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 | ||
160 | #define for_each_lock(inode, lockp) \ | |
161 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
162 | ||
1c8c601a | 163 | /* |
7012b02a JL |
164 | * The global file_lock_list is only used for displaying /proc/locks, so we |
165 | * keep a list on each CPU, with each list protected by its own spinlock via | |
166 | * the file_lock_lglock. Note that alterations to the list also require that | |
167 | * the relevant i_lock is held. | |
1c8c601a | 168 | */ |
7012b02a JL |
169 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
170 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
88974691 | 171 | |
1c8c601a | 172 | /* |
48f74186 | 173 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 174 | * It is protected by blocked_lock_lock. |
48f74186 JL |
175 | * |
176 | * We hash locks by lockowner in order to optimize searching for the lock a | |
177 | * particular lockowner is waiting on. | |
178 | * | |
179 | * FIXME: make this value scale via some heuristic? We generally will want more | |
180 | * buckets when we have more lockowners holding locks, but that's a little | |
181 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 182 | */ |
48f74186 JL |
183 | #define BLOCKED_HASH_BITS 7 |
184 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 185 | |
1c8c601a | 186 | /* |
7b2296af JL |
187 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
188 | * want to be holding this lock. | |
1c8c601a JL |
189 | * |
190 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
191 | * pointer for file_lock structures that are acting as lock requests (in | |
192 | * contrast to those that are acting as records of acquired locks). | |
193 | * | |
194 | * Note that when we acquire this lock in order to change the above fields, | |
195 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
196 | * protected by this lock, we can skip acquiring it iff we already hold the | |
197 | * i_lock. | |
198 | * | |
199 | * In particular, adding an entry to the fl_block list requires that you hold | |
200 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
201 | * an entry from the list however only requires the file_lock_lock. | |
202 | */ | |
7b2296af | 203 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 204 | |
4a075e39 | 205 | static struct kmem_cache *flctx_cache __read_mostly; |
e18b890b | 206 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 207 | |
4a075e39 JL |
208 | static struct file_lock_context * |
209 | locks_get_lock_context(struct inode *inode) | |
210 | { | |
211 | struct file_lock_context *new; | |
212 | ||
213 | if (likely(inode->i_flctx)) | |
214 | goto out; | |
215 | ||
216 | new = kmem_cache_alloc(flctx_cache, GFP_KERNEL); | |
217 | if (!new) | |
218 | goto out; | |
219 | ||
220 | INIT_LIST_HEAD(&new->flc_flock); | |
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)); | |
244 | kmem_cache_free(flctx_cache, ctx); | |
245 | } | |
246 | } | |
247 | ||
ee19cc40 | 248 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 249 | { |
139ca04e | 250 | INIT_HLIST_NODE(&fl->fl_link); |
6dee60f6 | 251 | INIT_LIST_HEAD(&fl->fl_list); |
ee19cc40 MS |
252 | INIT_LIST_HEAD(&fl->fl_block); |
253 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
254 | } |
255 | ||
1da177e4 | 256 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 257 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 258 | { |
ee19cc40 | 259 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
260 | |
261 | if (fl) | |
ee19cc40 | 262 | locks_init_lock_heads(fl); |
a51cb91d MS |
263 | |
264 | return fl; | |
1da177e4 | 265 | } |
c5b1f0d9 | 266 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 267 | |
a9e61e25 | 268 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
269 | { |
270 | if (fl->fl_ops) { | |
271 | if (fl->fl_ops->fl_release_private) | |
272 | fl->fl_ops->fl_release_private(fl); | |
273 | fl->fl_ops = NULL; | |
274 | } | |
47831f35 | 275 | |
5c97d7b1 KM |
276 | if (fl->fl_lmops) { |
277 | if (fl->fl_lmops->lm_put_owner) | |
278 | fl->fl_lmops->lm_put_owner(fl); | |
279 | fl->fl_lmops = NULL; | |
280 | } | |
47831f35 | 281 | } |
a9e61e25 | 282 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 283 | |
1da177e4 | 284 | /* Free a lock which is not in use. */ |
05fa3135 | 285 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 286 | { |
5ce29646 | 287 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
6dee60f6 | 288 | BUG_ON(!list_empty(&fl->fl_list)); |
5ce29646 | 289 | BUG_ON(!list_empty(&fl->fl_block)); |
139ca04e | 290 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 291 | |
47831f35 | 292 | locks_release_private(fl); |
1da177e4 LT |
293 | kmem_cache_free(filelock_cache, fl); |
294 | } | |
05fa3135 | 295 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 296 | |
ed9814d8 JL |
297 | static void |
298 | locks_dispose_list(struct list_head *dispose) | |
299 | { | |
300 | struct file_lock *fl; | |
301 | ||
302 | while (!list_empty(dispose)) { | |
6dee60f6 JL |
303 | fl = list_first_entry(dispose, struct file_lock, fl_list); |
304 | list_del_init(&fl->fl_list); | |
ed9814d8 JL |
305 | locks_free_lock(fl); |
306 | } | |
307 | } | |
308 | ||
1da177e4 LT |
309 | void locks_init_lock(struct file_lock *fl) |
310 | { | |
ee19cc40 MS |
311 | memset(fl, 0, sizeof(struct file_lock)); |
312 | locks_init_lock_heads(fl); | |
1da177e4 LT |
313 | } |
314 | ||
315 | EXPORT_SYMBOL(locks_init_lock); | |
316 | ||
1da177e4 LT |
317 | /* |
318 | * Initialize a new lock from an existing file_lock structure. | |
319 | */ | |
3fe0fff1 | 320 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
321 | { |
322 | new->fl_owner = fl->fl_owner; | |
323 | new->fl_pid = fl->fl_pid; | |
0996905f | 324 | new->fl_file = NULL; |
1da177e4 LT |
325 | new->fl_flags = fl->fl_flags; |
326 | new->fl_type = fl->fl_type; | |
327 | new->fl_start = fl->fl_start; | |
328 | new->fl_end = fl->fl_end; | |
f328296e | 329 | new->fl_lmops = fl->fl_lmops; |
0996905f | 330 | new->fl_ops = NULL; |
f328296e KM |
331 | |
332 | if (fl->fl_lmops) { | |
