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 | |
6109c850 | 164 | * the relevant flc_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, | |
6109c850 | 192 | * we often hold the flc_lock as well. In certain cases, when reading the fields |
1c8c601a | 193 | * protected by this lock, we can skip acquiring it iff we already hold the |
6109c850 | 194 | * flc_lock. |
1c8c601a JL |
195 | * |
196 | * In particular, adding an entry to the fl_block list requires that you hold | |
6109c850 JL |
197 | * both the flc_lock and the blocked_lock_lock (acquired in that order). |
198 | * Deleting an entry from the list however only requires the file_lock_lock. | |
1c8c601a | 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 | ||
6109c850 | 217 | spin_lock_init(&new->flc_lock); |
4a075e39 | 218 | INIT_LIST_HEAD(&new->flc_flock); |
bd61e0a9 | 219 | INIT_LIST_HEAD(&new->flc_posix); |
8634b51f | 220 | INIT_LIST_HEAD(&new->flc_lease); |
4a075e39 JL |
221 | |
222 | /* | |
223 | * Assign the pointer if it's not already assigned. If it is, then | |
224 | * free the context we just allocated. | |
225 | */ | |
226 | spin_lock(&inode->i_lock); | |
227 | if (likely(!inode->i_flctx)) { | |
228 | inode->i_flctx = new; | |
229 | new = NULL; | |
230 | } | |
231 | spin_unlock(&inode->i_lock); | |
232 | ||
233 | if (new) | |
234 | kmem_cache_free(flctx_cache, new); | |
235 | out: | |
236 | return inode->i_flctx; | |
237 | } | |
238 | ||
239 | void | |
240 | locks_free_lock_context(struct file_lock_context *ctx) | |
241 | { | |
242 | if (ctx) { | |
243 | WARN_ON_ONCE(!list_empty(&ctx->flc_flock)); | |
bd61e0a9 | 244 | WARN_ON_ONCE(!list_empty(&ctx->flc_posix)); |
8634b51f | 245 | WARN_ON_ONCE(!list_empty(&ctx->flc_lease)); |
4a075e39 JL |
246 | kmem_cache_free(flctx_cache, ctx); |
247 | } | |
248 | } | |
249 | ||
ee19cc40 | 250 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 251 | { |
139ca04e | 252 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 253 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
254 | INIT_LIST_HEAD(&fl->fl_block); |
255 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
256 | } |
257 | ||
1da177e4 | 258 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 259 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 260 | { |
ee19cc40 | 261 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
262 | |
263 | if (fl) | |
ee19cc40 | 264 | locks_init_lock_heads(fl); |
a51cb91d MS |
265 | |
266 | return fl; | |
1da177e4 | 267 | } |
c5b1f0d9 | 268 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 269 | |
a9e61e25 | 270 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
271 | { |
272 | if (fl->fl_ops) { | |
273 | if (fl->fl_ops->fl_release_private) | |
274 | fl->fl_ops->fl_release_private(fl); | |
275 | fl->fl_ops = NULL; | |
276 | } | |
47831f35 | 277 | |
5c97d7b1 KM |
278 | if (fl->fl_lmops) { |
279 | if (fl->fl_lmops->lm_put_owner) | |
280 | fl->fl_lmops->lm_put_owner(fl); | |
281 | fl->fl_lmops = NULL; | |
282 | } | |
47831f35 | 283 | } |
a9e61e25 | 284 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 285 | |
1da177e4 | 286 | /* Free a lock which is not in use. */ |
05fa3135 | 287 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 288 | { |
5ce29646 | 289 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 290 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 291 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 292 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 293 | |
47831f35 | 294 | locks_release_private(fl); |
1da177e4 LT |
295 | kmem_cache_free(filelock_cache, fl); |
296 | } | |
05fa3135 | 297 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 298 | |
ed9814d8 JL |
299 | static void |
300 | locks_dispose_list(struct list_head *dispose) | |
301 | { | |
302 | struct file_lock *fl; | |
303 | ||
304 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
305 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
306 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
307 | locks_free_lock(fl); |
308 | } | |
309 | } | |
310 | ||
1da177e4 LT |
311 | void locks_init_lock(struct file_lock *fl) |
312 | { | |
ee19cc40 MS |
313 | memset(fl, 0, sizeof(struct file_lock)); |
314 | locks_init_lock_heads(fl); | |
1da177e4 LT |
315 | } |
316 | ||
317 | EXPORT_SYMBOL(locks_init_lock); | |
318 | ||
1da177e4 LT |
319 | /* |
320 | * Initialize a new lock from an existing file_lock structure. | |
321 | */ | |
3fe0fff1 | 322 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
323 | { |
324 | new->fl_owner = fl->fl_owner; | |
325 | new->fl_pid = fl->fl_pid; | |
0996905f | 326 | new->fl_file = NULL; |
1da177e4 LT |
327 | new->fl_flags = fl->fl_flags; |
328 | new->fl_type = fl->fl_type; | |
329 | new->fl_start = fl->fl_start; | |
330 | new->fl_end = fl->fl_end; | |
f328296e | 331 | new->fl_lmops = fl->fl_lmops; |
0996905f | 332 | new->fl_ops = NULL; |
f328296e KM |
333 | |
334 | if (fl->fl_lmops) { | |
335 | if (fl->fl_lmops->lm_get_owner) | |
336 | fl->fl_lmops->lm_get_owner(new, fl); | |
337 | } | |
0996905f | 338 | } |
3fe0fff1 | 339 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
340 | |
341 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
342 | { | |
566709bd JL |
343 | /* "new" must be a freshly-initialized lock */ |
344 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 345 | |
3fe0fff1 | 346 | locks_copy_conflock(new, fl); |
f328296e | 347 | |
0996905f | 348 | new->fl_file = fl->fl_file; |
1da177e4 | 349 | new->fl_ops = fl->fl_ops; |
47831f35 | 350 | |
f328296e KM |
351 | if (fl->fl_ops) { |
352 | if (fl->fl_ops->fl_copy_lock) | |
353 | fl->fl_ops->fl_copy_lock(new, fl); | |
354 | } | |
1da177e4 LT |
355 | } |
356 | ||
357 | EXPORT_SYMBOL(locks_copy_lock); | |
358 | ||
359 | static inline int flock_translate_cmd(int cmd) { | |
360 | if (cmd & LOCK_MAND) | |
361 | return cmd & (LOCK_MAND | LOCK_RW); | |
362 | switch (cmd) { | |
363 | case LOCK_SH: | |
364 | return F_RDLCK; | |
365 | case LOCK_EX: | |
366 | return F_WRLCK; | |
367 | case LOCK_UN: | |
368 | return F_UNLCK; | |
369 | } | |
370 | return -EINVAL; | |
371 | } | |
372 | ||
373 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
374 | static struct file_lock * |
375 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
376 | { |
377 | struct file_lock *fl; | |
378 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 379 | |
1da177e4 | 380 | if (type < 0) |
6e129d00 | 381 | return ERR_PTR(type); |
1da177e4 LT |
382 | |
383 | fl = locks_alloc_lock(); | |
384 | if (fl == NULL) | |
6e129d00 | 385 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
386 | |
387 | fl->fl_file = filp; | |
73a8f5f7 | 388 | fl->fl_owner = filp; |
1da177e4 LT |
389 | fl->fl_pid = current->tgid; |
390 | fl->fl_flags = FL_FLOCK; | |
391 | fl->fl_type = type; | |
392 | fl->fl_end = OFFSET_MAX; | |
393 | ||
6e129d00 | 394 | return fl; |
1da177e4 LT |
395 | } |
396 | ||
0ec4f431 | 397 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
398 | { |
399 | switch (type) { | |
400 | case F_RDLCK: | |
401 | case F_WRLCK: | |
402 | case F_UNLCK: | |
403 | fl->fl_type = type; | |
404 | break; | |
405 | default: | |
406 | return -EINVAL; | |
407 | } | |
408 | return 0; | |
409 | } | |
410 | ||
ef12e72a BF |
411 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
412 | struct flock64 *l) | |
1da177e4 | 413 | { |
1da177e4 | 414 | switch (l->l_whence) { |
f5579f8c | 415 | case SEEK_SET: |
ef12e72a | 416 | fl->fl_start = 0; |
1da177e4 | 417 | break; |
f5579f8c | 418 | case SEEK_CUR: |
ef12e72a | 419 | fl->fl_start = filp->f_pos; |
1da177e4 | 420 | break; |
f5579f8c | 421 | case SEEK_END: |
ef12e72a | 422 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
423 | break; |
424 | default: | |
425 | return -EINVAL; | |
426 | } | |
ef12e72a BF |
427 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
428 | return -EOVERFLOW; | |
429 | fl->fl_start += l->l_start; | |
430 | if (fl->fl_start < 0) | |
431 | return -EINVAL; | |
1da177e4 LT |
432 | |
433 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
434 | POSIX-2001 defines it. */ | |
4c780a46 | 435 | if (l->l_len > 0) { |
ef12e72a BF |
436 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
437 | return -EOVERFLOW; | |
438 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
439 | ||
4c780a46 | 440 | } else if (l->l_len < 0) { |
ef12e72a | 441 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 442 | return -EINVAL; |
ef12e72a BF |
443 | fl->fl_end = fl->fl_start - 1; |
444 | fl->fl_start += l->l_len; | |
445 | } else | |
446 | fl->fl_end = OFFSET_MAX; | |
447 | ||
1da177e4 LT |
448 | fl->fl_owner = current->files; |
449 | fl->fl_pid = current->tgid; | |
450 | fl->fl_file = filp; | |
451 | fl->fl_flags = FL_POSIX; | |
452 | fl->fl_ops = NULL; | |
453 | fl->fl_lmops = NULL; | |
454 | ||
455 | return assign_type(fl, l->l_type); | |
456 | } | |
457 | ||
ef12e72a BF |
458 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
459 | * style lock. | |
460 | */ | |
461 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
462 | struct flock *l) | |
1da177e4 | 463 | { |
ef12e72a BF |
464 | struct flock64 ll = { |
465 | .l_type = l->l_type, | |
466 | .l_whence = l->l_whence, | |
467 | .l_start = l->l_start, | |
468 | .l_len = l->l_len, | |
469 | }; | |
470 | ||
471 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 472 | } |
1da177e4 LT |
473 | |
474 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
475 | static bool |
476 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
477 | { |
478 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 479 | return false; |
1da177e4 LT |
480 | } |
481 | ||
1c7dd2ff JL |
482 | static void |
483 | lease_setup(struct file_lock *fl, void **priv) | |
484 | { | |
485 | struct file *filp = fl->fl_file; | |
486 | struct fasync_struct *fa = *priv; | |
487 | ||
488 | /* | |
489 | * fasync_insert_entry() returns the old entry if any. If there was no | |
490 | * old entry, then it used "priv" and inserted it into the fasync list. | |
491 | * Clear the pointer to indicate that it shouldn't be freed. | |
492 | */ | |
493 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
494 | *priv = NULL; | |
495 | ||
496 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
497 | } | |
498 | ||
7b021967 | 499 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 500 | .lm_break = lease_break_callback, |
8fb47a4f | 501 | .lm_change = lease_modify, |
1c7dd2ff | 502 | .lm_setup = lease_setup, |
1da177e4 LT |
503 | }; |
504 | ||
505 | /* | |
506 | * Initialize a lease, use the default lock manager operations | |
507 | */ | |
0ec4f431 | 508 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 509 | { |
75dff55a TM |
510 | if (assign_type(fl, type) != 0) |
511 | return -EINVAL; | |
512 | ||
7ca76311 | 513 | fl->fl_owner = filp; |
1da177e4 LT |
514 | fl->fl_pid = current->tgid; |
515 | ||
516 | fl->fl_file = filp; | |
517 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
518 | fl->fl_start = 0; |
519 | fl->fl_end = OFFSET_MAX; | |
520 | fl->fl_ops = NULL; | |
521 | fl->fl_lmops = &lease_manager_ops; | |
522 | return 0; | |
523 | } | |
524 | ||
525 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 526 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
527 | { |
528 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 529 | int error = -ENOMEM; |
1da177e4 LT |
530 | |
531 | if (fl == NULL) | |
e32b8ee2 | 532 | return ERR_PTR(error); |
1da177e4 LT |
533 | |
