668 | struct file_lock *fl = *thisfl_p; | |
669 | ||
88974691 JL |
670 | locks_delete_global_locks(fl); |
671 | ||
1da177e4 LT |
672 | *thisfl_p = fl->fl_next; |
673 | fl->fl_next = NULL; | |
1da177e4 | 674 | |
ab1f1611 VG |
675 | if (fl->fl_nspid) { |
676 | put_pid(fl->fl_nspid); | |
677 | fl->fl_nspid = NULL; | |
678 | } | |
679 | ||
1da177e4 | 680 | locks_wake_up_blocks(fl); |
24cbe784 JL |
681 | } |
682 | ||
683 | /* | |
684 | * Unlink a lock from all lists and free it. | |
685 | * | |
686 | * Must be called with i_lock held! | |
687 | */ | |
688 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
689 | { | |
690 | struct file_lock *fl = *thisfl_p; | |
691 | ||
692 | locks_unlink_lock(thisfl_p); | |
1da177e4 LT |
693 | locks_free_lock(fl); |
694 | } | |
695 | ||
696 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
697 | * checks for shared/exclusive status of overlapping locks. | |
698 | */ | |
699 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
700 | { | |
701 | if (sys_fl->fl_type == F_WRLCK) | |
702 | return 1; | |
703 | if (caller_fl->fl_type == F_WRLCK) | |
704 | return 1; | |
705 | return 0; | |
706 | } | |
707 | ||
708 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
709 | * checking before calling the locks_conflict(). | |
710 | */ | |
711 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
712 | { | |
713 | /* POSIX locks owned by the same process do not conflict with | |
714 | * each other. | |
715 | */ | |
716 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
717 | return (0); | |
718 | ||
719 | /* Check whether they overlap */ | |
720 | if (!locks_overlap(caller_fl, sys_fl)) | |
721 | return 0; | |
722 | ||
723 | return (locks_conflict(caller_fl, sys_fl)); | |
724 | } | |
725 | ||
726 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
727 | * checking before calling the locks_conflict(). | |
728 | */ | |
729 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
730 | { | |
731 | /* FLOCK locks referring to the same filp do not conflict with | |
732 | * each other. | |
733 | */ | |
734 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
735 | return (0); | |
736 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
737 | return 0; | |
738 | ||
739 | return (locks_conflict(caller_fl, sys_fl)); | |
740 | } | |
741 | ||
6d34ac19 | 742 | void |
9d6a8c5c | 743 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
744 | { |
745 | struct file_lock *cfl; | |
1c8c601a | 746 | struct inode *inode = file_inode(filp); |
1da177e4 | 747 | |
1c8c601a | 748 | spin_lock(&inode->i_lock); |
496ad9aa | 749 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
750 | if (!IS_POSIX(cfl)) |
751 | continue; | |
b842e240 | 752 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
753 | break; |
754 | } | |
ab1f1611 | 755 | if (cfl) { |
9d6a8c5c | 756 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 757 | if (cfl->fl_nspid) |
6c5f3e7b | 758 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 759 | } else |
129a84de | 760 | fl->fl_type = F_UNLCK; |
1c8c601a | 761 | spin_unlock(&inode->i_lock); |
6d34ac19 | 762 | return; |
1da177e4 | 763 | } |
1da177e4 LT |
764 | EXPORT_SYMBOL(posix_test_lock); |
765 | ||
b533184f BF |
766 | /* |
767 | * Deadlock detection: | |
768 | * | |
769 | * We attempt to detect deadlocks that are due purely to posix file | |
770 | * locks. | |
1da177e4 | 771 | * |
b533184f BF |
772 | * We assume that a task can be waiting for at most one lock at a time. |
773 | * So for any acquired lock, the process holding that lock may be | |
774 | * waiting on at most one other lock. That lock in turns may be held by | |
775 | * someone waiting for at most one other lock. Given a requested lock | |
776 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
777 | * follow this chain of waiters to ensure we are not about to create a | |
778 | * cycle. | |
1da177e4 | 779 | * |
b533184f BF |
780 | * Since we do this before we ever put a process to sleep on a lock, we |
781 | * are ensured that there is never a cycle; that is what guarantees that | |
782 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 783 | * |
b533184f BF |
784 | * Note: the above assumption may not be true when handling lock |
785 | * requests from a broken NFS client. It may also fail in the presence | |
786 | * of tasks (such as posix threads) sharing the same open file table. | |
787 | * | |
788 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 789 | */ |
97855b49 BF |
790 | |
791 | #define MAX_DEADLK_ITERATIONS 10 | |
792 | ||
b533184f BF |
793 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
794 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
795 | { | |
796 | struct file_lock *fl; | |
797 | ||
3999e493 | 798 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
799 | if (posix_same_owner(fl, block_fl)) |
800 | return fl->fl_next; | |
801 | } | |
802 | return NULL; | |
803 | } | |
804 | ||
7b2296af | 805 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 806 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
807 | struct file_lock *block_fl) |
808 | { | |
97855b49 | 809 | int i = 0; |
1da177e4 | 810 | |
b533184f BF |
811 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
812 | if (i++ > MAX_DEADLK_ITERATIONS) | |
813 | return 0; | |
814 | if (posix_same_owner(caller_fl, block_fl)) | |
815 | return 1; | |
1da177e4 LT |
816 | } |
817 | return 0; | |
818 | } | |
819 | ||
1da177e4 | 820 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 821 | * after any leases, but before any posix locks. |
f475ae95 TM |
822 | * |
823 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
824 | * whether or not a lock was successfully freed by testing the return | |
825 | * value for -ENOENT. | |
1da177e4 | 826 | */ |
993dfa87 | 827 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 828 | { |
993dfa87 | 829 | struct file_lock *new_fl = NULL; |
1da177e4 | 830 | struct file_lock **before; |
496ad9aa | 831 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
832 | int error = 0; |
833 | int found = 0; | |
834 | ||
b89f4321 | 835 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 836 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
837 | if (!new_fl) |
838 | return -ENOMEM; | |
84d535ad PE |
839 | } |
840 | ||
1c8c601a | 841 | spin_lock(&inode->i_lock); |
b89f4321 AB |
842 | if (request->fl_flags & FL_ACCESS) |
843 | goto find_conflict; | |
844 | ||
1da177e4 LT |
845 | for_each_lock(inode, before) { |
846 | struct file_lock *fl = *before; | |
847 | if (IS_POSIX(fl)) | |
848 | break; | |
849 | if (IS_LEASE(fl)) | |
850 | continue; | |
851 | if (filp != fl->fl_file) | |
852 | continue; | |
993dfa87 | 853 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
854 | goto out; |
855 | found = 1; | |
856 | locks_delete_lock(before); | |
857 | break; | |
858 | } | |
1da177e4 | 859 | |
f475ae95 TM |
860 | if (request->fl_type == F_UNLCK) { |
861 | if ((request->fl_flags & FL_EXISTS) && !found) | |
862 | error = -ENOENT; | |
993dfa87 | 863 | goto out; |
f475ae95 | 864 | } |
1da177e4 LT |
865 | |
866 | /* | |
867 | * If a higher-priority process was blocked on the old file lock, | |
868 | * give it the opportunity to lock the file. | |
869 | */ | |
b89f4321 | 870 | if (found) { |
1c8c601a | 871 | spin_unlock(&inode->i_lock); |
def01bc5 | 872 | cond_resched(); |
1c8c601a | 873 | spin_lock(&inode->i_lock); |
b89f4321 | 874 | } |
1da177e4 | 875 | |
f07f18dd | 876 | find_conflict: |
1da177e4 LT |
877 | for_each_lock(inode, before) { |
878 | struct file_lock *fl = *before; | |
879 | if (IS_POSIX(fl)) | |
880 | break; | |
881 | if (IS_LEASE(fl)) | |
882 | continue; | |
993dfa87 | 883 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
884 | continue; |
885 | error = -EAGAIN; | |
bde74e4b MS |
886 | if (!(request->fl_flags & FL_SLEEP)) |
887 | goto out; | |
888 | error = FILE_LOCK_DEFERRED; | |
889 | locks_insert_block(fl, request); | |
1da177e4 LT |
890 | goto out; |
891 | } | |
f07f18dd TM |
892 | if (request->fl_flags & FL_ACCESS) |
893 | goto out; | |
993dfa87 | 894 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 895 | locks_insert_lock(before, new_fl); |
993dfa87 | 896 | new_fl = NULL; |
9cedc194 | 897 | error = 0; |
1da177e4 LT |
898 | |
899 | out: | |
1c8c601a | 900 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
901 | if (new_fl) |
902 | locks_free_lock(new_fl); | |
1da177e4 LT |
903 | return error; |
904 | } | |
905 | ||
150b3934 | 906 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
907 | { |
908 | struct file_lock *fl; | |
39005d02 MS |
909 | struct file_lock *new_fl = NULL; |