333 | if (fl->fl_lmops->lm_get_owner) | |
334 | fl->fl_lmops->lm_get_owner(new, fl); | |
335 | } | |
0996905f | 336 | } |
3fe0fff1 | 337 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
338 | |
339 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
340 | { | |
566709bd JL |
341 | /* "new" must be a freshly-initialized lock */ |
342 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 343 | |
3fe0fff1 | 344 | locks_copy_conflock(new, fl); |
f328296e | 345 | |
0996905f | 346 | new->fl_file = fl->fl_file; |
1da177e4 | 347 | new->fl_ops = fl->fl_ops; |
47831f35 | 348 | |
f328296e KM |
349 | if (fl->fl_ops) { |
350 | if (fl->fl_ops->fl_copy_lock) | |
351 | fl->fl_ops->fl_copy_lock(new, fl); | |
352 | } | |
1da177e4 LT |
353 | } |
354 | ||
355 | EXPORT_SYMBOL(locks_copy_lock); | |
356 | ||
357 | static inline int flock_translate_cmd(int cmd) { | |
358 | if (cmd & LOCK_MAND) | |
359 | return cmd & (LOCK_MAND | LOCK_RW); | |
360 | switch (cmd) { | |
361 | case LOCK_SH: | |
362 | return F_RDLCK; | |
363 | case LOCK_EX: | |
364 | return F_WRLCK; | |
365 | case LOCK_UN: | |
366 | return F_UNLCK; | |
367 | } | |
368 | return -EINVAL; | |
369 | } | |
370 | ||
371 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
6e129d00 JL |
372 | static struct file_lock * |
373 | flock_make_lock(struct file *filp, unsigned int cmd) | |
1da177e4 LT |
374 | { |
375 | struct file_lock *fl; | |
376 | int type = flock_translate_cmd(cmd); | |
6e129d00 | 377 | |
1da177e4 | 378 | if (type < 0) |
6e129d00 | 379 | return ERR_PTR(type); |
1da177e4 LT |
380 | |
381 | fl = locks_alloc_lock(); | |
382 | if (fl == NULL) | |
6e129d00 | 383 | return ERR_PTR(-ENOMEM); |
1da177e4 LT |
384 | |
385 | fl->fl_file = filp; | |
73a8f5f7 | 386 | fl->fl_owner = filp; |
1da177e4 LT |
387 | fl->fl_pid = current->tgid; |
388 | fl->fl_flags = FL_FLOCK; | |
389 | fl->fl_type = type; | |
390 | fl->fl_end = OFFSET_MAX; | |
391 | ||
6e129d00 | 392 | return fl; |
1da177e4 LT |
393 | } |
394 | ||
0ec4f431 | 395 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
396 | { |
397 | switch (type) { | |
398 | case F_RDLCK: | |
399 | case F_WRLCK: | |
400 | case F_UNLCK: | |
401 | fl->fl_type = type; | |
402 | break; | |
403 | default: | |
404 | return -EINVAL; | |
405 | } | |
406 | return 0; | |
407 | } | |
408 | ||
ef12e72a BF |
409 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
410 | struct flock64 *l) | |
1da177e4 | 411 | { |
1da177e4 | 412 | switch (l->l_whence) { |
f5579f8c | 413 | case SEEK_SET: |
ef12e72a | 414 | fl->fl_start = 0; |
1da177e4 | 415 | break; |
f5579f8c | 416 | case SEEK_CUR: |
ef12e72a | 417 | fl->fl_start = filp->f_pos; |
1da177e4 | 418 | break; |
f5579f8c | 419 | case SEEK_END: |
ef12e72a | 420 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
421 | break; |
422 | default: | |
423 | return -EINVAL; | |
424 | } | |
ef12e72a BF |
425 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
426 | return -EOVERFLOW; | |
427 | fl->fl_start += l->l_start; | |
428 | if (fl->fl_start < 0) | |
429 | return -EINVAL; | |
1da177e4 LT |
430 | |
431 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
432 | POSIX-2001 defines it. */ | |
4c780a46 | 433 | if (l->l_len > 0) { |
ef12e72a BF |
434 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
435 | return -EOVERFLOW; | |
436 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
437 | ||
4c780a46 | 438 | } else if (l->l_len < 0) { |
ef12e72a | 439 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 440 | return -EINVAL; |
ef12e72a BF |
441 | fl->fl_end = fl->fl_start - 1; |
442 | fl->fl_start += l->l_len; | |
443 | } else | |
444 | fl->fl_end = OFFSET_MAX; | |
445 | ||
1da177e4 LT |
446 | fl->fl_owner = current->files; |
447 | fl->fl_pid = current->tgid; | |
448 | fl->fl_file = filp; | |
449 | fl->fl_flags = FL_POSIX; | |
450 | fl->fl_ops = NULL; | |
451 | fl->fl_lmops = NULL; | |
452 | ||
453 | return assign_type(fl, l->l_type); | |
454 | } | |
455 | ||
ef12e72a BF |
456 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
457 | * style lock. | |
458 | */ | |
459 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
460 | struct flock *l) | |
1da177e4 | 461 | { |
ef12e72a BF |
462 | struct flock64 ll = { |
463 | .l_type = l->l_type, | |
464 | .l_whence = l->l_whence, | |
465 | .l_start = l->l_start, | |
466 | .l_len = l->l_len, | |
467 | }; | |
468 | ||
469 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 470 | } |
1da177e4 LT |
471 | |
472 | /* default lease lock manager operations */ | |
4d01b7f5 JL |
473 | static bool |
474 | lease_break_callback(struct file_lock *fl) | |
1da177e4 LT |
475 | { |
476 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
4d01b7f5 | 477 | return false; |
1da177e4 LT |
478 | } |
479 | ||
1c7dd2ff JL |
480 | static void |
481 | lease_setup(struct file_lock *fl, void **priv) | |
482 | { | |
483 | struct file *filp = fl->fl_file; | |
484 | struct fasync_struct *fa = *priv; | |
485 | ||
486 | /* | |
487 | * fasync_insert_entry() returns the old entry if any. If there was no | |
488 | * old entry, then it used "priv" and inserted it into the fasync list. | |
489 | * Clear the pointer to indicate that it shouldn't be freed. | |
490 | */ | |
491 | if (!fasync_insert_entry(fa->fa_fd, filp, &fl->fl_fasync, fa)) | |
492 | *priv = NULL; | |
493 | ||
494 | __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); | |
495 | } | |
496 | ||
7b021967 | 497 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 498 | .lm_break = lease_break_callback, |
8fb47a4f | 499 | .lm_change = lease_modify, |
1c7dd2ff | 500 | .lm_setup = lease_setup, |
1da177e4 LT |
501 | }; |
502 | ||
503 | /* | |
504 | * Initialize a lease, use the default lock manager operations | |
505 | */ | |
0ec4f431 | 506 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 507 | { |
75dff55a TM |
508 | if (assign_type(fl, type) != 0) |
509 | return -EINVAL; | |
510 | ||
7ca76311 | 511 | fl->fl_owner = filp; |
1da177e4 LT |
512 | fl->fl_pid = current->tgid; |
513 | ||
514 | fl->fl_file = filp; | |
515 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
