]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/locks.c | |
3 | * | |
4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
6 | * | |
7 | * Deadlock detection added. | |
8 | * FIXME: one thing isn't handled yet: | |
9 | * - mandatory locks (requires lots of changes elsewhere) | |
10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
11 | * | |
12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
14 | * | |
15 | * Converted file_lock_table to a linked list from an array, which eliminates | |
16 | * the limits on how many active file locks are open. | |
17 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
18 | * | |
19 | * Removed dependency on file descriptors. dup()'ed file descriptors now | |
20 | * get the same locks as the original file descriptors, and a close() on | |
21 | * any file descriptor removes ALL the locks on the file for the current | |
22 | * process. Since locks still depend on the process id, locks are inherited | |
23 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
24 | * BSD and SVR4 practice. | |
25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
26 | * | |
27 | * Scrapped free list which is redundant now that we allocate locks | |
28 | * dynamically with kmalloc()/kfree(). | |
29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
30 | * | |
31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
32 | * | |
33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
34 | * fcntl() system call. They have the semantics described above. | |
35 | * | |
36 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
37 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
38 | * and will continue to use the old, broken implementation. | |
39 | * | |
40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
41 | * with a file pointer (filp). As a result they can be shared by a parent | |
42 | * process and its children after a fork(). They are removed when the last | |
43 | * file descriptor referring to the file pointer is closed (unless explicitly | |
44 | * unlocked). | |
45 | * | |
46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
47 | * upgrading from shared to exclusive (or vice versa). When this happens | |
48 | * any processes blocked by the current lock are woken up and allowed to | |
49 | * run before the new lock is applied. | |
50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
51 | * | |
52 | * Removed some race conditions in flock_lock_file(), marked other possible | |
53 | * races. Just grep for FIXME to see them. | |
54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. | |
55 | * | |
56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
58 | * once we've checked for blocking and deadlocking. | |
59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
60 | * | |
61 | * Initial implementation of mandatory locks. SunOS turned out to be | |
62 | * a rotten model, so I implemented the "obvious" semantics. | |
395cf969 | 63 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * | |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
69 | * Manual, Section 2. | |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
71 | * | |
72 | * Tidied up block list handling. Added '/proc/locks' interface. | |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
74 | * | |
75 | * Fixed deadlock condition for pathological code that mixes calls to | |
76 | * flock() and fcntl(). | |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
78 | * | |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
81 | * guarantee sensible behaviour in the case where file system modules might | |
82 | * be compiled with different options than the kernel itself. | |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
84 | * | |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
88 | * | |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
91 | * have already been freed. | |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
93 | * | |
94 | * Made the block list a circular list to minimise searching in the list. | |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
96 | * | |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
98 | * locking. | |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
100 | * | |
101 | * Some adaptations for NFS support. | |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
103 | * | |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
106 | * | |
107 | * Use slab allocator instead of kmalloc/kfree. | |
108 | * Use generic list implementation from <linux/list.h>. | |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
111 | * | |
112 | * Leases and LOCK_MAND | |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
115 | */ | |
116 | ||
117 | #include <linux/capability.h> | |
118 | #include <linux/file.h> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
120 | #include <linux/fs.h> |
121 | #include <linux/init.h> | |
122 | #include <linux/module.h> | |
123 | #include <linux/security.h> | |
124 | #include <linux/slab.h> | |
1da177e4 LT |
125 | #include <linux/syscalls.h> |
126 | #include <linux/time.h> | |
4fb3a538 | 127 | #include <linux/rcupdate.h> |
ab1f1611 | 128 | #include <linux/pid_namespace.h> |
48f74186 | 129 | #include <linux/hashtable.h> |
7012b02a JL |
130 | #include <linux/percpu.h> |
131 | #include <linux/lglock.h> | |
1da177e4 | 132 | |
62af4f1f JL |
133 | #define CREATE_TRACE_POINTS |
134 | #include <trace/events/filelock.h> | |
135 | ||
1da177e4 LT |
136 | #include <asm/uaccess.h> |
137 | ||
138 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
139 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
617588d5 | 140 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG)) |
cff2fce5 | 141 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
1da177e4 | 142 | |
ab83fa4b BF |
143 | static bool lease_breaking(struct file_lock *fl) |
144 | { | |
778fc546 BF |
145 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
146 | } | |
147 | ||
148 | static int target_leasetype(struct file_lock *fl) | |
149 | { | |
150 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
151 | return F_UNLCK; | |
152 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
153 | return F_RDLCK; | |
154 | return fl->fl_type; | |
ab83fa4b BF |
155 | } |
156 | ||
1da177e4 LT |
157 | int leases_enable = 1; |
158 | int lease_break_time = 45; | |
159 | ||
1c8c601a | 160 | /* |
7012b02a JL |
161 | * The global file_lock_list is only used for displaying /proc/locks, so we |
162 | * keep a list on each CPU, with each list protected by its own spinlock via | |
163 | * the file_lock_lglock. Note that alterations to the list also require that | |
164 | * the relevant i_lock is held. | |
1c8c601a | 165 | */ |
7012b02a JL |
166 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
167 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
88974691 | 168 | |
1c8c601a | 169 | /* |
48f74186 | 170 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 171 | * It is protected by blocked_lock_lock. |
48f74186 JL |
172 | * |
173 | * We hash locks by lockowner in order to optimize searching for the lock a | |
174 | * particular lockowner is waiting on. | |
175 | * | |
176 | * FIXME: make this value scale via some heuristic? We generally will want more | |
177 | * buckets when we have more lockowners holding locks, but that's a little | |
178 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 179 | */ |
48f74186 JL |
180 | #define BLOCKED_HASH_BITS 7 |
181 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 182 | |
1c8c601a | 183 | /* |
7b2296af JL |
184 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
185 | * want to be holding this lock. | |
1c8c601a JL |
186 | * |
187 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
188 | * pointer for file_lock structures that are acting as lock requests (in | |
189 | * contrast to those that are acting as records of acquired locks). | |
190 | * | |
191 | * Note that when we acquire this lock in order to change the above fields, | |
192 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
193 | * protected by this lock, we can skip acquiring it iff we already hold the | |
194 | * i_lock. | |
195 | * | |
196 | * In particular, adding an entry to the fl_block list requires that you hold | |
197 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
198 | * an entry from the list however only requires the file_lock_lock. | |
199 | */ | |
7b2296af | 200 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 201 | |
4a075e39 | 202 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 203 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 204 | |
4a075e39 JL |
205 | static struct file_lock_context * |
206 | locks_get_lock_context(struct inode *inode) | |
207 | { | |
208 | struct file_lock_context *new; | |
209 | ||
210 | if (likely(inode->i_flctx)) | |
211 | goto out; | |
212 | ||
213 | new = kmem_cache_alloc(flctx_cache, GFP_KERNEL); | |
214 | if (!new) | |
215 | goto out; | |
216 | ||
217 | INIT_LIST_HEAD(&new->flc_flock); | |
bd61e0a9 | 218 | INIT_LIST_HEAD(&new->flc_posix); |
4a075e39 JL |
219 | |
220 | /* | |
221 | * Assign the pointer if it's not already assigned. If it is, then | |
222 | * free the context we just allocated. | |
223 | */ | |
224 | spin_lock(&inode->i_lock); | |
225 | if (likely(!inode->i_flctx)) { | |
226 | inode->i_flctx = new; | |
227 | new = NULL; | |
228 | } | |
229 | spin_unlock(&inode->i_lock); | |
230 | ||
231 | if (new) | |
232 | kmem_cache_free(flctx_cache, new); | |
233 | out: | |
234 | return inode->i_flctx; | |
235 | } | |
236 | ||
237 | void | |
238 | locks_free_lock_context(struct file_lock_context *ctx) | |
239 | { | |
240 | if (ctx) { | |
241 | WARN_ON_ONCE(!list_empty(&ctx->flc_flock)); | |
bd61e0a9 | 242 | WARN_ON_ONCE(!list_empty(&ctx->flc_posix)); |
4a075e39 JL |
243 | kmem_cache_free(flctx_cache, ctx); |
244 | } | |
245 | } | |
246 | ||
ee19cc40 | 247 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 248 | { |
139ca04e | 249 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 250 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
251 | INIT_LIST_HEAD(&fl->fl_block); |
252 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
253 | } |
254 | ||
1da177e4 | 255 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 256 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 257 | { |
ee19cc40 | 258 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
259 | |
260 | if (fl) | |
ee19cc40 | 261 | locks_init_lock_heads(fl); |
a51cb91d MS |
262 | |
263 | return fl; | |
1da177e4 | 264 | } |
c5b1f0d9 | 265 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 266 | |
a9e61e25 | 267 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
268 | { |
269 | if (fl->fl_ops) { | |
270 | if (fl->fl_ops->fl_release_private) | |
271 | fl->fl_ops->fl_release_private(fl); | |
272 | fl->fl_ops = NULL; | |
273 | } | |
47831f35 | 274 | |
5c97d7b1 KM |
275 | if (fl->fl_lmops) { |
276 | if (fl->fl_lmops->lm_put_owner) | |
277 | fl->fl_lmops->lm_put_owner(fl); | |
278 | fl->fl_lmops = NULL; | |
279 | } | |
47831f35 | 280 | } |
a9e61e25 | 281 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 282 | |
1da177e4 | 283 | /* Free a lock which is not in use. */ |
05fa3135 | 284 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 285 | { |
5ce29646 | 286 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 287 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 288 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 289 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 290 | |
47831f35 | 291 | locks_release_private(fl); |
1da177e4 LT |
292 | kmem_cache_free(filelock_cache, fl); |
293 | } | |
05fa3135 | 294 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 295 | |
ed9814d8 JL |
296 | static void |
297 | locks_dispose_list(struct list_head *dispose) | |
298 | { | |
299 | struct file_lock *fl; | |
300 | ||
301 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
302 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
303 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
304 | locks_free_lock(fl); |
305 | } | |
306 | } | |
307 | ||
1da177e4 LT |
308 | void locks_init_lock(struct file_lock *fl) |
309 | { | |
ee19cc40 MS |
310 | memset(fl, 0, sizeof(struct file_lock)); |
311 | locks_init_lock_heads(fl); | |
1da177e4 LT |
312 | } |
313 | ||
314 | EXPORT_SYMBOL(locks_init_lock); | |
315 | ||
1da177e4 LT |
316 | /* |
317 | * Initialize a new lock from an existing file_lock structure. | |
318 | */ | |
3fe0fff1 | 319 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
