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! */ |
6ca10ed8 | 514 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 515 | { |
7012b02a JL |
516 | lg_local_lock(&file_lock_lglock); |
517 | fl->fl_link_cpu = smp_processor_id(); | |
518 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
519 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
520 | } |
521 | ||
7012b02a | 522 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 523 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 524 | { |
7012b02a JL |
525 | /* |
526 | * Avoid taking lock if already unhashed. This is safe since this check | |
527 | * is done while holding the i_lock, and new insertions into the list | |
528 | * also require that it be held. | |
529 | */ | |
530 | if (hlist_unhashed(&fl->fl_link)) | |
531 | return; | |
532 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 533 | hlist_del_init(&fl->fl_link); |
7012b02a | 534 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
535 | } |
536 | ||
3999e493 JL |
537 | static unsigned long |
538 | posix_owner_key(struct file_lock *fl) | |
539 | { | |
540 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
541 | return fl->fl_lmops->lm_owner_key(fl); | |
542 | return (unsigned long)fl->fl_owner; | |
543 | } | |
544 | ||
6ca10ed8 | 545 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 546 | { |
3999e493 | 547 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
548 | } |
549 | ||
6ca10ed8 | 550 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 551 | { |
48f74186 | 552 | hash_del(&waiter->fl_link); |
88974691 JL |
553 | } |
554 | ||
1da177e4 LT |
555 | /* Remove waiter from blocker's block list. |
556 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 557 | * |
7b2296af | 558 | * Must be called with blocked_lock_lock held. |
1da177e4 | 559 | */ |
33443c42 | 560 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 561 | { |
88974691 | 562 | locks_delete_global_blocked(waiter); |
1da177e4 | 563 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
564 | waiter->fl_next = NULL; |
565 | } | |
566 | ||
1a9e64a7 | 567 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 568 | { |
7b2296af | 569 | spin_lock(&blocked_lock_lock); |
1da177e4 | 570 | __locks_delete_block(waiter); |
7b2296af | 571 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
572 | } |
573 | ||
574 | /* Insert waiter into blocker's block list. | |
575 | * We use a circular list so that processes can be easily woken up in | |
576 | * the order they blocked. The documentation doesn't require this but | |
577 | * it seems like the reasonable thing to do. | |
1c8c601a | 578 | * |
7b2296af | 579 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 580 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 581 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 582 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 583 | */ |
1c8c601a JL |
584 | static void __locks_insert_block(struct file_lock *blocker, |
585 | struct file_lock *waiter) | |
1da177e4 | 586 | { |
6dc0fe8f | 587 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 588 | waiter->fl_next = blocker; |
88974691 | 589 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
1da177e4 | 590 | if (IS_POSIX(blocker)) |
1c8c601a JL |
591 | locks_insert_global_blocked(waiter); |
592 | } | |
593 | ||
594 | /* Must be called with i_lock held. */ | |
595 | static void locks_insert_block(struct file_lock *blocker, | |
596 | struct file_lock *waiter) | |
597 | { | |
7b2296af | 598 | spin_lock(&blocked_lock_lock); |
1c8c601a | 599 | __locks_insert_block(blocker, waiter); |
7b2296af | 600 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
601 | } |
602 | ||
1cb36012 JL |
603 | /* |
604 | * Wake up processes blocked waiting for blocker. | |
605 | * | |
1c8c601a | 606 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
607 | */ |
608 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
609 | { | |
4e8c765d JL |
610 | /* |
611 | * Avoid taking global lock if list is empty. This is safe since new | |
612 | * blocked requests are only added to the list under the i_lock, and | |
613 | * the i_lock is always held here. Note that removal from the fl_block | |
614 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 615 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
616 | */ |
617 | if (list_empty(&blocker->fl_block)) | |
618 | return; | |
619 | ||
7b2296af | 620 | spin_lock(&blocked_lock_lock); |
1da177e4 | 621 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
622 | struct file_lock *waiter; |
623 | ||
624 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
625 | struct file_lock, fl_block); |
626 | __locks_delete_block(waiter); | |
8fb47a4f BF |
627 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
628 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
629 | else |
630 | wake_up(&waiter->fl_wait); | |
631 | } | |
7b2296af | 632 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
633 | } |
634 | ||
635 | /* Insert file lock fl into an inode's lock list at the position indicated | |
636 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
637 | * |
638 | * Must be called with the i_lock held! | |
1da177e4 LT |
639 | */ |
640 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
641 | { | |
ab1f1611 VG |
642 | fl->fl_nspid = get_pid(task_tgid(current)); |
643 | ||
1da177e4 LT |
644 | /* insert into file's list */ |
645 | fl->fl_next = *pos; | |
646 | *pos = fl; | |
88974691 JL |
647 | |
648 | locks_insert_global_locks(fl); | |
1da177e4 LT |
649 | } |
650 | ||
24cbe784 JL |
651 | /** |
652 | * locks_delete_lock - Delete a lock and then free it. | |
653 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
654 | * inode->i_flock list entry | |
655 | * | |
656 | * Unlink a lock from all lists and free the namespace reference, but don't | |
657 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
658 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
659 | * |
660 | * Must be called with the i_lock held! | |
1da177e4 | 661 | */ |
24cbe784 | 662 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
663 | { |
664 | struct file_lock *fl = *thisfl_p; | |
665 | ||
88974691 JL |
666 | locks_delete_global_locks(fl); |
667 | ||
1da177e4 LT |
668 | *thisfl_p = fl->fl_next; |
669 | fl->fl_next = NULL; | |
1da177e4 | 670 | |
ab1f1611 VG |
671 | if (fl->fl_nspid) { |
672 | put_pid(fl->fl_nspid); | |
673 | fl->fl_nspid = NULL; | |
674 | } | |
675 | ||
1da177e4 | 676 | locks_wake_up_blocks(fl); |
24cbe784 JL |
677 | } |
678 | ||
679 | /* | |
680 | * Unlink a lock from all lists and free it. | |
681 | * | |
682 | * Must be called with i_lock held! | |
683 | */ | |
684 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
685 | { | |
686 | struct file_lock *fl = *thisfl_p; | |
687 | ||
688 | locks_unlink_lock(thisfl_p); | |
1da177e4 LT |
689 | locks_free_lock(fl); |
690 | } | |
691 | ||
692 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
693 | * checks for shared/exclusive status of overlapping locks. | |
694 | */ | |
695 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
696 | { | |
697 | if (sys_fl->fl_type == F_WRLCK) | |
698 | return 1; | |
699 | if (caller_fl->fl_type == F_WRLCK) | |
700 | return 1; | |
701 | return 0; | |
702 | } | |
703 | ||
704 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
705 | * checking before calling the locks_conflict(). | |
706 | */ | |
707 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
708 | { | |
709 | /* POSIX locks owned by the same process do not conflict with | |
710 | * each other. | |
711 | */ | |
712 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
713 | return (0); | |
714 | ||
715 | /* Check whether they overlap */ | |
716 | if (!locks_overlap(caller_fl, sys_fl)) | |
717 | return 0; | |
718 | ||
719 | return (locks_conflict(caller_fl, sys_fl)); | |
720 | } | |
721 | ||
722 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
723 | * checking before calling the locks_conflict(). | |
724 | */ | |
725 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
726 | { | |
727 | /* FLOCK locks referring to the same filp do not conflict with | |
728 | * each other. | |
729 | */ | |
730 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
731 | return (0); | |
732 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
733 | return 0; | |
734 | ||
735 | return (locks_conflict(caller_fl, sys_fl)); | |
736 | } | |
737 | ||
6d34ac19 | 738 | void |
9d6a8c5c | 739 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
740 | { |
741 | struct file_lock *cfl; | |
1c8c601a | 742 | struct inode *inode = file_inode(filp); |
1da177e4 | 743 | |
1c8c601a | 744 | spin_lock(&inode->i_lock); |
496ad9aa | 745 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
746 | if (!IS_POSIX(cfl)) |
747 | continue; | |
b842e240 | 748 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
749 | break; |
750 | } | |
ab1f1611 | 751 | if (cfl) { |
9d6a8c5c | 752 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 753 | if (cfl->fl_nspid) |
6c5f3e7b | 754 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 755 | } else |
129a84de | 756 | fl->fl_type = F_UNLCK; |
1c8c601a | 757 | spin_unlock(&inode->i_lock); |
6d34ac19 | 758 | return; |
1da177e4 | 759 | } |
1da177e4 LT |
760 | EXPORT_SYMBOL(posix_test_lock); |
761 | ||
b533184f BF |
762 | /* |
763 | * Deadlock detection: | |
764 | * | |