516 | fl->fl_start = 0; |
517 | fl->fl_end = OFFSET_MAX; | |
518 | fl->fl_ops = NULL; | |
519 | fl->fl_lmops = &lease_manager_ops; | |
520 | return 0; | |
521 | } | |
522 | ||
523 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 524 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
525 | { |
526 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 527 | int error = -ENOMEM; |
1da177e4 LT |
528 | |
529 | if (fl == NULL) | |
e32b8ee2 | 530 | return ERR_PTR(error); |
1da177e4 LT |
531 | |
532 | error = lease_init(filp, type, fl); | |
75dff55a TM |
533 | if (error) { |
534 | locks_free_lock(fl); | |
e32b8ee2 | 535 | return ERR_PTR(error); |
75dff55a | 536 | } |
e32b8ee2 | 537 | return fl; |
1da177e4 LT |
538 | } |
539 | ||
540 | /* Check if two locks overlap each other. | |
541 | */ | |
542 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
543 | { | |
544 | return ((fl1->fl_end >= fl2->fl_start) && | |
545 | (fl2->fl_end >= fl1->fl_start)); | |
546 | } | |
547 | ||
548 | /* | |
549 | * Check whether two locks have the same owner. | |
550 | */ | |
33443c42 | 551 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 552 | { |
8fb47a4f | 553 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 554 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 555 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
556 | return fl1->fl_owner == fl2->fl_owner; |
557 | } | |
558 | ||
7012b02a | 559 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 560 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 561 | { |
7012b02a JL |
562 | lg_local_lock(&file_lock_lglock); |
563 | fl->fl_link_cpu = smp_processor_id(); | |
564 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
565 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
566 | } |
567 | ||
7012b02a | 568 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 569 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 570 | { |
7012b02a JL |
571 | /* |
572 | * Avoid taking lock if already unhashed. This is safe since this check | |
573 | * is done while holding the i_lock, and new insertions into the list | |
574 | * also require that it be held. | |
575 | */ | |
576 | if (hlist_unhashed(&fl->fl_link)) | |
577 | return; | |
578 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 579 | hlist_del_init(&fl->fl_link); |
7012b02a | 580 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
581 | } |
582 | ||
3999e493 JL |
583 | static unsigned long |
584 | posix_owner_key(struct file_lock *fl) | |
585 | { | |
586 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
587 | return fl->fl_lmops->lm_owner_key(fl); | |
588 | return (unsigned long)fl->fl_owner; | |
589 | } | |
590 | ||
6ca10ed8 | 591 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 592 | { |
3999e493 | 593 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
594 | } |
595 | ||
6ca10ed8 | 596 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 597 | { |
48f74186 | 598 | hash_del(&waiter->fl_link); |
88974691 JL |
599 | } |
600 | ||
1da177e4 LT |
601 | /* Remove waiter from blocker's block list. |
602 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 603 | * |
7b2296af | 604 | * Must be called with blocked_lock_lock held. |
1da177e4 | 605 | */ |
33443c42 | 606 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 607 | { |
88974691 | 608 | locks_delete_global_blocked(waiter); |
1da177e4 | 609 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
610 | waiter->fl_next = NULL; |
611 | } | |
612 | ||
1a9e64a7 | 613 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 614 | { |
7b2296af | 615 | spin_lock(&blocked_lock_lock); |
1da177e4 | 616 | __locks_delete_block(waiter); |
7b2296af | 617 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
618 | } |
619 | ||
620 | /* Insert waiter into blocker's block list. | |
621 | * We use a circular list so that processes can be easily woken up in | |
622 | * the order they blocked. The documentation doesn't require this but | |
623 | * it seems like the reasonable thing to do. | |
1c8c601a | 624 | * |
7b2296af | 625 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 626 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 627 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 628 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 629 | */ |
1c8c601a JL |
630 | static void __locks_insert_block(struct file_lock *blocker, |
631 | struct file_lock *waiter) | |
1da177e4 | 632 | { |
6dc0fe8f | 633 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 634 | waiter->fl_next = blocker; |
88974691 | 635 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 636 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
637 | locks_insert_global_blocked(waiter); |
638 | } | |
639 | ||
640 | /* Must be called with i_lock held. */ | |
641 | static void locks_insert_block(struct file_lock *blocker, | |
642 | struct file_lock *waiter) | |
643 | { | |
7b2296af | 644 | spin_lock(&blocked_lock_lock); |
1c8c601a | 645 | __locks_insert_block(blocker, waiter); |
7b2296af | 646 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
647 | } |
648 | ||
1cb36012 JL |
649 | /* |
650 | * Wake up processes blocked waiting for blocker. | |
651 | * | |
1c8c601a | 652 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
653 | */ |
654 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
655 | { | |
4e8c765d JL |
656 | /* |
657 | * Avoid taking global lock if list is empty. This is safe since new | |
658 | * blocked requests are only added to the list under the i_lock, and | |
659 | * the i_lock is always held here. Note that removal from the fl_block | |
660 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 661 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
662 | */ |
663 | if (list_empty(&blocker->fl_block)) | |
664 | return; | |
665 | ||
7b2296af | 666 | spin_lock(&blocked_lock_lock); |
1da177e4 | 667 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
668 | struct file_lock *waiter; |
669 | ||
670 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