320 | { |
321 | new->fl_owner = fl->fl_owner; | |
322 | new->fl_pid = fl->fl_pid; | |
0996905f | 323 | new->fl_file = NULL; |
1da177e4 LT |
324 | new->fl_flags = fl->fl_flags; |
325 | new->fl_type = fl->fl_type; | |
326 | new->fl_start = fl->fl_start; | |
327 | new->fl_end = fl->fl_end; | |
f328296e | 328 | new->fl_lmops = fl->fl_lmops; |
0996905f | 329 | new->fl_ops = NULL; |
f328296e KM |
330 | |
331 | if (fl->fl_lmops) { | |
332 | if (fl->fl_lmops->lm_get_owner) | |
333 | fl->fl_lmops->lm_get_owner(new, fl); | |
334 | } | |
0996905f | 335 | } |
3fe0fff1 | 336 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
337 | |
338 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
339 | { | |
566709bd JL |
340 | /* "new" must be a freshly-initialized lock */ |
341 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 342 | |
3fe0fff1 | 343 | locks_copy_conflock(new, fl); |
f328296e | 344 | |
0996905f | 345 | new->fl_file = fl->fl_file; |
1da177e4 | 346 | new->fl_ops = fl->fl_ops; |
47831f35 | 347 | |
f328296e KM |
348 | if (fl->fl_ops) { |
349 | if (fl->fl_ops->fl_copy_lock) | |
350 | fl->fl_ops->fl_copy_lock(new, fl); | |
351 | } | |
1da177e4 LT |
352 | } |
353 | ||
354 | EXPORT_SYMBOL(locks_copy_lock); | |
355 | ||
356 | static inline int flock_translate_cmd(int cmd) { | |
357 | if (cmd & LOCK_MAND) | |
358 | return cmd & (LOCK_MAND | LOCK_RW); | |
359 | switch (cmd) { | |
360 | case LOCK_SH: | |
361 | return F_RDLCK; | |
362 | case LOCK_EX: | |
363 | return F_WRLCK; | |
364 | case LOCK_UN: | |
365 | return F_UNLCK; | |
366 | } | |
367 | return -EINVAL; | |
368 | } | |
369 | ||
370 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
371 | static struct file_lock * |
372 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
373 | { |
374 | struct file_lock *fl; | |
375 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 376 | |
1da177e4 | 377 | if (type < 0) |
6e129d00 | 378 | return ERR_PTR(type); |
1da177e4 LT |
379 | |
380 | fl = locks_alloc_lock(); | |
381 | if (fl == NULL) | |
6e129d00 | 382 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
383 | |
384 | fl->fl_file = filp; | |
73a8f5f7 | 385 | fl->fl_owner = filp; |
1da177e4 LT |
386 | fl->fl_pid = current->tgid; |
387 | fl->fl_flags = FL_FLOCK; | |
388 | fl->fl_type = type; | |
389 | fl->fl_end = OFFSET_MAX; | |
390 | ||
6e129d00 | 391 | return fl; |
1da177e4 LT |
392 | } |
393 | ||
0ec4f431 | 394 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
395 | { |
396 | switch (type) { | |
397 | case F_RDLCK: | |
398 | case F_WRLCK: | |
399 | case F_UNLCK: | |
400 | fl->fl_type = type; | |
401 | break; | |
402 | default: | |
403 | return -EINVAL; | |
404 | } | |
405 | return 0; | |
406 | } | |
407 | ||
ef12e72a BF |
408 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
409 | struct flock64 *l) | |
1da177e4 | 410 | { |
1da177e4 | 411 | switch (l->l_whence) { |
f5579f8c | 412 | case SEEK_SET: |
ef12e72a | 413 | fl->fl_start = 0; |
1da177e4 | 414 | break; |
f5579f8c | 415 | case SEEK_CUR: |
ef12e72a | 416 | fl->fl_start = filp->f_pos; |
1da177e4 | 417 | break; |
f5579f8c | 418 | case SEEK_END: |
ef12e72a | 419 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
420 | break; |
421 | default: | |
422 | return -EINVAL; | |
423 | } | |
ef12e72a BF |
424 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
425 | return -EOVERFLOW; | |
426 | fl->fl_start += l->l_start; | |
427 | if (fl->fl_start < 0) | |
428 | return -EINVAL; | |
1da177e4 LT |
429 | |
430 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
431 | POSIX-2001 defines it. */ | |
4c780a46 | 432 | if (l->l_len > 0) { |
ef12e72a BF |
433 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
434 | return -EOVERFLOW; | |
435 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
436 | ||
4c780a46 | 437 | } else if (l->l_len < 0) { |
ef12e72a | 438 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 439 | return -EINVAL; |
ef12e72a BF |
440 | fl->fl_end = fl->fl_start - 1; |
441 | fl->fl_start += l->l_len; | |
442 | } else | |
443 | fl->fl_end = OFFSET_MAX; | |
444 | ||
1da177e4 LT |
445 | fl->fl_owner = current->files; |
446 | fl->fl_pid = current->tgid; | |
447 | fl->fl_file = filp; | |
448 | fl->fl_flags = FL_POSIX; | |
449 | fl->fl_ops = NULL; | |
450 | fl->fl_lmops = NULL; | |
451 | ||
452 | return assign_type(fl, l->l_type); | |
453 | } | |
454 | ||
ef12e72a BF |
455 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
456 | * style lock. | |
457 | */ | |
458 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
459 | struct flock *l) | |
1da177e4 | 460 | { |
ef12e72a BF |
461 | struct flock64 ll = { |
462 | .l_type = l->l_type, | |
463 | .l_whence = l->l_whence, | |
464 | .l_start = l->l_start, | |
465 | .l_len = l->l_len, | |
466 | }; | |
467 | ||
468 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 469 | } |
1da177e4 LT |
470 | |
471 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
472 | static bool |
473 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
474 | { |
475 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 476 | return false; |
1da177e4 LT |
477 | } |
478 | ||
1c7dd2ff JL |
479 | static void |
480 | lease_setup(struct file_lock *fl, void **priv) | |
481 | { | |
482 | struct file *filp = fl->fl_file; | |
483 | struct fasync_struct *fa = *priv; | |
484 | ||
485 | /* | |
486 | * fasync_insert_entry() returns the old entry if any. If there was no | |
487 | * old entry, then it used "priv" and inserted it into the fasync list. | |
488 | * Clear the pointer to indicate that it shouldn't be freed. | |
489 | */ | |
490 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
491 | *priv = NULL; | |
492 | ||
493 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
494 | } | |
495 | ||
7b021967 | 496 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 497 | .lm_break = lease_break_callback, |
8fb47a4f | 498 | .lm_change = lease_modify, |
1c7dd2ff | 499 | .lm_setup = lease_setup, |
1da177e4 LT |
500 | }; |
501 | ||
502 | /* | |
503 | * Initialize a lease, use the default lock manager operations | |
504 | */ | |
0ec4f431 | 505 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 506 | { |
75dff55a TM |
507 | if (assign_type(fl, type) != 0) |
508 | return -EINVAL; | |
509 | ||
7ca76311 | 510 | fl->fl_owner = filp; |
1da177e4 LT |
511 | fl->fl_pid = current->tgid; |
512 | ||
513 | fl->fl_file = filp; | |
514 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
515 | fl->fl_start = 0; |
516 | fl->fl_end = OFFSET_MAX; | |
517 | fl->fl_ops = NULL; | |
518 | fl->fl_lmops = &lease_manager_ops; | |
519 | return 0; | |
520 | } | |
521 | ||
522 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 523 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
524 | { |
525 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 526 | int error = -ENOMEM; |
1da177e4 LT |
527 | |
528 | if (fl == NULL) | |
e32b8ee2 | 529 | return ERR_PTR(error); |
1da177e4 LT |
530 | |
531 | error = lease_init(filp, type, fl); | |
75dff55a TM |
532 | if (error) { |
533 | locks_free_lock(fl); | |
e32b8ee2 | 534 | return ERR_PTR(error); |
75dff55a | 535 | } |
e32b8ee2 | 536 | return fl; |
1da177e4 LT |
537 | } |
538 | ||
539 | /* Check if two locks overlap each other. | |
540 | */ | |
541 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
542 | { | |
543 | return ((fl1->fl_end >= fl2->fl_start) && | |
544 | (fl2->fl_end >= fl1->fl_start)); | |
545 | } | |
546 | ||
547 | /* | |
548 | * Check whether two locks have the same owner. | |
549 | */ | |
33443c42 | 550 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 551 | { |
8fb47a4f | 552 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 553 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 554 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
555 | return fl1->fl_owner == fl2->fl_owner; |
556 | } | |
557 | ||
7012b02a | 558 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 559 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 560 | { |
7012b02a JL |
561 | lg_local_lock(&file_lock_lglock); |
562 | fl->fl_link_cpu = smp_processor_id(); | |
563 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
564 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
565 | } |
566 | ||
7012b02a | 567 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 568 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 569 | { |
7012b02a JL |
570 | /* |
571 | * Avoid taking lock if already unhashed. This is safe since this check | |
572 | * is done while holding the i_lock, and new insertions into the list | |
573 | * also require that it be held. | |
574 | */ | |
575 | if (hlist_unhashed(&fl->fl_link)) | |
576 | return; | |
577 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 578 | hlist_del_init(&fl->fl_link); |
7012b02a | 579 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
580 | } |
581 | ||
3999e493 JL |
582 | static unsigned long |
583 | posix_owner_key(struct file_lock *fl) | |
584 | { | |
585 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
586 | return fl->fl_lmops->lm_owner_key(fl); | |
587 | return (unsigned long)fl->fl_owner; | |
588 | } | |
589 | ||
6ca10ed8 | 590 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 591 | { |
3999e493 | 592 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
593 | } |
594 | ||
6ca10ed8 | 595 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 596 | { |
48f74186 | 597 | hash_del(&waiter->fl_link); |
88974691 JL |
598 | } |
599 | ||
1da177e4 LT |
600 | /* Remove waiter from blocker's block list. |
601 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 602 | * |
7b2296af | 603 | * Must be called with blocked_lock_lock held. |
1da177e4 | 604 | */ |
33443c42 | 605 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 606 | { |
88974691 | 607 | locks_delete_global_blocked(waiter); |
1da177e4 | 608 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
609 | waiter->fl_next = NULL; |
610 | } | |
611 | ||
1a9e64a7 | 612 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 613 | { |
7b2296af | 614 | spin_lock(&blocked_lock_lock); |
1da177e4 | 615 | __locks_delete_block(waiter); |
7b2296af | 616 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
617 | } |
618 | ||
619 | /* Insert waiter into blocker's block list. | |
620 | * We use a circular list so that processes can be easily woken up in | |
621 | * the order they blocked. The documentation doesn't require this but | |
622 | * it seems like the reasonable thing to do. | |
1c8c601a | 623 | * |
7b2296af | 624 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 625 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 626 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 627 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 628 | */ |
1c8c601a JL |
629 | static void __locks_insert_block(struct file_lock *blocker, |
630 | struct file_lock *waiter) | |
1da177e4 | 631 | { |
6dc0fe8f | 632 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 633 | waiter->fl_next = blocker; |
88974691 | 634 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 635 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
636 | locks_insert_global_blocked(waiter); |
637 | } | |
638 | ||
639 | /* Must be called with i_lock held. */ | |
640 | static void locks_insert_block(struct file_lock *blocker, | |
641 | struct file_lock *waiter) | |
642 | { | |
7b2296af | 643 | spin_lock(&blocked_lock_lock); |
1c8c601a | 644 | __locks_insert_block(blocker, waiter); |
7b2296af | 645 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
646 | } |
647 | ||
1cb36012 JL |
648 | /* |
649 | * Wake up processes blocked waiting for blocker. | |
650 | * | |
1c8c601a | 651 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
652 | */ |
653 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
654 | { | |
4e8c765d JL |
655 | /* |
656 | * Avoid taking global lock if list is empty. This is safe since new | |
657 | * blocked requests are only added to the list under the i_lock, and | |
658 | * the i_lock is always held here. Note that removal from the fl_block | |
659 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 660 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
661 | */ |
662 | if (list_empty(&blocker->fl_block)) | |
663 | return; | |
664 | ||
7b2296af | 665 | spin_lock(&blocked_lock_lock); |
1da177e4 | 666 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
667 | struct file_lock *waiter; |
668 | ||
669 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
670 | struct file_lock, fl_block); |
671 | __locks_delete_block(waiter); | |