765 | * We attempt to detect deadlocks that are due purely to posix file | |
766 | * locks. | |
1da177e4 | 767 | * |
b533184f BF |
768 | * We assume that a task can be waiting for at most one lock at a time. |
769 | * So for any acquired lock, the process holding that lock may be | |
770 | * waiting on at most one other lock. That lock in turns may be held by | |
771 | * someone waiting for at most one other lock. Given a requested lock | |
772 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
773 | * follow this chain of waiters to ensure we are not about to create a | |
774 | * cycle. | |
1da177e4 | 775 | * |
b533184f BF |
776 | * Since we do this before we ever put a process to sleep on a lock, we |
777 | * are ensured that there is never a cycle; that is what guarantees that | |
778 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 779 | * |
b533184f BF |
780 | * Note: the above assumption may not be true when handling lock |
781 | * requests from a broken NFS client. It may also fail in the presence | |
782 | * of tasks (such as posix threads) sharing the same open file table. | |
783 | * | |
784 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 785 | */ |
97855b49 BF |
786 | |
787 | #define MAX_DEADLK_ITERATIONS 10 | |
788 | ||
b533184f BF |
789 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
790 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
791 | { | |
792 | struct file_lock *fl; | |
793 | ||
3999e493 | 794 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
795 | if (posix_same_owner(fl, block_fl)) |
796 | return fl->fl_next; | |
797 | } | |
798 | return NULL; | |
799 | } | |
800 | ||
7b2296af | 801 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 802 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
803 | struct file_lock *block_fl) |
804 | { | |
97855b49 | 805 | int i = 0; |
1da177e4 | 806 | |
b533184f BF |
807 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
808 | if (i++ > MAX_DEADLK_ITERATIONS) | |
809 | return 0; | |
810 | if (posix_same_owner(caller_fl, block_fl)) | |
811 | return 1; | |
1da177e4 LT |
812 | } |
813 | return 0; | |
814 | } | |
815 | ||
1da177e4 | 816 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 817 | * after any leases, but before any posix locks. |
f475ae95 TM |
818 | * |
819 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
820 | * whether or not a lock was successfully freed by testing the return | |
821 | * value for -ENOENT. | |
1da177e4 | 822 | */ |
993dfa87 | 823 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 824 | { |
993dfa87 | 825 | struct file_lock *new_fl = NULL; |
1da177e4 | 826 | struct file_lock **before; |
496ad9aa | 827 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
828 | int error = 0; |
829 | int found = 0; | |
830 | ||
b89f4321 | 831 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 832 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
833 | if (!new_fl) |
834 | return -ENOMEM; | |
84d535ad PE |
835 | } |
836 | ||
1c8c601a | 837 | spin_lock(&inode->i_lock); |
b89f4321 AB |
838 | if (request->fl_flags & FL_ACCESS) |
839 | goto find_conflict; | |
840 | ||
1da177e4 LT |
841 | for_each_lock(inode, before) { |
842 | struct file_lock *fl = *before; | |
843 | if (IS_POSIX(fl)) | |
844 | break; | |
845 | if (IS_LEASE(fl)) | |
846 | continue; | |
847 | if (filp != fl->fl_file) | |
848 | continue; | |
993dfa87 | 849 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
850 | goto out; |
851 | found = 1; | |
852 | locks_delete_lock(before); | |
853 | break; | |
854 | } | |
1da177e4 | 855 | |
f475ae95 TM |
856 | if (request->fl_type == F_UNLCK) { |
857 | if ((request->fl_flags & FL_EXISTS) && !found) | |
858 | error = -ENOENT; | |
993dfa87 | 859 | goto out; |
f475ae95 | 860 | } |
1da177e4 LT |
861 | |
862 | /* | |
863 | * If a higher-priority process was blocked on the old file lock, | |
864 | * give it the opportunity to lock the file. | |
865 | */ | |
b89f4321 | 866 | if (found) { |
1c8c601a | 867 | spin_unlock(&inode->i_lock); |
def01bc5 | 868 | cond_resched(); |
1c8c601a | 869 | spin_lock(&inode->i_lock); |
b89f4321 | 870 | } |
1da177e4 | 871 | |
f07f18dd | 872 | find_conflict: |
1da177e4 LT |
873 | for_each_lock(inode, before) { |
874 | struct file_lock *fl = *before; | |
875 | if (IS_POSIX(fl)) | |
876 | break; | |
877 | if (IS_LEASE(fl)) | |
878 | continue; | |
993dfa87 | 879 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
880 | continue; |
881 | error = -EAGAIN; | |
bde74e4b MS |
882 | if (!(request->fl_flags & FL_SLEEP)) |
883 | goto out; | |
884 | error = FILE_LOCK_DEFERRED; | |
885 | locks_insert_block(fl, request); | |
1da177e4 LT |
886 | goto out; |
887 | } | |
f07f18dd TM |
888 | if (request->fl_flags & FL_ACCESS) |
889 | goto out; | |
993dfa87 | 890 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 891 | locks_insert_lock(before, new_fl); |
993dfa87 | 892 | new_fl = NULL; |
9cedc194 | 893 | error = 0; |
1da177e4 LT |
894 | |
895 | out: | |
1c8c601a | 896 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
897 | if (new_fl) |
898 | locks_free_lock(new_fl); | |
1da177e4 LT |
899 | return error; |
900 | } | |
901 | ||
150b3934 | 902 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
903 | { |
904 | struct file_lock *fl; | |
39005d02 MS |
905 | struct file_lock *new_fl = NULL; |
906 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
907 | struct file_lock *left = NULL; |
908 | struct file_lock *right = NULL; | |
909 | struct file_lock **before; | |
b9746ef8 JL |
910 | int error; |
911 | bool added = false; | |
1da177e4 LT |
912 | |
913 | /* | |
914 | * We may need two file_lock structures for this operation, | |
915 | * so we get them in advance to avoid races. | |
39005d02 MS |
916 | * |
917 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 918 | */ |
39005d02 MS |
919 | if (!(request->fl_flags & FL_ACCESS) && |
920 | (request->fl_type != F_UNLCK || | |
921 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
922 | new_fl = locks_alloc_lock(); | |
923 | new_fl2 = locks_alloc_lock(); | |
924 | } | |
1da177e4 | 925 | |
1c8c601a | 926 | spin_lock(&inode->i_lock); |
1cb36012 JL |
927 | /* |
928 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
929 | * there are any, either return error or put the request on the | |
48f74186 | 930 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 931 | */ |
1da177e4 LT |
932 | if (request->fl_type != F_UNLCK) { |
933 | for_each_lock(inode, before) { | |
526985b9 | 934 | fl = *before; |
1da177e4 LT |
935 | if (!IS_POSIX(fl)) |
936 | continue; | |
937 | if (!posix_locks_conflict(request, fl)) | |
938 | continue; | |
5842add2 | 939 | if (conflock) |
1a747ee0 | 940 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
941 | error = -EAGAIN; |
942 | if (!(request->fl_flags & FL_SLEEP)) | |
943 | goto out; | |
1c8c601a JL |
944 | /* |
945 | * Deadlock detection and insertion into the blocked | |
946 | * locks list must be done while holding the same lock! | |
947 | */ | |
1da177e4 | 948 | error = -EDEADLK; |
7b2296af | 949 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
950 | if (likely(!posix_locks_deadlock(request, fl))) { |
951 | error = FILE_LOCK_DEFERRED; | |
952 | __locks_insert_block(fl, request); | |
953 | } | |
7b2296af | 954 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
955 | goto out; |
956 | } | |
957 | } | |
958 | ||
959 | /* If we're just looking for a conflict, we're done. */ | |
960 | error = 0; | |
961 | if (request->fl_flags & FL_ACCESS) | |
962 | goto out; | |
963 | ||
1da177e4 | 964 | /* |
1da177e4 LT |
965 | * Find the first old lock with the same owner as the new lock. |
966 | */ | |
967 | ||
968 | before = &inode->i_flock; | |
969 | ||
970 | /* First skip locks owned by other processes. */ | |
971 | while ((fl = *before) && (!IS_POSIX(fl) || | |
972 | !posix_same_owner(request, fl))) { | |
973 | before = &fl->fl_next; | |
974 | } | |
975 | ||
1cb36012 | 976 | /* Process locks with this owner. */ |
1da177e4 LT |
977 | while ((fl = *before) && posix_same_owner(request, fl)) { |
978 | /* Detect adjacent or overlapping regions (if same lock type) | |
979 | */ | |
980 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
981 | /* In all comparisons of start vs end, use |
982 | * "start - 1" rather than "end + 1". If end | |
983 | * is OFFSET_MAX, end + 1 will become negative. | |
984 | */ | |
1da177e4 LT |
985 | if (fl->fl_end < request->fl_start - 1) |
986 | goto next_lock; | |
987 | /* If the next lock in the list has entirely bigger | |
988 | * addresses than the new one, insert the lock here. | |
989 | */ | |
449231d6 | 990 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
991 | break; |
992 | ||
993 | /* If we come here, the new and old lock are of the | |
994 | * same type and adjacent or overlapping. Make one | |
995 | * lock yielding from the lower start address of both | |
996 | * locks to the higher end address. | |
997 | */ | |
998 | if (fl->fl_start > request->fl_start) | |
999 | fl->fl_start = request->fl_start; | |
1000 | else | |
1001 | request->fl_start = fl->fl_start; | |
1002 | if (fl->fl_end < request->fl_end) | |
1003 | fl->fl_end = request->fl_end; | |
1004 | else | |
1005 | request->fl_end = fl->fl_end; | |
1006 | if (added) { | |
1007 | locks_delete_lock(before); | |
1008 | continue; | |
1009 | } | |
1010 | request = fl; | |
b9746ef8 | 1011 | added = true; |