671 | struct file_lock, fl_block); |
672 | __locks_delete_block(waiter); | |
8fb47a4f BF |
673 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
674 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
675 | else |
676 | wake_up(&waiter->fl_wait); | |
677 | } | |
7b2296af | 678 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
679 | } |
680 | ||
681 | /* Insert file lock fl into an inode's lock list at the position indicated | |
682 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
683 | * |
684 | * Must be called with the i_lock held! | |
1da177e4 LT |
685 | */ |
686 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
687 | { | |
ab1f1611 VG |
688 | fl->fl_nspid = get_pid(task_tgid(current)); |
689 | ||
1da177e4 LT |
690 | /* insert into file's list */ |
691 | fl->fl_next = *pos; | |
692 | *pos = fl; | |
88974691 JL |
693 | |
694 | locks_insert_global_locks(fl); | |
1da177e4 LT |
695 | } |
696 | ||
5263e31e JL |
697 | static void |
698 | locks_insert_lock_ctx(struct file_lock *fl, struct list_head *before) | |
699 | { | |
700 | fl->fl_nspid = get_pid(task_tgid(current)); | |
701 | list_add_tail(&fl->fl_list, before); | |
702 | locks_insert_global_locks(fl); | |
703 | } | |
704 | ||
24cbe784 JL |
705 | /** |
706 | * locks_delete_lock - Delete a lock and then free it. | |
707 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
708 | * inode->i_flock list entry | |
709 | * | |
710 | * Unlink a lock from all lists and free the namespace reference, but don't | |
711 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
712 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
713 | * |
714 | * Must be called with the i_lock held! | |
1da177e4 | 715 | */ |
24cbe784 | 716 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
717 | { |
718 | struct file_lock *fl = *thisfl_p; | |
719 | ||
88974691 JL |
720 | locks_delete_global_locks(fl); |
721 | ||
1da177e4 LT |
722 | *thisfl_p = fl->fl_next; |
723 | fl->fl_next = NULL; | |
1da177e4 | 724 | |
ab1f1611 VG |
725 | if (fl->fl_nspid) { |
726 | put_pid(fl->fl_nspid); | |
727 | fl->fl_nspid = NULL; | |
728 | } | |
729 | ||
1da177e4 | 730 | locks_wake_up_blocks(fl); |
24cbe784 JL |
731 | } |
732 | ||
733 | /* | |
734 | * Unlink a lock from all lists and free it. | |
735 | * | |
736 | * Must be called with i_lock held! | |
737 | */ | |
ed9814d8 JL |
738 | static void locks_delete_lock(struct file_lock **thisfl_p, |
739 | struct list_head *dispose) | |
24cbe784 JL |
740 | { |
741 | struct file_lock *fl = *thisfl_p; | |
742 | ||
743 | locks_unlink_lock(thisfl_p); | |
ed9814d8 | 744 | if (dispose) |
6dee60f6 | 745 | list_add(&fl->fl_list, dispose); |
ed9814d8 JL |
746 | else |
747 | locks_free_lock(fl); | |
1da177e4 LT |
748 | } |
749 | ||
5263e31e JL |
750 | static void |
751 | locks_delete_lock_ctx(struct file_lock *fl, struct list_head *dispose) | |
752 | { | |
753 | locks_delete_global_locks(fl); | |
754 | if (fl->fl_nspid) { | |
755 | put_pid(fl->fl_nspid); | |
756 | fl->fl_nspid = NULL; | |
757 | } | |
758 | locks_wake_up_blocks(fl); | |
759 | list_move(&fl->fl_list, dispose); | |
760 | } | |
761 | ||
1da177e4 LT |
762 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality |
763 | * checks for shared/exclusive status of overlapping locks. | |
764 | */ | |
765 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
766 | { | |
767 | if (sys_fl->fl_type == F_WRLCK) | |
768 | return 1; | |
769 | if (caller_fl->fl_type == F_WRLCK) | |
770 | return 1; | |
771 | return 0; | |
772 | } | |
773 | ||
774 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
775 | * checking before calling the locks_conflict(). | |
776 | */ | |
777 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
778 | { | |
779 | /* POSIX locks owned by the same process do not conflict with | |
780 | * each other. | |
781 | */ | |
782 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
783 | return (0); | |
784 | ||
785 | /* Check whether they overlap */ | |
786 | if (!locks_overlap(caller_fl, sys_fl)) | |
787 | return 0; | |
788 | ||
789 | return (locks_conflict(caller_fl, sys_fl)); | |
790 | } | |
791 | ||
792 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
793 | * checking before calling the locks_conflict(). | |
794 | */ | |
795 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
796 | { | |
797 | /* FLOCK locks referring to the same filp do not conflict with | |
798 | * each other. | |
799 | */ | |
800 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
801 | return (0); | |
802 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
803 | return 0; | |
804 | ||
805 | return (locks_conflict(caller_fl, sys_fl)); | |
806 | } | |
807 | ||
6d34ac19 | 808 | void |
9d6a8c5c | 809 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
810 | { |
811 | struct file_lock *cfl; | |
1c8c601a | 812 | struct inode *inode = file_inode(filp); |
1da177e4 | 813 | |
1c8c601a | 814 | spin_lock(&inode->i_lock); |
496ad9aa | 815 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
816 | if (!IS_POSIX(cfl)) |
817 | continue; | |
b842e240 | 818 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
819 | break; |
820 | } | |
ab1f1611 | 821 | if (cfl) { |
3fe0fff1 | 822 | locks_copy_conflock(fl, cfl); |
ab1f1611 | 823 | if (cfl->fl_nspid) |
6c5f3e7b | 824 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 825 | } else |
129a84de | 826 | fl->fl_type = F_UNLCK; |
1c8c601a | 827 | spin_unlock(&inode->i_lock); |
6d34ac19 | 828 | return; |
1da177e4 | 829 | } |
1da177e4 LT |
830 | EXPORT_SYMBOL(posix_test_lock); |
831 | ||
b533184f BF |
832 | /* |
833 | * Deadlock detection: | |
834 | * | |
835 | * We attempt to detect deadlocks that are due purely to posix file | |
836 | * locks. | |
1da177e4 | 837 | * |
b533184f BF |
838 | * We assume that a task can be waiting for at most one lock at a time. |
839 | * So for any acquired lock, the process holding that lock may be | |
840 | * waiting on at most one other lock. That lock in turns may be held by | |
841 | * someone waiting for at most one other lock. Given a requested lock | |
842 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