8fb47a4f BF |
672 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
673 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
674 | else |
675 | wake_up(&waiter->fl_wait); | |
676 | } | |
7b2296af | 677 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
678 | } |
679 | ||
680 | /* Insert file lock fl into an inode's lock list at the position indicated | |
681 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
682 | * |
683 | * Must be called with the i_lock held! | |
1da177e4 LT |
684 | */ |
685 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
686 | { | |
ab1f1611 VG |
687 | fl->fl_nspid = get_pid(task_tgid(current)); |
688 | ||
1da177e4 LT |
689 | /* insert into file's list */ |
690 | fl->fl_next = *pos; | |
691 | *pos = fl; | |
88974691 JL |
692 | |
693 | locks_insert_global_locks(fl); | |
1da177e4 LT |
694 | } |
695 | ||
5263e31e JL |
696 | static void |
697 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) | |
698 | { | |
699 | fl->fl_nspid = get_pid(task_tgid(current)); | |
700 | list_add_tail(&fl->fl_list, before); | |
701 | locks_insert_global_locks(fl); | |
702 | } | |
703 | ||
24cbe784 JL |
704 | /** |
705 | * locks_delete_lock - Delete a lock and then free it. | |
706 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
707 | * inode->i_flock list entry | |
708 | * | |
709 | * Unlink a lock from all lists and free the namespace reference, but don't | |
710 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
711 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
712 | * |
713 | * Must be called with the i_lock held! | |
1da177e4 | 714 | */ |
24cbe784 | 715 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
716 | { |
717 | struct file_lock *fl = *thisfl_p; | |
718 | ||
88974691 JL |
719 | locks_delete_global_locks(fl); |
720 | ||
1da177e4 LT |
721 | *thisfl_p = fl->fl_next; |
722 | fl->fl_next = NULL; | |
1da177e4 | 723 | |
ab1f1611 VG |
724 | if (fl->fl_nspid) { |
725 | put_pid(fl->fl_nspid); | |
726 | fl->fl_nspid = NULL; | |
727 | } | |
728 | ||
1da177e4 | 729 | locks_wake_up_blocks(fl); |
24cbe784 JL |
730 | } |
731 | ||
732 | /* | |
733 | * Unlink a lock from all lists and free it. | |
734 | * | |
735 | * Must be called with i_lock held! | |
736 | */ | |
ed9814d8 JL |
737 | static void locks_delete_lock(struct file_lock **thisfl_p, |
738 | struct list_head *dispose) | |
24cbe784 JL |
739 | { |
740 | struct file_lock *fl = *thisfl_p; | |
741 | ||
742 | locks_unlink_lock(thisfl_p); | |
ed9814d8 | 743 | if (dispose) |
6dee60f6 | 744 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
745 | else |
746 | locks_free_lock(fl); | |
1da177e4 LT |
747 | } |
748 | ||
5263e31e JL |
749 | static void |
750 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) | |
751 | { | |
752 | locks_delete_global_locks(fl); | |
753 | if (fl->fl_nspid) { | |
754 | put_pid(fl->fl_nspid); | |
755 | fl->fl_nspid = NULL; | |
756 | } | |
757 | locks_wake_up_blocks(fl); | |
758 | list_move(&fl->fl_list, dispose); | |
759 | } | |
760 | ||
1da177e4 LT |
761 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality |
762 | * checks for shared/exclusive status of overlapping locks. | |
763 | */ | |
764 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
765 | { | |
766 | if (sys_fl->fl_type == F_WRLCK) | |
767 | return 1; | |
768 | if (caller_fl->fl_type == F_WRLCK) | |
769 | return 1; | |
770 | return 0; | |
771 | } | |
772 | ||
773 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
774 | * checking before calling the locks_conflict(). | |
775 | */ | |
776 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
777 | { | |
778 | /* POSIX locks owned by the same process do not conflict with | |
779 | * each other. | |
780 | */ | |
781 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
782 | return (0); | |
783 | ||
784 | /* Check whether they overlap */ | |
785 | if (!locks_overlap(caller_fl, sys_fl)) | |
786 | return 0; | |
787 | ||
788 | return (locks_conflict(caller_fl, sys_fl)); | |
789 | } | |
790 | ||
791 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
792 | * checking before calling the locks_conflict(). | |
793 | */ | |
794 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
795 | { | |
796 | /* FLOCK locks referring to the same filp do not conflict with | |
797 | * each other. | |
798 | */ | |
799 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
800 | return (0); | |
801 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
802 | return 0; | |
803 | ||
804 | return (locks_conflict(caller_fl, sys_fl)); | |
805 | } | |
806 | ||
6d34ac19 | 807 | void |
9d6a8c5c | 808 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
809 | { |
810 | struct file_lock *cfl; | |
bd61e0a9 | 811 | struct file_lock_context *ctx; |
1c8c601a | 812 | struct inode *inode = file_inode(filp); |
1da177e4 | 813 | |
bd61e0a9 JL |
814 | ctx = inode->i_flctx; |
815 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { | |
816 | fl->fl_type = F_UNLCK; | |
817 | return; | |
818 | } | |
819 | ||
1c8c601a | 820 | spin_lock(&inode->i_lock); |
bd61e0a9 JL |
821 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
822 | if (posix_locks_conflict(fl, cfl)) { | |
823 | locks_copy_conflock(fl, cfl); | |
824 | if (cfl->fl_nspid) | |
825 | fl->fl_pid = pid_vnr(cfl->fl_nspid); | |
826 | goto out; | |
827 | } | |
1da177e4 | 828 | } |
bd61e0a9 JL |
829 | fl->fl_type = F_UNLCK; |
830 | out: | |
1c8c601a | 831 | spin_unlock(&inode->i_lock); |
6d34ac19 | 832 | return; |
1da177e4 | 833 | } |
1da177e4 LT |
834 | EXPORT_SYMBOL(posix_test_lock); |
835 | ||
b533184f BF |
836 | /* |
837 | * Deadlock detection: | |
838 | * | |
839 | * We attempt to detect deadlocks that are due purely to posix file | |
840 | * locks. | |
1da177e4 | 841 | * |
b533184f BF |
842 | * We assume that a task can be waiting for at most one lock at a time. |
843 | * So for any acquired lock, the process holding that lock may be | |
844 | * waiting on at most one other lock. That lock in turns may be held by | |
845 | * someone waiting for at most one other lock. Given a requested lock | |
846 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
847 | * follow this chain of waiters to ensure we are not about to create a | |
848 | * cycle. | |
1da177e4 | 849 | * |
b533184f BF |
850 | * Since we do this before we ever put a process to sleep on a lock, we |
851 | * are ensured that there is never a cycle; that is what guarantees that | |
852 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 853 | * |
b533184f BF |
854 | * Note: the above assumption may not be true when handling lock |
855 | * requests from a broken NFS client. It may also fail in the presence | |
856 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 857 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 858 | * |
cff2fce5 | 859 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
860 | * Because the owner is not even nominally tied to a thread of |
861 | * execution, the deadlock detection below can't reasonably work well. Just | |
862 | * skip it for those. | |
863 | * | |
cff2fce5 | 864 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
865 | * locks that just checks for the case where two tasks are attempting to |
866 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 867 | */ |
97855b49 BF |
868 | |
869 | #define MAX_DEADLK_ITERATIONS 10 | |
870 | ||
b533184f BF |
871 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
872 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
873 | { | |
874 | struct file_lock *fl; | |
875 | ||
3999e493 | 876 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
877 | if (posix_same_owner(fl, block_fl)) |
878 | return fl->fl_next; | |
879 | } | |
880 | return NULL; | |
881 | } | |
882 | ||
7b2296af | 883 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 884 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
885 | struct file_lock *block_fl) |
886 | { | |
97855b49 | 887 | int i = 0; |
1da177e4 | 888 | |
57b65325 JL |
889 | /* |
890 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 891 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 892 | */ |
cff2fce5 | 893 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
894 | return 0; |
895 | ||
b533184f BF |
896 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
897 | if (i++ > MAX_DEADLK_ITERATIONS) | |
898 | return 0; | |
899 | if (posix_same_owner(caller_fl, block_fl)) | |
900 | return 1; | |
1da177e4 LT |
901 | } |
902 | return 0; | |
903 | } | |
904 | ||
1da177e4 | 905 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 906 | * after any leases, but before any posix locks. |
f475ae95 TM |
907 | * |
908 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
909 | * whether or not a lock was successfully freed by testing the return | |
910 | * value for -ENOENT. | |
1da177e4 | 911 | */ |
993dfa87 | 912 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 913 | { |
993dfa87 | 914 | struct file_lock *new_fl = NULL; |
5263e31e JL |
915 | struct file_lock *fl; |
916 | struct file_lock_context *ctx; | |
917 | struct inode *inode = file_inode(filp); | |
1da177e4 | 918 | int error = 0; |
5263e31e | 919 | bool found = false; |
ed9814d8 | 920 | LIST_HEAD(dispose); |
1da177e4 | 921 | |
5263e31e JL |
922 | ctx = locks_get_lock_context(inode); |
923 | if (!ctx) | |
924 | return -ENOMEM; | |
925 | ||
b89f4321 | 926 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 927 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
928 | if (!new_fl) |
929 | return -ENOMEM; | |
84d535ad PE |
930 | } |
931 | ||
1c8c601a | 932 | spin_lock(&inode->i_lock); |
b89f4321 AB |
933 | if (request->fl_flags & FL_ACCESS) |
934 | goto find_conflict; | |
935 | ||
5263e31e | 936 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
1da177e4 LT |
937 | if (filp != fl->fl_file) |
938 | continue; | |
993dfa87 | 939 | if (request->fl_type == fl->fl_type) |
1da177e4 | 940 | goto out; |
5263e31e JL |
941 | found = true; |
942 | locks_delete_lock_ctx(fl, &dispose); | |
1da177e4 LT |
943 | break; |
944 | } | |
1da177e4 | 945 | |
f475ae95 TM |
946 | if (request->fl_type == F_UNLCK) { |
947 | if ((request->fl_flags & FL_EXISTS) && !found) | |
948 | error = -ENOENT; | |
993dfa87 | 949 | goto out; |
f475ae95 | 950 | } |
1da177e4 LT |
951 | |
952 | /* | |
953 | * If a higher-priority process was blocked on the old file lock, | |
954 | * give it the opportunity to lock the file. | |
955 | */ | |
b89f4321 | 956 | if (found) { |
1c8c601a | 957 | spin_unlock(&inode->i_lock); |
def01bc5 | 958 | cond_resched(); |
1c8c601a | 959 | spin_lock(&inode->i_lock); |
b89f4321 | 960 | } |
1da177e4 | 961 | |
f07f18dd | 962 | find_conflict: |
5263e31e | 963 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 964 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
965 | continue; |
966 | error = -EAGAIN; | |
bde74e4b MS |
967 | if (!(request->fl_flags & FL_SLEEP)) |
968 | goto out; | |
969 | error = FILE_LOCK_DEFERRED; | |
970 | locks_insert_block(fl, request); | |
1da177e4 LT |
971 | goto out; |
972 | } | |
f07f18dd TM |
973 | if (request->fl_flags & FL_ACCESS) |
974 | goto out; | |
993dfa87 | 975 | locks_copy_lock(new_fl, request); |
5263e31e | 976 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 977 | new_fl = NULL; |
9cedc194 | 978 | error = 0; |
1da177e4 LT |
979 | |
980 | out: | |
1c8c601a | 981 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
982 | if (new_fl) |
983 | locks_free_lock(new_fl); | |
ed9814d8 | 984 | locks_dispose_list(&dispose); |
1da177e4 LT |
985 | return error; |
986 | } | |
987 | ||
150b3934 | 988 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 | 989 | { |
bd61e0a9 | 990 | struct file_lock *fl, *tmp; |
39005d02 MS |
991 | struct file_lock *new_fl = NULL; |
992 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
993 | struct file_lock *left = NULL; |
994 | struct file_lock *right = NULL; | |
bd61e0a9 | 995 | struct file_lock_context *ctx; |
b9746ef8 JL |
996 | int error; |
997 | bool added = false; | |
ed9814d8 | 998 | LIST_HEAD(dispose); |
1da177e4 | 999 | |
bd61e0a9 JL |
1000 | ctx = locks_get_lock_context(inode); |
1001 | if (!ctx) | |
1002 | return -ENOMEM; | |
1003 | ||
1da177e4 LT |
1004 | /* |
1005 | * We may need two file_lock structures for this operation, | |
1006 | * so we get them in advance to avoid races. | |
39005d02 MS |
1007 | * |
1008 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1009 | */ |
39005d02 MS |
1010 | if (!(request->fl_flags & FL_ACCESS) && |
1011 | (request->fl_type != F_UNLCK || | |