1da177e4 LT |
1012 | } |
1013 | else { | |
1014 | /* Processing for different lock types is a bit | |
1015 | * more complex. | |
1016 | */ | |
1017 | if (fl->fl_end < request->fl_start) | |
1018 | goto next_lock; | |
1019 | if (fl->fl_start > request->fl_end) | |
1020 | break; | |
1021 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1022 | added = true; |
1da177e4 LT |
1023 | if (fl->fl_start < request->fl_start) |
1024 | left = fl; | |
1025 | /* If the next lock in the list has a higher end | |
1026 | * address than the new one, insert the new one here. | |
1027 | */ | |
1028 | if (fl->fl_end > request->fl_end) { | |
1029 | right = fl; | |
1030 | break; | |
1031 | } | |
1032 | if (fl->fl_start >= request->fl_start) { | |
1033 | /* The new lock completely replaces an old | |
1034 | * one (This may happen several times). | |
1035 | */ | |
1036 | if (added) { | |
1037 | locks_delete_lock(before); | |
1038 | continue; | |
1039 | } | |
1040 | /* Replace the old lock with the new one. | |
1041 | * Wake up anybody waiting for the old one, | |
1042 | * as the change in lock type might satisfy | |
1043 | * their needs. | |
1044 | */ | |
1045 | locks_wake_up_blocks(fl); | |
1046 | fl->fl_start = request->fl_start; | |
1047 | fl->fl_end = request->fl_end; | |
1048 | fl->fl_type = request->fl_type; | |
47831f35 TM |
1049 | locks_release_private(fl); |
1050 | locks_copy_private(fl, request); | |
1da177e4 | 1051 | request = fl; |
b9746ef8 | 1052 | added = true; |
1da177e4 LT |
1053 | } |
1054 | } | |
1055 | /* Go on to next lock. | |
1056 | */ | |
1057 | next_lock: | |
1058 | before = &fl->fl_next; | |
1059 | } | |
1060 | ||
0d9a490a | 1061 | /* |
1cb36012 JL |
1062 | * The above code only modifies existing locks in case of merging or |
1063 | * replacing. If new lock(s) need to be inserted all modifications are | |
1064 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1065 | */ |
1066 | error = -ENOLCK; /* "no luck" */ | |
1067 | if (right && left == right && !new_fl2) | |
1068 | goto out; | |
1069 | ||
1da177e4 LT |
1070 | error = 0; |
1071 | if (!added) { | |
f475ae95 TM |
1072 | if (request->fl_type == F_UNLCK) { |
1073 | if (request->fl_flags & FL_EXISTS) | |
1074 | error = -ENOENT; | |
1da177e4 | 1075 | goto out; |
f475ae95 | 1076 | } |
0d9a490a MS |
1077 | |
1078 | if (!new_fl) { | |
1079 | error = -ENOLCK; | |
1080 | goto out; | |
1081 | } | |
1da177e4 LT |
1082 | locks_copy_lock(new_fl, request); |
1083 | locks_insert_lock(before, new_fl); | |
1084 | new_fl = NULL; | |
1085 | } | |
1086 | if (right) { | |
1087 | if (left == right) { | |
1088 | /* The new lock breaks the old one in two pieces, | |
1089 | * so we have to use the second new lock. | |
1090 | */ | |
1091 | left = new_fl2; | |
1092 | new_fl2 = NULL; | |
1093 | locks_copy_lock(left, right); | |
1094 | locks_insert_lock(before, left); | |
1095 | } | |
1096 | right->fl_start = request->fl_end + 1; | |
1097 | locks_wake_up_blocks(right); | |
1098 | } | |
1099 | if (left) { | |
1100 | left->fl_end = request->fl_start - 1; | |
1101 | locks_wake_up_blocks(left); | |
1102 | } | |
1103 | out: | |
1c8c601a | 1104 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1105 | /* |
1106 | * Free any unused locks. | |
1107 | */ | |
1108 | if (new_fl) | |
1109 | locks_free_lock(new_fl); | |
1110 | if (new_fl2) | |
1111 | locks_free_lock(new_fl2); | |
1112 | return error; | |
1113 | } | |
1114 | ||
1115 | /** | |
1116 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1117 | * @filp: The file to apply the lock to | |
1118 | * @fl: The lock to be applied | |
150b3934 | 1119 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1120 | * |
1121 | * Add a POSIX style lock to a file. | |
1122 | * We merge adjacent & overlapping locks whenever possible. | |
1123 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1124 | * |
1125 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1126 | * whether or not a lock was successfully freed by testing the return | |
1127 | * value for -ENOENT. | |
1da177e4 | 1128 | */ |
150b3934 | 1129 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1130 | struct file_lock *conflock) |
1131 | { | |
496ad9aa | 1132 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1133 | } |
150b3934 | 1134 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1135 | |
1136 | /** | |
1137 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1138 | * @filp: The file to apply the lock to | |
1139 | * @fl: The lock to be applied | |
1140 | * | |
1141 | * Add a POSIX style lock to a file. | |
1142 | * We merge adjacent & overlapping locks whenever possible. | |
1143 | * POSIX locks are sorted by owner task, then by starting address | |
1144 | */ | |
1145 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1146 | { | |
1147 | int error; | |
1148 | might_sleep (); | |
1149 | for (;;) { | |
150b3934 | 1150 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1151 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1152 | break; |
1153 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1154 | if (!error) | |
1155 | continue; | |
1156 | ||
1157 | locks_delete_block(fl); | |
1158 | break; | |
1159 | } | |
1160 | return error; | |
1161 | } | |
1162 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1163 | ||
1164 | /** | |
1165 | * locks_mandatory_locked - Check for an active lock | |
1166 | * @inode: the file to check | |
1167 | * | |
1168 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1169 | * This function is called from locks_verify_locked() only. | |
1170 | */ | |
1171 | int locks_mandatory_locked(struct inode *inode) | |
1172 | { | |
1173 | fl_owner_t owner = current->files; | |
1174 | struct file_lock *fl; | |
1175 | ||
1176 | /* | |
1177 | * Search the lock list for this inode for any POSIX locks. | |
1178 | */ | |
1c8c601a | 1179 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1180 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1181 | if (!IS_POSIX(fl)) | |
1182 | continue; | |
1183 | if (fl->fl_owner != owner) | |
1184 | break; | |
1185 | } | |
1c8c601a | 1186 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1187 | return fl ? -EAGAIN : 0; |
1188 | } | |
1189 | ||
1190 | /** | |
1191 | * locks_mandatory_area - Check for a conflicting lock | |
1192 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1193 | * for shared | |
1194 | * @inode: the file to check | |
1195 | * @filp: how the file was opened (if it was) | |
1196 | * @offset: start of area to check | |
1197 | * @count: length of area to check | |
1198 | * | |
1199 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1200 | * This function is called from rw_verify_area() and | |
1201 | * locks_verify_truncate(). | |
1202 | */ | |
1203 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1204 | struct file *filp, loff_t offset, | |
1205 | size_t count) | |
1206 | { | |
1207 | struct file_lock fl; | |
1208 | int error; | |
1209 | ||
1210 | locks_init_lock(&fl); | |
1211 | fl.fl_owner = current->files; | |
1212 | fl.fl_pid = current->tgid; | |
1213 | fl.fl_file = filp; | |
1214 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1215 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1216 | fl.fl_flags |= FL_SLEEP; | |
1217 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1218 | fl.fl_start = offset; | |
1219 | fl.fl_end = offset + count - 1; | |
1220 | ||
1221 | for (;;) { | |
150b3934 | 1222 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1223 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1224 | break; |
1225 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1226 | if (!error) { | |
1227 | /* | |
1228 | * If we've been sleeping someone might have | |
1229 | * changed the permissions behind our back. | |
1230 | */ | |
a16877ca | 1231 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1232 | continue; |
1233 | } | |
1234 | ||
1235 | locks_delete_block(&fl); | |
1236 | break; | |
1237 | } | |
1238 | ||
1239 | return error; | |
1240 | } | |
1241 | ||
1242 | EXPORT_SYMBOL(locks_mandatory_area); | |
1243 | ||
778fc546 BF |
1244 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1245 | { | |
1246 | switch (arg) { | |
1247 | case F_UNLCK: | |
1248 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1249 | /* fall through: */ | |
1250 | case F_RDLCK: | |
1251 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1252 | } | |
1253 | } | |
1254 | ||
1da177e4 LT |
1255 | /* We already had a lease on this file; just change its type */ |
1256 | int lease_modify(struct file_lock **before, int arg) | |
1257 | { | |
1258 | struct file_lock *fl = *before; | |
1259 | int error = assign_type(fl, arg); | |
1260 | ||
1261 | if (error) | |
1262 | return error; | |
778fc546 | 1263 | lease_clear_pending(fl, arg); |
1da177e4 | 1264 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1265 | if (arg == F_UNLCK) { |
1266 | struct file *filp = fl->fl_file; | |
1267 | ||
1268 | f_delown(filp); | |
1269 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1270 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1271 | if (fl->fl_fasync != NULL) { | |
1272 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1273 | fl->fl_fasync = NULL; | |
1274 | } | |
1da177e4 | 1275 | locks_delete_lock(before); |
3b6e2723 | 1276 | } |
1da177e4 LT |
1277 | return 0; |
1278 | } | |
1279 | ||
1280 | EXPORT_SYMBOL(lease_modify); | |
1281 | ||
778fc546 BF |