534 | error = lease_init(filp, type, fl); | |
75dff55a TM |
535 | if (error) { |
536 | locks_free_lock(fl); | |
e32b8ee2 | 537 | return ERR_PTR(error); |
75dff55a | 538 | } |
e32b8ee2 | 539 | return fl; |
1da177e4 LT |
540 | } |
541 | ||
542 | /* Check if two locks overlap each other. | |
543 | */ | |
544 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
545 | { | |
546 | return ((fl1->fl_end >= fl2->fl_start) && | |
547 | (fl2->fl_end >= fl1->fl_start)); | |
548 | } | |
549 | ||
550 | /* | |
551 | * Check whether two locks have the same owner. | |
552 | */ | |
33443c42 | 553 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 554 | { |
8fb47a4f | 555 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 556 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 557 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
558 | return fl1->fl_owner == fl2->fl_owner; |
559 | } | |
560 | ||
6109c850 | 561 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 562 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 563 | { |
7012b02a JL |
564 | lg_local_lock(&file_lock_lglock); |
565 | fl->fl_link_cpu = smp_processor_id(); | |
566 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
567 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
568 | } |
569 | ||
6109c850 | 570 | /* Must be called with the flc_lock held! */ |
6ca10ed8 | 571 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 572 | { |
7012b02a JL |
573 | /* |
574 | * Avoid taking lock if already unhashed. This is safe since this check | |
6109c850 | 575 | * is done while holding the flc_lock, and new insertions into the list |
7012b02a JL |
576 | * also require that it be held. |
577 | */ | |
578 | if (hlist_unhashed(&fl->fl_link)) | |
579 | return; | |
580 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 581 | hlist_del_init(&fl->fl_link); |
7012b02a | 582 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
583 | } |
584 | ||
3999e493 JL |
585 | static unsigned long |
586 | posix_owner_key(struct file_lock *fl) | |
587 | { | |
588 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
589 | return fl->fl_lmops->lm_owner_key(fl); | |
590 | return (unsigned long)fl->fl_owner; | |
591 | } | |
592 | ||
6ca10ed8 | 593 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 594 | { |
3999e493 | 595 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
596 | } |
597 | ||
6ca10ed8 | 598 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 599 | { |
48f74186 | 600 | hash_del(&waiter->fl_link); |
88974691 JL |
601 | } |
602 | ||
1da177e4 LT |
603 | /* Remove waiter from blocker's block list. |
604 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 605 | * |
7b2296af | 606 | * Must be called with blocked_lock_lock held. |
1da177e4 | 607 | */ |
33443c42 | 608 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 609 | { |
88974691 | 610 | locks_delete_global_blocked(waiter); |
1da177e4 | 611 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
612 | waiter->fl_next = NULL; |
613 | } | |
614 | ||
1a9e64a7 | 615 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 616 | { |
7b2296af | 617 | spin_lock(&blocked_lock_lock); |
1da177e4 | 618 | __locks_delete_block(waiter); |
7b2296af | 619 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
620 | } |
621 | ||
622 | /* Insert waiter into blocker's block list. | |
623 | * We use a circular list so that processes can be easily woken up in | |
624 | * the order they blocked. The documentation doesn't require this but | |
625 | * it seems like the reasonable thing to do. | |
1c8c601a | 626 | * |
6109c850 JL |
627 | * Must be called with both the flc_lock and blocked_lock_lock held. The |
628 | * fl_block list itself is protected by the blocked_lock_lock, but by ensuring | |
629 | * that the flc_lock is also held on insertions we can avoid taking the | |
630 | * blocked_lock_lock in some cases when we see that the fl_block list is empty. | |
1da177e4 | 631 | */ |
1c8c601a JL |
632 | static void __locks_insert_block(struct file_lock *blocker, |
633 | struct file_lock *waiter) | |
1da177e4 | 634 | { |
6dc0fe8f | 635 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 636 | waiter->fl_next = blocker; |
88974691 | 637 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 638 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
639 | locks_insert_global_blocked(waiter); |
640 | } | |
641 | ||
6109c850 | 642 | /* Must be called with flc_lock held. */ |
1c8c601a JL |
643 | static void locks_insert_block(struct file_lock *blocker, |
644 | struct file_lock *waiter) | |
645 | { | |
7b2296af | 646 | spin_lock(&blocked_lock_lock); |
1c8c601a | 647 | __locks_insert_block(blocker, waiter); |
7b2296af | 648 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
649 | } |
650 | ||
1cb36012 JL |
651 | /* |
652 | * Wake up processes blocked waiting for blocker. | |
653 | * | |
6109c850 | 654 | * Must be called with the inode->flc_lock held! |
1da177e4 LT |
655 | */ |
656 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
657 | { | |
4e8c765d JL |
658 | /* |
659 | * Avoid taking global lock if list is empty. This is safe since new | |
6109c850 JL |
660 | * blocked requests are only added to the list under the flc_lock, and |
661 | * the flc_lock is always held here. Note that removal from the fl_block | |
662 | * list does not require the flc_lock, so we must recheck list_empty() | |
7b2296af | 663 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
664 | */ |
665 | if (list_empty(&blocker->fl_block)) | |
666 | return; | |
667 | ||
7b2296af | 668 | spin_lock(&blocked_lock_lock); |
1da177e4 | 669 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
670 | struct file_lock *waiter; |
671 | ||
672 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
673 | struct file_lock, fl_block); |
674 | __locks_delete_block(waiter); | |
8fb47a4f BF |
675 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
676 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
677 | else |
678 | wake_up(&waiter->fl_wait); | |
679 | } | |
7b2296af | 680 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
681 | } |
682 | ||
5263e31e | 683 | static void |
9bd0f45b JL |
684 | locks_insert_lock_ctx(struct file_lock *fl, int *counter, |
685 | struct list_head *before) | |
5263e31e JL |
686 | { |
687 | fl->fl_nspid = get_pid(task_tgid(current)); | |
688 | list_add_tail(&fl->fl_list, before); | |
9bd0f45b | 689 | ++*counter; |
5263e31e JL |
690 | locks_insert_global_locks(fl); |
691 | } | |
692 | ||
8634b51f | 693 | static void |
9bd0f45b | 694 | locks_unlink_lock_ctx(struct file_lock *fl, int *counter) |
1da177e4 | 695 | { |
88974691 | 696 | locks_delete_global_locks(fl); |
8634b51f | 697 | list_del_init(&fl->fl_list); |
9bd0f45b | 698 | --*counter; |
ab1f1611 VG |
699 | if (fl->fl_nspid) { |
700 | put_pid(fl->fl_nspid); | |
701 | fl->fl_nspid = NULL; | |
702 | } | |
1da177e4 | 703 | locks_wake_up_blocks(fl); |
24cbe784 JL |
704 | } |
705 | ||
8634b51f | 706 | static void |
9bd0f45b JL |
707 | locks_delete_lock_ctx(struct file_lock *fl, int *counter, |
708 | struct list_head *dispose) | |
24cbe784 | 709 | { |
9bd0f45b | 710 | locks_unlink_lock_ctx(fl, counter); |
ed9814d8 | 711 | if (dispose) |
6dee60f6 | 712 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
713 | else |
714 | locks_free_lock(fl); | |
1da177e4 LT |
715 | } |
716 | ||
717 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
718 | * checks for shared/exclusive status of overlapping locks. | |
719 | */ | |
720 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
721 | { | |
722 | if (sys_fl->fl_type == F_WRLCK) | |
723 | return 1; | |
724 | if (caller_fl->fl_type == F_WRLCK) | |
725 | return 1; | |
726 | return 0; | |
727 | } | |
728 | ||
729 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
730 | * checking before calling the locks_conflict(). | |
731 | */ | |
732 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
733 | { | |
734 | /* POSIX locks owned by the same process do not conflict with | |
735 | * each other. | |
736 | */ | |
737 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
738 | return (0); | |
739 | ||
740 | /* Check whether they overlap */ | |
741 | if (!locks_overlap(caller_fl, sys_fl)) | |
742 | return 0; | |
743 | ||
744 | return (locks_conflict(caller_fl, sys_fl)); | |
745 | } | |
746 | ||
747 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
748 | * checking before calling the locks_conflict(). | |
749 | */ | |
750 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
751 | { | |
752 | /* FLOCK locks referring to the same filp do not conflict with | |
753 | * each other. | |
754 | */ | |
755 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
756 | return (0); | |
757 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
758 | return 0; | |
759 | ||
760 | return (locks_conflict(caller_fl, sys_fl)); | |
761 | } | |
762 | ||
6d34ac19 | 763 | void |
9d6a8c5c | 764 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
765 | { |
766 | struct file_lock *cfl; | |
bd61e0a9 | 767 | struct file_lock_context *ctx; |
1c8c601a | 768 | struct inode *inode = file_inode(filp); |
1da177e4 | 769 | |
bd61e0a9 JL |
770 | ctx = inode->i_flctx; |
771 | if (!ctx || list_empty_careful(&ctx->flc_posix)) { | |
772 | fl->fl_type = F_UNLCK; | |
773 | return; | |
774 | } | |
775 | ||
6109c850 | 776 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
777 | list_for_each_entry(cfl, &ctx->flc_posix, fl_list) { |
778 | if (posix_locks_conflict(fl, cfl)) { | |
779 | locks_copy_conflock(fl, cfl); | |
780 | if (cfl->fl_nspid) | |
781 | fl->fl_pid = pid_vnr(cfl->fl_nspid); | |
782 | goto out; | |
783 | } | |
1da177e4 | 784 | } |
bd61e0a9 JL |
785 | fl->fl_type = F_UNLCK; |
786 | out: | |
6109c850 | 787 | spin_unlock(&ctx->flc_lock); |
6d34ac19 | 788 | return; |
1da177e4 | 789 | } |
1da177e4 LT |
790 | EXPORT_SYMBOL(posix_test_lock); |
791 | ||
b533184f BF |
792 | /* |
793 | * Deadlock detection: | |
794 | * | |
795 | * We attempt to detect deadlocks that are due purely to posix file | |
796 | * locks. | |
1da177e4 | 797 | * |
b533184f BF |
798 | * We assume that a task can be waiting for at most one lock at a time. |
799 | * So for any acquired lock, the process holding that lock may be | |
800 | * waiting on at most one other lock. That lock in turns may be held by | |
801 | * someone waiting for at most one other lock. Given a requested lock | |
802 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
803 | * follow this chain of waiters to ensure we are not about to create a | |
804 | * cycle. | |
1da177e4 | 805 | * |
b533184f BF |
806 | * Since we do this before we ever put a process to sleep on a lock, we |
807 | * are ensured that there is never a cycle; that is what guarantees that | |
808 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 809 | * |
b533184f BF |
810 | * Note: the above assumption may not be true when handling lock |
811 | * requests from a broken NFS client. It may also fail in the presence | |
812 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 813 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 814 | * |
cff2fce5 | 815 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
816 | * Because the owner is not even nominally tied to a thread of |
817 | * execution, the deadlock detection below can't reasonably work well. Just | |
818 | * skip it for those. | |
819 | * | |
cff2fce5 | 820 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