843 | * follow this chain of waiters to ensure we are not about to create a | |
844 | * cycle. | |
1da177e4 | 845 | * |
b533184f BF |
846 | * Since we do this before we ever put a process to sleep on a lock, we |
847 | * are ensured that there is never a cycle; that is what guarantees that | |
848 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 849 | * |
b533184f BF |
850 | * Note: the above assumption may not be true when handling lock |
851 | * requests from a broken NFS client. It may also fail in the presence | |
852 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 853 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 854 | * |
cff2fce5 | 855 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
856 | * Because the owner is not even nominally tied to a thread of |
857 | * execution, the deadlock detection below can't reasonably work well. Just | |
858 | * skip it for those. | |
859 | * | |
cff2fce5 | 860 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
861 | * locks that just checks for the case where two tasks are attempting to |
862 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 863 | */ |
97855b49 BF |
864 | |
865 | #define MAX_DEADLK_ITERATIONS 10 | |
866 | ||
b533184f BF |
867 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
868 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
869 | { | |
870 | struct file_lock *fl; | |
871 | ||
3999e493 | 872 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
873 | if (posix_same_owner(fl, block_fl)) |
874 | return fl->fl_next; | |
875 | } | |
876 | return NULL; | |
877 | } | |
878 | ||
7b2296af | 879 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 880 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
881 | struct file_lock *block_fl) |
882 | { | |
97855b49 | 883 | int i = 0; |
1da177e4 | 884 | |
57b65325 JL |
885 | /* |
886 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 887 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 888 | */ |
cff2fce5 | 889 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
890 | return 0; |
891 | ||
b533184f BF |
892 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
893 | if (i++ > MAX_DEADLK_ITERATIONS) | |
894 | return 0; | |
895 | if (posix_same_owner(caller_fl, block_fl)) | |
896 | return 1; | |
1da177e4 LT |
897 | } |
898 | return 0; | |
899 | } | |
900 | ||
1da177e4 | 901 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 902 | * after any leases, but before any posix locks. |
f475ae95 TM |
903 | * |
904 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
905 | * whether or not a lock was successfully freed by testing the return | |
906 | * value for -ENOENT. | |
1da177e4 | 907 | */ |
993dfa87 | 908 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 909 | { |
993dfa87 | 910 | struct file_lock *new_fl = NULL; |
5263e31e JL |
911 | struct file_lock *fl; |
912 | struct file_lock_context *ctx; | |
913 | struct inode *inode = file_inode(filp); | |
1da177e4 | 914 | int error = 0; |
5263e31e | 915 | bool found = false; |
ed9814d8 | 916 | LIST_HEAD(dispose); |
1da177e4 | 917 | |
5263e31e JL |
918 | ctx = locks_get_lock_context(inode); |
919 | if (!ctx) | |
920 | return -ENOMEM; | |
921 | ||
b89f4321 | 922 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 923 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
924 | if (!new_fl) |
925 | return -ENOMEM; | |
84d535ad PE |
926 | } |
927 | ||
1c8c601a | 928 | spin_lock(&inode->i_lock); |
b89f4321 AB |
929 | if (request->fl_flags & FL_ACCESS) |
930 | goto find_conflict; | |
931 | ||
5263e31e | 932 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
1da177e4 LT |
933 | if (filp != fl->fl_file) |
934 | continue; | |
993dfa87 | 935 | if (request->fl_type == fl->fl_type) |
1da177e4 | 936 | goto out; |
5263e31e JL |
937 | found = true; |
938 | locks_delete_lock_ctx(fl, &dispose); | |
1da177e4 LT |
939 | break; |
940 | } | |
1da177e4 | 941 | |
f475ae95 TM |
942 | if (request->fl_type == F_UNLCK) { |
943 | if ((request->fl_flags & FL_EXISTS) && !found) | |
944 | error = -ENOENT; | |
993dfa87 | 945 | goto out; |
f475ae95 | 946 | } |
1da177e4 LT |
947 | |
948 | /* | |
949 | * If a higher-priority process was blocked on the old file lock, | |
950 | * give it the opportunity to lock the file. | |
951 | */ | |
b89f4321 | 952 | if (found) { |
1c8c601a | 953 | spin_unlock(&inode->i_lock); |
def01bc5 | 954 | cond_resched(); |
1c8c601a | 955 | spin_lock(&inode->i_lock); |
b89f4321 | 956 | } |
1da177e4 | 957 | |
f07f18dd | 958 | find_conflict: |
5263e31e | 959 | list_for_each_entry(fl, &ctx->flc_flock, fl_list) { |
993dfa87 | 960 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
961 | continue; |
962 | error = -EAGAIN; | |
bde74e4b MS |
963 | if (!(request->fl_flags & FL_SLEEP)) |
964 | goto out; | |
965 | error = FILE_LOCK_DEFERRED; | |
966 | locks_insert_block(fl, request); | |
1da177e4 LT |
967 | goto out; |
968 | } | |
f07f18dd TM |
969 | if (request->fl_flags & FL_ACCESS) |
970 | goto out; | |
993dfa87 | 971 | locks_copy_lock(new_fl, request); |
5263e31e | 972 | locks_insert_lock_ctx(new_fl, &ctx->flc_flock); |
993dfa87 | 973 | new_fl = NULL; |
9cedc194 | 974 | error = 0; |
1da177e4 LT |
975 | |
976 | out: | |
1c8c601a | 977 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
978 | if (new_fl) |
979 | locks_free_lock(new_fl); | |
ed9814d8 | 980 | locks_dispose_list(&dispose); |
1da177e4 LT |
981 | return error; |
982 | } | |
983 | ||
150b3934 | 984 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
985 | { |
986 | struct file_lock *fl; | |
39005d02 MS |
987 | struct file_lock *new_fl = NULL; |
988 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
989 | struct file_lock *left = NULL; |
990 | struct file_lock *right = NULL; | |
991 | struct file_lock **before; | |
b9746ef8 JL |
992 | int error; |
993 | bool added = false; | |
ed9814d8 | 994 | LIST_HEAD(dispose); |
1da177e4 LT |
995 | |
996 | /* | |
997 | * We may need two file_lock structures for this operation, | |
998 | * so we get them in advance to avoid races. | |
39005d02 MS |