1282 | static bool past_time(unsigned long then) |
1283 | { | |
1284 | if (!then) | |
1285 | /* 0 is a special value meaning "this never expires": */ | |
1286 | return false; | |
1287 | return time_after(jiffies, then); | |
1288 | } | |
1289 | ||
1da177e4 LT |
1290 | static void time_out_leases(struct inode *inode) |
1291 | { | |
1292 | struct file_lock **before; | |
1293 | struct file_lock *fl; | |
1294 | ||
1295 | before = &inode->i_flock; | |
ab83fa4b | 1296 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
778fc546 BF |
1297 | if (past_time(fl->fl_downgrade_time)) |
1298 | lease_modify(before, F_RDLCK); | |
1299 | if (past_time(fl->fl_break_time)) | |
1300 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1301 | if (fl == *before) /* lease_modify may have freed fl */ |
1302 | before = &fl->fl_next; | |
1303 | } | |
1304 | } | |
1305 | ||
df4e8d2c BF |
1306 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1307 | { | |
1308 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1309 | return false; | |
1310 | return locks_conflict(breaker, lease); | |
1311 | } | |
1312 | ||
1da177e4 LT |
1313 | /** |
1314 | * __break_lease - revoke all outstanding leases on file | |
1315 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1316 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1317 | * break all leases | |
1318 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1319 | * only delegations | |
1da177e4 | 1320 | * |
87250dd2 | 1321 | * break_lease (inlined for speed) has checked there already is at least |
1322 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1323 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1324 | * specified %O_NONBLOCK to your open(). |
1325 | */ | |
df4e8d2c | 1326 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1327 | { |
778fc546 | 1328 | int error = 0; |
1da177e4 LT |
1329 | struct file_lock *new_fl, *flock; |
1330 | struct file_lock *fl; | |
1da177e4 LT |
1331 | unsigned long break_time; |
1332 | int i_have_this_lease = 0; | |
df4e8d2c | 1333 | bool lease_conflict = false; |
8737c930 | 1334 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1335 | |
8737c930 | 1336 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1337 | if (IS_ERR(new_fl)) |
1338 | return PTR_ERR(new_fl); | |
df4e8d2c | 1339 | new_fl->fl_flags = type; |
1da177e4 | 1340 | |
1c8c601a | 1341 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1342 | |
1343 | time_out_leases(inode); | |
1344 | ||
1345 | flock = inode->i_flock; | |
1346 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1347 | goto out; | |
1348 | ||
df4e8d2c BF |
1349 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1350 | if (leases_conflict(fl, new_fl)) { | |
1351 | lease_conflict = true; | |
1352 | if (fl->fl_owner == current->files) | |
1353 | i_have_this_lease = 1; | |
1354 | } | |
1355 | } | |
1356 | if (!lease_conflict) | |
778fc546 BF |
1357 | goto out; |
1358 | ||
1da177e4 LT |
1359 | break_time = 0; |
1360 | if (lease_break_time > 0) { | |
1361 | break_time = jiffies + lease_break_time * HZ; | |
1362 | if (break_time == 0) | |
1363 | break_time++; /* so that 0 means no break time */ | |
1364 | } | |
1365 | ||
1366 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
df4e8d2c BF |
1367 | if (!leases_conflict(fl, new_fl)) |
1368 | continue; | |
778fc546 BF |
1369 | if (want_write) { |
1370 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1371 | continue; | |
1372 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1373 | fl->fl_break_time = break_time; |
778fc546 BF |
1374 | } else { |
1375 | if (lease_breaking(flock)) | |
1376 | continue; | |
1377 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1378 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1379 | } |
778fc546 | 1380 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1381 | } |
1382 | ||
1383 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1384 | error = -EWOULDBLOCK; | |
1385 | goto out; | |
1386 | } | |
1387 | ||
1388 | restart: | |
1389 | break_time = flock->fl_break_time; | |
1390 | if (break_time != 0) { | |
1391 | break_time -= jiffies; | |
1392 | if (break_time == 0) | |
1393 | break_time++; | |
1394 | } | |
4321e01e | 1395 | locks_insert_block(flock, new_fl); |
1c8c601a | 1396 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1397 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1398 | !new_fl->fl_next, break_time); | |
1c8c601a JL |
1399 | spin_lock(&inode->i_lock); |
1400 | locks_delete_block(new_fl); | |
1da177e4 LT |
1401 | if (error >= 0) { |
1402 | if (error == 0) | |
1403 | time_out_leases(inode); | |
778fc546 BF |
1404 | /* |
1405 | * Wait for the next conflicting lease that has not been | |
1406 | * broken yet | |
1407 | */ | |
1da177e4 LT |
1408 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1409 | flock = flock->fl_next) { | |
df4e8d2c | 1410 | if (leases_conflict(new_fl, flock)) |
1da177e4 LT |
1411 | goto restart; |
1412 | } | |
1413 | error = 0; | |
1414 | } | |
1415 | ||
1416 | out: | |
1c8c601a | 1417 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1418 | locks_free_lock(new_fl); |
1da177e4 LT |
1419 | return error; |
1420 | } | |
1421 | ||
1422 | EXPORT_SYMBOL(__break_lease); | |
1423 | ||
1424 | /** | |
a6b91919 | 1425 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1426 | * @inode: the inode |
1427 | * @time: pointer to a timespec which will contain the last modified time | |
1428 | * | |
1429 | * This is to force NFS clients to flush their caches for files with | |
1430 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1431 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1432 | */ |
1433 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1434 | { | |
1435 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1436 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1437 | *time = current_fs_time(inode->i_sb); |
1438 | else | |
1439 | *time = inode->i_mtime; | |
1440 | } | |
1441 | ||
1442 | EXPORT_SYMBOL(lease_get_mtime); | |
1443 | ||
1444 | /** | |
1445 | * fcntl_getlease - Enquire what lease is currently active | |
1446 | * @filp: the file | |
1447 | * | |
1448 | * The value returned by this function will be one of | |
1449 | * (if no lease break is pending): | |
1450 | * | |
1451 | * %F_RDLCK to indicate a shared lease is held. | |
1452 | * | |
1453 | * %F_WRLCK to indicate an exclusive lease is held. | |
1454 | * | |
1455 | * %F_UNLCK to indicate no lease is held. | |
1456 | * | |
1457 | * (if a lease break is pending): | |
1458 | * | |
1459 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1460 | * changed to a shared lease (or removed). | |
1461 | * | |
1462 | * %F_UNLCK to indicate the lease needs to be removed. | |
1463 | * | |
1464 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1465 | * should be returned to userspace. | |
1466 | */ | |
1467 | int fcntl_getlease(struct file *filp) | |
1468 | { | |
1469 | struct file_lock *fl; | |
1c8c601a | 1470 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1471 | int type = F_UNLCK; |
1472 | ||
1c8c601a | 1473 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1474 | time_out_leases(file_inode(filp)); |
1475 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1476 | fl = fl->fl_next) { |
1477 | if (fl->fl_file == filp) { | |
778fc546 | 1478 | type = target_leasetype(fl); |
1da177e4 LT |
1479 | break; |
1480 | } | |
1481 | } | |
1c8c601a | 1482 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1483 | return type; |
1484 | } | |
1485 | ||
24cbe784 JL |
1486 | /** |
1487 | * check_conflicting_open - see if the given dentry points to a file that has | |
1488 | * an existing open that would conflict with the | |
1489 | * desired lease. | |
1490 | * @dentry: dentry to check | |
1491 | * @arg: type of lease that we're trying to acquire | |
1492 | * | |
1493 | * Check to see if there's an existing open fd on this file that would | |
1494 | * conflict with the lease we're trying to set. | |
1495 | */ | |
1496 | static int | |
1497 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1498 | { | |
1499 | int ret = 0; | |
1500 | struct inode *inode = dentry->d_inode; | |
1501 | ||
1502 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1503 | return -EAGAIN; | |
1504 | ||
1505 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1506 | (atomic_read(&inode->i_count) > 1))) | |
1507 | ret = -EAGAIN; | |
1508 | ||
1509 | return ret; | |
1510 | } | |
1511 | ||
d4f22d19 | 1512 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1513 | { |
7eaae282 | 1514 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1515 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1516 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1517 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1518 | int error; |
1da177e4 | 1519 | |
096657b6 | 1520 | lease = *flp; |
df4e8d2c BF |
1521 | /* |
1522 | * In the delegation case we need mutual exclusion with | |
1523 | * a number of operations that take the i_mutex. We trylock | |
1524 | * because delegations are an optional optimization, and if | |
1525 | * there's some chance of a conflict--we'd rather not | |
1526 | * bother, maybe that's a sign this just isn't a good file to | |
1527 | * hand out a delegation on. | |
1528 | */ | |
1529 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1530 | return -EAGAIN; | |
1531 | ||
1532 | if (is_deleg && arg == F_WRLCK) { | |