821 | * locks that just checks for the case where two tasks are attempting to |
822 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 823 | */ |
97855b49 BF |
824 | |
825 | #define MAX_DEADLK_ITERATIONS 10 | |
826 | ||
b533184f BF |
827 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
828 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
829 | { | |
830 | struct file_lock *fl; | |
831 | ||
3999e493 | 832 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
833 | if (posix_same_owner(fl, block_fl)) |
834 | return fl->fl_next; | |
835 | } | |
836 | return NULL; | |
837 | } | |
838 | ||
7b2296af | 839 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 840 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
841 | struct file_lock *block_fl) |
842 | { | |
97855b49 | 843 | int i = 0; |
1da177e4 | 844 | |
57b65325 JL |
845 | /* |
846 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 847 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 848 | */ |
cff2fce5 | 849 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
850 | return 0; |
851 | ||
b533184f BF |
852 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
853 | if (i++ > MAX_DEADLK_ITERATIONS) | |
854 | return 0; | |
855 | if (posix_same_owner(caller_fl, block_fl)) | |
856 | return 1; | |
1da177e4 LT |
857 | } |
858 | return 0; | |
859 | } | |
860 | ||
1da177e4 | 861 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 862 | * after any leases, but before any posix locks. |
f475ae95 TM |
863 | * |
864 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
865 | * whether or not a lock was successfully freed by testing the return | |
866 | * value for -ENOENT. | |
1da177e4 | 867 | */ |
993dfa87 | 868 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 869 | { |
993dfa87 | 870 | struct file_lock *new_fl = NULL; |
5263e31e JL |
871 | struct file_lock *fl; |
872 | struct file_lock_context *ctx; | |
873 | struct inode *inode = file_inode(filp); | |
1da177e4 | 874 | int error = 0; |
5263e31e | 875 | bool found = false; |
ed9814d8 | 876 | LIST_HEAD(dispose); |
1da177e4 | 877 | |
5263e31e JL |
878 | ctx = locks_get_lock_context(inode); |
879 | if (!ctx) | |
880 | return -ENOMEM; | |
881 | ||
b89f4321 | 882 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 883 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
884 | if (!new_fl) |
885 | return -ENOMEM; | |
84d535ad PE |
886 | } |
887 | ||
6109c850 | 888 | spin_lock(&ctx->flc_lock); |
b89f4321 AB |
889 | if (request->fl_flags & FL_ACCESS) |
890 | goto find_conflict; | |
891 | ||
5263e31e | 892 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
1da177e4 LT |
893 | if (filp != fl->fl_file) |
894 | continue; | |
993dfa87 | 895 | if (request->fl_type == fl->fl_type) |
1da177e4 | 896 | goto out; |
5263e31e | 897 | found = true; |
9bd0f45b | 898 | locks_delete_lock_ctx(fl, &ctx->flc_flock_cnt, &dispose); |
1da177e4 LT |
899 | break; |
900 | } | |
1da177e4 | 901 | |
f475ae95 TM |
902 | if (request->fl_type == F_UNLCK) { |
903 | if ((request->fl_flags & FL_EXISTS) && !found) | |
904 | error = -ENOENT; | |
993dfa87 | 905 | goto out; |
f475ae95 | 906 | } |
1da177e4 LT |
907 | |
908 | /* | |
909 | * If a higher-priority process was blocked on the old file lock, | |
910 | * give it the opportunity to lock the file. | |
911 | */ | |
b89f4321 | 912 | if (found) { |
6109c850 | 913 | spin_unlock(&ctx->flc_lock); |
def01bc5 | 914 | cond_resched(); |
6109c850 | 915 | spin_lock(&ctx->flc_lock); |
b89f4321 | 916 | } |
1da177e4 | 917 | |
f07f18dd | 918 | find_conflict: |
5263e31e | 919 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 920 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
921 | continue; |
922 | error = -EAGAIN; | |
bde74e4b MS |
923 | if (!(request->fl_flags & FL_SLEEP)) |
924 | goto out; | |
925 | error = FILE_LOCK_DEFERRED; | |
926 | locks_insert_block(fl, request); | |
1da177e4 LT |
927 | goto out; |
928 | } | |
f07f18dd TM |
929 | if (request->fl_flags & FL_ACCESS) |
930 | goto out; | |
993dfa87 | 931 | locks_copy_lock(new_fl, request); |
9bd0f45b | 932 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock_cnt, &ctx->flc_flock); |
993dfa87 | 933 | new_fl = NULL; |
9cedc194 | 934 | error = 0; |
1da177e4 LT |
935 | |
936 | out: | |
6109c850 | 937 | spin_unlock(&ctx->flc_lock); |
993dfa87 TM |
938 | if (new_fl) |
939 | locks_free_lock(new_fl); | |
ed9814d8 | 940 | locks_dispose_list(&dispose); |
1da177e4 LT |
941 | return error; |
942 | } | |
943 | ||
150b3934 | 944 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 | 945 | { |
bd61e0a9 | 946 | struct file_lock *fl, *tmp; |
39005d02 MS |
947 | struct file_lock *new_fl = NULL; |
948 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
949 | struct file_lock *left = NULL; |
950 | struct file_lock *right = NULL; | |
bd61e0a9 | 951 | struct file_lock_context *ctx; |
b9746ef8 JL |
952 | int error; |
953 | bool added = false; | |
ed9814d8 | 954 | LIST_HEAD(dispose); |
1da177e4 | 955 | |
bd61e0a9 JL |
956 | ctx = locks_get_lock_context(inode); |
957 | if (!ctx) | |
958 | return -ENOMEM; | |
959 | ||
1da177e4 LT |
960 | /* |
961 | * We may need two file_lock structures for this operation, | |
962 | * so we get them in advance to avoid races. | |
39005d02 MS |
963 | * |
964 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 965 | */ |
39005d02 MS |
966 | if (!(request->fl_flags & FL_ACCESS) && |
967 | (request->fl_type != F_UNLCK || | |
968 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
969 | new_fl = locks_alloc_lock(); | |
970 | new_fl2 = locks_alloc_lock(); | |
971 | } | |
1da177e4 | 972 | |
6109c850 | 973 | spin_lock(&ctx->flc_lock); |
1cb36012 JL |
974 | /* |
975 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
976 | * there are any, either return error or put the request on the | |
48f74186 | 977 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 978 | */ |
1da177e4 | 979 | if (request->fl_type != F_UNLCK) { |
bd61e0a9 | 980 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { |
1da177e4 LT |
981 | if (!IS_POSIX(fl)) |
982 | continue; | |
983 | if (!posix_locks_conflict(request, fl)) | |
984 | continue; | |
5842add2 | 985 | if (conflock) |
3fe0fff1 | 986 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
987 | error = -EAGAIN; |
988 | if (!(request->fl_flags & FL_SLEEP)) | |
989 | goto out; | |
1c8c601a JL |
990 | /* |
991 | * Deadlock detection and insertion into the blocked | |
992 | * locks list must be done while holding the same lock! | |
993 | */ | |
1da177e4 | 994 | error = -EDEADLK; |
7b2296af | 995 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
996 | if (likely(!posix_locks_deadlock(request, fl))) { |
997 | error = FILE_LOCK_DEFERRED; | |
998 | __locks_insert_block(fl, request); | |
999 | } | |
7b2296af | 1000 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1001 | goto out; |
1002 | } | |
1003 | } | |
1004 | ||
1005 | /* If we're just looking for a conflict, we're done. */ | |
1006 | error = 0; | |
1007 | if (request->fl_flags & FL_ACCESS) | |
1008 | goto out; | |
1009 | ||
bd61e0a9 JL |
1010 | /* Find the first old lock with the same owner as the new lock */ |
1011 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
1012 | if (posix_same_owner(request, fl)) | |
1013 | break; | |
1da177e4 LT |
1014 | } |
1015 | ||
1cb36012 | 1016 | /* Process locks with this owner. */ |
bd61e0a9 JL |
1017 | list_for_each_entry_safe_from(fl, tmp, &ctx->flc_posix, fl_list) { |
1018 | if (!posix_same_owner(request, fl)) | |
1019 | break; | |
1020 | ||
1021 | /* Detect adjacent or overlapping regions (if same lock type) */ | |
1da177e4 | 1022 | if (request->fl_type == fl->fl_type) { |
449231d6 OK |
1023 | /* In all comparisons of start vs end, use |
1024 | * "start - 1" rather than "end + 1". If end | |
1025 | * is OFFSET_MAX, end + 1 will become negative. | |
1026 | */ | |
1da177e4 | 1027 | if (fl->fl_end < request->fl_start - 1) |
bd61e0a9 | 1028 | continue; |
1da177e4 LT |
1029 | /* If the next lock in the list has entirely bigger |
1030 | * addresses than the new one, insert the lock here. | |
1031 | */ | |
449231d6 | 1032 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1033 | break; |
1034 | ||
1035 | /* If we come here, the new and old lock are of the | |
1036 | * same type and adjacent or overlapping. Make one | |
1037 | * lock yielding from the lower start address of both | |
1038 | * locks to the higher end address. | |
1039 | */ | |
1040 | if (fl->fl_start > request->fl_start) | |
1041 | fl->fl_start = request->fl_start; | |
1042 | else | |
1043 | request->fl_start = fl->fl_start; | |
1044 | if (fl->fl_end < request->fl_end) | |
1045 | fl->fl_end = request->fl_end; | |
1046 | else | |
1047 | request->fl_end = fl->fl_end; | |
1048 | if (added) { | |
9bd0f45b JL |
1049 | locks_delete_lock_ctx(fl, &ctx->flc_posix_cnt, |
1050 | &dispose); | |
1da177e4 LT |
1051 | continue; |
1052 | } | |
1053 | request = fl; | |
b9746ef8 | 1054 | added = true; |
bd61e0a9 | 1055 | } else { |
1da177e4 LT |
1056 | /* Processing for different lock types is a bit |
1057 | * more complex. | |
1058 | */ | |
1059 | if (fl->fl_end < request->fl_start) | |
bd61e0a9 | 1060 | continue; |
1da177e4 LT |
1061 | if (fl->fl_start > request->fl_end) |
1062 | break; | |
1063 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1064 | added = true; |
1da177e4 LT |
1065 | if (fl->fl_start < request->fl_start) |
1066 | left = fl; | |
1067 | /* If the next lock in the list has a higher end | |
1068 | * address than the new one, insert the new one here. | |
1069 | */ | |
1070 | if (fl->fl_end > request->fl_end) { | |
1071 | right = fl; | |
1072 | break; | |
1073 | } | |
1074 | if (fl->fl_start >= request->fl_start) { | |
1075 | /* The new lock completely replaces an old | |
1076 | * one (This may happen several times). | |
1077 | */ | |
1078 | if (added) { | |
9bd0f45b JL |
1079 | locks_delete_lock_ctx(fl, |
1080 | &ctx->flc_posix_cnt, &dispose); | |
1da177e4 LT |
1081 | continue; |
1082 | } | |
b84d49f9 JL |
1083 | /* |
1084 | * Replace the old lock with new_fl, and | |
1085 | * remove the old one. It's safe to do the | |
1086 | * insert here since we know that we won't be | |
1087 | * using new_fl later, and that the lock is | |
1088 | * just replacing an existing lock. | |
1da177e4 | 1089 | */ |
b84d49f9 JL |
1090 | error = -ENOLCK; |
1091 | if (!new_fl) | |
1092 | goto out; | |
1093 | locks_copy_lock(new_fl, request); | |
1094 | request = new_fl; | |
1095 | new_fl = NULL; | |
9bd0f45b JL |
1096 | locks_insert_lock_ctx(request, |
1097 | &ctx->flc_posix_cnt, &fl->fl_list); | |
1098 | locks_delete_lock_ctx(fl, | |
1099 | &ctx->flc_posix_cnt, &dispose); | |
b9746ef8 | 1100 | added = true; |
1da177e4 LT |
1101 | } |
1102 | } | |
1da177e4 LT |
1103 | } |
1104 | ||
0d9a490a | 1105 | /* |
1cb36012 JL |
1106 | * The above code only modifies existing locks in case of merging or |
1107 | * replacing. If new lock(s) need to be inserted all modifications are | |
1108 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1109 | */ |
1110 | error = -ENOLCK; /* "no luck" */ | |
1111 | if (right && left == right && !new_fl2) | |
1112 | goto out; | |
1113 | ||
1da177e4 LT |
1114 | error = 0; |
1115 | if (!added) { | |
f475ae95 TM |
1116 | if (request->fl_type == F_UNLCK) { |
1117 | if (request->fl_flags & FL_EXISTS) | |
1118 | error = -ENOENT; | |
1da177e4 | 1119 | goto out; |
f475ae95 | 1120 | } |
0d9a490a MS |
1121 | |