910 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
911 | struct file_lock *left = NULL; |
912 | struct file_lock *right = NULL; | |
913 | struct file_lock **before; | |
b9746ef8 JL |
914 | int error; |
915 | bool added = false; | |
1da177e4 LT |
916 | |
917 | /* | |
918 | * We may need two file_lock structures for this operation, | |
919 | * so we get them in advance to avoid races. | |
39005d02 MS |
920 | * |
921 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 922 | */ |
39005d02 MS |
923 | if (!(request->fl_flags & FL_ACCESS) && |
924 | (request->fl_type != F_UNLCK || | |
925 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
926 | new_fl = locks_alloc_lock(); | |
927 | new_fl2 = locks_alloc_lock(); | |
928 | } | |
1da177e4 | 929 | |
1c8c601a | 930 | spin_lock(&inode->i_lock); |
1cb36012 JL |
931 | /* |
932 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
933 | * there are any, either return error or put the request on the | |
48f74186 | 934 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 935 | */ |
1da177e4 LT |
936 | if (request->fl_type != F_UNLCK) { |
937 | for_each_lock(inode, before) { | |
526985b9 | 938 | fl = *before; |
1da177e4 LT |
939 | if (!IS_POSIX(fl)) |
940 | continue; | |
941 | if (!posix_locks_conflict(request, fl)) | |
942 | continue; | |
5842add2 | 943 | if (conflock) |
1a747ee0 | 944 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
945 | error = -EAGAIN; |
946 | if (!(request->fl_flags & FL_SLEEP)) | |
947 | goto out; | |
1c8c601a JL |
948 | /* |
949 | * Deadlock detection and insertion into the blocked | |
950 | * locks list must be done while holding the same lock! | |
951 | */ | |
1da177e4 | 952 | error = -EDEADLK; |
7b2296af | 953 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
954 | if (likely(!posix_locks_deadlock(request, fl))) { |
955 | error = FILE_LOCK_DEFERRED; | |
956 | __locks_insert_block(fl, request); | |
957 | } | |
7b2296af | 958 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
959 | goto out; |
960 | } | |
961 | } | |
962 | ||
963 | /* If we're just looking for a conflict, we're done. */ | |
964 | error = 0; | |
965 | if (request->fl_flags & FL_ACCESS) | |
966 | goto out; | |
967 | ||
1da177e4 | 968 | /* |
1da177e4 LT |
969 | * Find the first old lock with the same owner as the new lock. |
970 | */ | |
971 | ||
972 | before = &inode->i_flock; | |
973 | ||
974 | /* First skip locks owned by other processes. */ | |
975 | while ((fl = *before) && (!IS_POSIX(fl) || | |
976 | !posix_same_owner(request, fl))) { | |
977 | before = &fl->fl_next; | |
978 | } | |
979 | ||
1cb36012 | 980 | /* Process locks with this owner. */ |
1da177e4 LT |
981 | while ((fl = *before) && posix_same_owner(request, fl)) { |
982 | /* Detect adjacent or overlapping regions (if same lock type) | |
983 | */ | |
984 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
985 | /* In all comparisons of start vs end, use |
986 | * "start - 1" rather than "end + 1". If end | |
987 | * is OFFSET_MAX, end + 1 will become negative. | |
988 | */ | |
1da177e4 LT |
989 | if (fl->fl_end < request->fl_start - 1) |
990 | goto next_lock; | |
991 | /* If the next lock in the list has entirely bigger | |
992 | * addresses than the new one, insert the lock here. | |
993 | */ | |
449231d6 | 994 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
995 | break; |
996 | ||
997 | /* If we come here, the new and old lock are of the | |
998 | * same type and adjacent or overlapping. Make one | |
999 | * lock yielding from the lower start address of both | |
1000 | * locks to the higher end address. | |
1001 | */ | |
1002 | if (fl->fl_start > request->fl_start) | |
1003 | fl->fl_start = request->fl_start; | |
1004 | else | |
1005 | request->fl_start = fl->fl_start; | |
1006 | if (fl->fl_end < request->fl_end) | |
1007 | fl->fl_end = request->fl_end; | |
1008 | else | |
1009 | request->fl_end = fl->fl_end; | |
1010 | if (added) { | |
1011 | locks_delete_lock(before); | |
1012 | continue; | |
1013 | } | |
1014 | request = fl; | |
b9746ef8 | 1015 | added = true; |
1da177e4 LT |
1016 | } |
1017 | else { | |
1018 | /* Processing for different lock types is a bit | |
1019 | * more complex. | |
1020 | */ | |
1021 | if (fl->fl_end < request->fl_start) | |
1022 | goto next_lock; | |
1023 | if (fl->fl_start > request->fl_end) | |
1024 | break; | |
1025 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1026 | added = true; |
1da177e4 LT |
1027 | if (fl->fl_start < request->fl_start) |
1028 | left = fl; | |
1029 | /* If the next lock in the list has a higher end | |
1030 | * address than the new one, insert the new one here. | |
1031 | */ | |
1032 | if (fl->fl_end > request->fl_end) { | |
1033 | right = fl; | |
1034 | break; | |
1035 | } | |
1036 | if (fl->fl_start >= request->fl_start) { | |
1037 | /* The new lock completely replaces an old | |
1038 | * one (This may happen several times). | |
1039 | */ | |
1040 | if (added) { | |
1041 | locks_delete_lock(before); | |
1042 | continue; | |
1043 | } | |
1044 | /* Replace the old lock with the new one. | |
1045 | * Wake up anybody waiting for the old one, | |
1046 | * as the change in lock type might satisfy | |
1047 | * their needs. | |
1048 | */ | |
1049 | locks_wake_up_blocks(fl); | |
1050 | fl->fl_start = request->fl_start; | |
1051 | fl->fl_end = request->fl_end; | |
1052 | fl->fl_type = request->fl_type; | |
47831f35 TM |
1053 | locks_release_private(fl); |
1054 | locks_copy_private(fl, request); | |
1da177e4 | 1055 | request = fl; |
b9746ef8 | 1056 | added = true; |
1da177e4 LT |
1057 | } |
1058 | } | |
1059 | /* Go on to next lock. | |
1060 | */ | |
1061 | next_lock: | |
1062 | before = &fl->fl_next; | |
1063 | } | |
1064 | ||
0d9a490a | 1065 | /* |
1cb36012 JL |
1066 | * The above code only modifies existing locks in case of merging or |
1067 | * replacing. If new lock(s) need to be inserted all modifications are | |
1068 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1069 | */ |
1070 | error = -ENOLCK; /* "no luck" */ | |
1071 | if (right && left == right && !new_fl2) | |
1072 | goto out; | |
1073 | ||
1da177e4 LT |
1074 | error = 0; |
1075 | if (!added) { | |
f475ae95 TM |
1076 | if (request->fl_type == F_UNLCK) { |
1077 | if (request->fl_flags & FL_EXISTS) | |
1078 | error = -ENOENT; | |
1da177e4 | 1079 | goto out; |
f475ae95 | 1080 | } |
0d9a490a MS |
1081 | |
1082 | if (!new_fl) { | |
1083 | error = -ENOLCK; | |
1084 | goto out; | |
1085 | } | |
1da177e4 LT |
1086 | locks_copy_lock(new_fl, request); |
1087 | locks_insert_lock(before, new_fl); | |
1088 | new_fl = NULL; | |
1089 | } | |
1090 | if (right) { | |
1091 | if (left == right) { | |
1092 | /* The new lock breaks the old one in two pieces, | |
1093 | * so we have to use the second new lock. | |
1094 | */ | |
1095 | left = new_fl2; | |
1096 | new_fl2 = NULL; | |
1097 | locks_copy_lock(left, right); | |
1098 | locks_insert_lock(before, left); | |
1099 | } | |
1100 | right->fl_start = request->fl_end + 1; | |
1101 | locks_wake_up_blocks(right); | |
1102 | } | |
1103 | if (left) { | |
1104 | left->fl_end = request->fl_start - 1; | |
1105 | locks_wake_up_blocks(left); | |
1106 | } | |
1107 | out: | |
1c8c601a | 1108 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1109 | /* |
1110 | * Free any unused locks. | |
1111 | */ | |
1112 | if (new_fl) | |
1113 | locks_free_lock(new_fl); | |
1114 | if (new_fl2) | |
1115 | locks_free_lock(new_fl2); | |
1116 | return error; | |
1117 | } | |
1118 | ||
1119 | /** | |
1120 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1121 | * @filp: The file to apply the lock to | |
1122 | * @fl: The lock to be applied | |
150b3934 | 1123 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1124 | * |
1125 | * Add a POSIX style lock to a file. | |
1126 | * We merge adjacent & overlapping locks whenever possible. | |
1127 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1128 | * |
1129 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1130 | * whether or not a lock was successfully freed by testing the return | |
1131 | * value for -ENOENT. | |
1da177e4 | 1132 | */ |
150b3934 | 1133 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1134 | struct file_lock *conflock) |
1135 | { | |
496ad9aa | 1136 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1137 | } |
150b3934 | 1138 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1139 | |
1140 | /** | |
1141 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1142 | * @filp: The file to apply the lock to | |
1143 | * @fl: The lock to be applied | |
1144 | * | |
1145 | * Add a POSIX style lock to a file. | |