1533 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1534 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1535 | WARN_ON_ONCE(1); |
1536 | return -EINVAL; | |
1537 | } | |
096657b6 | 1538 | |
24cbe784 JL |
1539 | error = check_conflicting_open(dentry, arg); |
1540 | if (error) | |
096657b6 | 1541 | goto out; |
6d5e8b05 | 1542 | |
1da177e4 LT |
1543 | /* |
1544 | * At this point, we know that if there is an exclusive | |
1545 | * lease on this file, then we hold it on this filp | |
1546 | * (otherwise our open of this file would have blocked). | |
1547 | * And if we are trying to acquire an exclusive lease, | |
1548 | * then the file is not open by anyone (including us) | |
1549 | * except for this filp. | |
1550 | */ | |
c1f24ef4 | 1551 | error = -EAGAIN; |
1da177e4 LT |
1552 | for (before = &inode->i_flock; |
1553 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1554 | before = &fl->fl_next) { | |
c1f24ef4 | 1555 | if (fl->fl_file == filp) { |
1da177e4 | 1556 | my_before = before; |
c1f24ef4 BF |
1557 | continue; |
1558 | } | |
1559 | /* | |
1560 | * No exclusive leases if someone else has a lease on | |
1561 | * this file: | |
1562 | */ | |
1563 | if (arg == F_WRLCK) | |
1564 | goto out; | |
1565 | /* | |
1566 | * Modifying our existing lease is OK, but no getting a | |
1567 | * new lease if someone else is opening for write: | |
1568 | */ | |
1569 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1570 | goto out; | |
1da177e4 LT |
1571 | } |
1572 | ||
1da177e4 | 1573 | if (my_before != NULL) { |
8fb47a4f | 1574 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1575 | if (!error) |
1576 | *flp = *my_before; | |
1da177e4 LT |
1577 | goto out; |
1578 | } | |
1579 | ||
1da177e4 LT |
1580 | error = -EINVAL; |
1581 | if (!leases_enable) | |
1582 | goto out; | |
1583 | ||
c5b1f0d9 | 1584 | locks_insert_lock(before, lease); |
24cbe784 JL |
1585 | /* |
1586 | * The check in break_lease() is lockless. It's possible for another | |
1587 | * open to race in after we did the earlier check for a conflicting | |
1588 | * open but before the lease was inserted. Check again for a | |
1589 | * conflicting open and cancel the lease if there is one. | |
1590 | * | |
1591 | * We also add a barrier here to ensure that the insertion of the lock | |
1592 | * precedes these checks. | |
1593 | */ | |
1594 | smp_mb(); | |
1595 | error = check_conflicting_open(dentry, arg); | |
1596 | if (error) | |
1597 | locks_unlink_lock(flp); | |
1da177e4 | 1598 | out: |
df4e8d2c BF |
1599 | if (is_deleg) |
1600 | mutex_unlock(&inode->i_mutex); | |
1da177e4 LT |
1601 | return error; |
1602 | } | |
8335ebd9 | 1603 | |
d4f22d19 | 1604 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1605 | { |
1606 | struct file_lock *fl, **before; | |
1607 | struct dentry *dentry = filp->f_path.dentry; | |
1608 | struct inode *inode = dentry->d_inode; | |
1609 | ||
1610 | for (before = &inode->i_flock; | |
1611 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1612 | before = &fl->fl_next) { | |
1613 | if (fl->fl_file != filp) | |
1614 | continue; | |
1615 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1616 | } | |
1617 | return -EAGAIN; | |
1618 | } | |
1619 | ||
1620 | /** | |
1621 | * generic_setlease - sets a lease on an open file | |
1622 | * @filp: file pointer | |
1623 | * @arg: type of lease to obtain | |
1624 | * @flp: input - file_lock to use, output - file_lock inserted | |
1625 | * | |
1626 | * The (input) flp->fl_lmops->lm_break function is required | |
1627 | * by break_lease(). | |
1628 | * | |
1c8c601a | 1629 | * Called with inode->i_lock held. |
8335ebd9 BF |
1630 | */ |
1631 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1632 | { | |
1633 | struct dentry *dentry = filp->f_path.dentry; | |
1634 | struct inode *inode = dentry->d_inode; | |
1635 | int error; | |
1636 | ||
8e96e3b7 | 1637 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1638 | return -EACCES; |
1639 | if (!S_ISREG(inode->i_mode)) | |
1640 | return -EINVAL; | |
1641 | error = security_file_lock(filp, arg); | |
1642 | if (error) | |
1643 | return error; | |
1644 | ||
1645 | time_out_leases(inode); | |
1646 | ||
1647 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1648 | ||
1649 | switch (arg) { | |
1650 | case F_UNLCK: | |
1651 | return generic_delete_lease(filp, flp); | |
1652 | case F_RDLCK: | |
1653 | case F_WRLCK: | |
1654 | return generic_add_lease(filp, arg, flp); | |
1655 | default: | |
8d657eb3 | 1656 | return -EINVAL; |
8335ebd9 BF |
1657 | } |
1658 | } | |
0af1a450 | 1659 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1660 | |
b89f4321 AB |
1661 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1662 | { | |
72c2d531 | 1663 | if (filp->f_op->setlease) |
b89f4321 AB |
1664 | return filp->f_op->setlease(filp, arg, lease); |
1665 | else | |
1666 | return generic_setlease(filp, arg, lease); | |
1667 | } | |
1668 | ||
1669 | /** | |
a9933cea | 1670 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1671 | * @filp: file pointer |
1672 | * @arg: type of lease to obtain | |
1673 | * @lease: file_lock to use | |
1674 | * | |
1675 | * Call this to establish a lease on the file. | |
8fb47a4f | 1676 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1677 | * break_lease will oops! |
1678 | * | |
1679 | * This will call the filesystem's setlease file method, if | |
1680 | * defined. Note that there is no getlease method; instead, the | |
1681 | * filesystem setlease method should call back to setlease() to | |
1682 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1683 | * find it. Since fcntl_getlease() only reports whether the current | |
1684 | * task holds a lease, a cluster filesystem need only do this for | |
1685 | * leases held by processes on this node. | |
1686 | * | |
1687 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1688 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1689 | * filesystem's open, create, and (on truncate) setattr methods. |
1690 | * | |
1691 | * Warning: the only current setlease methods exist only to disable | |
1692 | * leases in certain cases. More vfs changes may be required to | |
1693 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1694 | */ |
1695 | ||
a9933cea | 1696 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1697 | { |
1c8c601a | 1698 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1699 | int error; |
1700 | ||
1c8c601a | 1701 | spin_lock(&inode->i_lock); |
b89f4321 | 1702 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1703 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1704 | |
1705 | return error; | |
1706 | } | |
a9933cea | 1707 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1708 | |
0ceaf6c7 BF |
1709 | static int do_fcntl_delete_lease(struct file *filp) |
1710 | { | |
1711 | struct file_lock fl, *flp = &fl; | |
1712 | ||
1713 | lease_init(filp, F_UNLCK, flp); | |
1714 | ||
1715 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1716 | } | |
1717 | ||
1718 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1719 | { |
3df057ac | 1720 | struct file_lock *fl, *ret; |
1c8c601a | 1721 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1722 | struct fasync_struct *new; |
1da177e4 LT |
1723 | int error; |
1724 | ||
c5b1f0d9 AB |
1725 | fl = lease_alloc(filp, arg); |
1726 | if (IS_ERR(fl)) | |
1727 | return PTR_ERR(fl); | |
1da177e4 | 1728 | |
f7347ce4 LT |
1729 | new = fasync_alloc(); |
1730 | if (!new) { | |
1731 | locks_free_lock(fl); | |
1732 | return -ENOMEM; | |
1733 | } | |
3df057ac | 1734 | ret = fl; |
1c8c601a | 1735 | spin_lock(&inode->i_lock); |
8896b93f | 1736 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 | 1737 | if (error) { |
1c8c601a | 1738 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1739 | locks_free_lock(fl); |
1740 | goto out_free_fasync; | |
1741 | } | |
3df057ac BF |
1742 | if (ret != fl) |
1743 | locks_free_lock(fl); | |
1da177e4 | 1744 | |
f7347ce4 LT |
1745 | /* |
1746 | * fasync_insert_entry() returns the old entry if any. | |
1747 | * If there was no old entry, then it used 'new' and | |
1748 | * inserted it into the fasync list. Clear new so that | |
1749 | * we don't release it here. | |
1750 | */ | |
3df057ac | 1751 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1752 | new = NULL; |
1753 | ||
8896b93f | 1754 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1c8c601a | 1755 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1756 | |
1757 | out_free_fasync: | |
f7347ce4 LT |
1758 | if (new) |
1759 | fasync_free(new); | |
1da177e4 LT |
1760 | return error; |
1761 | } | |
1762 | ||
0ceaf6c7 BF |
1763 | /** |
1764 | * fcntl_setlease - sets a lease on an open file | |
1765 | * @fd: open file descriptor | |
1766 | * @filp: file pointer | |
1767 | * @arg: type of lease to obtain | |
1768 | * | |
1769 | * Call this fcntl to establish a lease on the file. | |
1770 | * Note that you also need to call %F_SETSIG to | |
1771 | * receive a signal when the lease is broken. | |
1772 | */ | |
1773 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1774 | { | |
1775 | if (arg == F_UNLCK) | |
1776 | return do_fcntl_delete_lease(filp); | |
1777 | return do_fcntl_add_lease(fd, filp, arg); | |
1778 | } | |
1779 | ||
1da177e4 LT |
1780 | /** |
1781 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1782 | * @filp: The file to apply the lock to | |
1783 | * @fl: The lock to be applied | |