1122 | if (!new_fl) { | |
1123 | error = -ENOLCK; | |
1124 | goto out; | |
1125 | } | |
1da177e4 | 1126 | locks_copy_lock(new_fl, request); |
9bd0f45b JL |
1127 | locks_insert_lock_ctx(new_fl, &ctx->flc_posix_cnt, |
1128 | &fl->fl_list); | |
1da177e4 LT |
1129 | new_fl = NULL; |
1130 | } | |
1131 | if (right) { | |
1132 | if (left == right) { | |
1133 | /* The new lock breaks the old one in two pieces, | |
1134 | * so we have to use the second new lock. | |
1135 | */ | |
1136 | left = new_fl2; | |
1137 | new_fl2 = NULL; | |
1138 | locks_copy_lock(left, right); | |
9bd0f45b JL |
1139 | locks_insert_lock_ctx(left, &ctx->flc_posix_cnt, |
1140 | &fl->fl_list); | |
1da177e4 LT |
1141 | } |
1142 | right->fl_start = request->fl_end + 1; | |
1143 | locks_wake_up_blocks(right); | |
1144 | } | |
1145 | if (left) { | |
1146 | left->fl_end = request->fl_start - 1; | |
1147 | locks_wake_up_blocks(left); | |
1148 | } | |
1149 | out: | |
6109c850 | 1150 | spin_unlock(&ctx->flc_lock); |
1da177e4 LT |
1151 | /* |
1152 | * Free any unused locks. | |
1153 | */ | |
1154 | if (new_fl) | |
1155 | locks_free_lock(new_fl); | |
1156 | if (new_fl2) | |
1157 | locks_free_lock(new_fl2); | |
ed9814d8 | 1158 | locks_dispose_list(&dispose); |
1da177e4 LT |
1159 | return error; |
1160 | } | |
1161 | ||
1162 | /** | |
1163 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1164 | * @filp: The file to apply the lock to | |
1165 | * @fl: The lock to be applied | |
150b3934 | 1166 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1167 | * |
1168 | * Add a POSIX style lock to a file. | |
1169 | * We merge adjacent & overlapping locks whenever possible. | |
1170 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1171 | * |
1172 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1173 | * whether or not a lock was successfully freed by testing the return | |
1174 | * value for -ENOENT. | |
1da177e4 | 1175 | */ |
150b3934 | 1176 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1177 | struct file_lock *conflock) |
1178 | { | |
496ad9aa | 1179 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1180 | } |
150b3934 | 1181 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1182 | |
1183 | /** | |
1184 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1185 | * @filp: The file to apply the lock to | |
1186 | * @fl: The lock to be applied | |
1187 | * | |
1188 | * Add a POSIX style lock to a file. | |
1189 | * We merge adjacent & overlapping locks whenever possible. | |
1190 | * POSIX locks are sorted by owner task, then by starting address | |
1191 | */ | |
1192 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1193 | { | |
1194 | int error; | |
1195 | might_sleep (); | |
1196 | for (;;) { | |
150b3934 | 1197 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1198 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1199 | break; |
1200 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1201 | if (!error) | |
1202 | continue; | |
1203 | ||
1204 | locks_delete_block(fl); | |
1205 | break; | |
1206 | } | |
1207 | return error; | |
1208 | } | |
1209 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1210 | ||
1211 | /** | |
1212 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1213 | * @file: the file to check |
1da177e4 LT |
1214 | * |
1215 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1216 | * This function is called from locks_verify_locked() only. | |
1217 | */ | |
d7a06983 | 1218 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1219 | { |
bd61e0a9 | 1220 | int ret; |
d7a06983 | 1221 | struct inode *inode = file_inode(file); |
bd61e0a9 | 1222 | struct file_lock_context *ctx; |
1da177e4 LT |
1223 | struct file_lock *fl; |
1224 | ||
bd61e0a9 JL |
1225 | ctx = inode->i_flctx; |
1226 | if (!ctx || list_empty_careful(&ctx->flc_posix)) | |
1227 | return 0; | |
1228 | ||
1da177e4 LT |
1229 | /* |
1230 | * Search the lock list for this inode for any POSIX locks. | |
1231 | */ | |
6109c850 | 1232 | spin_lock(&ctx->flc_lock); |
bd61e0a9 JL |
1233 | ret = 0; |
1234 | list_for_each_entry(fl, &ctx->flc_posix, fl_list) { | |
73a8f5f7 | 1235 | if (fl->fl_owner != current->files && |
bd61e0a9 JL |
1236 | fl->fl_owner != file) { |
1237 | ret = -EAGAIN; | |
1da177e4 | 1238 | break; |
bd61e0a9 | 1239 | } |
1da177e4 | 1240 | } |
6109c850 | 1241 | spin_unlock(&ctx->flc_lock); |
bd61e0a9 | 1242 | return ret; |
1da177e4 LT |
1243 | } |
1244 | ||
1245 | /** | |
1246 | * locks_mandatory_area - Check for a conflicting lock | |
1247 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1248 | * for shared | |
1249 | * @inode: the file to check | |
1250 | * @filp: how the file was opened (if it was) | |
1251 | * @offset: start of area to check | |
1252 | * @count: length of area to check | |
1253 | * | |
1254 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1255 | * This function is called from rw_verify_area() and | |
1256 | * locks_verify_truncate(). | |
1257 | */ | |
1258 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1259 | struct file *filp, loff_t offset, | |
1260 | size_t count) | |
1261 | { | |
1262 | struct file_lock fl; | |
1263 | int error; | |
29723ade | 1264 | bool sleep = false; |
1da177e4 LT |
1265 | |
1266 | locks_init_lock(&fl); | |
1da177e4 LT |
1267 | fl.fl_pid = current->tgid; |
1268 | fl.fl_file = filp; | |
1269 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1270 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1271 | sleep = true; |
1da177e4 LT |
1272 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1273 | fl.fl_start = offset; | |
1274 | fl.fl_end = offset + count - 1; | |
1275 | ||
1276 | for (;;) { | |
29723ade | 1277 | if (filp) { |
73a8f5f7 | 1278 | fl.fl_owner = filp; |
29723ade JL |
1279 | fl.fl_flags &= ~FL_SLEEP; |
1280 | error = __posix_lock_file(inode, &fl, NULL); | |
1281 | if (!error) | |
1282 | break; | |
1283 | } | |
1284 | ||
1285 | if (sleep) | |
1286 | fl.fl_flags |= FL_SLEEP; | |
1287 | fl.fl_owner = current->files; | |
150b3934 | 1288 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1289 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1290 | break; |
1291 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1292 | if (!error) { | |
1293 | /* | |
1294 | * If we've been sleeping someone might have | |
1295 | * changed the permissions behind our back. | |
1296 | */ | |
a16877ca | 1297 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1298 | continue; |
1299 | } | |
1300 | ||
1301 | locks_delete_block(&fl); | |
1302 | break; | |
1303 | } | |
1304 | ||
1305 | return error; | |
1306 | } | |
1307 | ||
1308 | EXPORT_SYMBOL(locks_mandatory_area); | |
1309 | ||
778fc546 BF |
1310 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1311 | { | |
1312 | switch (arg) { | |
1313 | case F_UNLCK: | |
1314 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1315 | /* fall through: */ | |
1316 | case F_RDLCK: | |
1317 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1318 | } | |
1319 | } | |
1320 | ||
1da177e4 | 1321 | /* We already had a lease on this file; just change its type */ |
7448cc37 | 1322 | int lease_modify(struct file_lock *fl, int arg, struct list_head *dispose) |
1da177e4 | 1323 | { |
9bd0f45b | 1324 | struct file_lock_context *flctx; |
1da177e4 LT |
1325 | int error = assign_type(fl, arg); |
1326 | ||
1327 | if (error) | |
1328 | return error; | |
778fc546 | 1329 | lease_clear_pending(fl, arg); |
1da177e4 | 1330 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1331 | if (arg == F_UNLCK) { |
1332 | struct file *filp = fl->fl_file; | |
1333 | ||
9bd0f45b | 1334 | flctx = file_inode(filp)->i_flctx; |
3b6e2723 FB |
1335 | f_delown(filp); |
1336 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1337 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1338 | if (fl->fl_fasync != NULL) { | |
1339 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1340 | fl->fl_fasync = NULL; | |
1341 | } | |
9bd0f45b | 1342 | locks_delete_lock_ctx(fl, &flctx->flc_lease_cnt, dispose); |
3b6e2723 | 1343 | } |
1da177e4 LT |
1344 | return 0; |
1345 | } | |
1da177e4 LT |
1346 | EXPORT_SYMBOL(lease_modify); |
1347 | ||
778fc546 BF |
1348 | static bool past_time(unsigned long then) |
1349 | { | |
1350 | if (!then) | |
1351 | /* 0 is a special value meaning "this never expires": */ | |
1352 | return false; | |
1353 | return time_after(jiffies, then); | |
1354 | } | |
1355 | ||
c45198ed | 1356 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 | 1357 | { |
8634b51f JL |
1358 | struct file_lock_context *ctx = inode->i_flctx; |
1359 | struct file_lock *fl, *tmp; | |
1da177e4 | 1360 | |
6109c850 | 1361 | lockdep_assert_held(&ctx->flc_lock); |
f82b4b67 | 1362 | |
8634b51f | 1363 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) { |
62af4f1f | 1364 | trace_time_out_leases(inode, fl); |
778fc546 | 1365 | if (past_time(fl->fl_downgrade_time)) |
7448cc37 | 1366 | lease_modify(fl, F_RDLCK, dispose); |
778fc546 | 1367 | if (past_time(fl->fl_break_time)) |
7448cc37 | 1368 | lease_modify(fl, F_UNLCK, dispose); |
1da177e4 LT |
1369 | } |
1370 | } | |
1371 | ||
df4e8d2c BF |
1372 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1373 | { | |
1374 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1375 | return false; | |
1376 | return locks_conflict(breaker, lease); | |
1377 | } | |
1378 | ||
03d12ddf JL |
1379 | static bool |
1380 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1381 | { | |
8634b51f | 1382 | struct file_lock_context *ctx = inode->i_flctx; |
03d12ddf JL |
1383 | struct file_lock *fl; |
1384 | ||
6109c850 | 1385 | lockdep_assert_held(&ctx->flc_lock); |
03d12ddf | 1386 | |
8634b51f | 1387 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
03d12ddf JL |
1388 | if (leases_conflict(fl, breaker)) |
1389 | return true; | |
1390 | } | |
1391 | return false; | |
1392 | } | |
1393 | ||
1da177e4 LT |
1394 | /** |
1395 | * __break_lease - revoke all outstanding leases on file | |
1396 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1397 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1398 | * break all leases | |
1399 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1400 | * only delegations | |
1da177e4 | 1401 | * |
87250dd2 | 1402 | * break_lease (inlined for speed) has checked there already is at least |
1403 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1404 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1405 | * specified %O_NONBLOCK to your open(). |
1406 | */ | |
df4e8d2c | 1407 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1408 | { |
778fc546 | 1409 | int error = 0; |
03d12ddf | 1410 | struct file_lock *new_fl; |
8634b51f JL |
1411 | struct file_lock_context *ctx = inode->i_flctx; |
1412 | struct file_lock *fl; | |
1da177e4 | 1413 | unsigned long break_time; |
8737c930 | 1414 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1415 | LIST_HEAD(dispose); |
1da177e4 | 1416 | |
8737c930 | 1417 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1418 | if (IS_ERR(new_fl)) |
1419 | return PTR_ERR(new_fl); | |
df4e8d2c | 1420 | new_fl->fl_flags = type; |
1da177e4 | 1421 | |
8634b51f JL |
1422 | /* typically we will check that ctx is non-NULL before calling */ |
1423 | if (!ctx) { | |
1424 | WARN_ON_ONCE(1); | |
1425 | return error; | |
1426 | } | |
1427 | ||
6109c850 | 1428 | spin_lock(&ctx->flc_lock); |
1da177e4 | 1429 | |
c45198ed | 1430 | time_out_leases(inode, &dispose); |
1da177e4 | 1431 | |