1012 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1013 | new_fl = locks_alloc_lock(); | |
1014 | new_fl2 = locks_alloc_lock(); | |
1015 | } | |
1da177e4 | 1016 | |
1c8c601a | 1017 | spin_lock(&inode->i_lock); |
1cb36012 JL |
1018 | /* |
1019 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1020 | * there are any, either return error or put the request on the | |
48f74186 | 1021 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1022 | */ |
1da177e4 | 1023 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 1024 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
1025 | if (!IS_POSIX(fl)) |
1026 | continue; | |
1027 | if (!posix_locks_conflict(request, fl)) | |
1028 | continue; | |
5842add2 | 1029 | if (conflock) |
3fe0fff1 | 1030 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1031 | error = -EAGAIN; |
1032 | if (!(request->fl_flags & FL_SLEEP)) | |
1033 | goto out; | |
1c8c601a JL |
1034 | /* |
1035 | * Deadlock detection and insertion into the blocked | |
1036 | * locks list must be done while holding the same lock! | |
1037 | */ | |
1da177e4 | 1038 | error = -EDEADLK; |
7b2296af | 1039 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1040 | if (likely(!posix_locks_deadlock(request, fl))) { |
1041 | error = FILE_LOCK_DEFERRED; | |
1042 | __locks_insert_block(fl, request); | |
1043 | } | |
7b2296af | 1044 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1045 | goto out; |
1046 | } | |
1047 | } | |
1048 | ||
1049 | /* If we're just looking for a conflict, we're done. */ | |
1050 | error = 0; | |
1051 | if (request->fl_flags & FL_ACCESS) | |
1052 | goto out; | |
1053 | ||
bd61e0a9 JL |
1054 | /* Find the first old lock with the same owner as the new lock */ |
1055 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1056 | if (posix_same_owner(request, fl)) | |
1057 | break; | |
1da177e4 LT |
1058 | } |
1059 | ||
1cb36012 | 1060 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1061 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1062 | if (!posix_same_owner(request, fl)) | |
1063 | break; | |
1064 | ||
1065 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1066 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1067 | /* In all comparisons of start vs end, use |
1068 | * "start - 1" rather than "end + 1". If end | |
1069 | * is OFFSET_MAX, end + 1 will become negative. | |
1070 | */ | |
1da177e4 | 1071 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1072 | continue; |
1da177e4 LT |
1073 | /* If the next lock in the list has entirely bigger |
1074 | * addresses than the new one, insert the lock here. | |
1075 | */ | |
449231d6 | 1076 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1077 | break; |
1078 | ||
1079 | /* If we come here, the new and old lock are of the | |
1080 | * same type and adjacent or overlapping. Make one | |
1081 | * lock yielding from the lower start address of both | |
1082 | * locks to the higher end address. | |
1083 | */ | |
1084 | if (fl->fl_start > request->fl_start) | |
1085 | fl->fl_start = request->fl_start; | |
1086 | else | |
1087 | request->fl_start = fl->fl_start; | |
1088 | if (fl->fl_end < request->fl_end) | |
1089 | fl->fl_end = request->fl_end; | |
1090 | else | |
1091 | request->fl_end = fl->fl_end; | |
1092 | if (added) { | |
bd61e0a9 | 1093 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1094 | continue; |
1095 | } | |
1096 | request = fl; | |
b9746ef8 | 1097 | added = true; |
bd61e0a9 | 1098 | } else { |
1da177e4 LT |
1099 | /* Processing for different lock types is a bit |
1100 | * more complex. | |
1101 | */ | |
1102 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1103 | continue; |
1da177e4 LT |
1104 | if (fl->fl_start > request->fl_end) |
1105 | break; | |
1106 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1107 | added = true; |
1da177e4 LT |
1108 | if (fl->fl_start < request->fl_start) |
1109 | left = fl; | |
1110 | /* If the next lock in the list has a higher end | |
1111 | * address than the new one, insert the new one here. | |
1112 | */ | |
1113 | if (fl->fl_end > request->fl_end) { | |
1114 | right = fl; | |
1115 | break; | |
1116 | } | |
1117 | if (fl->fl_start >= request->fl_start) { | |
1118 | /* The new lock completely replaces an old | |
1119 | * one (This may happen several times). | |
1120 | */ | |
1121 | if (added) { | |
bd61e0a9 | 1122 | locks_delete_lock_ctx(fl, &dispose); |
1da177e4 LT |
1123 | continue; |
1124 | } | |
b84d49f9 JL |
1125 | /* |
1126 | * Replace the old lock with new_fl, and | |
1127 | * remove the old one. It's safe to do the | |
1128 | * insert here since we know that we won't be | |
1129 | * using new_fl later, and that the lock is | |
1130 | * just replacing an existing lock. | |
1da177e4 | 1131 | */ |
b84d49f9 JL |
1132 | error = -ENOLCK; |
1133 | if (!new_fl) | |
1134 | goto out; | |
1135 | locks_copy_lock(new_fl, request); | |
1136 | request = new_fl; | |
1137 | new_fl = NULL; | |
bd61e0a9 JL |
1138 | locks_insert_lock_ctx(request, &fl->fl_list); |
1139 | locks_delete_lock_ctx(fl, &dispose); | |
b9746ef8 | 1140 | added = true; |
1da177e4 LT |
1141 | } |
1142 | } | |
1da177e4 LT |
1143 | } |
1144 | ||
0d9a490a | 1145 | /* |
1cb36012 JL |
1146 | * The above code only modifies existing locks in case of merging or |
1147 | * replacing. If new lock(s) need to be inserted all modifications are | |
1148 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1149 | */ |
1150 | error = -ENOLCK; /* "no luck" */ | |
1151 | if (right && left == right && !new_fl2) | |
1152 | goto out; | |
1153 | ||
1da177e4 LT |
1154 | error = 0; |
1155 | if (!added) { | |
f475ae95 TM |
1156 | if (request->fl_type == F_UNLCK) { |
1157 | if (request->fl_flags & FL_EXISTS) | |
1158 | error = -ENOENT; | |
1da177e4 | 1159 | goto out; |
f475ae95 | 1160 | } |
0d9a490a MS |
1161 | |
1162 | if (!new_fl) { | |
1163 | error = -ENOLCK; | |
1164 | goto out; | |
1165 | } | |
1da177e4 | 1166 | locks_copy_lock(new_fl, request); |
bd61e0a9 | 1167 | locks_insert_lock_ctx(new_fl, &fl->fl_list); |
1da177e4 LT |
1168 | new_fl = NULL; |
1169 | } | |
1170 | if (right) { | |
1171 | if (left == right) { | |
1172 | /* The new lock breaks the old one in two pieces, | |
1173 | * so we have to use the second new lock. | |
1174 | */ | |
1175 | left = new_fl2; | |
1176 | new_fl2 = NULL; | |
1177 | locks_copy_lock(left, right); | |
bd61e0a9 | 1178 | locks_insert_lock_ctx(left, &fl->fl_list); |
1da177e4 LT |
1179 | } |
1180 | right->fl_start = request->fl_end + 1; | |
1181 | locks_wake_up_blocks(right); | |
1182 | } | |
1183 | if (left) { | |
1184 | left->fl_end = request->fl_start - 1; | |
1185 | locks_wake_up_blocks(left); | |
1186 | } | |
1187 | out: | |
1c8c601a | 1188 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1189 | /* |
1190 | * Free any unused locks. | |
1191 | */ | |
1192 | if (new_fl) | |
1193 | locks_free_lock(new_fl); | |
1194 | if (new_fl2) | |
1195 | locks_free_lock(new_fl2); | |
ed9814d8 | 1196 | locks_dispose_list(&dispose); |
1da177e4 LT |
1197 | return error; |
1198 | } | |
1199 | ||
1200 | /** | |
1201 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1202 | * @filp: The file to apply the lock to | |
1203 | * @fl: The lock to be applied | |
150b3934 | 1204 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1205 | * |
1206 | * Add a POSIX style lock to a file. | |
1207 | * We merge adjacent & overlapping locks whenever possible. | |
1208 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1209 | * |
1210 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1211 | * whether or not a lock was successfully freed by testing the return | |
1212 | * value for -ENOENT. | |
1da177e4 | 1213 | */ |
150b3934 | 1214 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1215 | struct file_lock *conflock) |
1216 | { | |
496ad9aa | 1217 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1218 | } |
150b3934 | 1219 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1220 | |
1221 | /** | |
1222 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1223 | * @filp: The file to apply the lock to | |
1224 | * @fl: The lock to be applied | |
1225 | * | |
1226 | * Add a POSIX style lock to a file. | |
1227 | * We merge adjacent & overlapping locks whenever possible. | |
1228 | * POSIX locks are sorted by owner task, then by starting address | |
1229 | */ | |
1230 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1231 | { | |
1232 | int error; | |
1233 | might_sleep (); | |
1234 | for (;;) { | |
150b3934 | 1235 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1236 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1237 | break; |
1238 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1239 | if (!error) | |
1240 | continue; | |
1241 | ||
1242 | locks_delete_block(fl); | |
1243 | break; | |
1244 | } | |
1245 | return error; | |
1246 | } | |
1247 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1248 | ||
1249 | /** | |
1250 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1251 | * @file: the file to check |
1da177e4 LT |
1252 | * |
1253 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1254 | * This function is called from locks_verify_locked() only. | |
1255 | */ | |
d7a06983 | 1256 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1257 | { |
bd61e0a9 | 1258 | int ret; |
d7a06983 | 1259 | struct inode *inode = file_inode(file); |
bd61e0a9 | 1260 | struct file_lock_context *ctx; |
1da177e4 LT |
1261 | struct file_lock *fl; |
1262 | ||
bd61e0a9 JL |
1263 | ctx = inode->i_flctx; |
1264 | if (!ctx || list_empty_careful(&ctx->flc_posix)) | |
1265 | return 0; | |
1266 | ||
1da177e4 LT |
1267 | /* |
1268 | * Search the lock list for this inode for any POSIX locks. | |
1269 | */ | |
1c8c601a | 1270 | spin_lock(&inode->i_lock); |
bd61e0a9 JL |
1271 | ret = 0; |
1272 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1273 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1274 | fl->fl_owner != file) { |
1275 | ret = -EAGAIN; | |
1da177e4 | 1276 | break; |
bd61e0a9 | 1277 | } |
1da177e4 | 1278 | } |
1c8c601a | 1279 | spin_unlock(&inode->i_lock); |
bd61e0a9 | 1280 | return ret; |
1da177e4 LT |
1281 | } |
1282 | ||
1283 | /** | |
1284 | * locks_mandatory_area - Check for a conflicting lock | |
1285 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1286 | * for shared | |
1287 | * @inode: the file to check | |
1288 | * @filp: how the file was opened (if it was) | |
1289 | * @offset: start of area to check | |
1290 | * @count: length of area to check | |
1291 | * | |
1292 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1293 | * This function is called from rw_verify_area() and | |
1294 | * locks_verify_truncate(). | |
1295 | */ | |
1296 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1297 | struct file *filp, loff_t offset, | |
1298 | size_t count) | |
1299 | { | |
1300 | struct file_lock fl; | |
1301 | int error; | |
29723ade | 1302 | bool sleep = false; |
1da177e4 LT |
1303 | |
1304 | locks_init_lock(&fl); | |
1da177e4 LT |
1305 | fl.fl_pid = current->tgid; |
1306 | fl.fl_file = filp; | |
1307 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1308 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1309 | sleep = true; |
1da177e4 LT |
1310 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1311 | fl.fl_start = offset; | |
1312 | fl.fl_end = offset + count - 1; | |
1313 | ||
1314 | for (;;) { | |
29723ade | 1315 | if (filp) { |
73a8f5f7 | 1316 | fl.fl_owner = filp; |
29723ade JL |
1317 | fl.fl_flags &= ~FL_SLEEP; |
1318 | error = __posix_lock_file(inode, &fl, NULL); | |
1319 | if (!error) | |
1320 | break; | |
1321 | } | |
1322 | ||
1323 | if (sleep) | |
1324 | fl.fl_flags |= FL_SLEEP; | |
1325 | fl.fl_owner = current->files; | |
150b3934 | 1326 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1327 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1328 | break; |
1329 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1330 | if (!error) { | |
1331 | /* | |
1332 | * If we've been sleeping someone might have | |
1333 | * changed the permissions behind our back. | |
1334 | */ | |
a16877ca | 1335 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1336 | continue; |
1337 | } | |
1338 | ||
1339 | locks_delete_block(&fl); | |
1340 | break; | |
1341 | } | |
1342 | ||
1343 | return error; | |
1344 | } | |
1345 | ||
1346 | EXPORT_SYMBOL(locks_mandatory_area); | |
1347 | ||
778fc546 BF |
1348 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1349 | { | |
1350 | switch (arg) { | |
1351 | case F_UNLCK: | |
1352 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1353 | /* fall through: */ | |