1146 | * We merge adjacent & overlapping locks whenever possible. | |
1147 | * POSIX locks are sorted by owner task, then by starting address | |
1148 | */ | |
1149 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1150 | { | |
1151 | int error; | |
1152 | might_sleep (); | |
1153 | for (;;) { | |
150b3934 | 1154 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1155 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1156 | break; |
1157 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1158 | if (!error) | |
1159 | continue; | |
1160 | ||
1161 | locks_delete_block(fl); | |
1162 | break; | |
1163 | } | |
1164 | return error; | |
1165 | } | |
1166 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1167 | ||
1168 | /** | |
1169 | * locks_mandatory_locked - Check for an active lock | |
1170 | * @inode: the file to check | |
1171 | * | |
1172 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1173 | * This function is called from locks_verify_locked() only. | |
1174 | */ | |
1175 | int locks_mandatory_locked(struct inode *inode) | |
1176 | { | |
1177 | fl_owner_t owner = current->files; | |
1178 | struct file_lock *fl; | |
1179 | ||
1180 | /* | |
1181 | * Search the lock list for this inode for any POSIX locks. | |
1182 | */ | |
1c8c601a | 1183 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1184 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1185 | if (!IS_POSIX(fl)) | |
1186 | continue; | |
1187 | if (fl->fl_owner != owner) | |
1188 | break; | |
1189 | } | |
1c8c601a | 1190 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1191 | return fl ? -EAGAIN : 0; |
1192 | } | |
1193 | ||
1194 | /** | |
1195 | * locks_mandatory_area - Check for a conflicting lock | |
1196 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1197 | * for shared | |
1198 | * @inode: the file to check | |
1199 | * @filp: how the file was opened (if it was) | |
1200 | * @offset: start of area to check | |
1201 | * @count: length of area to check | |
1202 | * | |
1203 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1204 | * This function is called from rw_verify_area() and | |
1205 | * locks_verify_truncate(). | |
1206 | */ | |
1207 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1208 | struct file *filp, loff_t offset, | |
1209 | size_t count) | |
1210 | { | |
1211 | struct file_lock fl; | |
1212 | int error; | |
1213 | ||
1214 | locks_init_lock(&fl); | |
1215 | fl.fl_owner = current->files; | |
1216 | fl.fl_pid = current->tgid; | |
1217 | fl.fl_file = filp; | |
1218 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1219 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1220 | fl.fl_flags |= FL_SLEEP; | |
1221 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1222 | fl.fl_start = offset; | |
1223 | fl.fl_end = offset + count - 1; | |
1224 | ||
1225 | for (;;) { | |
150b3934 | 1226 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1227 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1228 | break; |
1229 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1230 | if (!error) { | |
1231 | /* | |
1232 | * If we've been sleeping someone might have | |
1233 | * changed the permissions behind our back. | |
1234 | */ | |
a16877ca | 1235 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1236 | continue; |
1237 | } | |
1238 | ||
1239 | locks_delete_block(&fl); | |
1240 | break; | |
1241 | } | |
1242 | ||
1243 | return error; | |
1244 | } | |
1245 | ||
1246 | EXPORT_SYMBOL(locks_mandatory_area); | |
1247 | ||
778fc546 BF |
1248 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1249 | { | |
1250 | switch (arg) { | |
1251 | case F_UNLCK: | |
1252 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1253 | /* fall through: */ | |
1254 | case F_RDLCK: | |
1255 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1256 | } | |
1257 | } | |
1258 | ||
1da177e4 LT |
1259 | /* We already had a lease on this file; just change its type */ |
1260 | int lease_modify(struct file_lock **before, int arg) | |
1261 | { | |
1262 | struct file_lock *fl = *before; | |
1263 | int error = assign_type(fl, arg); | |
1264 | ||
1265 | if (error) | |
1266 | return error; | |
778fc546 | 1267 | lease_clear_pending(fl, arg); |
1da177e4 | 1268 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1269 | if (arg == F_UNLCK) { |
1270 | struct file *filp = fl->fl_file; | |
1271 | ||
1272 | f_delown(filp); | |
1273 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1274 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1275 | if (fl->fl_fasync != NULL) { | |
1276 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1277 | fl->fl_fasync = NULL; | |
1278 | } | |
1da177e4 | 1279 | locks_delete_lock(before); |
3b6e2723 | 1280 | } |
1da177e4 LT |
1281 | return 0; |
1282 | } | |
1283 | ||
1284 | EXPORT_SYMBOL(lease_modify); | |
1285 | ||
778fc546 BF |
1286 | static bool past_time(unsigned long then) |
1287 | { | |
1288 | if (!then) | |
1289 | /* 0 is a special value meaning "this never expires": */ | |
1290 | return false; | |
1291 | return time_after(jiffies, then); | |
1292 | } | |
1293 | ||
1da177e4 LT |
1294 | static void time_out_leases(struct inode *inode) |
1295 | { | |
1296 | struct file_lock **before; | |
1297 | struct file_lock *fl; | |
1298 | ||
1299 | before = &inode->i_flock; | |
ab83fa4b | 1300 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
778fc546 BF |
1301 | if (past_time(fl->fl_downgrade_time)) |
1302 | lease_modify(before, F_RDLCK); | |
1303 | if (past_time(fl->fl_break_time)) | |
1304 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1305 | if (fl == *before) /* lease_modify may have freed fl */ |
1306 | before = &fl->fl_next; | |
1307 | } | |
1308 | } | |
1309 | ||
df4e8d2c BF |
1310 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1311 | { | |
1312 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1313 | return false; | |
1314 | return locks_conflict(breaker, lease); | |
1315 | } | |
1316 | ||
1da177e4 LT |
1317 | /** |
1318 | * __break_lease - revoke all outstanding leases on file | |
1319 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1320 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1321 | * break all leases | |
1322 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1323 | * only delegations | |
1da177e4 | 1324 | * |
87250dd2 | 1325 | * break_lease (inlined for speed) has checked there already is at least |
1326 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1327 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1328 | * specified %O_NONBLOCK to your open(). |
1329 | */ | |
df4e8d2c | 1330 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1331 | { |
778fc546 | 1332 | int error = 0; |
1da177e4 LT |
1333 | struct file_lock *new_fl, *flock; |
1334 | struct file_lock *fl; | |
1da177e4 LT |
1335 | unsigned long break_time; |
1336 | int i_have_this_lease = 0; | |
df4e8d2c | 1337 | bool lease_conflict = false; |
8737c930 | 1338 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1339 | |
8737c930 | 1340 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1341 | if (IS_ERR(new_fl)) |
1342 | return PTR_ERR(new_fl); | |
df4e8d2c | 1343 | new_fl->fl_flags = type; |
1da177e4 | 1344 | |
1c8c601a | 1345 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1346 | |
1347 | time_out_leases(inode); | |
1348 | ||
1349 | flock = inode->i_flock; | |
1350 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1351 | goto out; | |
1352 | ||
df4e8d2c BF |
1353 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1354 | if (leases_conflict(fl, new_fl)) { | |
1355 | lease_conflict = true; | |
1356 | if (fl->fl_owner == current->files) | |
1357 | i_have_this_lease = 1; | |
1358 | } | |
1359 | } | |
1360 | if (!lease_conflict) | |
778fc546 BF |
1361 | goto out; |
1362 | ||
1da177e4 LT |
1363 | break_time = 0; |
1364 | if (lease_break_time > 0) { | |
1365 | break_time = jiffies + lease_break_time * HZ; | |
1366 | if (break_time == 0) | |
1367 | break_time++; /* so that 0 means no break time */ | |
1368 | } | |
1369 | ||
1370 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
df4e8d2c BF |
1371 | if (!leases_conflict(fl, new_fl)) |
1372 | continue; | |
778fc546 BF |
1373 | if (want_write) { |
1374 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1375 | continue; | |
1376 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1377 | fl->fl_break_time = break_time; |
778fc546 BF |
1378 | } else { |
1379 | if (lease_breaking(flock)) | |
1380 | continue; | |
1381 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1382 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1383 | } |