03d12ddf | 1432 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1433 | goto out; |
1434 | ||
1da177e4 LT |
1435 | break_time = 0; |
1436 | if (lease_break_time > 0) { | |
1437 | break_time = jiffies + lease_break_time * HZ; | |
1438 | if (break_time == 0) | |
1439 | break_time++; /* so that 0 means no break time */ | |
1440 | } | |
1441 | ||
8634b51f | 1442 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
df4e8d2c BF |
1443 | if (!leases_conflict(fl, new_fl)) |
1444 | continue; | |
778fc546 BF |
1445 | if (want_write) { |
1446 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1447 | continue; | |
1448 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1449 | fl->fl_break_time = break_time; |
778fc546 | 1450 | } else { |
8634b51f | 1451 | if (lease_breaking(fl)) |
778fc546 BF |
1452 | continue; |
1453 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1454 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1455 | } |
4d01b7f5 | 1456 | if (fl->fl_lmops->lm_break(fl)) |
9bd0f45b JL |
1457 | locks_delete_lock_ctx(fl, &ctx->flc_lease_cnt, |
1458 | &dispose); | |
1da177e4 LT |
1459 | } |
1460 | ||
8634b51f | 1461 | if (list_empty(&ctx->flc_lease)) |
4d01b7f5 JL |
1462 | goto out; |
1463 | ||
843c6b2f | 1464 | if (mode & O_NONBLOCK) { |
62af4f1f | 1465 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1466 | error = -EWOULDBLOCK; |
1467 | goto out; | |
1468 | } | |
1469 | ||
1470 | restart: | |
8634b51f JL |
1471 | fl = list_first_entry(&ctx->flc_lease, struct file_lock, fl_list); |
1472 | break_time = fl->fl_break_time; | |
f1c6bb2c | 1473 | if (break_time != 0) |
1da177e4 | 1474 | break_time -= jiffies; |
f1c6bb2c JL |
1475 | if (break_time == 0) |
1476 | break_time++; | |
8634b51f | 1477 | locks_insert_block(fl, new_fl); |
62af4f1f | 1478 | trace_break_lease_block(inode, new_fl); |
6109c850 | 1479 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1480 | locks_dispose_list(&dispose); |
4321e01e MW |
1481 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1482 | !new_fl->fl_next, break_time); | |
6109c850 | 1483 | spin_lock(&ctx->flc_lock); |
62af4f1f | 1484 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1485 | locks_delete_block(new_fl); |
1da177e4 | 1486 | if (error >= 0) { |
778fc546 BF |
1487 | /* |
1488 | * Wait for the next conflicting lease that has not been | |
1489 | * broken yet | |
1490 | */ | |
03d12ddf JL |
1491 | if (error == 0) |
1492 | time_out_leases(inode, &dispose); | |
1493 | if (any_leases_conflict(inode, new_fl)) | |
1494 | goto restart; | |
1da177e4 LT |
1495 | error = 0; |
1496 | } | |
1da177e4 | 1497 | out: |
6109c850 | 1498 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1499 | locks_dispose_list(&dispose); |
6d4b9e38 | 1500 | locks_free_lock(new_fl); |
1da177e4 LT |
1501 | return error; |
1502 | } | |
1503 | ||
1504 | EXPORT_SYMBOL(__break_lease); | |
1505 | ||
1506 | /** | |
a6b91919 | 1507 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1508 | * @inode: the inode |
1509 | * @time: pointer to a timespec which will contain the last modified time | |
1510 | * | |
1511 | * This is to force NFS clients to flush their caches for files with | |
1512 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1513 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1514 | */ |
1515 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1516 | { | |
bfe86024 | 1517 | bool has_lease = false; |
8634b51f JL |
1518 | struct file_lock_context *ctx = inode->i_flctx; |
1519 | struct file_lock *fl; | |
bfe86024 | 1520 | |
8634b51f | 1521 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1522 | spin_lock(&ctx->flc_lock); |
8634b51f JL |
1523 | if (!list_empty(&ctx->flc_lease)) { |
1524 | fl = list_first_entry(&ctx->flc_lease, | |
1525 | struct file_lock, fl_list); | |
1526 | if (fl->fl_type == F_WRLCK) | |
1527 | has_lease = true; | |
1528 | } | |
6109c850 | 1529 | spin_unlock(&ctx->flc_lock); |
bfe86024 JL |
1530 | } |
1531 | ||
1532 | if (has_lease) | |
1da177e4 LT |
1533 | *time = current_fs_time(inode->i_sb); |
1534 | else | |
1535 | *time = inode->i_mtime; | |
1536 | } | |
1537 | ||
1538 | EXPORT_SYMBOL(lease_get_mtime); | |
1539 | ||
1540 | /** | |
1541 | * fcntl_getlease - Enquire what lease is currently active | |
1542 | * @filp: the file | |
1543 | * | |
1544 | * The value returned by this function will be one of | |
1545 | * (if no lease break is pending): | |
1546 | * | |
1547 | * %F_RDLCK to indicate a shared lease is held. | |
1548 | * | |
1549 | * %F_WRLCK to indicate an exclusive lease is held. | |
1550 | * | |
1551 | * %F_UNLCK to indicate no lease is held. | |
1552 | * | |
1553 | * (if a lease break is pending): | |
1554 | * | |
1555 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1556 | * changed to a shared lease (or removed). | |
1557 | * | |
1558 | * %F_UNLCK to indicate the lease needs to be removed. | |
1559 | * | |
1560 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1561 | * should be returned to userspace. | |
1562 | */ | |
1563 | int fcntl_getlease(struct file *filp) | |
1564 | { | |
1565 | struct file_lock *fl; | |
1c8c601a | 1566 | struct inode *inode = file_inode(filp); |
8634b51f | 1567 | struct file_lock_context *ctx = inode->i_flctx; |
1da177e4 | 1568 | int type = F_UNLCK; |
c45198ed | 1569 | LIST_HEAD(dispose); |
1da177e4 | 1570 | |
8634b51f | 1571 | if (ctx && !list_empty_careful(&ctx->flc_lease)) { |
6109c850 | 1572 | spin_lock(&ctx->flc_lock); |
8634b51f JL |
1573 | time_out_leases(file_inode(filp), &dispose); |
1574 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { | |
1575 | if (fl->fl_file != filp) | |
1576 | continue; | |
778fc546 | 1577 | type = target_leasetype(fl); |
1da177e4 LT |
1578 | break; |
1579 | } | |
6109c850 | 1580 | spin_unlock(&ctx->flc_lock); |
8634b51f | 1581 | locks_dispose_list(&dispose); |
1da177e4 | 1582 | } |
1da177e4 LT |
1583 | return type; |
1584 | } | |
1585 | ||
24cbe784 JL |
1586 | /** |
1587 | * check_conflicting_open - see if the given dentry points to a file that has | |
1588 | * an existing open that would conflict with the | |
1589 | * desired lease. | |
1590 | * @dentry: dentry to check | |
1591 | * @arg: type of lease that we're trying to acquire | |
1592 | * | |
1593 | * Check to see if there's an existing open fd on this file that would | |
1594 | * conflict with the lease we're trying to set. | |
1595 | */ | |
1596 | static int | |
1597 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1598 | { | |
1599 | int ret = 0; | |
1600 | struct inode *inode = dentry->d_inode; | |
1601 | ||
1602 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1603 | return -EAGAIN; | |
1604 | ||
1605 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1606 | (atomic_read(&inode->i_count) > 1))) | |
1607 | ret = -EAGAIN; | |
1608 | ||
1609 | return ret; | |
1610 | } | |
1611 | ||
e6f5c789 JL |
1612 | static int |
1613 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1614 | { |
8634b51f | 1615 | struct file_lock *fl, *my_fl = NULL, *lease; |
0f7fc9e4 | 1616 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1617 | struct inode *inode = dentry->d_inode; |
8634b51f | 1618 | struct file_lock_context *ctx; |
df4e8d2c | 1619 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1620 | int error; |
c45198ed | 1621 | LIST_HEAD(dispose); |
1da177e4 | 1622 | |
096657b6 | 1623 | lease = *flp; |
62af4f1f JL |
1624 | trace_generic_add_lease(inode, lease); |
1625 | ||
8634b51f JL |
1626 | ctx = locks_get_lock_context(inode); |
1627 | if (!ctx) | |
1628 | return -ENOMEM; | |
1629 | ||
df4e8d2c BF |
1630 | /* |
1631 | * In the delegation case we need mutual exclusion with | |
1632 | * a number of operations that take the i_mutex. We trylock | |
1633 | * because delegations are an optional optimization, and if | |
1634 | * there's some chance of a conflict--we'd rather not | |
1635 | * bother, maybe that's a sign this just isn't a good file to | |
1636 | * hand out a delegation on. | |
1637 | */ | |
1638 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1639 | return -EAGAIN; | |
1640 | ||
1641 | if (is_deleg && arg == F_WRLCK) { | |
1642 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1643 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1644 | WARN_ON_ONCE(1); |
1645 | return -EINVAL; | |
1646 | } | |
096657b6 | 1647 | |
6109c850 | 1648 | spin_lock(&ctx->flc_lock); |
c45198ed | 1649 | time_out_leases(inode, &dispose); |
24cbe784 JL |
1650 | error = check_conflicting_open(dentry, arg); |
1651 | if (error) | |
096657b6 | 1652 | goto out; |
6d5e8b05 | 1653 | |
1da177e4 LT |
1654 | /* |
1655 | * At this point, we know that if there is an exclusive | |
1656 | * lease on this file, then we hold it on this filp | |
1657 | * (otherwise our open of this file would have blocked). | |
1658 | * And if we are trying to acquire an exclusive lease, | |
1659 | * then the file is not open by anyone (including us) | |
1660 | * except for this filp. | |
1661 | */ | |
c1f24ef4 | 1662 | error = -EAGAIN; |
8634b51f | 1663 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
c1f24ef4 | 1664 | if (fl->fl_file == filp) { |
8634b51f | 1665 | my_fl = fl; |
c1f24ef4 BF |
1666 | continue; |
1667 | } | |
8634b51f | 1668 | |
c1f24ef4 BF |
1669 | /* |
1670 | * No exclusive leases if someone else has a lease on | |
1671 | * this file: | |
1672 | */ | |
1673 | if (arg == F_WRLCK) | |
1674 | goto out; | |
1675 | /* | |
1676 | * Modifying our existing lease is OK, but no getting a | |
1677 | * new lease if someone else is opening for write: | |
1678 | */ | |
1679 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1680 | goto out; | |
1da177e4 LT |
1681 | } |
1682 | ||
8634b51f | 1683 | if (my_fl != NULL) { |
7448cc37 | 1684 | error = lease->fl_lmops->lm_change(my_fl, arg, &dispose); |
1c7dd2ff JL |
1685 | if (error) |
1686 | goto out; | |
1687 | goto out_setup; | |
1da177e4 LT |
1688 | } |
1689 | ||
1da177e4 LT |
1690 | error = -EINVAL; |
1691 | if (!leases_enable) | |
1692 | goto out; | |
1693 | ||
9bd0f45b | 1694 | locks_insert_lock_ctx(lease, &ctx->flc_lease_cnt, &ctx->flc_lease); |
24cbe784 JL |
1695 | /* |
1696 | * The check in break_lease() is lockless. It's possible for another | |
1697 | * open to race in after we did the earlier check for a conflicting | |
1698 | * open but before the lease was inserted. Check again for a | |
1699 | * conflicting open and cancel the lease if there is one. | |
1700 | * | |
1701 | * We also add a barrier here to ensure that the insertion of the lock | |
1702 | * precedes these checks. | |
1703 | */ | |
1704 | smp_mb(); | |
1705 | error = check_conflicting_open(dentry, arg); | |
8634b51f | 1706 | if (error) { |
9bd0f45b | 1707 | locks_unlink_lock_ctx(lease, &ctx->flc_lease_cnt); |
8634b51f JL |
1708 | goto out; |
1709 | } | |
1c7dd2ff JL |
1710 | |
1711 | out_setup: | |
1712 | if (lease->fl_lmops->lm_setup) | |
1713 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1714 | out: |
6109c850 | 1715 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1716 | locks_dispose_list(&dispose); |
df4e8d2c BF |
1717 | if (is_deleg) |
1718 | mutex_unlock(&inode->i_mutex); | |
8634b51f | 1719 | if (!error && !my_fl) |
1c7dd2ff | 1720 | *flp = NULL; |
1da177e4 LT |
1721 | return error; |
1722 | } | |
8335ebd9 | 1723 | |
0efaa7e8 | 1724 | static int generic_delete_lease(struct file *filp) |
8335ebd9 | 1725 | { |
0efaa7e8 | 1726 | int error = -EAGAIN; |
8634b51f | 1727 | struct file_lock *fl, *victim = NULL; |
8335ebd9 BF |