1354 | case F_RDLCK: | |
1355 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1356 | } | |
1357 | } | |
1358 | ||
1da177e4 | 1359 | /* We already had a lease on this file; just change its type */ |
c45198ed | 1360 | int lease_modify(struct file_lock **before, int arg, struct list_head *dispose) |
1da177e4 LT |
1361 | { |
1362 | struct file_lock *fl = *before; | |
1363 | int error = assign_type(fl, arg); | |
1364 | ||
1365 | if (error) | |
1366 | return error; | |
778fc546 | 1367 | lease_clear_pending(fl, arg); |
1da177e4 | 1368 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1369 | if (arg == F_UNLCK) { |
1370 | struct file *filp = fl->fl_file; | |
1371 | ||
1372 | f_delown(filp); | |
1373 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1374 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1375 | if (fl->fl_fasync != NULL) { | |
1376 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1377 | fl->fl_fasync = NULL; | |
1378 | } | |
c45198ed | 1379 | locks_delete_lock(before, dispose); |
3b6e2723 | 1380 | } |
1da177e4 LT |
1381 | return 0; |
1382 | } | |
1da177e4 LT |
1383 | EXPORT_SYMBOL(lease_modify); |
1384 | ||
778fc546 BF |
1385 | static bool past_time(unsigned long then) |
1386 | { | |
1387 | if (!then) | |
1388 | /* 0 is a special value meaning "this never expires": */ | |
1389 | return false; | |
1390 | return time_after(jiffies, then); | |
1391 | } | |
1392 | ||
c45198ed | 1393 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 LT |
1394 | { |
1395 | struct file_lock **before; | |
1396 | struct file_lock *fl; | |
1397 | ||
f82b4b67 JL |
1398 | lockdep_assert_held(&inode->i_lock); |
1399 | ||
1da177e4 | 1400 | before = &inode->i_flock; |
ab83fa4b | 1401 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1402 | trace_time_out_leases(inode, fl); |
778fc546 | 1403 | if (past_time(fl->fl_downgrade_time)) |
c45198ed | 1404 | lease_modify(before, F_RDLCK, dispose); |
778fc546 | 1405 | if (past_time(fl->fl_break_time)) |
c45198ed | 1406 | lease_modify(before, F_UNLCK, dispose); |
1da177e4 LT |
1407 | if (fl == *before) /* lease_modify may have freed fl */ |
1408 | before = &fl->fl_next; | |
1409 | } | |
1410 | } | |
1411 | ||
df4e8d2c BF |
1412 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1413 | { | |
1414 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1415 | return false; | |
1416 | return locks_conflict(breaker, lease); | |
1417 | } | |
1418 | ||
03d12ddf JL |
1419 | static bool |
1420 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1421 | { | |
1422 | struct file_lock *fl; | |
1423 | ||
1424 | lockdep_assert_held(&inode->i_lock); | |
1425 | ||
1426 | for (fl = inode->i_flock ; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1427 | if (leases_conflict(fl, breaker)) | |
1428 | return true; | |
1429 | } | |
1430 | return false; | |
1431 | } | |
1432 | ||
1da177e4 LT |
1433 | /** |
1434 | * __break_lease - revoke all outstanding leases on file | |
1435 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1436 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1437 | * break all leases | |
1438 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1439 | * only delegations | |
1da177e4 | 1440 | * |
87250dd2 | 1441 | * break_lease (inlined for speed) has checked there already is at least |
1442 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1443 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1444 | * specified %O_NONBLOCK to your open(). |
1445 | */ | |
df4e8d2c | 1446 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1447 | { |
778fc546 | 1448 | int error = 0; |
03d12ddf | 1449 | struct file_lock *new_fl; |
4d01b7f5 | 1450 | struct file_lock *fl, **before; |
1da177e4 | 1451 | unsigned long break_time; |
8737c930 | 1452 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1453 | LIST_HEAD(dispose); |
1da177e4 | 1454 | |
8737c930 | 1455 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1456 | if (IS_ERR(new_fl)) |
1457 | return PTR_ERR(new_fl); | |
df4e8d2c | 1458 | new_fl->fl_flags = type; |
1da177e4 | 1459 | |
1c8c601a | 1460 | spin_lock(&inode->i_lock); |
1da177e4 | 1461 | |
c45198ed | 1462 | time_out_leases(inode, &dispose); |
1da177e4 | 1463 | |
03d12ddf | 1464 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1465 | goto out; |
1466 | ||
1da177e4 LT |
1467 | break_time = 0; |
1468 | if (lease_break_time > 0) { | |
1469 | break_time = jiffies + lease_break_time * HZ; | |
1470 | if (break_time == 0) | |
1471 | break_time++; /* so that 0 means no break time */ | |
1472 | } | |
1473 | ||
4d01b7f5 JL |
1474 | for (before = &inode->i_flock; |
1475 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1476 | before = &fl->fl_next) { | |
df4e8d2c BF |
1477 | if (!leases_conflict(fl, new_fl)) |
1478 | continue; | |
778fc546 BF |
1479 | if (want_write) { |
1480 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1481 | continue; | |
1482 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1483 | fl->fl_break_time = break_time; |
778fc546 | 1484 | } else { |
03d12ddf | 1485 | if (lease_breaking(inode->i_flock)) |
778fc546 BF |
1486 | continue; |
1487 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1488 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1489 | } |
4d01b7f5 JL |
1490 | if (fl->fl_lmops->lm_break(fl)) |
1491 | locks_delete_lock(before, &dispose); | |
1da177e4 LT |
1492 | } |
1493 | ||
4d01b7f5 JL |
1494 | fl = inode->i_flock; |
1495 | if (!fl || !IS_LEASE(fl)) | |
1496 | goto out; | |
1497 | ||
843c6b2f | 1498 | if (mode & O_NONBLOCK) { |
62af4f1f | 1499 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1500 | error = -EWOULDBLOCK; |
1501 | goto out; | |
1502 | } | |
1503 | ||
1504 | restart: | |
03d12ddf | 1505 | break_time = inode->i_flock->fl_break_time; |
f1c6bb2c | 1506 | if (break_time != 0) |
1da177e4 | 1507 | break_time -= jiffies; |
f1c6bb2c JL |
1508 | if (break_time == 0) |
1509 | break_time++; | |
03d12ddf | 1510 | locks_insert_block(inode->i_flock, new_fl); |
62af4f1f | 1511 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1512 | spin_unlock(&inode->i_lock); |
c45198ed | 1513 | locks_dispose_list(&dispose); |
4321e01e MW |
1514 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1515 | !new_fl->fl_next, break_time); | |
1c8c601a | 1516 | spin_lock(&inode->i_lock); |
62af4f1f | 1517 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1518 | locks_delete_block(new_fl); |
1da177e4 | 1519 | if (error >= 0) { |
778fc546 BF |
1520 | /* |
1521 | * Wait for the next conflicting lease that has not been | |
1522 | * broken yet | |
1523 | */ | |
03d12ddf JL |
1524 | if (error == 0) |
1525 | time_out_leases(inode, &dispose); | |
1526 | if (any_leases_conflict(inode, new_fl)) | |
1527 | goto restart; | |
1528 | ||
1da177e4 LT |
1529 | error = 0; |
1530 | } | |
1531 | ||
1532 | out: | |
1c8c601a | 1533 | spin_unlock(&inode->i_lock); |
c45198ed | 1534 | locks_dispose_list(&dispose); |
6d4b9e38 | 1535 | locks_free_lock(new_fl); |
1da177e4 LT |
1536 | return error; |
1537 | } | |
1538 | ||
1539 | EXPORT_SYMBOL(__break_lease); | |
1540 | ||
1541 | /** | |
a6b91919 | 1542 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1543 | * @inode: the inode |
1544 | * @time: pointer to a timespec which will contain the last modified time | |
1545 | * | |
1546 | * This is to force NFS clients to flush their caches for files with | |
1547 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1548 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1549 | */ |
1550 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1551 | { | |
bfe86024 JL |
1552 | bool has_lease = false; |
1553 | struct file_lock *flock; | |
1554 | ||
1555 | if (inode->i_flock) { | |
1556 | spin_lock(&inode->i_lock); | |
1557 | flock = inode->i_flock; | |
1558 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) | |
1559 | has_lease = true; | |
1560 | spin_unlock(&inode->i_lock); | |
1561 | } | |
1562 | ||
1563 | if (has_lease) | |
1da177e4 LT |
1564 | *time = current_fs_time(inode->i_sb); |
1565 | else | |
1566 | *time = inode->i_mtime; | |
1567 | } | |
1568 | ||
1569 | EXPORT_SYMBOL(lease_get_mtime); | |
1570 | ||
1571 | /** | |
1572 | * fcntl_getlease - Enquire what lease is currently active | |
1573 | * @filp: the file | |
1574 | * | |
1575 | * The value returned by this function will be one of | |
1576 | * (if no lease break is pending): | |
1577 | * | |
1578 | * %F_RDLCK to indicate a shared lease is held. | |
1579 | * | |
1580 | * %F_WRLCK to indicate an exclusive lease is held. | |
1581 | * | |
1582 | * %F_UNLCK to indicate no lease is held. | |
1583 | * | |
1584 | * (if a lease break is pending): | |
1585 | * | |
1586 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1587 | * changed to a shared lease (or removed). | |
1588 | * | |
1589 | * %F_UNLCK to indicate the lease needs to be removed. | |
1590 | * | |
1591 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1592 | * should be returned to userspace. | |
1593 | */ | |
1594 | int fcntl_getlease(struct file *filp) | |
1595 | { | |
1596 | struct file_lock *fl; | |
1c8c601a | 1597 | struct inode *inode = file_inode(filp); |
1da177e4 | 1598 | int type = F_UNLCK; |
c45198ed | 1599 | LIST_HEAD(dispose); |
1da177e4 | 1600 | |
1c8c601a | 1601 | spin_lock(&inode->i_lock); |
c45198ed | 1602 | time_out_leases(file_inode(filp), &dispose); |
496ad9aa | 1603 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); |
1da177e4 LT |
1604 | fl = fl->fl_next) { |
1605 | if (fl->fl_file == filp) { | |
778fc546 | 1606 | type = target_leasetype(fl); |
1da177e4 LT |
1607 | break; |
1608 | } | |
1609 | } | |
1c8c601a | 1610 | spin_unlock(&inode->i_lock); |
c45198ed | 1611 | locks_dispose_list(&dispose); |
1da177e4 LT |
1612 | return type; |
1613 | } | |
1614 | ||
24cbe784 JL |
1615 | /** |
1616 | * check_conflicting_open - see if the given dentry points to a file that has | |
1617 | * an existing open that would conflict with the | |
1618 | * desired lease. | |
1619 | * @dentry: dentry to check | |
1620 | * @arg: type of lease that we're trying to acquire | |
1621 | * | |
1622 | * Check to see if there's an existing open fd on this file that would | |
1623 | * conflict with the lease we're trying to set. | |
1624 | */ | |
1625 | static int | |
1626 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1627 | { | |
1628 | int ret = 0; | |
1629 | struct inode *inode = dentry->d_inode; | |
1630 | ||
1631 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1632 | return -EAGAIN; | |
1633 | ||
1634 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1635 | (atomic_read(&inode->i_count) > 1))) | |
1636 | ret = -EAGAIN; | |
1637 | ||
1638 | return ret; | |
1639 | } | |
1640 | ||
e6f5c789 JL |
1641 | static int |
1642 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1643 | { |
7eaae282 | 1644 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1645 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1646 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1647 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1648 | int error; |
c45198ed | 1649 | LIST_HEAD(dispose); |
1da177e4 | 1650 | |
096657b6 | 1651 | lease = *flp; |
62af4f1f JL |
1652 | trace_generic_add_lease(inode, lease); |
1653 | ||
df4e8d2c BF |
1654 | /* |
1655 | * In the delegation case we need mutual exclusion with | |
1656 | * a number of operations that take the i_mutex. We trylock | |
1657 | * because delegations are an optional optimization, and if | |
1658 | * there's some chance of a conflict--we'd rather not | |
1659 | * bother, maybe that's a sign this just isn't a good file to | |
1660 | * hand out a delegation on. | |
1661 | */ | |
1662 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1663 | return -EAGAIN; | |
1664 | ||
1665 | if (is_deleg && arg == F_WRLCK) { | |
1666 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1667 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1668 | WARN_ON_ONCE(1); |
1669 | return -EINVAL; | |
1670 | } | |
096657b6 | 1671 | |
f82b4b67 | 1672 | spin_lock(&inode->i_lock); |
c45198ed | 1673 | time_out_leases(inode, &dispose); |
24cbe784 JL |
1674 | error = check_conflicting_open(dentry, arg); |
1675 | if (error) | |
096657b6 | 1676 | goto out; |
6d5e8b05 | 1677 | |
1da177e4 LT |
1678 | /* |
1679 | * At this point, we know that if there is an exclusive | |
1680 | * lease on this file, then we hold it on this filp | |
1681 | * (otherwise our open of this file would have blocked). | |
1682 | * And if we are trying to acquire an exclusive lease, | |
1683 | * then the file is not open by anyone (including us) | |