999 | * |
1000 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 1001 | */ |
39005d02 MS |
1002 | if (!(request->fl_flags & FL_ACCESS) && |
1003 | (request->fl_type != F_UNLCK || | |
1004 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
1005 | new_fl = locks_alloc_lock(); | |
1006 | new_fl2 = locks_alloc_lock(); | |
1007 | } | |
1da177e4 | 1008 | |
1c8c601a | 1009 | spin_lock(&inode->i_lock); |
1cb36012 JL |
1010 | /* |
1011 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
1012 | * there are any, either return error or put the request on the | |
48f74186 | 1013 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 1014 | */ |
1da177e4 LT |
1015 | if (request->fl_type != F_UNLCK) { |
1016 | for_each_lock(inode, before) { | |
526985b9 | 1017 | fl = *before; |
1da177e4 LT |
1018 | if (!IS_POSIX(fl)) |
1019 | continue; | |
1020 | if (!posix_locks_conflict(request, fl)) | |
1021 | continue; | |
5842add2 | 1022 | if (conflock) |
3fe0fff1 | 1023 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
1024 | error = -EAGAIN; |
1025 | if (!(request->fl_flags & FL_SLEEP)) | |
1026 | goto out; | |
1c8c601a JL |
1027 | /* |
1028 | * Deadlock detection and insertion into the blocked | |
1029 | * locks list must be done while holding the same lock! | |
1030 | */ | |
1da177e4 | 1031 | error = -EDEADLK; |
7b2296af | 1032 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
1033 | if (likely(!posix_locks_deadlock(request, fl))) { |
1034 | error = FILE_LOCK_DEFERRED; | |
1035 | __locks_insert_block(fl, request); | |
1036 | } | |
7b2296af | 1037 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
1038 | goto out; |
1039 | } | |
1040 | } | |
1041 | ||
1042 | /* If we're just looking for a conflict, we're done. */ | |
1043 | error = 0; | |
1044 | if (request->fl_flags & FL_ACCESS) | |
1045 | goto out; | |
1046 | ||
1da177e4 | 1047 | /* |
1da177e4 LT |
1048 | * Find the first old lock with the same owner as the new lock. |
1049 | */ | |
1050 | ||
1051 | before = &inode->i_flock; | |
1052 | ||
1053 | /* First skip locks owned by other processes. */ | |
1054 | while ((fl = *before) && (!IS_POSIX(fl) || | |
1055 | !posix_same_owner(request, fl))) { | |
1056 | before = &fl->fl_next; | |
1057 | } | |
1058 | ||
1cb36012 | 1059 | /* Process locks with this owner. */ |
1da177e4 LT |
1060 | while ((fl = *before) && posix_same_owner(request, fl)) { |
1061 | /* Detect adjacent or overlapping regions (if same lock type) | |
1062 | */ | |
1063 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
1064 | /* In all comparisons of start vs end, use |
1065 | * "start - 1" rather than "end + 1". If end | |
1066 | * is OFFSET_MAX, end + 1 will become negative. | |
1067 | */ | |
1da177e4 LT |
1068 | if (fl->fl_end < request->fl_start - 1) |
1069 | goto next_lock; | |
1070 | /* If the next lock in the list has entirely bigger | |
1071 | * addresses than the new one, insert the lock here. | |
1072 | */ | |
449231d6 | 1073 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1074 | break; |
1075 | ||
1076 | /* If we come here, the new and old lock are of the | |
1077 | * same type and adjacent or overlapping. Make one | |
1078 | * lock yielding from the lower start address of both | |
1079 | * locks to the higher end address. | |
1080 | */ | |
1081 | if (fl->fl_start > request->fl_start) | |
1082 | fl->fl_start = request->fl_start; | |
1083 | else | |
1084 | request->fl_start = fl->fl_start; | |
1085 | if (fl->fl_end < request->fl_end) | |
1086 | fl->fl_end = request->fl_end; | |
1087 | else | |
1088 | request->fl_end = fl->fl_end; | |
1089 | if (added) { | |
ed9814d8 | 1090 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1091 | continue; |
1092 | } | |
1093 | request = fl; | |
b9746ef8 | 1094 | added = true; |
1da177e4 LT |
1095 | } |
1096 | else { | |
1097 | /* Processing for different lock types is a bit | |
1098 | * more complex. | |
1099 | */ | |
1100 | if (fl->fl_end < request->fl_start) | |
1101 | goto next_lock; | |
1102 | if (fl->fl_start > request->fl_end) | |
1103 | break; | |
1104 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1105 | added = true; |
1da177e4 LT |
1106 | if (fl->fl_start < request->fl_start) |
1107 | left = fl; | |
1108 | /* If the next lock in the list has a higher end | |
1109 | * address than the new one, insert the new one here. | |
1110 | */ | |
1111 | if (fl->fl_end > request->fl_end) { | |
1112 | right = fl; | |
1113 | break; | |
1114 | } | |
1115 | if (fl->fl_start >= request->fl_start) { | |
1116 | /* The new lock completely replaces an old | |
1117 | * one (This may happen several times). | |
1118 | */ | |
1119 | if (added) { | |
ed9814d8 | 1120 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1121 | continue; |
1122 | } | |
b84d49f9 JL |
1123 | /* |
1124 | * Replace the old lock with new_fl, and | |
1125 | * remove the old one. It's safe to do the | |
1126 | * insert here since we know that we won't be | |
1127 | * using new_fl later, and that the lock is | |
1128 | * just replacing an existing lock. | |
1da177e4 | 1129 | */ |
b84d49f9 JL |
1130 | error = -ENOLCK; |
1131 | if (!new_fl) | |
1132 | goto out; | |
1133 | locks_copy_lock(new_fl, request); | |
1134 | request = new_fl; | |
1135 | new_fl = NULL; | |
ed9814d8 | 1136 | locks_delete_lock(before, &dispose); |
b84d49f9 | 1137 | locks_insert_lock(before, request); |
b9746ef8 | 1138 | added = true; |
1da177e4 LT |
1139 | } |
1140 | } | |
1141 | /* Go on to next lock. | |
1142 | */ | |
1143 | next_lock: | |
1144 | before = &fl->fl_next; | |
1145 | } | |
1146 | ||
0d9a490a | 1147 | /* |
1cb36012 JL |
1148 | * The above code only modifies existing locks in case of merging or |
1149 | * replacing. If new lock(s) need to be inserted all modifications are | |
1150 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1151 | */ |
1152 | error = -ENOLCK; /* "no luck" */ | |
1153 | if (right && left == right && !new_fl2) | |
1154 | goto out; | |
1155 | ||
1da177e4 LT |
1156 | error = 0; |
1157 | if (!added) { | |
f475ae95 TM |
1158 | if (request->fl_type == F_UNLCK) { |