1728 | struct dentry *dentry = filp->f_path.dentry; |
1729 | struct inode *inode = dentry->d_inode; | |
8634b51f | 1730 | struct file_lock_context *ctx = inode->i_flctx; |
c45198ed | 1731 | LIST_HEAD(dispose); |
8335ebd9 | 1732 | |
8634b51f JL |
1733 | if (!ctx) { |
1734 | trace_generic_delete_lease(inode, NULL); | |
1735 | return error; | |
1736 | } | |
1737 | ||
6109c850 | 1738 | spin_lock(&ctx->flc_lock); |
8634b51f JL |
1739 | list_for_each_entry(fl, &ctx->flc_lease, fl_list) { |
1740 | if (fl->fl_file == filp) { | |
1741 | victim = fl; | |
0efaa7e8 | 1742 | break; |
8634b51f | 1743 | } |
8335ebd9 | 1744 | } |
0efaa7e8 | 1745 | trace_generic_delete_lease(inode, fl); |
8634b51f | 1746 | if (victim) |
7448cc37 | 1747 | error = fl->fl_lmops->lm_change(victim, F_UNLCK, &dispose); |
6109c850 | 1748 | spin_unlock(&ctx->flc_lock); |
c45198ed | 1749 | locks_dispose_list(&dispose); |
0efaa7e8 | 1750 | return error; |
8335ebd9 BF |
1751 | } |
1752 | ||
1753 | /** | |
1754 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1755 | * @filp: file pointer |
1756 | * @arg: type of lease to obtain | |
1757 | * @flp: input - file_lock to use, output - file_lock inserted | |
1758 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1759 | * doesn't require it) | |
8335ebd9 BF |
1760 | * |
1761 | * The (input) flp->fl_lmops->lm_break function is required | |
1762 | * by break_lease(). | |
8335ebd9 | 1763 | */ |
e6f5c789 JL |
1764 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1765 | void **priv) | |
8335ebd9 BF |
1766 | { |
1767 | struct dentry *dentry = filp->f_path.dentry; | |
1768 | struct inode *inode = dentry->d_inode; | |
1769 | int error; | |
1770 | ||
8e96e3b7 | 1771 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1772 | return -EACCES; |
1773 | if (!S_ISREG(inode->i_mode)) | |
1774 | return -EINVAL; | |
1775 | error = security_file_lock(filp, arg); | |
1776 | if (error) | |
1777 | return error; | |
1778 | ||
8335ebd9 BF |
1779 | switch (arg) { |
1780 | case F_UNLCK: | |
0efaa7e8 | 1781 | return generic_delete_lease(filp); |
8335ebd9 BF |
1782 | case F_RDLCK: |
1783 | case F_WRLCK: | |
0efaa7e8 JL |
1784 | if (!(*flp)->fl_lmops->lm_break) { |
1785 | WARN_ON_ONCE(1); | |
1786 | return -ENOLCK; | |
1787 | } | |
e6f5c789 | 1788 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1789 | default: |
8d657eb3 | 1790 | return -EINVAL; |
8335ebd9 BF |
1791 | } |
1792 | } | |
0af1a450 | 1793 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1794 | |
b89f4321 | 1795 | /** |
e51673aa | 1796 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1797 | * @filp: file pointer |
1798 | * @arg: type of lease to obtain | |
1799 | * @lease: file_lock to use when adding a lease | |
1800 | * @priv: private info for lm_setup when adding a lease (may be | |
1801 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1802 | * |
1803 | * Call this to establish a lease on the file. The "lease" argument is not | |
1804 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1805 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1806 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1807 | * stack trace). | |
1c7dd2ff JL |
1808 | * |
1809 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1810 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1811 | */ |
e6f5c789 JL |
1812 | int |
1813 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1814 | { |
1c7dd2ff | 1815 | if (filp->f_op->setlease) |
f82b4b67 | 1816 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1817 | else |
f82b4b67 | 1818 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1819 | } |
a9933cea | 1820 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1821 | |
0ceaf6c7 | 1822 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1823 | { |
1c7dd2ff | 1824 | struct file_lock *fl; |
f7347ce4 | 1825 | struct fasync_struct *new; |
1da177e4 LT |
1826 | int error; |
1827 | ||
c5b1f0d9 AB |
1828 | fl = lease_alloc(filp, arg); |
1829 | if (IS_ERR(fl)) | |
1830 | return PTR_ERR(fl); | |
1da177e4 | 1831 | |
f7347ce4 LT |
1832 | new = fasync_alloc(); |
1833 | if (!new) { | |
1834 | locks_free_lock(fl); | |
1835 | return -ENOMEM; | |
1836 | } | |
1c7dd2ff | 1837 | new->fa_fd = fd; |
f7347ce4 | 1838 | |
1c7dd2ff | 1839 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1840 | if (fl) |
1841 | locks_free_lock(fl); | |
f7347ce4 LT |
1842 | if (new) |
1843 | fasync_free(new); | |
1da177e4 LT |
1844 | return error; |
1845 | } | |
1846 | ||
0ceaf6c7 BF |
1847 | /** |
1848 | * fcntl_setlease - sets a lease on an open file | |
1849 | * @fd: open file descriptor | |
1850 | * @filp: file pointer | |
1851 | * @arg: type of lease to obtain | |
1852 | * | |
1853 | * Call this fcntl to establish a lease on the file. | |
1854 | * Note that you also need to call %F_SETSIG to | |
1855 | * receive a signal when the lease is broken. | |
1856 | */ | |
1857 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1858 | { | |
1859 | if (arg == F_UNLCK) | |
e6f5c789 | 1860 | return vfs_setlease(filp, F_UNLCK, NULL, NULL); |
0ceaf6c7 BF |
1861 | return do_fcntl_add_lease(fd, filp, arg); |
1862 | } | |
1863 | ||
1da177e4 LT |
1864 | /** |
1865 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1866 | * @filp: The file to apply the lock to | |
1867 | * @fl: The lock to be applied | |
1868 | * | |
1869 | * Add a FLOCK style lock to a file. | |
1870 | */ | |
1871 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1872 | { | |
1873 | int error; | |
1874 | might_sleep(); | |
1875 | for (;;) { | |
1876 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1877 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1878 | break; |
1879 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1880 | if (!error) | |
1881 | continue; | |
1882 | ||
1883 | locks_delete_block(fl); | |
1884 | break; | |
1885 | } | |
1886 | return error; | |
1887 | } | |
1888 | ||
1889 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1890 | ||
1891 | /** | |
1892 | * sys_flock: - flock() system call. | |
1893 | * @fd: the file descriptor to lock. | |
1894 | * @cmd: the type of lock to apply. | |
1895 | * | |
1896 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1897 | * The @cmd can be one of | |
1898 | * | |
1899 | * %LOCK_SH -- a shared lock. | |
1900 | * | |
1901 | * %LOCK_EX -- an exclusive lock. | |
1902 | * | |
1903 | * %LOCK_UN -- remove an existing lock. | |
1904 | * | |
1905 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1906 | * | |
1907 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1908 | * processes read and write access respectively. | |
1909 | */ | |
002c8976 | 1910 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1911 | { |
2903ff01 | 1912 | struct fd f = fdget(fd); |
1da177e4 LT |
1913 | struct file_lock *lock; |
1914 | int can_sleep, unlock; | |
1915 | int error; | |
1916 | ||
1917 | error = -EBADF; | |
2903ff01 | 1918 | if (!f.file) |
1da177e4 LT |
1919 | goto out; |
1920 | ||
1921 | can_sleep = !(cmd & LOCK_NB); | |
1922 | cmd &= ~LOCK_NB; | |
1923 | unlock = (cmd == LOCK_UN); | |
1924 | ||
aeb5d727 | 1925 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1926 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1927 | goto out_putf; |
1928 | ||
6e129d00 JL |
1929 | lock = flock_make_lock(f.file, cmd); |
1930 | if (IS_ERR(lock)) { | |
1931 | error = PTR_ERR(lock); | |
1da177e4 | 1932 | goto out_putf; |
6e129d00 JL |
1933 | } |
1934 | ||
1da177e4 LT |
1935 | if (can_sleep) |
1936 | lock->fl_flags |= FL_SLEEP; | |
1937 | ||
2903ff01 | 1938 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1939 | if (error) |
1940 | goto out_free; | |
1941 | ||
72c2d531 | 1942 | if (f.file->f_op->flock) |
2903ff01 | 1943 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1944 | (can_sleep) ? F_SETLKW : F_SETLK, |
1945 | lock); | |
1946 | else | |
2903ff01 | 1947 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1948 | |
1949 | out_free: | |
993dfa87 | 1950 | locks_free_lock(lock); |
1da177e4 LT |
1951 | |
1952 | out_putf: | |
2903ff01 | 1953 | fdput(f); |
1da177e4 LT |
1954 | out: |
1955 | return error; | |
1956 | } | |
1957 | ||
3ee17abd BF |
1958 | /** |
1959 | * vfs_test_lock - test file byte range lock | |
1960 | * @filp: The file to test lock for | |
6924c554 | 1961 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1962 | * |
1963 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1964 | * setting conf->fl_type to something other than F_UNLCK. | |
1965 | */ | |
1966 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1967 | { | |
72c2d531 | 1968 | if (filp->f_op->lock) |
3ee17abd BF |
1969 | return filp->f_op->lock(filp, F_GETLK, fl); |
1970 | posix_test_lock(filp, fl); | |
1971 | return 0; | |
1972 | } | |
1973 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1974 | ||
c2fa1b8a BF |
1975 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1976 | { | |
cff2fce5 | 1977 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1978 | #if BITS_PER_LONG == 32 |
1979 | /* | |
1980 | * Make sure we can represent the posix lock via | |
1981 | * legacy 32bit flock. | |
1982 | */ | |
1983 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1984 | return -EOVERFLOW; | |
1985 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1986 | return -EOVERFLOW; | |
1987 | #endif | |
1988 | flock->l_start = fl->fl_start; | |
1989 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1990 | fl->fl_end - fl->fl_start + 1; | |
1991 | flock->l_whence = 0; | |
129a84de | 1992 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1993 | return 0; |
1994 | } | |
1995 | ||
1996 | #if BITS_PER_LONG == 32 | |
1997 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1998 | { | |
cff2fce5 | 1999 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2000 | flock->l_start = fl->fl_start; |
2001 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2002 | fl->fl_end - fl->fl_start + 1; | |
2003 | flock->l_whence = 0; | |
2004 | flock->l_type = fl->fl_type; | |
2005 | } | |
2006 | #endif | |
2007 | ||
1da177e4 LT |
2008 | /* Report the first existing lock that would conflict with l. |
2009 | * This implements the F_GETLK command of fcntl(). | |
2010 | */ | |
c1e62b8f | 2011 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 2012 | { |
9d6a8c5c | 2013 | struct file_lock file_lock; |
1da177e4 LT |
2014 | struct flock flock; |
2015 | int error; | |
2016 | ||
2017 | error = -EFAULT; | |
2018 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2019 | goto out; | |
2020 | error = -EINVAL; | |
2021 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2022 | goto out; | |
2023 | ||
2024 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
2025 | if (error) | |
2026 | goto out; | |
2027 | ||
0d3f7a2d | 2028 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2029 | error = -EINVAL; |
2030 | if (flock.l_pid != 0) | |
2031 | goto out; | |
2032 | ||
5d50ffd7 | 2033 | cmd = F_GETLK; |
cff2fce5 | 2034 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2035 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2036 | } |
2037 | ||
3ee17abd BF |
2038 | error = vfs_test_lock(filp, &file_lock); |
2039 | if (error) | |
2040 | goto out; | |
1da177e4 | 2041 | |
9d6a8c5c ME |
2042 | flock.l_type = file_lock.fl_type; |
2043 | if (file_lock.fl_type != F_UNLCK) { | |
2044 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2045 | if (error) |
f328296e | 2046 | goto rel_priv; |
1da177e4 LT |
2047 | } |
2048 | error = -EFAULT; | |