1784 | * | |
1785 | * Add a FLOCK style lock to a file. | |
1786 | */ | |
1787 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1788 | { | |
1789 | int error; | |
1790 | might_sleep(); | |
1791 | for (;;) { | |
1792 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1793 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1794 | break; |
1795 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1796 | if (!error) | |
1797 | continue; | |
1798 | ||
1799 | locks_delete_block(fl); | |
1800 | break; | |
1801 | } | |
1802 | return error; | |
1803 | } | |
1804 | ||
1805 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1806 | ||
1807 | /** | |
1808 | * sys_flock: - flock() system call. | |
1809 | * @fd: the file descriptor to lock. | |
1810 | * @cmd: the type of lock to apply. | |
1811 | * | |
1812 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1813 | * The @cmd can be one of | |
1814 | * | |
1815 | * %LOCK_SH -- a shared lock. | |
1816 | * | |
1817 | * %LOCK_EX -- an exclusive lock. | |
1818 | * | |
1819 | * %LOCK_UN -- remove an existing lock. | |
1820 | * | |
1821 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1822 | * | |
1823 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1824 | * processes read and write access respectively. | |
1825 | */ | |
002c8976 | 1826 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1827 | { |
2903ff01 | 1828 | struct fd f = fdget(fd); |
1da177e4 LT |
1829 | struct file_lock *lock; |
1830 | int can_sleep, unlock; | |
1831 | int error; | |
1832 | ||
1833 | error = -EBADF; | |
2903ff01 | 1834 | if (!f.file) |
1da177e4 LT |
1835 | goto out; |
1836 | ||
1837 | can_sleep = !(cmd & LOCK_NB); | |
1838 | cmd &= ~LOCK_NB; | |
1839 | unlock = (cmd == LOCK_UN); | |
1840 | ||
aeb5d727 | 1841 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1842 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1843 | goto out_putf; |
1844 | ||
2903ff01 | 1845 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1846 | if (error) |
1847 | goto out_putf; | |
1848 | if (can_sleep) | |
1849 | lock->fl_flags |= FL_SLEEP; | |
1850 | ||
2903ff01 | 1851 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1852 | if (error) |
1853 | goto out_free; | |
1854 | ||
72c2d531 | 1855 | if (f.file->f_op->flock) |
2903ff01 | 1856 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1857 | (can_sleep) ? F_SETLKW : F_SETLK, |
1858 | lock); | |
1859 | else | |
2903ff01 | 1860 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1861 | |
1862 | out_free: | |
993dfa87 | 1863 | locks_free_lock(lock); |
1da177e4 LT |
1864 | |
1865 | out_putf: | |
2903ff01 | 1866 | fdput(f); |
1da177e4 LT |
1867 | out: |
1868 | return error; | |
1869 | } | |
1870 | ||
3ee17abd BF |
1871 | /** |
1872 | * vfs_test_lock - test file byte range lock | |
1873 | * @filp: The file to test lock for | |
6924c554 | 1874 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1875 | * |
1876 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1877 | * setting conf->fl_type to something other than F_UNLCK. | |
1878 | */ | |
1879 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1880 | { | |
72c2d531 | 1881 | if (filp->f_op->lock) |
3ee17abd BF |
1882 | return filp->f_op->lock(filp, F_GETLK, fl); |
1883 | posix_test_lock(filp, fl); | |
1884 | return 0; | |
1885 | } | |
1886 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1887 | ||
c2fa1b8a BF |
1888 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1889 | { | |
1890 | flock->l_pid = fl->fl_pid; | |
1891 | #if BITS_PER_LONG == 32 | |
1892 | /* | |
1893 | * Make sure we can represent the posix lock via | |
1894 | * legacy 32bit flock. | |
1895 | */ | |
1896 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1897 | return -EOVERFLOW; | |
1898 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1899 | return -EOVERFLOW; | |
1900 | #endif | |
1901 | flock->l_start = fl->fl_start; | |
1902 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1903 | fl->fl_end - fl->fl_start + 1; | |
1904 | flock->l_whence = 0; | |
129a84de | 1905 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1906 | return 0; |
1907 | } | |
1908 | ||
1909 | #if BITS_PER_LONG == 32 | |
1910 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1911 | { | |
1912 | flock->l_pid = fl->fl_pid; | |
1913 | flock->l_start = fl->fl_start; | |
1914 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1915 | fl->fl_end - fl->fl_start + 1; | |
1916 | flock->l_whence = 0; | |
1917 | flock->l_type = fl->fl_type; | |
1918 | } | |
1919 | #endif | |
1920 | ||
1da177e4 LT |
1921 | /* Report the first existing lock that would conflict with l. |
1922 | * This implements the F_GETLK command of fcntl(). | |
1923 | */ | |
1924 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1925 | { | |
9d6a8c5c | 1926 | struct file_lock file_lock; |
1da177e4 LT |
1927 | struct flock flock; |
1928 | int error; | |
1929 | ||
1930 | error = -EFAULT; | |
1931 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1932 | goto out; | |
1933 | error = -EINVAL; | |
1934 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1935 | goto out; | |
1936 | ||
1937 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1938 | if (error) | |
1939 | goto out; | |
1940 | ||
3ee17abd BF |
1941 | error = vfs_test_lock(filp, &file_lock); |
1942 | if (error) | |
1943 | goto out; | |
1da177e4 | 1944 | |
9d6a8c5c ME |
1945 | flock.l_type = file_lock.fl_type; |
1946 | if (file_lock.fl_type != F_UNLCK) { | |
1947 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1948 | if (error) |
1da177e4 | 1949 | goto out; |
1da177e4 LT |
1950 | } |
1951 | error = -EFAULT; | |
1952 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1953 | error = 0; | |
1954 | out: | |
1955 | return error; | |
1956 | } | |
1957 | ||
7723ec97 ME |
1958 | /** |
1959 | * vfs_lock_file - file byte range lock | |
1960 | * @filp: The file to apply the lock to | |
1961 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1962 | * @fl: The lock to be applied | |
150b3934 ME |
1963 | * @conf: Place to return a copy of the conflicting lock, if found. |
1964 | * | |
1965 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1966 | * as the final argument. | |
1967 | * | |
1968 | * If the filesystem defines a private ->lock() method, then @conf will | |
1969 | * be left unchanged; so a caller that cares should initialize it to | |
1970 | * some acceptable default. | |
2beb6614 ME |
1971 | * |
1972 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1973 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1974 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1975 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1976 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1977 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1978 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1979 | * request completes. |
1980 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1981 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1982 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1983 | * nonzero return code and the file system should release the lock. The file |
1984 | * system is also responsible to keep a corresponding posix lock when it | |
1985 | * grants a lock so the VFS can find out which locks are locally held and do | |
1986 | * the correct lock cleanup when required. | |
1987 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1988 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1989 | * return code. |
7723ec97 | 1990 | */ |
150b3934 | 1991 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 1992 | { |
72c2d531 | 1993 | if (filp->f_op->lock) |
7723ec97 ME |
1994 | return filp->f_op->lock(filp, cmd, fl); |
1995 | else | |
150b3934 | 1996 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1997 | } |
1998 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1999 | ||
b648a6de MS |
2000 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2001 | struct file_lock *fl) | |
2002 | { | |
2003 | int error; | |
2004 | ||
2005 | error = security_file_lock(filp, fl->fl_type); | |
2006 | if (error) | |
2007 | return error; | |
2008 | ||
764c76b3 MS |
2009 | for (;;) { |
2010 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2011 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2012 | break; |
764c76b3 MS |
2013 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2014 | if (!error) | |
2015 | continue; | |
2016 | ||
2017 | locks_delete_block(fl); | |
2018 | break; | |
b648a6de MS |
2019 | } |
2020 | ||
2021 | return error; | |
2022 | } | |
2023 | ||
1da177e4 LT |
2024 | /* Apply the lock described by l to an open file descriptor. |
2025 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2026 | */ | |
c293621b PS |
2027 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2028 | struct flock __user *l) | |
1da177e4 LT |
2029 | { |
2030 | struct file_lock *file_lock = locks_alloc_lock(); | |
2031 | struct flock flock; | |
2032 | struct inode *inode; | |
0b2bac2f | 2033 | struct file *f; |
1da177e4 LT |
2034 | int error; |
2035 | ||
2036 | if (file_lock == NULL) | |
2037 | return -ENOLCK; | |
2038 | ||
2039 | /* | |
2040 | * This might block, so we do it before checking the inode. | |
2041 | */ | |
2042 | error = -EFAULT; | |
2043 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2044 | goto out; | |
2045 | ||