1159 | if (request->fl_flags & FL_EXISTS) | |
1160 | error = -ENOENT; | |
1da177e4 | 1161 | goto out; |
f475ae95 | 1162 | } |
0d9a490a MS |
1163 | |
1164 | if (!new_fl) { | |
1165 | error = -ENOLCK; | |
1166 | goto out; | |
1167 | } | |
1da177e4 LT |
1168 | locks_copy_lock(new_fl, request); |
1169 | locks_insert_lock(before, new_fl); | |
1170 | new_fl = NULL; | |
1171 | } | |
1172 | if (right) { | |
1173 | if (left == right) { | |
1174 | /* The new lock breaks the old one in two pieces, | |
1175 | * so we have to use the second new lock. | |
1176 | */ | |
1177 | left = new_fl2; | |
1178 | new_fl2 = NULL; | |
1179 | locks_copy_lock(left, right); | |
1180 | locks_insert_lock(before, left); | |
1181 | } | |
1182 | right->fl_start = request->fl_end + 1; | |
1183 | locks_wake_up_blocks(right); | |
1184 | } | |
1185 | if (left) { | |
1186 | left->fl_end = request->fl_start - 1; | |
1187 | locks_wake_up_blocks(left); | |
1188 | } | |
1189 | out: | |
1c8c601a | 1190 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1191 | /* |
1192 | * Free any unused locks. | |
1193 | */ | |
1194 | if (new_fl) | |
1195 | locks_free_lock(new_fl); | |
1196 | if (new_fl2) | |
1197 | locks_free_lock(new_fl2); | |
ed9814d8 | 1198 | locks_dispose_list(&dispose); |
1da177e4 LT |
1199 | return error; |
1200 | } | |
1201 | ||
1202 | /** | |
1203 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1204 | * @filp: The file to apply the lock to | |
1205 | * @fl: The lock to be applied | |
150b3934 | 1206 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1207 | * |
1208 | * Add a POSIX style lock to a file. | |
1209 | * We merge adjacent & overlapping locks whenever possible. | |
1210 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1211 | * |
1212 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1213 | * whether or not a lock was successfully freed by testing the return | |
1214 | * value for -ENOENT. | |
1da177e4 | 1215 | */ |
150b3934 | 1216 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1217 | struct file_lock *conflock) |
1218 | { | |
496ad9aa | 1219 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1220 | } |
150b3934 | 1221 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1222 | |
1223 | /** | |
1224 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1225 | * @filp: The file to apply the lock to | |
1226 | * @fl: The lock to be applied | |
1227 | * | |
1228 | * Add a POSIX style lock to a file. | |
1229 | * We merge adjacent & overlapping locks whenever possible. | |
1230 | * POSIX locks are sorted by owner task, then by starting address | |
1231 | */ | |
1232 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1233 | { | |
1234 | int error; | |
1235 | might_sleep (); | |
1236 | for (;;) { | |
150b3934 | 1237 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1238 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1239 | break; |
1240 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1241 | if (!error) | |
1242 | continue; | |
1243 | ||
1244 | locks_delete_block(fl); | |
1245 | break; | |
1246 | } | |
1247 | return error; | |
1248 | } | |
1249 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1250 | ||
1251 | /** | |
1252 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1253 | * @file: the file to check |
1da177e4 LT |
1254 | * |
1255 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1256 | * This function is called from locks_verify_locked() only. | |
1257 | */ | |
d7a06983 | 1258 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1259 | { |
d7a06983 | 1260 | struct inode *inode = file_inode(file); |
1da177e4 LT |
1261 | struct file_lock *fl; |
1262 | ||
1263 | /* | |
1264 | * Search the lock list for this inode for any POSIX locks. | |
1265 | */ | |
1c8c601a | 1266 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1267 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1268 | if (!IS_POSIX(fl)) | |
1269 | continue; | |
73a8f5f7 CH |
1270 | if (fl->fl_owner != current->files && |
1271 | fl->fl_owner != file) | |
1da177e4 LT |
1272 | break; |
1273 | } | |
1c8c601a | 1274 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1275 | return fl ? -EAGAIN : 0; |
1276 | } | |
1277 | ||
1278 | /** | |
1279 | * locks_mandatory_area - Check for a conflicting lock | |
1280 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1281 | * for shared | |
1282 | * @inode: the file to check | |
1283 | * @filp: how the file was opened (if it was) | |
1284 | * @offset: start of area to check | |
1285 | * @count: length of area to check | |
1286 | * | |
1287 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1288 | * This function is called from rw_verify_area() and | |
1289 | * locks_verify_truncate(). | |
1290 | */ | |
1291 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1292 | struct file *filp, loff_t offset, | |
1293 | size_t count) | |
1294 | { | |
1295 | struct file_lock fl; | |
1296 | int error; | |
29723ade | 1297 | bool sleep = false; |
1da177e4 LT |
1298 | |
1299 | locks_init_lock(&fl); | |
1da177e4 LT |
1300 | fl.fl_pid = current->tgid; |
1301 | fl.fl_file = filp; | |
1302 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1303 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1304 | sleep = true; |
1da177e4 LT |
1305 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1306 | fl.fl_start = offset; | |
1307 | fl.fl_end = offset + count - 1; | |
1308 | ||
1309 | for (;;) { | |
29723ade | 1310 | if (filp) { |
73a8f5f7 | 1311 | fl.fl_owner = filp; |
29723ade JL |
1312 | fl.fl_flags &= ~FL_SLEEP; |
1313 | error = __posix_lock_file(inode, &fl, NULL); | |
1314 | if (!error) | |
1315 | break; | |
1316 | } | |
1317 | ||
1318 | if (sleep) | |
1319 | fl.fl_flags |= FL_SLEEP; | |
1320 | fl.fl_owner = current->files; | |
150b3934 | 1321 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1322 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1323 | break; |
1324 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1325 | if (!error) { | |
1326 | /* | |
1327 | * If we've been sleeping someone might have | |