2049 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2050 | error = 0; | |
f328296e KM |
2051 | rel_priv: |
2052 | locks_release_private(&file_lock); | |
1da177e4 LT |
2053 | out: |
2054 | return error; | |
2055 | } | |
2056 | ||
7723ec97 ME |
2057 | /** |
2058 | * vfs_lock_file - file byte range lock | |
2059 | * @filp: The file to apply the lock to | |
2060 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2061 | * @fl: The lock to be applied | |
150b3934 ME |
2062 | * @conf: Place to return a copy of the conflicting lock, if found. |
2063 | * | |
2064 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2065 | * as the final argument. | |
2066 | * | |
2067 | * If the filesystem defines a private ->lock() method, then @conf will | |
2068 | * be left unchanged; so a caller that cares should initialize it to | |
2069 | * some acceptable default. | |
2beb6614 ME |
2070 | * |
2071 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2072 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2073 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2074 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2075 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2076 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2077 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2078 | * request completes. |
2079 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2080 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2081 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2082 | * nonzero return code and the file system should release the lock. The file |
2083 | * system is also responsible to keep a corresponding posix lock when it | |
2084 | * grants a lock so the VFS can find out which locks are locally held and do | |
2085 | * the correct lock cleanup when required. | |
2086 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2087 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2088 | * return code. |
7723ec97 | 2089 | */ |
150b3934 | 2090 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2091 | { |
72c2d531 | 2092 | if (filp->f_op->lock) |
7723ec97 ME |
2093 | return filp->f_op->lock(filp, cmd, fl); |
2094 | else | |
150b3934 | 2095 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2096 | } |
2097 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2098 | ||
b648a6de MS |
2099 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2100 | struct file_lock *fl) | |
2101 | { | |
2102 | int error; | |
2103 | ||
2104 | error = security_file_lock(filp, fl->fl_type); | |
2105 | if (error) | |
2106 | return error; | |
2107 | ||
764c76b3 MS |
2108 | for (;;) { |
2109 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2110 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2111 | break; |
764c76b3 MS |
2112 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2113 | if (!error) | |
2114 | continue; | |
2115 | ||
2116 | locks_delete_block(fl); | |
2117 | break; | |
b648a6de MS |
2118 | } |
2119 | ||
2120 | return error; | |
2121 | } | |
2122 | ||
cf01f4ee JL |
2123 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2124 | static int | |
2125 | check_fmode_for_setlk(struct file_lock *fl) | |
2126 | { | |
2127 | switch (fl->fl_type) { | |
2128 | case F_RDLCK: | |
2129 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2130 | return -EBADF; | |
2131 | break; | |
2132 | case F_WRLCK: | |
2133 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2134 | return -EBADF; | |
2135 | } | |
2136 | return 0; | |
2137 | } | |
2138 | ||
1da177e4 LT |
2139 | /* Apply the lock described by l to an open file descriptor. |
2140 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2141 | */ | |
c293621b PS |
2142 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2143 | struct flock __user *l) | |
1da177e4 LT |
2144 | { |
2145 | struct file_lock *file_lock = locks_alloc_lock(); | |
2146 | struct flock flock; | |
2147 | struct inode *inode; | |
0b2bac2f | 2148 | struct file *f; |
1da177e4 LT |
2149 | int error; |
2150 | ||
2151 | if (file_lock == NULL) | |
2152 | return -ENOLCK; | |
2153 | ||
2154 | /* | |
2155 | * This might block, so we do it before checking the inode. | |
2156 | */ | |
2157 | error = -EFAULT; | |
2158 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2159 | goto out; | |
2160 | ||
496ad9aa | 2161 | inode = file_inode(filp); |
1da177e4 LT |
2162 | |
2163 | /* Don't allow mandatory locks on files that may be memory mapped | |
2164 | * and shared. | |
2165 | */ | |
a16877ca | 2166 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2167 | error = -EAGAIN; |
2168 | goto out; | |
2169 | } | |
2170 | ||
c293621b | 2171 | again: |
1da177e4 LT |
2172 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2173 | if (error) | |
2174 | goto out; | |
5d50ffd7 | 2175 | |
cf01f4ee JL |
2176 | error = check_fmode_for_setlk(file_lock); |
2177 | if (error) | |
2178 | goto out; | |
2179 | ||
5d50ffd7 JL |
2180 | /* |
2181 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2182 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2183 | */ |
2184 | switch (cmd) { | |
0d3f7a2d | 2185 | case F_OFD_SETLK: |
90478939 JL |
2186 | error = -EINVAL; |
2187 | if (flock.l_pid != 0) | |
2188 | goto out; | |
2189 | ||
5d50ffd7 | 2190 | cmd = F_SETLK; |
cff2fce5 | 2191 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2192 | file_lock->fl_owner = filp; |
5d50ffd7 | 2193 | break; |
0d3f7a2d | 2194 | case F_OFD_SETLKW: |
90478939 JL |
2195 | error = -EINVAL; |
2196 | if (flock.l_pid != 0) | |
2197 | goto out; | |
2198 | ||
5d50ffd7 | 2199 | cmd = F_SETLKW; |
cff2fce5 | 2200 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2201 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2202 | /* Fallthrough */ |
2203 | case F_SETLKW: | |
1da177e4 LT |
2204 | file_lock->fl_flags |= FL_SLEEP; |
2205 | } | |
5d50ffd7 | 2206 | |
b648a6de | 2207 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2208 | |
c293621b PS |
2209 | /* |
2210 | * Attempt to detect a close/fcntl race and recover by | |
2211 | * releasing the lock that was just acquired. | |
2212 | */ | |
0b2bac2f AV |
2213 | /* |
2214 | * we need that spin_lock here - it prevents reordering between | |
2215 | * update of inode->i_flock and check for it done in close(). | |
2216 | * rcu_read_lock() wouldn't do. | |
2217 | */ | |
2218 | spin_lock(¤t->files->file_lock); | |
2219 | f = fcheck(fd); | |
2220 | spin_unlock(¤t->files->file_lock); | |
2221 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2222 | flock.l_type = F_UNLCK; |
2223 | goto again; | |
1da177e4 LT |
2224 | } |
2225 | ||
c293621b | 2226 | out: |
1da177e4 LT |
2227 | locks_free_lock(file_lock); |
2228 | return error; | |
2229 | } | |
2230 | ||
2231 | #if BITS_PER_LONG == 32 | |
2232 | /* Report the first existing lock that would conflict with l. | |
2233 | * This implements the F_GETLK command of fcntl(). | |
2234 | */ | |
c1e62b8f | 2235 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2236 | { |
9d6a8c5c | 2237 | struct file_lock file_lock; |
1da177e4 LT |
2238 | struct flock64 flock; |
2239 | int error; | |
2240 | ||
2241 | error = -EFAULT; | |
2242 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2243 | goto out; | |
2244 | error = -EINVAL; | |
2245 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2246 | goto out; | |
2247 | ||
2248 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2249 | if (error) | |
2250 | goto out; | |
2251 | ||
0d3f7a2d | 2252 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2253 | error = -EINVAL; |
2254 | if (flock.l_pid != 0) | |
2255 | goto out; | |
2256 | ||
5d50ffd7 | 2257 | cmd = F_GETLK64; |
cff2fce5 | 2258 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2259 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2260 | } |
2261 | ||
3ee17abd BF |
2262 | error = vfs_test_lock(filp, &file_lock); |
2263 | if (error) | |
2264 | goto out; | |
2265 | ||
9d6a8c5c ME |
2266 | flock.l_type = file_lock.fl_type; |
2267 | if (file_lock.fl_type != F_UNLCK) | |
2268 | posix_lock_to_flock64(&flock, &file_lock); | |
2269 | ||
1da177e4 LT |
2270 | error = -EFAULT; |
2271 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2272 | error = 0; | |
f328296e KM |
2273 | |
2274 | locks_release_private(&file_lock); | |
1da177e4 LT |
2275 | out: |
2276 | return error; | |
2277 | } | |
2278 | ||
2279 | /* Apply the lock described by l to an open file descriptor. | |
2280 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2281 | */ | |
c293621b PS |
2282 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2283 | struct flock64 __user *l) | |
1da177e4 LT |
2284 | { |
2285 | struct file_lock *file_lock = locks_alloc_lock(); | |
2286 | struct flock64 flock; | |
2287 | struct inode *inode; | |
0b2bac2f | 2288 | struct file *f; |
1da177e4 LT |
2289 | int error; |
2290 | ||
2291 | if (file_lock == NULL) | |
2292 | return -ENOLCK; | |
2293 | ||
2294 | /* | |
2295 | * This might block, so we do it before checking the inode. | |
2296 | */ | |
2297 | error = -EFAULT; | |
2298 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2299 | goto out; | |
2300 | ||
496ad9aa | 2301 | inode = file_inode(filp); |
1da177e4 LT |
2302 | |
2303 | /* Don't allow mandatory locks on files that may be memory mapped | |
2304 | * and shared. | |
2305 | */ | |
a16877ca | 2306 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2307 | error = -EAGAIN; |
2308 | goto out; | |
2309 | } | |
2310 | ||
c293621b | 2311 | again: |
1da177e4 LT |
2312 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2313 | if (error) | |
2314 | goto out; | |
5d50ffd7 | 2315 | |
cf01f4ee JL |
2316 | error = check_fmode_for_setlk(file_lock); |
2317 | if (error) | |
2318 | goto out; | |
2319 | ||
5d50ffd7 JL |
2320 | /* |
2321 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2322 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2323 | */ |
2324 | switch (cmd) { | |
0d3f7a2d | 2325 | case F_OFD_SETLK: |
90478939 JL |
2326 | error = -EINVAL; |
2327 | if (flock.l_pid != 0) | |
2328 | goto out; | |
2329 | ||
5d50ffd7 | 2330 | cmd = F_SETLK64; |
cff2fce5 | 2331 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2332 | file_lock->fl_owner = filp; |
5d50ffd7 | 2333 | break; |
0d3f7a2d | 2334 | case F_OFD_SETLKW: |
90478939 JL |
2335 | error = -EINVAL; |
2336 | if (flock.l_pid != 0) | |
2337 | goto out; | |
2338 | ||
5d50ffd7 | 2339 | cmd = F_SETLKW64; |
cff2fce5 | 2340 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2341 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2342 | /* Fallthrough */ |
2343 | case F_SETLKW64: | |
1da177e4 LT |
2344 | file_lock->fl_flags |= FL_SLEEP; |
2345 | } | |
5d50ffd7 | 2346 | |
b648a6de | 2347 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2348 | |
c293621b PS |
2349 | /* |
2350 | * Attempt to detect a close/fcntl race and recover by | |
2351 | * releasing the lock that was just acquired. | |
2352 | */ | |
0b2bac2f AV |
2353 | spin_lock(¤t->files->file_lock); |
2354 | f = fcheck(fd); | |
2355 | spin_unlock(¤t->files->file_lock); | |
2356 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2357 | flock.l_type = F_UNLCK; |
2358 | goto again; | |
1da177e4 LT |
2359 | } |
2360 | ||
2361 | out: | |
2362 | locks_free_lock(file_lock); | |
2363 | return error; | |
2364 | } | |
2365 | #endif /* BITS_PER_LONG == 32 */ | |
2366 | ||
2367 | /* | |
2368 | * This function is called when the file is being removed | |
2369 | * from the task's fd array. POSIX locks belonging to this task | |