496ad9aa | 2046 | inode = file_inode(filp); |
1da177e4 LT |
2047 | |
2048 | /* Don't allow mandatory locks on files that may be memory mapped | |
2049 | * and shared. | |
2050 | */ | |
a16877ca | 2051 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2052 | error = -EAGAIN; |
2053 | goto out; | |
2054 | } | |
2055 | ||
c293621b | 2056 | again: |
1da177e4 LT |
2057 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2058 | if (error) | |
2059 | goto out; | |
2060 | if (cmd == F_SETLKW) { | |
2061 | file_lock->fl_flags |= FL_SLEEP; | |
2062 | } | |
2063 | ||
2064 | error = -EBADF; | |
2065 | switch (flock.l_type) { | |
2066 | case F_RDLCK: | |
2067 | if (!(filp->f_mode & FMODE_READ)) | |
2068 | goto out; | |
2069 | break; | |
2070 | case F_WRLCK: | |
2071 | if (!(filp->f_mode & FMODE_WRITE)) | |
2072 | goto out; | |
2073 | break; | |
2074 | case F_UNLCK: | |
2075 | break; | |
2076 | default: | |
2077 | error = -EINVAL; | |
2078 | goto out; | |
2079 | } | |
2080 | ||
b648a6de | 2081 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2082 | |
c293621b PS |
2083 | /* |
2084 | * Attempt to detect a close/fcntl race and recover by | |
2085 | * releasing the lock that was just acquired. | |
2086 | */ | |
0b2bac2f AV |
2087 | /* |
2088 | * we need that spin_lock here - it prevents reordering between | |
2089 | * update of inode->i_flock and check for it done in close(). | |
2090 | * rcu_read_lock() wouldn't do. | |
2091 | */ | |
2092 | spin_lock(¤t->files->file_lock); | |
2093 | f = fcheck(fd); | |
2094 | spin_unlock(¤t->files->file_lock); | |
2095 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2096 | flock.l_type = F_UNLCK; |
2097 | goto again; | |
1da177e4 LT |
2098 | } |
2099 | ||
c293621b | 2100 | out: |
1da177e4 LT |
2101 | locks_free_lock(file_lock); |
2102 | return error; | |
2103 | } | |
2104 | ||
2105 | #if BITS_PER_LONG == 32 | |
2106 | /* Report the first existing lock that would conflict with l. | |
2107 | * This implements the F_GETLK command of fcntl(). | |
2108 | */ | |
2109 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
2110 | { | |
9d6a8c5c | 2111 | struct file_lock file_lock; |
1da177e4 LT |
2112 | struct flock64 flock; |
2113 | int error; | |
2114 | ||
2115 | error = -EFAULT; | |
2116 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2117 | goto out; | |
2118 | error = -EINVAL; | |
2119 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2120 | goto out; | |
2121 | ||
2122 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2123 | if (error) | |
2124 | goto out; | |
2125 | ||
3ee17abd BF |
2126 | error = vfs_test_lock(filp, &file_lock); |
2127 | if (error) | |
2128 | goto out; | |
2129 | ||
9d6a8c5c ME |
2130 | flock.l_type = file_lock.fl_type; |
2131 | if (file_lock.fl_type != F_UNLCK) | |
2132 | posix_lock_to_flock64(&flock, &file_lock); | |
2133 | ||
1da177e4 LT |
2134 | error = -EFAULT; |
2135 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2136 | error = 0; | |
2137 | ||
2138 | out: | |
2139 | return error; | |
2140 | } | |
2141 | ||
2142 | /* Apply the lock described by l to an open file descriptor. | |
2143 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2144 | */ | |
c293621b PS |
2145 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2146 | struct flock64 __user *l) | |
1da177e4 LT |
2147 | { |
2148 | struct file_lock *file_lock = locks_alloc_lock(); | |
2149 | struct flock64 flock; | |
2150 | struct inode *inode; | |
0b2bac2f | 2151 | struct file *f; |
1da177e4 LT |
2152 | int error; |
2153 | ||
2154 | if (file_lock == NULL) | |
2155 | return -ENOLCK; | |
2156 | ||
2157 | /* | |
2158 | * This might block, so we do it before checking the inode. | |
2159 | */ | |
2160 | error = -EFAULT; | |
2161 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2162 | goto out; | |
2163 | ||
496ad9aa | 2164 | inode = file_inode(filp); |
1da177e4 LT |
2165 | |
2166 | /* Don't allow mandatory locks on files that may be memory mapped | |
2167 | * and shared. | |
2168 | */ | |
a16877ca | 2169 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2170 | error = -EAGAIN; |
2171 | goto out; | |
2172 | } | |
2173 | ||
c293621b | 2174 | again: |
1da177e4 LT |
2175 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2176 | if (error) | |
2177 | goto out; | |
2178 | if (cmd == F_SETLKW64) { | |
2179 | file_lock->fl_flags |= FL_SLEEP; | |
2180 | } | |
2181 | ||
2182 | error = -EBADF; | |
2183 | switch (flock.l_type) { | |
2184 | case F_RDLCK: | |
2185 | if (!(filp->f_mode & FMODE_READ)) | |
2186 | goto out; | |
2187 | break; | |
2188 | case F_WRLCK: | |
2189 | if (!(filp->f_mode & FMODE_WRITE)) | |
2190 | goto out; | |
2191 | break; | |
2192 | case F_UNLCK: | |
2193 | break; | |
2194 | default: | |
2195 | error = -EINVAL; | |
2196 | goto out; | |
2197 | } | |
2198 | ||
b648a6de | 2199 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2200 | |
c293621b PS |
2201 | /* |
2202 | * Attempt to detect a close/fcntl race and recover by | |
2203 | * releasing the lock that was just acquired. | |
2204 | */ | |
0b2bac2f AV |
2205 | spin_lock(¤t->files->file_lock); |
2206 | f = fcheck(fd); | |
2207 | spin_unlock(¤t->files->file_lock); | |
2208 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2209 | flock.l_type = F_UNLCK; |
2210 | goto again; | |
1da177e4 LT |
2211 | } |
2212 | ||
2213 | out: | |
2214 | locks_free_lock(file_lock); | |
2215 | return error; | |
2216 | } | |
2217 | #endif /* BITS_PER_LONG == 32 */ | |
2218 | ||
2219 | /* | |
2220 | * This function is called when the file is being removed | |
2221 | * from the task's fd array. POSIX locks belonging to this task | |
2222 | * are deleted at this time. | |
2223 | */ | |
2224 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2225 | { | |
ff7b86b8 | 2226 | struct file_lock lock; |
1da177e4 LT |
2227 | |
2228 | /* | |
2229 | * If there are no locks held on this file, we don't need to call | |
2230 | * posix_lock_file(). Another process could be setting a lock on this | |
2231 | * file at the same time, but we wouldn't remove that lock anyway. | |
2232 | */ | |
496ad9aa | 2233 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2234 | return; |
2235 | ||
2236 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2237 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2238 | lock.fl_start = 0; |
2239 | lock.fl_end = OFFSET_MAX; | |
2240 | lock.fl_owner = owner; | |
2241 | lock.fl_pid = current->tgid; | |
2242 | lock.fl_file = filp; | |
2243 | lock.fl_ops = NULL; | |
2244 | lock.fl_lmops = NULL; | |
2245 | ||
150b3934 | 2246 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2247 | |
1da177e4 LT |
2248 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2249 | lock.fl_ops->fl_release_private(&lock); | |
2250 | } | |
2251 | ||
2252 | EXPORT_SYMBOL(locks_remove_posix); | |
2253 | ||
2254 | /* | |
2255 | * This function is called on the last close of an open file. | |
2256 | */ | |
2257 | void locks_remove_flock(struct file *filp) | |
2258 | { | |
496ad9aa | 2259 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2260 | struct file_lock *fl; |
2261 | struct file_lock **before; | |
2262 | ||
2263 | if (!inode->i_flock) | |
2264 | return; | |
2265 | ||
72c2d531 | 2266 | if (filp->f_op->flock) { |
1da177e4 LT |
2267 | struct file_lock fl = { |
2268 | .fl_pid = current->tgid, | |
2269 | .fl_file = filp, | |
2270 | .fl_flags = FL_FLOCK, | |
2271 | .fl_type = F_UNLCK, | |
2272 | .fl_end = OFFSET_MAX, | |
2273 | }; | |
2274 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2275 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2276 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2277 | } |
2278 | ||
1c8c601a | 2279 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2280 | before = &inode->i_flock; |
2281 | ||
2282 | while ((fl = *before) != NULL) { | |
2283 | if (fl->fl_file == filp) { | |
c293621b | 2284 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2285 | locks_delete_lock(before); |
2286 | continue; | |
2287 | } | |
2288 | if (IS_LEASE(fl)) { | |
2289 | lease_modify(before, F_UNLCK); | |
2290 | continue; | |
2291 | } | |
2292 | /* What? */ | |
2293 | BUG(); | |
2294 | } | |
2295 | before = &fl->fl_next; | |
2296 | } | |
1c8c601a | 2297 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2298 | } |
2299 | ||
1da177e4 LT |
2300 | /** |
2301 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2302 | * @waiter: the lock which was waiting |
2303 | * | |
2304 | * lockd needs to block waiting for locks. | |
2305 | */ | |
64a318ee | 2306 | int |
f891a29f | 2307 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2308 | { |
64a318ee BF |
2309 | int status = 0; |
2310 | ||
7b2296af | 2311 | spin_lock(&blocked_lock_lock); |
5996a298 | 2312 | if (waiter->fl_next) |
1da177e4 | 2313 | __locks_delete_block(waiter); |
64a318ee BF |
2314 | else |
2315 | status = -ENOENT; | |
7b2296af | 2316 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2317 | return status; |
1da177e4 | 2318 | } |
1da177e4 LT |
2319 | EXPORT_SYMBOL(posix_unblock_lock); |
2320 | ||
9b9d2ab4 ME |
2321 | /** |
2322 | * vfs_cancel_lock - file byte range unblock lock | |
2323 | * @filp: The file to apply the unblock to | |
2324 | * @fl: The lock to be unblocked | |
2325 | * | |
2326 | * Used by lock managers to cancel blocked requests | |
2327 | */ | |
2328 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2329 | { | |