778fc546 | 1384 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1385 | } |
1386 | ||
1387 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1388 | error = -EWOULDBLOCK; | |
1389 | goto out; | |
1390 | } | |
1391 | ||
1392 | restart: | |
1393 | break_time = flock->fl_break_time; | |
1394 | if (break_time != 0) { | |
1395 | break_time -= jiffies; | |
1396 | if (break_time == 0) | |
1397 | break_time++; | |
1398 | } | |
4321e01e | 1399 | locks_insert_block(flock, new_fl); |
1c8c601a | 1400 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1401 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1402 | !new_fl->fl_next, break_time); | |
1c8c601a JL |
1403 | spin_lock(&inode->i_lock); |
1404 | locks_delete_block(new_fl); | |
1da177e4 LT |
1405 | if (error >= 0) { |
1406 | if (error == 0) | |
1407 | time_out_leases(inode); | |
778fc546 BF |
1408 | /* |
1409 | * Wait for the next conflicting lease that has not been | |
1410 | * broken yet | |
1411 | */ | |
1da177e4 LT |
1412 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1413 | flock = flock->fl_next) { | |
df4e8d2c | 1414 | if (leases_conflict(new_fl, flock)) |
1da177e4 LT |
1415 | goto restart; |
1416 | } | |
1417 | error = 0; | |
1418 | } | |
1419 | ||
1420 | out: | |
1c8c601a | 1421 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1422 | locks_free_lock(new_fl); |
1da177e4 LT |
1423 | return error; |
1424 | } | |
1425 | ||
1426 | EXPORT_SYMBOL(__break_lease); | |
1427 | ||
1428 | /** | |
a6b91919 | 1429 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1430 | * @inode: the inode |
1431 | * @time: pointer to a timespec which will contain the last modified time | |
1432 | * | |
1433 | * This is to force NFS clients to flush their caches for files with | |
1434 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1435 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1436 | */ |
1437 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1438 | { | |
1439 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1440 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1441 | *time = current_fs_time(inode->i_sb); |
1442 | else | |
1443 | *time = inode->i_mtime; | |
1444 | } | |
1445 | ||
1446 | EXPORT_SYMBOL(lease_get_mtime); | |
1447 | ||
1448 | /** | |
1449 | * fcntl_getlease - Enquire what lease is currently active | |
1450 | * @filp: the file | |
1451 | * | |
1452 | * The value returned by this function will be one of | |
1453 | * (if no lease break is pending): | |
1454 | * | |
1455 | * %F_RDLCK to indicate a shared lease is held. | |
1456 | * | |
1457 | * %F_WRLCK to indicate an exclusive lease is held. | |
1458 | * | |
1459 | * %F_UNLCK to indicate no lease is held. | |
1460 | * | |
1461 | * (if a lease break is pending): | |
1462 | * | |
1463 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1464 | * changed to a shared lease (or removed). | |
1465 | * | |
1466 | * %F_UNLCK to indicate the lease needs to be removed. | |
1467 | * | |
1468 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1469 | * should be returned to userspace. | |
1470 | */ | |
1471 | int fcntl_getlease(struct file *filp) | |
1472 | { | |
1473 | struct file_lock *fl; | |
1c8c601a | 1474 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1475 | int type = F_UNLCK; |
1476 | ||
1c8c601a | 1477 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1478 | time_out_leases(file_inode(filp)); |
1479 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1480 | fl = fl->fl_next) { |
1481 | if (fl->fl_file == filp) { | |
778fc546 | 1482 | type = target_leasetype(fl); |
1da177e4 LT |
1483 | break; |
1484 | } | |
1485 | } | |
1c8c601a | 1486 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1487 | return type; |
1488 | } | |
1489 | ||
24cbe784 JL |
1490 | /** |
1491 | * check_conflicting_open - see if the given dentry points to a file that has | |
1492 | * an existing open that would conflict with the | |
1493 | * desired lease. | |
1494 | * @dentry: dentry to check | |
1495 | * @arg: type of lease that we're trying to acquire | |
1496 | * | |
1497 | * Check to see if there's an existing open fd on this file that would | |
1498 | * conflict with the lease we're trying to set. | |
1499 | */ | |
1500 | static int | |
1501 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1502 | { | |
1503 | int ret = 0; | |
1504 | struct inode *inode = dentry->d_inode; | |
1505 | ||
1506 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1507 | return -EAGAIN; | |
1508 | ||
1509 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1510 | (atomic_read(&inode->i_count) > 1))) | |
1511 | ret = -EAGAIN; | |
1512 | ||
1513 | return ret; | |
1514 | } | |
1515 | ||
d4f22d19 | 1516 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1517 | { |
7eaae282 | 1518 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1519 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1520 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1521 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1522 | int error; |
1da177e4 | 1523 | |
096657b6 | 1524 | lease = *flp; |
df4e8d2c BF |
1525 | /* |
1526 | * In the delegation case we need mutual exclusion with | |
1527 | * a number of operations that take the i_mutex. We trylock | |
1528 | * because delegations are an optional optimization, and if | |
1529 | * there's some chance of a conflict--we'd rather not | |
1530 | * bother, maybe that's a sign this just isn't a good file to | |
1531 | * hand out a delegation on. | |
1532 | */ | |
1533 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1534 | return -EAGAIN; | |
1535 | ||
1536 | if (is_deleg && arg == F_WRLCK) { | |
1537 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1538 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1539 | WARN_ON_ONCE(1); |
1540 | return -EINVAL; | |
1541 | } | |
096657b6 | 1542 | |
24cbe784 JL |
1543 | error = check_conflicting_open(dentry, arg); |
1544 | if (error) | |
096657b6 | 1545 | goto out; |
6d5e8b05 | 1546 | |
1da177e4 LT |
1547 | /* |
1548 | * At this point, we know that if there is an exclusive | |
1549 | * lease on this file, then we hold it on this filp | |
1550 | * (otherwise our open of this file would have blocked). | |
1551 | * And if we are trying to acquire an exclusive lease, | |
1552 | * then the file is not open by anyone (including us) | |
1553 | * except for this filp. | |
1554 | */ | |
c1f24ef4 | 1555 | error = -EAGAIN; |
1da177e4 LT |
1556 | for (before = &inode->i_flock; |
1557 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1558 | before = &fl->fl_next) { | |
c1f24ef4 | 1559 | if (fl->fl_file == filp) { |
1da177e4 | 1560 | my_before = before; |
c1f24ef4 BF |
1561 | continue; |
1562 | } | |
1563 | /* | |
1564 | * No exclusive leases if someone else has a lease on | |
1565 | * this file: | |
1566 | */ | |
1567 | if (arg == F_WRLCK) | |
1568 | goto out; | |
1569 | /* | |
1570 | * Modifying our existing lease is OK, but no getting a | |
1571 | * new lease if someone else is opening for write: | |
1572 | */ | |
1573 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1574 | goto out; | |
1da177e4 LT |
1575 | } |
1576 | ||
1da177e4 | 1577 | if (my_before != NULL) { |
8fb47a4f | 1578 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1579 | if (!error) |
1580 | *flp = *my_before; | |
1da177e4 LT |
1581 | goto out; |
1582 | } | |
1583 | ||
1da177e4 LT |
1584 | error = -EINVAL; |
1585 | if (!leases_enable) | |
1586 | goto out; | |
1587 | ||
c5b1f0d9 | 1588 | locks_insert_lock(before, lease); |
24cbe784 JL |
1589 | /* |
1590 | * The check in break_lease() is lockless. It's possible for another | |
1591 | * open to race in after we did the earlier check for a conflicting | |
1592 | * open but before the lease was inserted. Check again for a | |
1593 | * conflicting open and cancel the lease if there is one. | |
1594 | * | |
1595 | * We also add a barrier here to ensure that the insertion of the lock | |
1596 | * precedes these checks. | |
1597 | */ | |
1598 | smp_mb(); | |
1599 | error = check_conflicting_open(dentry, arg); | |
1600 | if (error) | |
1601 | locks_unlink_lock(flp); | |
1da177e4 | 1602 | out: |
df4e8d2c BF |
1603 | if (is_deleg) |
1604 | mutex_unlock(&inode->i_mutex); | |
1da177e4 LT |
1605 | return error; |
1606 | } | |
8335ebd9 | 1607 | |
d4f22d19 | 1608 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1609 | { |
1610 | struct file_lock *fl, **before; | |
1611 | struct dentry *dentry = filp->f_path.dentry; | |
1612 | struct inode *inode = dentry->d_inode; | |
1613 | ||
1614 | for (before = &inode->i_flock; | |
1615 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1616 | before = &fl->fl_next) { | |