1684 | * except for this filp. | |
1685 | */ | |
c1f24ef4 | 1686 | error = -EAGAIN; |
1da177e4 LT |
1687 | for (before = &inode->i_flock; |
1688 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1689 | before = &fl->fl_next) { | |
c1f24ef4 | 1690 | if (fl->fl_file == filp) { |
1da177e4 | 1691 | my_before = before; |
c1f24ef4 BF |
1692 | continue; |
1693 | } | |
1694 | /* | |
1695 | * No exclusive leases if someone else has a lease on | |
1696 | * this file: | |
1697 | */ | |
1698 | if (arg == F_WRLCK) | |
1699 | goto out; | |
1700 | /* | |
1701 | * Modifying our existing lease is OK, but no getting a | |
1702 | * new lease if someone else is opening for write: | |
1703 | */ | |
1704 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1705 | goto out; | |
1da177e4 LT |
1706 | } |
1707 | ||
1da177e4 | 1708 | if (my_before != NULL) { |
1c7dd2ff | 1709 | lease = *my_before; |
c45198ed | 1710 | error = lease->fl_lmops->lm_change(my_before, arg, &dispose); |
1c7dd2ff JL |
1711 | if (error) |
1712 | goto out; | |
1713 | goto out_setup; | |
1da177e4 LT |
1714 | } |
1715 | ||
1da177e4 LT |
1716 | error = -EINVAL; |
1717 | if (!leases_enable) | |
1718 | goto out; | |
1719 | ||
c5b1f0d9 | 1720 | locks_insert_lock(before, lease); |
24cbe784 JL |
1721 | /* |
1722 | * The check in break_lease() is lockless. It's possible for another | |
1723 | * open to race in after we did the earlier check for a conflicting | |
1724 | * open but before the lease was inserted. Check again for a | |
1725 | * conflicting open and cancel the lease if there is one. | |
1726 | * | |
1727 | * We also add a barrier here to ensure that the insertion of the lock | |
1728 | * precedes these checks. | |
1729 | */ | |
1730 | smp_mb(); | |
1731 | error = check_conflicting_open(dentry, arg); | |
1732 | if (error) | |
e6f5c789 | 1733 | goto out_unlink; |
1c7dd2ff JL |
1734 | |
1735 | out_setup: | |
1736 | if (lease->fl_lmops->lm_setup) | |
1737 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1738 | out: |
f82b4b67 | 1739 | spin_unlock(&inode->i_lock); |
c45198ed | 1740 | locks_dispose_list(&dispose); |
df4e8d2c BF |
1741 | if (is_deleg) |
1742 | mutex_unlock(&inode->i_mutex); | |
1c7dd2ff JL |
1743 | if (!error && !my_before) |
1744 | *flp = NULL; | |
1da177e4 | 1745 | return error; |
e6f5c789 JL |
1746 | out_unlink: |
1747 | locks_unlink_lock(before); | |
1748 | goto out; | |
1da177e4 | 1749 | } |
8335ebd9 | 1750 | |
0efaa7e8 | 1751 | static int generic_delete_lease(struct file *filp) |
8335ebd9 | 1752 | { |
0efaa7e8 | 1753 | int error = -EAGAIN; |
8335ebd9 BF |
1754 | struct file_lock *fl, **before; |
1755 | struct dentry *dentry = filp->f_path.dentry; | |
1756 | struct inode *inode = dentry->d_inode; | |
c45198ed | 1757 | LIST_HEAD(dispose); |
8335ebd9 | 1758 | |
f82b4b67 | 1759 | spin_lock(&inode->i_lock); |
c45198ed | 1760 | time_out_leases(inode, &dispose); |
8335ebd9 BF |
1761 | for (before = &inode->i_flock; |
1762 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1763 | before = &fl->fl_next) { | |
0efaa7e8 JL |
1764 | if (fl->fl_file == filp) |
1765 | break; | |
8335ebd9 | 1766 | } |
0efaa7e8 | 1767 | trace_generic_delete_lease(inode, fl); |
52d304eb | 1768 | if (fl && IS_LEASE(fl)) |
c45198ed | 1769 | error = fl->fl_lmops->lm_change(before, F_UNLCK, &dispose); |
f82b4b67 | 1770 | spin_unlock(&inode->i_lock); |
c45198ed | 1771 | locks_dispose_list(&dispose); |
0efaa7e8 | 1772 | return error; |
8335ebd9 BF |
1773 | } |
1774 | ||
1775 | /** | |
1776 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1777 | * @filp: file pointer |
1778 | * @arg: type of lease to obtain | |
1779 | * @flp: input - file_lock to use, output - file_lock inserted | |
1780 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1781 | * doesn't require it) | |
8335ebd9 BF |
1782 | * |
1783 | * The (input) flp->fl_lmops->lm_break function is required | |
1784 | * by break_lease(). | |
8335ebd9 | 1785 | */ |
e6f5c789 JL |
1786 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1787 | void **priv) | |
8335ebd9 BF |
1788 | { |
1789 | struct dentry *dentry = filp->f_path.dentry; | |
1790 | struct inode *inode = dentry->d_inode; | |
1791 | int error; | |
1792 | ||
8e96e3b7 | 1793 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1794 | return -EACCES; |
1795 | if (!S_ISREG(inode->i_mode)) | |
1796 | return -EINVAL; | |
1797 | error = security_file_lock(filp, arg); | |
1798 | if (error) | |
1799 | return error; | |
1800 | ||
8335ebd9 BF |
1801 | switch (arg) { |
1802 | case F_UNLCK: | |
0efaa7e8 | 1803 | return generic_delete_lease(filp); |
8335ebd9 BF |
1804 | case F_RDLCK: |
1805 | case F_WRLCK: | |
0efaa7e8 JL |
1806 | if (!(*flp)->fl_lmops->lm_break) { |
1807 | WARN_ON_ONCE(1); | |
1808 | return -ENOLCK; | |
1809 | } | |
e6f5c789 | 1810 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1811 | default: |
8d657eb3 | 1812 | return -EINVAL; |
8335ebd9 BF |
1813 | } |
1814 | } | |
0af1a450 | 1815 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1816 | |
b89f4321 | 1817 | /** |
e51673aa | 1818 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1819 | * @filp: file pointer |
1820 | * @arg: type of lease to obtain | |
1821 | * @lease: file_lock to use when adding a lease | |
1822 | * @priv: private info for lm_setup when adding a lease (may be | |
1823 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1824 | * |
1825 | * Call this to establish a lease on the file. The "lease" argument is not | |
1826 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1827 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1828 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1829 | * stack trace). | |
1c7dd2ff JL |
1830 | * |
1831 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1832 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1833 | */ |
e6f5c789 JL |
1834 | int |
1835 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1836 | { |
1c7dd2ff | 1837 | if (filp->f_op->setlease) |
f82b4b67 | 1838 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1839 | else |
f82b4b67 | 1840 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1841 | } |
a9933cea | 1842 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1843 | |
0ceaf6c7 | 1844 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1845 | { |
1c7dd2ff | 1846 | struct file_lock *fl; |
f7347ce4 | 1847 | struct fasync_struct *new; |
1da177e4 LT |
1848 | int error; |
1849 | ||
c5b1f0d9 AB |
1850 | fl = lease_alloc(filp, arg); |
1851 | if (IS_ERR(fl)) | |
1852 | return PTR_ERR(fl); | |
1da177e4 | 1853 | |
f7347ce4 LT |
1854 | new = fasync_alloc(); |
1855 | if (!new) { | |
1856 | locks_free_lock(fl); | |
1857 | return -ENOMEM; | |
1858 | } | |
1c7dd2ff | 1859 | new->fa_fd = fd; |
f7347ce4 | 1860 | |
1c7dd2ff | 1861 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1862 | if (fl) |
1863 | locks_free_lock(fl); | |
f7347ce4 LT |
1864 | if (new) |
1865 | fasync_free(new); | |
1da177e4 LT |
1866 | return error; |
1867 | } | |
1868 | ||
0ceaf6c7 BF |
1869 | /** |
1870 | * fcntl_setlease - sets a lease on an open file | |
1871 | * @fd: open file descriptor | |
1872 | * @filp: file pointer | |
1873 | * @arg: type of lease to obtain | |
1874 | * | |
1875 | * Call this fcntl to establish a lease on the file. | |
1876 | * Note that you also need to call %F_SETSIG to | |
1877 | * receive a signal when the lease is broken. | |
1878 | */ | |
1879 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1880 | { | |
1881 | if (arg == F_UNLCK) | |
e6f5c789 | 1882 | return vfs_setlease(filp, F_UNLCK, NULL, NULL); |
0ceaf6c7 BF |
1883 | return do_fcntl_add_lease(fd, filp, arg); |
1884 | } | |
1885 | ||
1da177e4 LT |
1886 | /** |
1887 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1888 | * @filp: The file to apply the lock to | |
1889 | * @fl: The lock to be applied | |
1890 | * | |
1891 | * Add a FLOCK style lock to a file. | |
1892 | */ | |
1893 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1894 | { | |
1895 | int error; | |
1896 | might_sleep(); | |
1897 | for (;;) { | |
1898 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1899 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1900 | break; |
1901 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1902 | if (!error) | |
1903 | continue; | |
1904 | ||
1905 | locks_delete_block(fl); | |
1906 | break; | |
1907 | } | |
1908 | return error; | |
1909 | } | |
1910 | ||
1911 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1912 | ||
1913 | /** | |
1914 | * sys_flock: - flock() system call. | |
1915 | * @fd: the file descriptor to lock. | |
1916 | * @cmd: the type of lock to apply. | |
1917 | * | |
1918 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1919 | * The @cmd can be one of | |
1920 | * | |
1921 | * %LOCK_SH -- a shared lock. | |
1922 | * | |
1923 | * %LOCK_EX -- an exclusive lock. | |
1924 | * | |
1925 | * %LOCK_UN -- remove an existing lock. | |
1926 | * | |
1927 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1928 | * | |
1929 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1930 | * processes read and write access respectively. | |
1931 | */ | |
002c8976 | 1932 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1933 | { |
2903ff01 | 1934 | struct fd f = fdget(fd); |
1da177e4 LT |
1935 | struct file_lock *lock; |
1936 | int can_sleep, unlock; | |
1937 | int error; | |
1938 | ||
1939 | error = -EBADF; | |
2903ff01 | 1940 | if (!f.file) |
1da177e4 LT |
1941 | goto out; |
1942 | ||
1943 | can_sleep = !(cmd & LOCK_NB); | |
1944 | cmd &= ~LOCK_NB; | |
1945 | unlock = (cmd == LOCK_UN); | |
1946 | ||
aeb5d727 | 1947 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1948 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1949 | goto out_putf; |
1950 | ||
6e129d00 JL |
1951 | lock = flock_make_lock(f.file, cmd); |
1952 | if (IS_ERR(lock)) { | |
1953 | error = PTR_ERR(lock); | |
1da177e4 | 1954 | goto out_putf; |
6e129d00 JL |
1955 | } |
1956 | ||
1da177e4 LT |
1957 | if (can_sleep) |
1958 | lock->fl_flags |= FL_SLEEP; | |
1959 | ||
2903ff01 | 1960 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1961 | if (error) |
1962 | goto out_free; | |
1963 | ||
72c2d531 | 1964 | if (f.file->f_op->flock) |
2903ff01 | 1965 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1966 | (can_sleep) ? F_SETLKW : F_SETLK, |
1967 | lock); | |
1968 | else | |
2903ff01 | 1969 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1970 | |
1971 | out_free: | |
993dfa87 | 1972 | locks_free_lock(lock); |
1da177e4 LT |
1973 | |
1974 | out_putf: | |
2903ff01 | 1975 | fdput(f); |
1da177e4 LT |
1976 | out: |
1977 | return error; | |
1978 | } | |
1979 | ||
3ee17abd BF |
1980 | /** |
1981 | * vfs_test_lock - test file byte range lock | |
1982 | * @filp: The file to test lock for | |
6924c554 | 1983 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1984 | * |
1985 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1986 | * setting conf->fl_type to something other than F_UNLCK. | |
1987 | */ | |
1988 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1989 | { | |
72c2d531 | 1990 | if (filp->f_op->lock) |
3ee17abd BF |
1991 | return filp->f_op->lock(filp, F_GETLK, fl); |
1992 | posix_test_lock(filp, fl); | |
1993 | return 0; | |
1994 | } | |
1995 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1996 | ||
c2fa1b8a BF |
1997 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1998 | { | |
cff2fce5 | 1999 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2000 | #if BITS_PER_LONG == 32 |
2001 | /* | |
2002 | * Make sure we can represent the posix lock via | |
2003 | * legacy 32bit flock. | |
2004 | */ | |
2005 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2006 | return -EOVERFLOW; | |
2007 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2008 | return -EOVERFLOW; | |
2009 | #endif | |
2010 | flock->l_start = fl->fl_start; | |
2011 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2012 | fl->fl_end - fl->fl_start + 1; | |
2013 | flock->l_whence = 0; | |
129a84de | 2014 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2015 | return 0; |
2016 | } | |
2017 | ||
2018 | #if BITS_PER_LONG == 32 | |
2019 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2020 | { | |
cff2fce5 | 2021 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2022 | flock->l_start = fl->fl_start; |
2023 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2024 | fl->fl_end - fl->fl_start + 1; | |
2025 | flock->l_whence = 0; | |
2026 | flock->l_type = fl->fl_type; | |
2027 | } | |
2028 | #endif | |
2029 | ||
1da177e4 LT |