2370 | * are deleted at this time. | |
2371 | */ | |
2372 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2373 | { | |
ff7b86b8 | 2374 | struct file_lock lock; |
bd61e0a9 | 2375 | struct file_lock_context *ctx = file_inode(filp)->i_flctx; |
1da177e4 LT |
2376 | |
2377 | /* | |
2378 | * If there are no locks held on this file, we don't need to call | |
2379 | * posix_lock_file(). Another process could be setting a lock on this | |
2380 | * file at the same time, but we wouldn't remove that lock anyway. | |
2381 | */ | |
bd61e0a9 | 2382 | if (!ctx || list_empty(&ctx->flc_posix)) |
1da177e4 LT |
2383 | return; |
2384 | ||
2385 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2386 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2387 | lock.fl_start = 0; |
2388 | lock.fl_end = OFFSET_MAX; | |
2389 | lock.fl_owner = owner; | |
2390 | lock.fl_pid = current->tgid; | |
2391 | lock.fl_file = filp; | |
2392 | lock.fl_ops = NULL; | |
2393 | lock.fl_lmops = NULL; | |
2394 | ||
150b3934 | 2395 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2396 | |
1da177e4 LT |
2397 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2398 | lock.fl_ops->fl_release_private(&lock); | |
2399 | } | |
2400 | ||
2401 | EXPORT_SYMBOL(locks_remove_posix); | |
2402 | ||
dd459bb1 JL |
2403 | static void |
2404 | locks_remove_flock(struct file *filp) | |
2405 | { | |
2406 | struct file_lock fl = { | |
2407 | .fl_owner = filp, | |
2408 | .fl_pid = current->tgid, | |
2409 | .fl_file = filp, | |
2410 | .fl_flags = FL_FLOCK, | |
2411 | .fl_type = F_UNLCK, | |
2412 | .fl_end = OFFSET_MAX, | |
2413 | }; | |
5263e31e | 2414 | struct file_lock_context *flctx = file_inode(filp)->i_flctx; |
dd459bb1 | 2415 | |
5263e31e | 2416 | if (!flctx || list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2417 | return; |
2418 | ||
2419 | if (filp->f_op->flock) | |
2420 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2421 | else | |
2422 | flock_lock_file(filp, &fl); | |
2423 | ||
2424 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2425 | fl.fl_ops->fl_release_private(&fl); | |
2426 | } | |
2427 | ||
8634b51f JL |
2428 | static void |
2429 | locks_remove_lease(struct file *filp) | |
2430 | { | |
2431 | struct inode *inode = file_inode(filp); | |
2432 | struct file_lock_context *ctx = inode->i_flctx; | |
2433 | struct file_lock *fl, *tmp; | |
2434 | LIST_HEAD(dispose); | |
2435 | ||
2436 | if (!ctx || list_empty(&ctx->flc_lease)) | |
2437 | return; | |
2438 | ||
6109c850 | 2439 | spin_lock(&ctx->flc_lock); |
8634b51f | 2440 | list_for_each_entry_safe(fl, tmp, &ctx->flc_lease, fl_list) |
7448cc37 | 2441 | lease_modify(fl, F_UNLCK, &dispose); |
6109c850 | 2442 | spin_unlock(&ctx->flc_lock); |
8634b51f JL |
2443 | locks_dispose_list(&dispose); |
2444 | } | |
2445 | ||
1da177e4 LT |
2446 | /* |
2447 | * This function is called on the last close of an open file. | |
2448 | */ | |
78ed8a13 | 2449 | void locks_remove_file(struct file *filp) |
1da177e4 | 2450 | { |
dd459bb1 | 2451 | /* remove any OFD locks */ |
73a8f5f7 | 2452 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2453 | |
dd459bb1 JL |
2454 | /* remove flock locks */ |
2455 | locks_remove_flock(filp); | |
2456 | ||
8634b51f JL |
2457 | /* remove any leases */ |
2458 | locks_remove_lease(filp); | |
1da177e4 LT |
2459 | } |
2460 | ||
1da177e4 LT |
2461 | /** |
2462 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2463 | * @waiter: the lock which was waiting |
2464 | * | |
2465 | * lockd needs to block waiting for locks. | |
2466 | */ | |
64a318ee | 2467 | int |
f891a29f | 2468 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2469 | { |
64a318ee BF |
2470 | int status = 0; |
2471 | ||
7b2296af | 2472 | spin_lock(&blocked_lock_lock); |
5996a298 | 2473 | if (waiter->fl_next) |
1da177e4 | 2474 | __locks_delete_block(waiter); |
64a318ee BF |
2475 | else |
2476 | status = -ENOENT; | |
7b2296af | 2477 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2478 | return status; |
1da177e4 | 2479 | } |
1da177e4 LT |
2480 | EXPORT_SYMBOL(posix_unblock_lock); |
2481 | ||
9b9d2ab4 ME |
2482 | /** |
2483 | * vfs_cancel_lock - file byte range unblock lock | |
2484 | * @filp: The file to apply the unblock to | |
2485 | * @fl: The lock to be unblocked | |
2486 | * | |
2487 | * Used by lock managers to cancel blocked requests | |
2488 | */ | |
2489 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2490 | { | |
72c2d531 | 2491 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2492 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2493 | return 0; | |
2494 | } | |
2495 | ||
2496 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2497 | ||
7f8ada98 | 2498 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2499 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2500 | #include <linux/seq_file.h> |
2501 | ||
7012b02a JL |
2502 | struct locks_iterator { |
2503 | int li_cpu; | |
2504 | loff_t li_pos; | |
2505 | }; | |
2506 | ||
7f8ada98 | 2507 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2508 | loff_t id, char *pfx) |
1da177e4 LT |
2509 | { |
2510 | struct inode *inode = NULL; | |
ab1f1611 VG |
2511 | unsigned int fl_pid; |
2512 | ||
2513 | if (fl->fl_nspid) | |
6c5f3e7b | 2514 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2515 | else |
2516 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2517 | |
2518 | if (fl->fl_file != NULL) | |
496ad9aa | 2519 | inode = file_inode(fl->fl_file); |
1da177e4 | 2520 | |
99dc8292 | 2521 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2522 | if (IS_POSIX(fl)) { |
c918d42a | 2523 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2524 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2525 | else if (IS_OFDLCK(fl)) |
5315c26a | 2526 | seq_puts(f, "OFDLCK"); |
c918d42a | 2527 | else |
5315c26a | 2528 | seq_puts(f, "POSIX "); |
c918d42a JL |
2529 | |
2530 | seq_printf(f, " %s ", | |
1da177e4 | 2531 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2532 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2533 | } else if (IS_FLOCK(fl)) { |
2534 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2535 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2536 | } else { |
5315c26a | 2537 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2538 | } |
2539 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2540 | if (fl->fl_flags & FL_DELEG) |
2541 | seq_puts(f, "DELEG "); | |
2542 | else | |
2543 | seq_puts(f, "LEASE "); | |
2544 | ||
ab83fa4b | 2545 | if (lease_breaking(fl)) |
5315c26a | 2546 | seq_puts(f, "BREAKING "); |
1da177e4 | 2547 | else if (fl->fl_file) |
5315c26a | 2548 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2549 | else |
5315c26a | 2550 | seq_puts(f, "BREAKER "); |
1da177e4 | 2551 | } else { |
5315c26a | 2552 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2553 | } |
2554 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2555 | seq_printf(f, "%s ", |
1da177e4 LT |
2556 | (fl->fl_type & LOCK_READ) |
2557 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2558 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2559 | } else { | |
7f8ada98 | 2560 | seq_printf(f, "%s ", |
ab83fa4b | 2561 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2562 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2563 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2564 | } |
2565 | if (inode) { | |
2566 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2567 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2568 | inode->i_sb->s_id, inode->i_ino); |
2569 | #else | |
2570 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2571 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2572 | MAJOR(inode->i_sb->s_dev), |
2573 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2574 | #endif | |
2575 | } else { | |
ab1f1611 | 2576 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2577 | } |
2578 | if (IS_POSIX(fl)) { | |
2579 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2580 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2581 | else |
7f8ada98 | 2582 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2583 | } else { |
5315c26a | 2584 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2585 | } |
2586 | } | |
2587 | ||
7f8ada98 | 2588 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2589 | { |
7012b02a | 2590 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2591 | struct file_lock *fl, *bfl; |
1da177e4 | 2592 | |
139ca04e | 2593 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2594 | |
7012b02a | 2595 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2596 | |
7f8ada98 | 2597 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2598 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2599 | |
7f8ada98 PE |
2600 | return 0; |
2601 | } | |
1da177e4 | 2602 | |
7f8ada98 | 2603 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2604 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2605 | { |
7012b02a | 2606 | struct locks_iterator *iter = f->private; |
99dc8292 | 2607 | |
7012b02a JL |
2608 | iter->li_pos = *pos + 1; |
2609 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2610 | spin_lock(&blocked_lock_lock); |
7012b02a | 2611 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2612 | } |
1da177e4 | 2613 | |
7f8ada98 PE |
2614 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2615 | { | |
7012b02a JL |
2616 | struct locks_iterator *iter = f->private; |
2617 | ||
2618 | ++iter->li_pos; | |
2619 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2620 | } |
1da177e4 | 2621 | |
7f8ada98 | 2622 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2623 | __releases(&blocked_lock_lock) |
7f8ada98 | 2624 | { |
7b2296af | 2625 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2626 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2627 | } |
2628 | ||
d8ba7a36 | 2629 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2630 | .start = locks_start, |
2631 | .next = locks_next, | |
2632 | .stop = locks_stop, | |
2633 | .show = locks_show, | |
2634 | }; | |
d8ba7a36 AD |
2635 | |
2636 | static int locks_open(struct inode *inode, struct file *filp) | |
2637 | { | |
7012b02a JL |
2638 | return seq_open_private(filp, &locks_seq_operations, |
2639 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2640 | } |
2641 | ||
2642 | static const struct file_operations proc_locks_operations = { | |
2643 | .open = locks_open, | |
2644 | .read = seq_read, | |
2645 | .llseek = seq_lseek, | |
99dc8292 | 2646 | .release = seq_release_private, |
d8ba7a36 AD |
2647 | }; |
2648 | ||
2649 | static int __init proc_locks_init(void) | |
2650 | { | |
2651 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2652 | return 0; | |
2653 | } | |
2654 | module_init(proc_locks_init); | |
7f8ada98 PE |
2655 | #endif |
2656 | ||
1da177e4 LT |
2657 | static int __init filelock_init(void) |
2658 | { | |
7012b02a JL |
2659 | int i; |
2660 | ||
4a075e39 JL |
2661 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2662 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2663 | ||
1da177e4 | 2664 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2665 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2666 | ||
7012b02a JL |
2667 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2668 | ||
2669 | for_each_possible_cpu(i) | |
2670 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2671 | ||
1da177e4 LT |
2672 | return 0; |
2673 | } | |
2674 | ||
2675 | core_initcall(filelock_init); |