1617 | if (fl->fl_file != filp) | |
1618 | continue; | |
1619 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1620 | } | |
1621 | return -EAGAIN; | |
1622 | } | |
1623 | ||
1624 | /** | |
1625 | * generic_setlease - sets a lease on an open file | |
1626 | * @filp: file pointer | |
1627 | * @arg: type of lease to obtain | |
1628 | * @flp: input - file_lock to use, output - file_lock inserted | |
1629 | * | |
1630 | * The (input) flp->fl_lmops->lm_break function is required | |
1631 | * by break_lease(). | |
1632 | * | |
1c8c601a | 1633 | * Called with inode->i_lock held. |
8335ebd9 BF |
1634 | */ |
1635 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1636 | { | |
1637 | struct dentry *dentry = filp->f_path.dentry; | |
1638 | struct inode *inode = dentry->d_inode; | |
1639 | int error; | |
1640 | ||
8e96e3b7 | 1641 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1642 | return -EACCES; |
1643 | if (!S_ISREG(inode->i_mode)) | |
1644 | return -EINVAL; | |
1645 | error = security_file_lock(filp, arg); | |
1646 | if (error) | |
1647 | return error; | |
1648 | ||
1649 | time_out_leases(inode); | |
1650 | ||
1651 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1652 | ||
1653 | switch (arg) { | |
1654 | case F_UNLCK: | |
1655 | return generic_delete_lease(filp, flp); | |
1656 | case F_RDLCK: | |
1657 | case F_WRLCK: | |
1658 | return generic_add_lease(filp, arg, flp); | |
1659 | default: | |
8d657eb3 | 1660 | return -EINVAL; |
8335ebd9 BF |
1661 | } |
1662 | } | |
0af1a450 | 1663 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1664 | |
b89f4321 AB |
1665 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1666 | { | |
72c2d531 | 1667 | if (filp->f_op->setlease) |
b89f4321 AB |
1668 | return filp->f_op->setlease(filp, arg, lease); |
1669 | else | |
1670 | return generic_setlease(filp, arg, lease); | |
1671 | } | |
1672 | ||
1673 | /** | |
a9933cea | 1674 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1675 | * @filp: file pointer |
1676 | * @arg: type of lease to obtain | |
1677 | * @lease: file_lock to use | |
1678 | * | |
1679 | * Call this to establish a lease on the file. | |
8fb47a4f | 1680 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1681 | * break_lease will oops! |
1682 | * | |
1683 | * This will call the filesystem's setlease file method, if | |
1684 | * defined. Note that there is no getlease method; instead, the | |
1685 | * filesystem setlease method should call back to setlease() to | |
1686 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1687 | * find it. Since fcntl_getlease() only reports whether the current | |
1688 | * task holds a lease, a cluster filesystem need only do this for | |
1689 | * leases held by processes on this node. | |
1690 | * | |
1691 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1692 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1693 | * filesystem's open, create, and (on truncate) setattr methods. |
1694 | * | |
1695 | * Warning: the only current setlease methods exist only to disable | |
1696 | * leases in certain cases. More vfs changes may be required to | |
1697 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1698 | */ |
1699 | ||
a9933cea | 1700 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1701 | { |
1c8c601a | 1702 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1703 | int error; |
1704 | ||
1c8c601a | 1705 | spin_lock(&inode->i_lock); |
b89f4321 | 1706 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1707 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1708 | |
1709 | return error; | |
1710 | } | |
a9933cea | 1711 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1712 | |
0ceaf6c7 BF |
1713 | static int do_fcntl_delete_lease(struct file *filp) |
1714 | { | |
1715 | struct file_lock fl, *flp = &fl; | |
1716 | ||
1717 | lease_init(filp, F_UNLCK, flp); | |
1718 | ||
1719 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1720 | } | |
1721 | ||
1722 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1723 | { |
3df057ac | 1724 | struct file_lock *fl, *ret; |
1c8c601a | 1725 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1726 | struct fasync_struct *new; |
1da177e4 LT |
1727 | int error; |
1728 | ||
c5b1f0d9 AB |
1729 | fl = lease_alloc(filp, arg); |
1730 | if (IS_ERR(fl)) | |
1731 | return PTR_ERR(fl); | |
1da177e4 | 1732 | |
f7347ce4 LT |
1733 | new = fasync_alloc(); |
1734 | if (!new) { | |
1735 | locks_free_lock(fl); | |
1736 | return -ENOMEM; | |
1737 | } | |
3df057ac | 1738 | ret = fl; |
1c8c601a | 1739 | spin_lock(&inode->i_lock); |
8896b93f | 1740 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 | 1741 | if (error) { |
1c8c601a | 1742 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1743 | locks_free_lock(fl); |
1744 | goto out_free_fasync; | |
1745 | } | |
3df057ac BF |
1746 | if (ret != fl) |
1747 | locks_free_lock(fl); | |
1da177e4 | 1748 | |
f7347ce4 LT |
1749 | /* |
1750 | * fasync_insert_entry() returns the old entry if any. | |
1751 | * If there was no old entry, then it used 'new' and | |
1752 | * inserted it into the fasync list. Clear new so that | |
1753 | * we don't release it here. | |
1754 | */ | |
3df057ac | 1755 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1756 | new = NULL; |
1757 | ||
8896b93f | 1758 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1c8c601a | 1759 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1760 | |
1761 | out_free_fasync: | |
f7347ce4 LT |
1762 | if (new) |
1763 | fasync_free(new); | |
1da177e4 LT |
1764 | return error; |
1765 | } | |
1766 | ||
0ceaf6c7 BF |
1767 | /** |
1768 | * fcntl_setlease - sets a lease on an open file | |
1769 | * @fd: open file descriptor | |
1770 | * @filp: file pointer | |
1771 | * @arg: type of lease to obtain | |
1772 | * | |
1773 | * Call this fcntl to establish a lease on the file. | |
1774 | * Note that you also need to call %F_SETSIG to | |
1775 | * receive a signal when the lease is broken. | |
1776 | */ | |
1777 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1778 | { | |
1779 | if (arg == F_UNLCK) | |
1780 | return do_fcntl_delete_lease(filp); | |
1781 | return do_fcntl_add_lease(fd, filp, arg); | |
1782 | } | |
1783 | ||
1da177e4 LT |
1784 | /** |
1785 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1786 | * @filp: The file to apply the lock to | |
1787 | * @fl: The lock to be applied | |
1788 | * | |
1789 | * Add a FLOCK style lock to a file. | |
1790 | */ | |
1791 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1792 | { | |
1793 | int error; | |
1794 | might_sleep(); | |
1795 | for (;;) { | |
1796 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1797 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1798 | break; |
1799 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1800 | if (!error) | |
1801 | continue; | |
1802 | ||
1803 | locks_delete_block(fl); | |
1804 | break; | |
1805 | } | |
1806 | return error; | |
1807 | } | |
1808 | ||
1809 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1810 | ||
1811 | /** | |
1812 | * sys_flock: - flock() system call. | |
1813 | * @fd: the file descriptor to lock. | |
1814 | * @cmd: the type of lock to apply. | |
1815 | * | |
1816 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1817 | * The @cmd can be one of | |
1818 | * | |
1819 | * %LOCK_SH -- a shared lock. | |
1820 | * | |
1821 | * %LOCK_EX -- an exclusive lock. | |
1822 | * | |
1823 | * %LOCK_UN -- remove an existing lock. | |
1824 | * | |
1825 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1826 | * | |
1827 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1828 | * processes read and write access respectively. | |
1829 | */ | |
002c8976 | 1830 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1831 | { |
2903ff01 | 1832 | struct fd f = fdget(fd); |
1da177e4 LT |
1833 | struct file_lock *lock; |
1834 | int can_sleep, unlock; | |
1835 | int error; | |
1836 | ||
1837 | error = -EBADF; | |
2903ff01 | 1838 | if (!f.file) |
1da177e4 LT |
1839 | goto out; |
1840 | ||
1841 | can_sleep = !(cmd & LOCK_NB); | |
1842 | cmd &= ~LOCK_NB; | |
1843 | unlock = (cmd == LOCK_UN); | |
1844 | ||
aeb5d727 | 1845 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1846 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1847 | goto out_putf; |
1848 | ||
2903ff01 | 1849 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1850 | if (error) |
1851 | goto out_putf; | |
1852 | if (can_sleep) | |
1853 | lock->fl_flags |= FL_SLEEP; | |
1854 | ||
2903ff01 | 1855 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1856 | if (error) |
1857 | goto out_free; | |