2030 | /* Report the first existing lock that would conflict with l. |
2031 | * This implements the F_GETLK command of fcntl(). | |
2032 | */ | |
c1e62b8f | 2033 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 2034 | { |
9d6a8c5c | 2035 | struct file_lock file_lock; |
1da177e4 LT |
2036 | struct flock flock; |
2037 | int error; | |
2038 | ||
2039 | error = -EFAULT; | |
2040 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2041 | goto out; | |
2042 | error = -EINVAL; | |
2043 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2044 | goto out; | |
2045 | ||
2046 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
2047 | if (error) | |
2048 | goto out; | |
2049 | ||
0d3f7a2d | 2050 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2051 | error = -EINVAL; |
2052 | if (flock.l_pid != 0) | |
2053 | goto out; | |
2054 | ||
5d50ffd7 | 2055 | cmd = F_GETLK; |
cff2fce5 | 2056 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2057 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2058 | } |
2059 | ||
3ee17abd BF |
2060 | error = vfs_test_lock(filp, &file_lock); |
2061 | if (error) | |
2062 | goto out; | |
1da177e4 | 2063 | |
9d6a8c5c ME |
2064 | flock.l_type = file_lock.fl_type; |
2065 | if (file_lock.fl_type != F_UNLCK) { | |
2066 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2067 | if (error) |
f328296e | 2068 | goto rel_priv; |
1da177e4 LT |
2069 | } |
2070 | error = -EFAULT; | |
2071 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2072 | error = 0; | |
f328296e KM |
2073 | rel_priv: |
2074 | locks_release_private(&file_lock); | |
1da177e4 LT |
2075 | out: |
2076 | return error; | |
2077 | } | |
2078 | ||
7723ec97 ME |
2079 | /** |
2080 | * vfs_lock_file - file byte range lock | |
2081 | * @filp: The file to apply the lock to | |
2082 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2083 | * @fl: The lock to be applied | |
150b3934 ME |
2084 | * @conf: Place to return a copy of the conflicting lock, if found. |
2085 | * | |
2086 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2087 | * as the final argument. | |
2088 | * | |
2089 | * If the filesystem defines a private ->lock() method, then @conf will | |
2090 | * be left unchanged; so a caller that cares should initialize it to | |
2091 | * some acceptable default. | |
2beb6614 ME |
2092 | * |
2093 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2094 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2095 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2096 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2097 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2098 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2099 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2100 | * request completes. |
2101 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2102 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2103 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2104 | * nonzero return code and the file system should release the lock. The file |
2105 | * system is also responsible to keep a corresponding posix lock when it | |
2106 | * grants a lock so the VFS can find out which locks are locally held and do | |
2107 | * the correct lock cleanup when required. | |
2108 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2109 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2110 | * return code. |
7723ec97 | 2111 | */ |
150b3934 | 2112 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2113 | { |
72c2d531 | 2114 | if (filp->f_op->lock) |
7723ec97 ME |
2115 | return filp->f_op->lock(filp, cmd, fl); |
2116 | else | |
150b3934 | 2117 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2118 | } |
2119 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2120 | ||
b648a6de MS |
2121 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2122 | struct file_lock *fl) | |
2123 | { | |
2124 | int error; | |
2125 | ||
2126 | error = security_file_lock(filp, fl->fl_type); | |
2127 | if (error) | |
2128 | return error; | |
2129 | ||
764c76b3 MS |
2130 | for (;;) { |
2131 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2132 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2133 | break; |
764c76b3 MS |
2134 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2135 | if (!error) | |
2136 | continue; | |
2137 | ||
2138 | locks_delete_block(fl); | |
2139 | break; | |
b648a6de MS |
2140 | } |
2141 | ||
2142 | return error; | |
2143 | } | |
2144 | ||
cf01f4ee JL |
2145 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2146 | static int | |
2147 | check_fmode_for_setlk(struct file_lock *fl) | |
2148 | { | |
2149 | switch (fl->fl_type) { | |
2150 | case F_RDLCK: | |
2151 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2152 | return -EBADF; | |
2153 | break; | |
2154 | case F_WRLCK: | |
2155 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2156 | return -EBADF; | |
2157 | } | |
2158 | return 0; | |
2159 | } | |
2160 | ||
1da177e4 LT |
2161 | /* Apply the lock described by l to an open file descriptor. |
2162 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2163 | */ | |
c293621b PS |
2164 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2165 | struct flock __user *l) | |
1da177e4 LT |
2166 | { |
2167 | struct file_lock *file_lock = locks_alloc_lock(); | |
2168 | struct flock flock; | |
2169 | struct inode *inode; | |
0b2bac2f | 2170 | struct file *f; |
1da177e4 LT |
2171 | int error; |
2172 | ||
2173 | if (file_lock == NULL) | |
2174 | return -ENOLCK; | |
2175 | ||
2176 | /* | |
2177 | * This might block, so we do it before checking the inode. | |
2178 | */ | |
2179 | error = -EFAULT; | |
2180 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2181 | goto out; | |
2182 | ||
496ad9aa | 2183 | inode = file_inode(filp); |
1da177e4 LT |
2184 | |
2185 | /* Don't allow mandatory locks on files that may be memory mapped | |
2186 | * and shared. | |
2187 | */ | |
a16877ca | 2188 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2189 | error = -EAGAIN; |
2190 | goto out; | |
2191 | } | |
2192 | ||
c293621b | 2193 | again: |
1da177e4 LT |
2194 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2195 | if (error) | |
2196 | goto out; | |
5d50ffd7 | 2197 | |
cf01f4ee JL |
2198 | error = check_fmode_for_setlk(file_lock); |
2199 | if (error) | |
2200 | goto out; | |
2201 | ||
5d50ffd7 JL |
2202 | /* |
2203 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2204 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2205 | */ |
2206 | switch (cmd) { | |
0d3f7a2d | 2207 | case F_OFD_SETLK: |
90478939 JL |
2208 | error = -EINVAL; |
2209 | if (flock.l_pid != 0) | |
2210 | goto out; | |
2211 | ||
5d50ffd7 | 2212 | cmd = F_SETLK; |
cff2fce5 | 2213 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2214 | file_lock->fl_owner = filp; |
5d50ffd7 | 2215 | break; |
0d3f7a2d | 2216 | case F_OFD_SETLKW: |
90478939 JL |
2217 | error = -EINVAL; |
2218 | if (flock.l_pid != 0) | |
2219 | goto out; | |
2220 | ||
5d50ffd7 | 2221 | cmd = F_SETLKW; |
cff2fce5 | 2222 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2223 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2224 | /* Fallthrough */ |
2225 | case F_SETLKW: | |
1da177e4 LT |
2226 | file_lock->fl_flags |= FL_SLEEP; |
2227 | } | |
5d50ffd7 | 2228 | |
b648a6de | 2229 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2230 | |
c293621b PS |
2231 | /* |
2232 | * Attempt to detect a close/fcntl race and recover by | |
2233 | * releasing the lock that was just acquired. | |
2234 | */ | |
0b2bac2f AV |
2235 | /* |
2236 | * we need that spin_lock here - it prevents reordering between | |
2237 | * update of inode->i_flock and check for it done in close(). | |
2238 | * rcu_read_lock() wouldn't do. | |
2239 | */ | |
2240 | spin_lock(¤t->files->file_lock); | |
2241 | f = fcheck(fd); | |
2242 | spin_unlock(¤t->files->file_lock); | |
2243 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2244 | flock.l_type = F_UNLCK; |
2245 | goto again; | |
1da177e4 LT |
2246 | } |
2247 | ||
c293621b | 2248 | out: |
1da177e4 LT |
2249 | locks_free_lock(file_lock); |
2250 | return error; | |
2251 | } | |
2252 | ||
2253 | #if BITS_PER_LONG == 32 | |
2254 | /* Report the first existing lock that would conflict with l. | |
2255 | * This implements the F_GETLK command of fcntl(). | |
2256 | */ | |
c1e62b8f | 2257 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2258 | { |
9d6a8c5c | 2259 | struct file_lock file_lock; |
1da177e4 LT |
2260 | struct flock64 flock; |
2261 | int error; | |
2262 | ||
2263 | error = -EFAULT; | |
2264 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2265 | goto out; | |
2266 | error = -EINVAL; | |
2267 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2268 | goto out; | |
2269 | ||
2270 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2271 | if (error) | |
2272 | goto out; | |
2273 | ||
0d3f7a2d | 2274 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2275 | error = -EINVAL; |
2276 | if (flock.l_pid != 0) | |
2277 | goto out; | |
2278 | ||
5d50ffd7 | 2279 | cmd = F_GETLK64; |
cff2fce5 | 2280 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2281 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2282 | } |
2283 | ||
3ee17abd BF |
2284 | error = vfs_test_lock(filp, &file_lock); |
2285 | if (error) | |
2286 | goto out; | |
2287 | ||
9d6a8c5c ME |
2288 | flock.l_type = file_lock.fl_type; |
2289 | if (file_lock.fl_type != F_UNLCK) | |
2290 | posix_lock_to_flock64(&flock, &file_lock); | |
2291 | ||
1da177e4 LT |
2292 | error = -EFAULT; |
2293 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2294 | error = 0; | |
f328296e KM |
2295 | |
2296 | locks_release_private(&file_lock); | |
1da177e4 LT |
2297 | out: |
2298 | return error; | |
2299 | } | |
2300 | ||
2301 | /* Apply the lock described by l to an open file descriptor. | |
2302 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2303 | */ | |
c293621b PS |
2304 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2305 | struct flock64 __user *l) | |
1da177e4 LT |
2306 | { |
2307 | struct file_lock *file_lock = locks_alloc_lock(); | |
2308 | struct flock64 flock; | |
2309 | struct inode *inode; | |
0b2bac2f | 2310 | struct file *f; |
1da177e4 LT |
2311 | int error; |
2312 | ||
2313 | if (file_lock == NULL) | |
2314 | return -ENOLCK; | |
2315 | ||
2316 | /* | |
2317 | * This might block, so we do it before checking the inode. | |
2318 | */ | |
2319 | error = -EFAULT; | |
2320 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2321 | goto out; | |
2322 | ||
496ad9aa | 2323 | inode = file_inode(filp); |
1da177e4 LT |
2324 | |
2325 | /* Don't allow mandatory locks on files that may be memory mapped | |
2326 | * and shared. | |
2327 | */ | |
a16877ca | 2328 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2329 | error = -EAGAIN; |
2330 | goto out; | |
2331 | } | |
2332 | ||
c293621b | 2333 | again: |
1da177e4 LT |
2334 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2335 | if (error) | |
2336 | goto out; | |
5d50ffd7 | 2337 | |
cf01f4ee JL |
2338 | error = check_fmode_for_setlk(file_lock); |
2339 | if (error) | |
2340 | goto out; | |
2341 | ||
5d50ffd7 JL |
2342 | /* |
2343 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2344 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2345 | */ |
2346 | switch (cmd) { | |
0d3f7a2d | 2347 | case F_OFD_SETLK: |
90478939 JL |
2348 | error = -EINVAL; |
2349 | if (flock.l_pid != 0) | |
2350 | goto out; | |
2351 | ||
5d50ffd7 | 2352 | cmd = F_SETLK64; |
cff2fce5 | 2353 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2354 | file_lock->fl_owner = filp; |
5d50ffd7 | 2355 | break; |
0d3f7a2d | 2356 | case F_OFD_SETLKW: |
90478939 JL |
2357 | error = -EINVAL; |
2358 | if (flock.l_pid != 0) | |
2359 | goto out; | |
2360 | ||
5d50ffd7 | 2361 | cmd = F_SETLKW64; |
cff2fce5 | 2362 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2363 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2364 | /* Fallthrough */ |
2365 | case F_SETLKW64: | |
1da177e4 LT |
2366 | file_lock->fl_flags |= FL_SLEEP; |
2367 | } | |
5d50ffd7 | 2368 | |
b648a6de | 2369 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2370 | |
c293621b PS |
2371 | /* |
2372 | * Attempt to detect a close/fcntl race and recover by | |
2373 | * releasing the lock that was just acquired. | |
2374 | */ | |
0b2bac2f AV |
2375 | spin_lock(¤t->files->file_lock); |
2376 | f = fcheck(fd); | |
2377 | spin_unlock(¤t->files->file_lock); | |
2378 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2379 | flock.l_type = F_UNLCK; |
2380 | goto again; | |