72c2d531 | 2330 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2331 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2332 | return 0; | |
2333 | } | |
2334 | ||
2335 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2336 | ||
7f8ada98 | 2337 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2338 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2339 | #include <linux/seq_file.h> |
2340 | ||
7012b02a JL |
2341 | struct locks_iterator { |
2342 | int li_cpu; | |
2343 | loff_t li_pos; | |
2344 | }; | |
2345 | ||
7f8ada98 | 2346 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2347 | loff_t id, char *pfx) |
1da177e4 LT |
2348 | { |
2349 | struct inode *inode = NULL; | |
ab1f1611 VG |
2350 | unsigned int fl_pid; |
2351 | ||
2352 | if (fl->fl_nspid) | |
6c5f3e7b | 2353 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2354 | else |
2355 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2356 | |
2357 | if (fl->fl_file != NULL) | |
496ad9aa | 2358 | inode = file_inode(fl->fl_file); |
1da177e4 | 2359 | |
99dc8292 | 2360 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2361 | if (IS_POSIX(fl)) { |
7f8ada98 | 2362 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2363 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2364 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2365 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2366 | } else if (IS_FLOCK(fl)) { |
2367 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2368 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2369 | } else { |
7f8ada98 | 2370 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2371 | } |
2372 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2373 | seq_printf(f, "LEASE "); |
ab83fa4b | 2374 | if (lease_breaking(fl)) |
7f8ada98 | 2375 | seq_printf(f, "BREAKING "); |
1da177e4 | 2376 | else if (fl->fl_file) |
7f8ada98 | 2377 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2378 | else |
7f8ada98 | 2379 | seq_printf(f, "BREAKER "); |
1da177e4 | 2380 | } else { |
7f8ada98 | 2381 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2382 | } |
2383 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2384 | seq_printf(f, "%s ", |
1da177e4 LT |
2385 | (fl->fl_type & LOCK_READ) |
2386 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2387 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2388 | } else { | |
7f8ada98 | 2389 | seq_printf(f, "%s ", |
ab83fa4b | 2390 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2391 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2392 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2393 | } |
2394 | if (inode) { | |
2395 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2396 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2397 | inode->i_sb->s_id, inode->i_ino); |
2398 | #else | |
2399 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2400 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2401 | MAJOR(inode->i_sb->s_dev), |
2402 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2403 | #endif | |
2404 | } else { | |
ab1f1611 | 2405 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2406 | } |
2407 | if (IS_POSIX(fl)) { | |
2408 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2409 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2410 | else |
7f8ada98 | 2411 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2412 | } else { |
7f8ada98 | 2413 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2414 | } |
2415 | } | |
2416 | ||
7f8ada98 | 2417 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2418 | { |
7012b02a | 2419 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2420 | struct file_lock *fl, *bfl; |
1da177e4 | 2421 | |
139ca04e | 2422 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2423 | |
7012b02a | 2424 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2425 | |
7f8ada98 | 2426 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2427 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2428 | |
7f8ada98 PE |
2429 | return 0; |
2430 | } | |
1da177e4 | 2431 | |
7f8ada98 | 2432 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2433 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2434 | { |
7012b02a | 2435 | struct locks_iterator *iter = f->private; |
99dc8292 | 2436 | |
7012b02a JL |
2437 | iter->li_pos = *pos + 1; |
2438 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2439 | spin_lock(&blocked_lock_lock); |
7012b02a | 2440 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2441 | } |
1da177e4 | 2442 | |
7f8ada98 PE |
2443 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2444 | { | |
7012b02a JL |
2445 | struct locks_iterator *iter = f->private; |
2446 | ||
2447 | ++iter->li_pos; | |
2448 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2449 | } |
1da177e4 | 2450 | |
7f8ada98 | 2451 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2452 | __releases(&blocked_lock_lock) |
7f8ada98 | 2453 | { |
7b2296af | 2454 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2455 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2456 | } |
2457 | ||
d8ba7a36 | 2458 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2459 | .start = locks_start, |
2460 | .next = locks_next, | |
2461 | .stop = locks_stop, | |
2462 | .show = locks_show, | |
2463 | }; | |
d8ba7a36 AD |
2464 | |
2465 | static int locks_open(struct inode *inode, struct file *filp) | |
2466 | { | |
7012b02a JL |
2467 | return seq_open_private(filp, &locks_seq_operations, |
2468 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2469 | } |
2470 | ||
2471 | static const struct file_operations proc_locks_operations = { | |
2472 | .open = locks_open, | |
2473 | .read = seq_read, | |
2474 | .llseek = seq_lseek, | |
99dc8292 | 2475 | .release = seq_release_private, |
d8ba7a36 AD |
2476 | }; |
2477 | ||
2478 | static int __init proc_locks_init(void) | |
2479 | { | |
2480 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2481 | return 0; | |
2482 | } | |
2483 | module_init(proc_locks_init); | |
7f8ada98 PE |
2484 | #endif |
2485 | ||
1da177e4 LT |
2486 | /** |
2487 | * lock_may_read - checks that the region is free of locks | |
2488 | * @inode: the inode that is being read | |
2489 | * @start: the first byte to read | |
2490 | * @len: the number of bytes to read | |
2491 | * | |
2492 | * Emulates Windows locking requirements. Whole-file | |
2493 | * mandatory locks (share modes) can prohibit a read and | |
2494 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2495 | * | |
2496 | * N.B. this function is only ever called | |
2497 | * from knfsd and ownership of locks is never checked. | |
2498 | */ | |
2499 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2500 | { | |
2501 | struct file_lock *fl; | |
2502 | int result = 1; | |
1c8c601a JL |
2503 | |
2504 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2505 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2506 | if (IS_POSIX(fl)) { | |
2507 | if (fl->fl_type == F_RDLCK) | |
2508 | continue; | |
2509 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2510 | continue; | |
2511 | } else if (IS_FLOCK(fl)) { | |
2512 | if (!(fl->fl_type & LOCK_MAND)) | |
2513 | continue; | |
2514 | if (fl->fl_type & LOCK_READ) | |
2515 | continue; | |
2516 | } else | |
2517 | continue; | |
2518 | result = 0; | |
2519 | break; | |
2520 | } | |
1c8c601a | 2521 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2522 | return result; |
2523 | } | |
2524 | ||
2525 | EXPORT_SYMBOL(lock_may_read); | |
2526 | ||
2527 | /** | |
2528 | * lock_may_write - checks that the region is free of locks | |
2529 | * @inode: the inode that is being written | |
2530 | * @start: the first byte to write | |
2531 | * @len: the number of bytes to write | |
2532 | * | |
2533 | * Emulates Windows locking requirements. Whole-file | |
2534 | * mandatory locks (share modes) can prohibit a write and | |
2535 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2536 | * | |
2537 | * N.B. this function is only ever called | |
2538 | * from knfsd and ownership of locks is never checked. | |
2539 | */ | |
2540 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2541 | { | |
2542 | struct file_lock *fl; | |
2543 | int result = 1; | |
1c8c601a JL |
2544 | |
2545 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2546 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2547 | if (IS_POSIX(fl)) { | |
2548 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2549 | continue; | |
2550 | } else if (IS_FLOCK(fl)) { | |
2551 | if (!(fl->fl_type & LOCK_MAND)) | |
2552 | continue; | |
2553 | if (fl->fl_type & LOCK_WRITE) | |
2554 | continue; | |
2555 | } else | |
2556 | continue; | |
2557 | result = 0; | |
2558 | break; | |
2559 | } | |
1c8c601a | 2560 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2561 | return result; |
2562 | } | |
2563 | ||
2564 | EXPORT_SYMBOL(lock_may_write); | |
2565 | ||
1da177e4 LT |
2566 | static int __init filelock_init(void) |
2567 | { | |
7012b02a JL |
2568 | int i; |
2569 | ||
1da177e4 | 2570 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2571 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2572 | ||
7012b02a JL |
2573 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2574 | ||
2575 | for_each_possible_cpu(i) | |
2576 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2577 | ||
1da177e4 LT |
2578 | return 0; |
2579 | } | |
2580 | ||
2581 | core_initcall(filelock_init); |