1858 | ||
72c2d531 | 1859 | if (f.file->f_op->flock) |
2903ff01 | 1860 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1861 | (can_sleep) ? F_SETLKW : F_SETLK, |
1862 | lock); | |
1863 | else | |
2903ff01 | 1864 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1865 | |
1866 | out_free: | |
993dfa87 | 1867 | locks_free_lock(lock); |
1da177e4 LT |
1868 | |
1869 | out_putf: | |
2903ff01 | 1870 | fdput(f); |
1da177e4 LT |
1871 | out: |
1872 | return error; | |
1873 | } | |
1874 | ||
3ee17abd BF |
1875 | /** |
1876 | * vfs_test_lock - test file byte range lock | |
1877 | * @filp: The file to test lock for | |
6924c554 | 1878 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1879 | * |
1880 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1881 | * setting conf->fl_type to something other than F_UNLCK. | |
1882 | */ | |
1883 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1884 | { | |
72c2d531 | 1885 | if (filp->f_op->lock) |
3ee17abd BF |
1886 | return filp->f_op->lock(filp, F_GETLK, fl); |
1887 | posix_test_lock(filp, fl); | |
1888 | return 0; | |
1889 | } | |
1890 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1891 | ||
c2fa1b8a BF |
1892 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1893 | { | |
1894 | flock->l_pid = fl->fl_pid; | |
1895 | #if BITS_PER_LONG == 32 | |
1896 | /* | |
1897 | * Make sure we can represent the posix lock via | |
1898 | * legacy 32bit flock. | |
1899 | */ | |
1900 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1901 | return -EOVERFLOW; | |
1902 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1903 | return -EOVERFLOW; | |
1904 | #endif | |
1905 | flock->l_start = fl->fl_start; | |
1906 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1907 | fl->fl_end - fl->fl_start + 1; | |
1908 | flock->l_whence = 0; | |
129a84de | 1909 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1910 | return 0; |
1911 | } | |
1912 | ||
1913 | #if BITS_PER_LONG == 32 | |
1914 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1915 | { | |
1916 | flock->l_pid = fl->fl_pid; | |
1917 | flock->l_start = fl->fl_start; | |
1918 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1919 | fl->fl_end - fl->fl_start + 1; | |
1920 | flock->l_whence = 0; | |
1921 | flock->l_type = fl->fl_type; | |
1922 | } | |
1923 | #endif | |
1924 | ||
1da177e4 LT |
1925 | /* Report the first existing lock that would conflict with l. |
1926 | * This implements the F_GETLK command of fcntl(). | |
1927 | */ | |
1928 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1929 | { | |
9d6a8c5c | 1930 | struct file_lock file_lock; |
1da177e4 LT |
1931 | struct flock flock; |
1932 | int error; | |
1933 | ||
1934 | error = -EFAULT; | |
1935 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1936 | goto out; | |
1937 | error = -EINVAL; | |
1938 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1939 | goto out; | |
1940 | ||
1941 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1942 | if (error) | |
1943 | goto out; | |
1944 | ||
3ee17abd BF |
1945 | error = vfs_test_lock(filp, &file_lock); |
1946 | if (error) | |
1947 | goto out; | |
1da177e4 | 1948 | |
9d6a8c5c ME |
1949 | flock.l_type = file_lock.fl_type; |
1950 | if (file_lock.fl_type != F_UNLCK) { | |
1951 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1952 | if (error) |
1da177e4 | 1953 | goto out; |
1da177e4 LT |
1954 | } |
1955 | error = -EFAULT; | |
1956 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1957 | error = 0; | |
1958 | out: | |
1959 | return error; | |
1960 | } | |
1961 | ||
7723ec97 ME |
1962 | /** |
1963 | * vfs_lock_file - file byte range lock | |
1964 | * @filp: The file to apply the lock to | |
1965 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1966 | * @fl: The lock to be applied | |
150b3934 ME |
1967 | * @conf: Place to return a copy of the conflicting lock, if found. |
1968 | * | |
1969 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1970 | * as the final argument. | |
1971 | * | |
1972 | * If the filesystem defines a private ->lock() method, then @conf will | |
1973 | * be left unchanged; so a caller that cares should initialize it to | |
1974 | * some acceptable default. | |
2beb6614 ME |
1975 | * |
1976 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1977 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1978 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1979 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1980 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1981 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1982 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1983 | * request completes. |
1984 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1985 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1986 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1987 | * nonzero return code and the file system should release the lock. The file |
1988 | * system is also responsible to keep a corresponding posix lock when it | |
1989 | * grants a lock so the VFS can find out which locks are locally held and do | |
1990 | * the correct lock cleanup when required. | |
1991 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1992 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1993 | * return code. |
7723ec97 | 1994 | */ |
150b3934 | 1995 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 1996 | { |
72c2d531 | 1997 | if (filp->f_op->lock) |
7723ec97 ME |
1998 | return filp->f_op->lock(filp, cmd, fl); |
1999 | else | |
150b3934 | 2000 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2001 | } |
2002 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2003 | ||
b648a6de MS |
2004 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2005 | struct file_lock *fl) | |
2006 | { | |
2007 | int error; | |
2008 | ||
2009 | error = security_file_lock(filp, fl->fl_type); | |
2010 | if (error) | |
2011 | return error; | |
2012 | ||
764c76b3 MS |
2013 | for (;;) { |
2014 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2015 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2016 | break; |
764c76b3 MS |
2017 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2018 | if (!error) | |
2019 | continue; | |
2020 | ||
2021 | locks_delete_block(fl); | |
2022 | break; | |
b648a6de MS |
2023 | } |
2024 | ||
2025 | return error; | |
2026 | } | |
2027 | ||
1da177e4 LT |
2028 | /* Apply the lock described by l to an open file descriptor. |
2029 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2030 | */ | |
c293621b PS |
2031 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2032 | struct flock __user *l) | |
1da177e4 LT |
2033 | { |
2034 | struct file_lock *file_lock = locks_alloc_lock(); | |
2035 | struct flock flock; | |
2036 | struct inode *inode; | |
0b2bac2f | 2037 | struct file *f; |
1da177e4 LT |
2038 | int error; |
2039 | ||
2040 | if (file_lock == NULL) | |
2041 | return -ENOLCK; | |
2042 | ||
2043 | /* | |
2044 | * This might block, so we do it before checking the inode. | |
2045 | */ | |
2046 | error = -EFAULT; | |
2047 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2048 | goto out; | |
2049 | ||
496ad9aa | 2050 | inode = file_inode(filp); |
1da177e4 LT |
2051 | |
2052 | /* Don't allow mandatory locks on files that may be memory mapped | |
2053 | * and shared. | |
2054 | */ | |
a16877ca | 2055 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2056 | error = -EAGAIN; |
2057 | goto out; | |
2058 | } | |
2059 | ||
c293621b | 2060 | again: |
1da177e4 LT |
2061 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2062 | if (error) | |
2063 | goto out; | |
2064 | if (cmd == F_SETLKW) { | |
2065 | file_lock->fl_flags |= FL_SLEEP; | |
2066 | } | |
2067 | ||
2068 | error = -EBADF; | |
2069 | switch (flock.l_type) { | |
2070 | case F_RDLCK: | |
2071 | if (!(filp->f_mode & FMODE_READ)) | |
2072 | goto out; | |
2073 | break; | |
2074 | case F_WRLCK: | |
2075 | if (!(filp->f_mode & FMODE_WRITE)) | |
2076 | goto out; | |
2077 | break; | |
2078 | case F_UNLCK: | |
2079 | break; | |
2080 | default: | |
2081 | error = -EINVAL; | |
2082 | goto out; | |
2083 | } | |
2084 | ||
b648a6de | 2085 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2086 | |
c293621b PS |
2087 | /* |
2088 | * Attempt to detect a close/fcntl race and recover by | |
2089 | * releasing the lock that was just acquired. | |
2090 | */ | |
0b2bac2f AV |
2091 | /* |
2092 | * we need that spin_lock here - it prevents reordering between | |
2093 | * update of inode->i_flock and check for it done in close(). | |
2094 | * rcu_read_lock() wouldn't do. | |
2095 | */ | |
2096 | spin_lock(¤t->files->file_lock); | |