1da177e4 LT |
2381 | } |
2382 | ||
2383 | out: | |
2384 | locks_free_lock(file_lock); | |
2385 | return error; | |
2386 | } | |
2387 | #endif /* BITS_PER_LONG == 32 */ | |
2388 | ||
2389 | /* | |
2390 | * This function is called when the file is being removed | |
2391 | * from the task's fd array. POSIX locks belonging to this task | |
2392 | * are deleted at this time. | |
2393 | */ | |
2394 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2395 | { | |
ff7b86b8 | 2396 | struct file_lock lock; |
bd61e0a9 | 2397 | struct file_lock_context *ctx = file_inode(filp)->i_flctx; |
1da177e4 LT |
2398 | |
2399 | /* | |
2400 | * If there are no locks held on this file, we don't need to call | |
2401 | * posix_lock_file(). Another process could be setting a lock on this | |
2402 | * file at the same time, but we wouldn't remove that lock anyway. | |
2403 | */ | |
bd61e0a9 | 2404 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2405 | return; |
2406 | ||
2407 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2408 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2409 | lock.fl_start = 0; |
2410 | lock.fl_end = OFFSET_MAX; | |
2411 | lock.fl_owner = owner; | |
2412 | lock.fl_pid = current->tgid; | |
2413 | lock.fl_file = filp; | |
2414 | lock.fl_ops = NULL; | |
2415 | lock.fl_lmops = NULL; | |
2416 | ||
150b3934 | 2417 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2418 | |
1da177e4 LT |
2419 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2420 | lock.fl_ops->fl_release_private(&lock); | |
2421 | } | |
2422 | ||
2423 | EXPORT_SYMBOL(locks_remove_posix); | |
2424 | ||
dd459bb1 JL |
2425 | static void |
2426 | locks_remove_flock(struct file *filp) | |
2427 | { | |
2428 | struct file_lock fl = { | |
2429 | .fl_owner = filp, | |
2430 | .fl_pid = current->tgid, | |
2431 | .fl_file = filp, | |
2432 | .fl_flags = FL_FLOCK, | |
2433 | .fl_type = F_UNLCK, | |
2434 | .fl_end = OFFSET_MAX, | |
2435 | }; | |
5263e31e | 2436 | struct file_lock_context *flctx = file_inode(filp)->i_flctx; |
dd459bb1 | 2437 | |
5263e31e | 2438 | if (!flctx || list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2439 | return; |
2440 | ||
2441 | if (filp->f_op->flock) | |
2442 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2443 | else | |
2444 | flock_lock_file(filp, &fl); | |
2445 | ||
2446 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2447 | fl.fl_ops->fl_release_private(&fl); | |
2448 | } | |
2449 | ||
1da177e4 LT |
2450 | /* |
2451 | * This function is called on the last close of an open file. | |
2452 | */ | |
78ed8a13 | 2453 | void locks_remove_file(struct file *filp) |
1da177e4 | 2454 | { |
496ad9aa | 2455 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2456 | struct file_lock *fl; |
2457 | struct file_lock **before; | |
ed9814d8 | 2458 | LIST_HEAD(dispose); |
1da177e4 | 2459 | |
dd459bb1 | 2460 | /* remove any OFD locks */ |
73a8f5f7 | 2461 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2462 | |
dd459bb1 JL |
2463 | /* remove flock locks */ |
2464 | locks_remove_flock(filp); | |
2465 | ||
2466 | if (!inode->i_flock) | |
2467 | return; | |
1da177e4 | 2468 | |
1c8c601a | 2469 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2470 | before = &inode->i_flock; |
2471 | ||
2472 | while ((fl = *before) != NULL) { | |
2473 | if (fl->fl_file == filp) { | |
1da177e4 | 2474 | if (IS_LEASE(fl)) { |
c45198ed | 2475 | lease_modify(before, F_UNLCK, &dispose); |
1da177e4 LT |
2476 | continue; |
2477 | } | |
8c3cac5e JL |
2478 | |
2479 | /* | |
2480 | * There's a leftover lock on the list of a type that | |
2481 | * we didn't expect to see. Most likely a classic | |
2482 | * POSIX lock that ended up not getting released | |
2483 | * properly, or that raced onto the list somehow. Log | |
2484 | * some info about it and then just remove it from | |
2485 | * the list. | |
2486 | */ | |
dd459bb1 | 2487 | WARN(1, "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", |
8c3cac5e JL |
2488 | MAJOR(inode->i_sb->s_dev), |
2489 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2490 | fl->fl_type, fl->fl_flags, | |
2491 | fl->fl_start, fl->fl_end); | |
2492 | ||
ed9814d8 | 2493 | locks_delete_lock(before, &dispose); |
8c3cac5e | 2494 | continue; |
1da177e4 LT |
2495 | } |
2496 | before = &fl->fl_next; | |
2497 | } | |
1c8c601a | 2498 | spin_unlock(&inode->i_lock); |
ed9814d8 | 2499 | locks_dispose_list(&dispose); |
1da177e4 LT |
2500 | } |
2501 | ||
1da177e4 LT |
2502 | /** |
2503 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2504 | * @waiter: the lock which was waiting |
2505 | * | |
2506 | * lockd needs to block waiting for locks. | |
2507 | */ | |
64a318ee | 2508 | int |
f891a29f | 2509 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2510 | { |
64a318ee BF |
2511 | int status = 0; |
2512 | ||
7b2296af | 2513 | spin_lock(&blocked_lock_lock); |
5996a298 | 2514 | if (waiter->fl_next) |
1da177e4 | 2515 | __locks_delete_block(waiter); |
64a318ee BF |
2516 | else |
2517 | status = -ENOENT; | |
7b2296af | 2518 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2519 | return status; |
1da177e4 | 2520 | } |
1da177e4 LT |
2521 | EXPORT_SYMBOL(posix_unblock_lock); |
2522 | ||
9b9d2ab4 ME |
2523 | /** |
2524 | * vfs_cancel_lock - file byte range unblock lock | |
2525 | * @filp: The file to apply the unblock to | |
2526 | * @fl: The lock to be unblocked | |
2527 | * | |
2528 | * Used by lock managers to cancel blocked requests | |
2529 | */ | |
2530 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2531 | { | |
72c2d531 | 2532 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2533 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2534 | return 0; | |
2535 | } | |
2536 | ||
2537 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2538 | ||
7f8ada98 | 2539 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2540 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2541 | #include <linux/seq_file.h> |
2542 | ||
7012b02a JL |
2543 | struct locks_iterator { |
2544 | int li_cpu; | |
2545 | loff_t li_pos; | |
2546 | }; | |
2547 | ||
7f8ada98 | 2548 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2549 | loff_t id, char *pfx) |
1da177e4 LT |
2550 | { |
2551 | struct inode *inode = NULL; | |
ab1f1611 VG |
2552 | unsigned int fl_pid; |
2553 | ||
2554 | if (fl->fl_nspid) | |
6c5f3e7b | 2555 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2556 | else |
2557 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2558 | |
2559 | if (fl->fl_file != NULL) | |
496ad9aa | 2560 | inode = file_inode(fl->fl_file); |
1da177e4 | 2561 | |
99dc8292 | 2562 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2563 | if (IS_POSIX(fl)) { |
c918d42a | 2564 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2565 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2566 | else if (IS_OFDLCK(fl)) |
5315c26a | 2567 | seq_puts(f, "OFDLCK"); |
c918d42a | 2568 | else |
5315c26a | 2569 | seq_puts(f, "POSIX "); |
c918d42a JL |
2570 | |
2571 | seq_printf(f, " %s ", | |
1da177e4 | 2572 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2573 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2574 | } else if (IS_FLOCK(fl)) { |
2575 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2576 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2577 | } else { |
5315c26a | 2578 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2579 | } |
2580 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2581 | if (fl->fl_flags & FL_DELEG) |
2582 | seq_puts(f, "DELEG "); | |
2583 | else | |
2584 | seq_puts(f, "LEASE "); | |
2585 | ||
ab83fa4b | 2586 | if (lease_breaking(fl)) |
5315c26a | 2587 | seq_puts(f, "BREAKING "); |
1da177e4 | 2588 | else if (fl->fl_file) |
5315c26a | 2589 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2590 | else |
5315c26a | 2591 | seq_puts(f, "BREAKER "); |
1da177e4 | 2592 | } else { |
5315c26a | 2593 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2594 | } |
2595 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2596 | seq_printf(f, "%s ", |
1da177e4 LT |
2597 | (fl->fl_type & LOCK_READ) |
2598 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2599 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2600 | } else { | |
7f8ada98 | 2601 | seq_printf(f, "%s ", |
ab83fa4b | 2602 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2603 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2604 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2605 | } |
2606 | if (inode) { | |
2607 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2608 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2609 | inode->i_sb->s_id, inode->i_ino); |
2610 | #else | |
2611 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2612 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2613 | MAJOR(inode->i_sb->s_dev), |
2614 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2615 | #endif | |
2616 | } else { | |
ab1f1611 | 2617 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2618 | } |
2619 | if (IS_POSIX(fl)) { | |
2620 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2621 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2622 | else |
7f8ada98 | 2623 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2624 | } else { |
5315c26a | 2625 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2626 | } |
2627 | } | |
2628 | ||
7f8ada98 | 2629 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2630 | { |
7012b02a | 2631 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2632 | struct file_lock *fl, *bfl; |
1da177e4 | 2633 | |
139ca04e | 2634 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2635 | |
7012b02a | 2636 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2637 | |
7f8ada98 | 2638 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2639 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2640 | |
7f8ada98 PE |
2641 | return 0; |
2642 | } | |
1da177e4 | 2643 | |
7f8ada98 | 2644 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2645 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2646 | { |
7012b02a | 2647 | struct locks_iterator *iter = f->private; |
99dc8292 | 2648 | |
7012b02a JL |
2649 | iter->li_pos = *pos + 1; |
2650 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2651 | spin_lock(&blocked_lock_lock); |
7012b02a | 2652 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2653 | } |
1da177e4 | 2654 | |
7f8ada98 PE |
2655 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2656 | { | |
7012b02a JL |
2657 | struct locks_iterator *iter = f->private; |
2658 | ||
2659 | ++iter->li_pos; | |
2660 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2661 | } |
1da177e4 | 2662 | |
7f8ada98 | 2663 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2664 | __releases(&blocked_lock_lock) |
7f8ada98 | 2665 | { |
7b2296af | 2666 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2667 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2668 | } |
2669 | ||
d8ba7a36 | 2670 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2671 | .start = locks_start, |
2672 | .next = locks_next, | |
2673 | .stop = locks_stop, | |
2674 | .show = locks_show, | |
2675 | }; | |
d8ba7a36 AD |
2676 | |
2677 | static int locks_open(struct inode *inode, struct file *filp) | |
2678 | { | |
7012b02a JL |
2679 | return seq_open_private(filp, &locks_seq_operations, |
2680 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2681 | } |
2682 | ||
2683 | static const struct file_operations proc_locks_operations = { | |
2684 | .open = locks_open, | |
2685 | .read = seq_read, | |
2686 | .llseek = seq_lseek, | |
99dc8292 | 2687 | .release = seq_release_private, |
d8ba7a36 AD |
2688 | }; |
2689 | ||
2690 | static int __init proc_locks_init(void) | |
2691 | { | |
2692 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2693 | return 0; | |
2694 | } | |
2695 | module_init(proc_locks_init); | |
7f8ada98 PE |
2696 | #endif |
2697 | ||
1da177e4 LT |
2698 | static int __init filelock_init(void) |
2699 | { | |
7012b02a JL |
2700 | int i; |
2701 | ||
4a075e39 JL |
2702 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2703 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2704 | ||
1da177e4 | 2705 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2706 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2707 | ||
7012b02a JL |
2708 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2709 | ||
2710 | for_each_possible_cpu(i) | |
2711 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2712 | ||
1da177e4 LT |
2713 | return 0; |
2714 | } | |
2715 | ||
2716 | core_initcall(filelock_init); |