2097 | f = fcheck(fd); | |
2098 | spin_unlock(¤t->files->file_lock); | |
2099 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2100 | flock.l_type = F_UNLCK; |
2101 | goto again; | |
1da177e4 LT |
2102 | } |
2103 | ||
c293621b | 2104 | out: |
1da177e4 LT |
2105 | locks_free_lock(file_lock); |
2106 | return error; | |
2107 | } | |
2108 | ||
2109 | #if BITS_PER_LONG == 32 | |
2110 | /* Report the first existing lock that would conflict with l. | |
2111 | * This implements the F_GETLK command of fcntl(). | |
2112 | */ | |
2113 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
2114 | { | |
9d6a8c5c | 2115 | struct file_lock file_lock; |
1da177e4 LT |
2116 | struct flock64 flock; |
2117 | int error; | |
2118 | ||
2119 | error = -EFAULT; | |
2120 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2121 | goto out; | |
2122 | error = -EINVAL; | |
2123 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2124 | goto out; | |
2125 | ||
2126 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2127 | if (error) | |
2128 | goto out; | |
2129 | ||
3ee17abd BF |
2130 | error = vfs_test_lock(filp, &file_lock); |
2131 | if (error) | |
2132 | goto out; | |
2133 | ||
9d6a8c5c ME |
2134 | flock.l_type = file_lock.fl_type; |
2135 | if (file_lock.fl_type != F_UNLCK) | |
2136 | posix_lock_to_flock64(&flock, &file_lock); | |
2137 | ||
1da177e4 LT |
2138 | error = -EFAULT; |
2139 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2140 | error = 0; | |
2141 | ||
2142 | out: | |
2143 | return error; | |
2144 | } | |
2145 | ||
2146 | /* Apply the lock described by l to an open file descriptor. | |
2147 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2148 | */ | |
c293621b PS |
2149 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2150 | struct flock64 __user *l) | |
1da177e4 LT |
2151 | { |
2152 | struct file_lock *file_lock = locks_alloc_lock(); | |
2153 | struct flock64 flock; | |
2154 | struct inode *inode; | |
0b2bac2f | 2155 | struct file *f; |
1da177e4 LT |
2156 | int error; |
2157 | ||
2158 | if (file_lock == NULL) | |
2159 | return -ENOLCK; | |
2160 | ||
2161 | /* | |
2162 | * This might block, so we do it before checking the inode. | |
2163 | */ | |
2164 | error = -EFAULT; | |
2165 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2166 | goto out; | |
2167 | ||
496ad9aa | 2168 | inode = file_inode(filp); |
1da177e4 LT |
2169 | |
2170 | /* Don't allow mandatory locks on files that may be memory mapped | |
2171 | * and shared. | |
2172 | */ | |
a16877ca | 2173 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2174 | error = -EAGAIN; |
2175 | goto out; | |
2176 | } | |
2177 | ||
c293621b | 2178 | again: |
1da177e4 LT |
2179 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2180 | if (error) | |
2181 | goto out; | |
2182 | if (cmd == F_SETLKW64) { | |
2183 | file_lock->fl_flags |= FL_SLEEP; | |
2184 | } | |
2185 | ||
2186 | error = -EBADF; | |
2187 | switch (flock.l_type) { | |
2188 | case F_RDLCK: | |
2189 | if (!(filp->f_mode & FMODE_READ)) | |
2190 | goto out; | |
2191 | break; | |
2192 | case F_WRLCK: | |
2193 | if (!(filp->f_mode & FMODE_WRITE)) | |
2194 | goto out; | |
2195 | break; | |
2196 | case F_UNLCK: | |
2197 | break; | |
2198 | default: | |
2199 | error = -EINVAL; | |
2200 | goto out; | |
2201 | } | |
2202 | ||
b648a6de | 2203 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2204 | |
c293621b PS |
2205 | /* |
2206 | * Attempt to detect a close/fcntl race and recover by | |
2207 | * releasing the lock that was just acquired. | |
2208 | */ | |
0b2bac2f AV |
2209 | spin_lock(¤t->files->file_lock); |
2210 | f = fcheck(fd); | |
2211 | spin_unlock(¤t->files->file_lock); | |
2212 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2213 | flock.l_type = F_UNLCK; |
2214 | goto again; | |
1da177e4 LT |
2215 | } |
2216 | ||
2217 | out: | |
2218 | locks_free_lock(file_lock); | |
2219 | return error; | |
2220 | } | |
2221 | #endif /* BITS_PER_LONG == 32 */ | |
2222 | ||
2223 | /* | |
2224 | * This function is called when the file is being removed | |
2225 | * from the task's fd array. POSIX locks belonging to this task | |
2226 | * are deleted at this time. | |
2227 | */ | |
2228 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2229 | { | |
ff7b86b8 | 2230 | struct file_lock lock; |
1da177e4 LT |
2231 | |
2232 | /* | |
2233 | * If there are no locks held on this file, we don't need to call | |
2234 | * posix_lock_file(). Another process could be setting a lock on this | |
2235 | * file at the same time, but we wouldn't remove that lock anyway. | |
2236 | */ | |
496ad9aa | 2237 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2238 | return; |
2239 | ||
2240 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2241 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2242 | lock.fl_start = 0; |
2243 | lock.fl_end = OFFSET_MAX; | |
2244 | lock.fl_owner = owner; | |
2245 | lock.fl_pid = current->tgid; | |
2246 | lock.fl_file = filp; | |
2247 | lock.fl_ops = NULL; | |
2248 | lock.fl_lmops = NULL; | |
2249 | ||
150b3934 | 2250 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2251 | |
1da177e4 LT |
2252 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2253 | lock.fl_ops->fl_release_private(&lock); | |
2254 | } | |
2255 | ||
2256 | EXPORT_SYMBOL(locks_remove_posix); | |
2257 | ||
2258 | /* | |
2259 | * This function is called on the last close of an open file. | |
2260 | */ | |
2261 | void locks_remove_flock(struct file *filp) | |
2262 | { | |
496ad9aa | 2263 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2264 | struct file_lock *fl; |
2265 | struct file_lock **before; | |
2266 | ||
2267 | if (!inode->i_flock) | |
2268 | return; | |
2269 | ||
72c2d531 | 2270 | if (filp->f_op->flock) { |
1da177e4 LT |
2271 | struct file_lock fl = { |
2272 | .fl_pid = current->tgid, | |
2273 | .fl_file = filp, | |
2274 | .fl_flags = FL_FLOCK, | |
2275 | .fl_type = F_UNLCK, | |
2276 | .fl_end = OFFSET_MAX, | |
2277 | }; | |
2278 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2279 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2280 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2281 | } |
2282 | ||
1c8c601a | 2283 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2284 | before = &inode->i_flock; |
2285 | ||
2286 | while ((fl = *before) != NULL) { | |
2287 | if (fl->fl_file == filp) { | |
c293621b | 2288 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2289 | locks_delete_lock(before); |
2290 | continue; | |
2291 | } | |
2292 | if (IS_LEASE(fl)) { | |
2293 | lease_modify(before, F_UNLCK); | |
2294 | continue; | |
2295 | } | |
2296 | /* What? */ | |
2297 | BUG(); | |
2298 | } | |
2299 | before = &fl->fl_next; | |
2300 | } | |
1c8c601a | 2301 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2302 | } |
2303 | ||
1da177e4 LT |
2304 | /** |
2305 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2306 | * @waiter: the lock which was waiting |
2307 | * | |
2308 | * lockd needs to block waiting for locks. | |
2309 | */ | |
64a318ee | 2310 | int |
f891a29f | 2311 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2312 | { |
64a318ee BF |
2313 | int status = 0; |
2314 | ||
7b2296af | 2315 | spin_lock(&blocked_lock_lock); |
5996a298 | 2316 | if (waiter->fl_next) |
1da177e4 | 2317 | __locks_delete_block(waiter); |
64a318ee BF |
2318 | else |
2319 | status = -ENOENT; | |
7b2296af | 2320 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2321 | return status; |
1da177e4 | 2322 | } |
1da177e4 LT |
2323 | EXPORT_SYMBOL(posix_unblock_lock); |
2324 | ||
9b9d2ab4 ME |
2325 | /** |
2326 | * vfs_cancel_lock - file byte range unblock lock | |
2327 | * @filp: The file to apply the unblock to | |
2328 | * @fl: The lock to be unblocked | |
2329 | * | |
2330 | * Used by lock managers to cancel blocked requests | |
2331 | */ | |
2332 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2333 | { | |
72c2d531 | 2334 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2335 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2336 | return 0; | |
2337 | } | |
2338 | ||
2339 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2340 | ||
7f8ada98 | 2341 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2342 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2343 | #include <linux/seq_file.h> |
2344 | ||
7012b02a JL |
2345 | struct locks_iterator { |
2346 | int li_cpu; | |
2347 | loff_t li_pos; | |
2348 | }; | |
2349 | ||
7f8ada98 | 2350 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2351 | loff_t id, char *pfx) |
1da177e4 LT |
2352 | { |
2353 | struct inode *inode = NULL; | |
ab1f1611 VG |
2354 | unsigned int fl_pid; |
2355 | ||
2356 | if (fl->fl_nspid) | |
6c5f3e7b | 2357 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2358 | else |
2359 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2360 | |
2361 | if (fl->fl_file != NULL) | |
496ad9aa | 2362 | inode = file_inode(fl->fl_file); |
1da177e4 | 2363 | |
99dc8292 | 2364 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2365 | if (IS_POSIX(fl)) { |
7f8ada98 | 2366 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2367 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2368 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2369 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2370 | } else if (IS_FLOCK(fl)) { |
2371 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2372 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2373 | } else { |
7f8ada98 | 2374 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2375 | } |
2376 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2377 | seq_printf(f, "LEASE "); |
ab83fa4b | 2378 | if (lease_breaking(fl)) |
7f8ada98 | 2379 | seq_printf(f, "BREAKING "); |
1da177e4 | 2380 | else if (fl->fl_file) |
7f8ada98 | 2381 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2382 | else |
7f8ada98 | 2383 | seq_printf(f, "BREAKER "); |
1da177e4 | 2384 | } else { |
7f8ada98 | 2385 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2386 | } |
2387 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2388 | seq_printf(f, "%s ", |
1da177e4 LT |
2389 | (fl->fl_type & LOCK_READ) |
2390 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2391 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2392 | } else { | |
7f8ada98 | 2393 | seq_printf(f, "%s ", |
ab83fa4b | 2394 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2395 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2396 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2397 | } |
2398 | if (inode) { | |
2399 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2400 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2401 | inode->i_sb->s_id, inode->i_ino); |
2402 | #else | |
2403 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2404 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2405 | MAJOR(inode->i_sb->s_dev), |
2406 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2407 | #endif | |
2408 | } else { | |
ab1f1611 | 2409 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2410 | } |
2411 | if (IS_POSIX(fl)) { | |
2412 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2413 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2414 | else |
7f8ada98 | 2415 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2416 | } else { |
7f8ada98 | 2417 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2418 | } |
2419 | } | |
2420 | ||
7f8ada98 | 2421 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2422 | { |
7012b02a | 2423 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2424 | struct file_lock *fl, *bfl; |
1da177e4 | 2425 | |
139ca04e | 2426 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2427 | |
7012b02a | 2428 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2429 | |
7f8ada98 | 2430 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2431 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2432 | |
7f8ada98 PE |
2433 | return 0; |
2434 | } | |
1da177e4 | 2435 | |
7f8ada98 PE |
2436 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2437 | { | |
7012b02a | 2438 | struct locks_iterator *iter = f->private; |
99dc8292 | 2439 | |
7012b02a JL |
2440 | iter->li_pos = *pos + 1; |
2441 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2442 | spin_lock(&blocked_lock_lock); |
7012b02a | 2443 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2444 | } |
1da177e4 | 2445 | |
7f8ada98 PE |
2446 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2447 | { | |
7012b02a JL |
2448 | struct locks_iterator *iter = f->private; |
2449 | ||
2450 | ++iter->li_pos; | |
2451 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2452 | } |
1da177e4 | 2453 | |
7f8ada98 PE |
2454 | static void locks_stop(struct seq_file *f, void *v) |
2455 | { | |
7b2296af | 2456 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2457 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2458 | } |
2459 | ||
d8ba7a36 | 2460 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2461 | .start = locks_start, |
2462 | .next = locks_next, | |
2463 | .stop = locks_stop, | |
2464 | .show = locks_show, | |
2465 | }; | |
d8ba7a36 AD |
2466 | |
2467 | static int locks_open(struct inode *inode, struct file *filp) | |
2468 | { | |
7012b02a JL |
2469 | return seq_open_private(filp, &locks_seq_operations, |
2470 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2471 | } |
2472 | ||
2473 | static const struct file_operations proc_locks_operations = { | |
2474 | .open = locks_open, | |
2475 | .read = seq_read, | |
2476 | .llseek = seq_lseek, | |
99dc8292 | 2477 | .release = seq_release_private, |
d8ba7a36 AD |
2478 | }; |
2479 | ||
2480 | static int __init proc_locks_init(void) | |
2481 | { | |
2482 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2483 | return 0; | |
2484 | } | |
2485 | module_init(proc_locks_init); | |
7f8ada98 PE |
2486 | #endif |
2487 | ||
1da177e4 LT |
2488 | /** |
2489 | * lock_may_read - checks that the region is free of locks | |
2490 | * @inode: the inode that is being read | |
2491 | * @start: the first byte to read | |
2492 | * @len: the number of bytes to read | |
2493 | * | |
2494 | * Emulates Windows locking requirements. Whole-file | |
2495 | * mandatory locks (share modes) can prohibit a read and | |
2496 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2497 | * | |
2498 | * N.B. this function is only ever called | |
2499 | * from knfsd and ownership of locks is never checked. | |
2500 | */ | |
2501 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2502 | { | |
2503 | struct file_lock *fl; | |
2504 | int result = 1; | |
1c8c601a JL |
2505 | |
2506 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2507 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2508 | if (IS_POSIX(fl)) { | |
2509 | if (fl->fl_type == F_RDLCK) | |
2510 | continue; | |
2511 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2512 | continue; | |
2513 | } else if (IS_FLOCK(fl)) { | |
2514 | if (!(fl->fl_type & LOCK_MAND)) | |
2515 | continue; | |
2516 | if (fl->fl_type & LOCK_READ) | |
2517 | continue; | |
2518 | } else | |
2519 | continue; | |
2520 | result = 0; | |
2521 | break; | |
2522 | } | |
1c8c601a | 2523 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2524 | return result; |
2525 | } | |
2526 | ||
2527 | EXPORT_SYMBOL(lock_may_read); | |
2528 | ||
2529 | /** | |
2530 | * lock_may_write - checks that the region is free of locks | |
2531 | * @inode: the inode that is being written | |
2532 | * @start: the first byte to write | |
2533 | * @len: the number of bytes to write | |
2534 | * | |
2535 | * Emulates Windows locking requirements. Whole-file | |
2536 | * mandatory locks (share modes) can prohibit a write and | |
2537 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2538 | * | |
2539 | * N.B. this function is only ever called | |
2540 | * from knfsd and ownership of locks is never checked. | |
2541 | */ | |
2542 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2543 | { | |
2544 | struct file_lock *fl; | |
2545 | int result = 1; | |
1c8c601a JL |
2546 | |
2547 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2548 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2549 | if (IS_POSIX(fl)) { | |
2550 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2551 | continue; | |
2552 | } else if (IS_FLOCK(fl)) { | |
2553 | if (!(fl->fl_type & LOCK_MAND)) | |
2554 | continue; | |
2555 | if (fl->fl_type & LOCK_WRITE) | |
2556 | continue; | |
2557 | } else | |
2558 | continue; | |
2559 | result = 0; | |
2560 | break; | |
2561 | } | |
1c8c601a | 2562 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2563 | return result; |
2564 | } | |
2565 | ||
2566 | EXPORT_SYMBOL(lock_may_write); | |
2567 | ||
1da177e4 LT |
2568 | static int __init filelock_init(void) |
2569 | { | |
7012b02a JL |
2570 | int i; |
2571 | ||
1da177e4 | 2572 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2573 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2574 | ||
7012b02a JL |
2575 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2576 | ||
2577 | for_each_possible_cpu(i) | |
2578 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2579 | ||
1da177e4 LT |
2580 | return 0; |
2581 | } | |
2582 | ||
2583 | core_initcall(filelock_init); |