1328 | * changed the permissions behind our back. | |
1329 | */ | |
a16877ca | 1330 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1331 | continue; |
1332 | } | |
1333 | ||
1334 | locks_delete_block(&fl); | |
1335 | break; | |
1336 | } | |
1337 | ||
1338 | return error; | |
1339 | } | |
1340 | ||
1341 | EXPORT_SYMBOL(locks_mandatory_area); | |
1342 | ||
778fc546 BF |
1343 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1344 | { | |
1345 | switch (arg) { | |
1346 | case F_UNLCK: | |
1347 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1348 | /* fall through: */ | |
1349 | case F_RDLCK: | |
1350 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1351 | } | |
1352 | } | |
1353 | ||
1da177e4 | 1354 | /* We already had a lease on this file; just change its type */ |
c45198ed | 1355 | int lease_modify(struct file_lock **before, int arg, struct list_head *dispose) |
1da177e4 LT |
1356 | { |
1357 | struct file_lock *fl = *before; | |
1358 | int error = assign_type(fl, arg); | |
1359 | ||
1360 | if (error) | |
1361 | return error; | |
778fc546 | 1362 | lease_clear_pending(fl, arg); |
1da177e4 | 1363 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1364 | if (arg == F_UNLCK) { |
1365 | struct file *filp = fl->fl_file; | |
1366 | ||
1367 | f_delown(filp); | |
1368 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1369 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1370 | if (fl->fl_fasync != NULL) { | |
1371 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1372 | fl->fl_fasync = NULL; | |
1373 | } | |
c45198ed | 1374 | locks_delete_lock(before, dispose); |
3b6e2723 | 1375 | } |
1da177e4 LT |
1376 | return 0; |
1377 | } | |
1da177e4 LT |
1378 | EXPORT_SYMBOL(lease_modify); |
1379 | ||
778fc546 BF |
1380 | static bool past_time(unsigned long then) |
1381 | { | |
1382 | if (!then) | |
1383 | /* 0 is a special value meaning "this never expires": */ | |
1384 | return false; | |
1385 | return time_after(jiffies, then); | |
1386 | } | |
1387 | ||
c45198ed | 1388 | static void time_out_leases(struct inode *inode, struct list_head *dispose) |
1da177e4 LT |
1389 | { |
1390 | struct file_lock **before; | |
1391 | struct file_lock *fl; | |
1392 | ||
f82b4b67 JL |
1393 | lockdep_assert_held(&inode->i_lock); |
1394 | ||
1da177e4 | 1395 | before = &inode->i_flock; |
ab83fa4b | 1396 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1397 | trace_time_out_leases(inode, fl); |
778fc546 | 1398 | if (past_time(fl->fl_downgrade_time)) |
c45198ed | 1399 | lease_modify(before, F_RDLCK, dispose); |
778fc546 | 1400 | if (past_time(fl->fl_break_time)) |
c45198ed | 1401 | lease_modify(before, F_UNLCK, dispose); |
1da177e4 LT |
1402 | if (fl == *before) /* lease_modify may have freed fl */ |
1403 | before = &fl->fl_next; | |
1404 | } | |
1405 | } | |
1406 | ||
df4e8d2c BF |
1407 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1408 | { | |
1409 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1410 | return false; | |
1411 | return locks_conflict(breaker, lease); | |
1412 | } | |
1413 | ||
03d12ddf JL |
1414 | static bool |
1415 | any_leases_conflict(struct inode *inode, struct file_lock *breaker) | |
1416 | { | |
1417 | struct file_lock *fl; | |
1418 | ||
1419 | lockdep_assert_held(&inode->i_lock); | |
1420 | ||
1421 | for (fl = inode->i_flock ; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
1422 | if (leases_conflict(fl, breaker)) | |
1423 | return true; | |
1424 | } | |
1425 | return false; | |
1426 | } | |
1427 | ||
1da177e4 LT |
1428 | /** |
1429 | * __break_lease - revoke all outstanding leases on file | |
1430 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1431 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1432 | * break all leases | |
1433 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1434 | * only delegations | |
1da177e4 | 1435 | * |
87250dd2 | 1436 | * break_lease (inlined for speed) has checked there already is at least |
1437 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1438 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1439 | * specified %O_NONBLOCK to your open(). |
1440 | */ | |
df4e8d2c | 1441 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1442 | { |
778fc546 | 1443 | int error = 0; |
03d12ddf | 1444 | struct file_lock *new_fl; |
4d01b7f5 | 1445 | struct file_lock *fl, **before; |
1da177e4 | 1446 | unsigned long break_time; |
8737c930 | 1447 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
c45198ed | 1448 | LIST_HEAD(dispose); |
1da177e4 | 1449 | |
8737c930 | 1450 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1451 | if (IS_ERR(new_fl)) |
1452 | return PTR_ERR(new_fl); | |
df4e8d2c | 1453 | new_fl->fl_flags = type; |
1da177e4 | 1454 | |
1c8c601a | 1455 | spin_lock(&inode->i_lock); |
1da177e4 | 1456 | |
c45198ed | 1457 | time_out_leases(inode, &dispose); |
1da177e4 | 1458 | |
03d12ddf | 1459 | if (!any_leases_conflict(inode, new_fl)) |
778fc546 BF |
1460 | goto out; |
1461 | ||
1da177e4 LT |
1462 | break_time = 0; |
1463 | if (lease_break_time > 0) { | |
1464 | break_time = jiffies + lease_break_time * HZ; | |
1465 | if (break_time == 0) | |
1466 | break_time++; /* so that 0 means no break time */ | |
1467 | } | |
1468 | ||
4d01b7f5 JL |
1469 | for (before = &inode->i_flock; |
1470 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1471 | before = &fl->fl_next) { | |
df4e8d2c BF |
1472 | if (!leases_conflict(fl, new_fl)) |
1473 | continue; | |
778fc546 BF |
1474 | if (want_write) { |
1475 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1476 | continue; | |
1477 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1478 | fl->fl_break_time = break_time; |
778fc546 | 1479 | } else { |
03d12ddf | 1480 | if (lease_breaking(inode->i_flock)) |
778fc546 BF |
1481 | continue; |
1482 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1483 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1484 | } |
4d01b7f5 JL |
1485 | if (fl->fl_lmops->lm_break(fl)) |
1486 | locks_delete_lock(before, &dispose); | |
1da177e4 LT |
1487 | } |
1488 | ||
4d01b7f5 JL |
1489 | fl = inode->i_flock; |
1490 | if (!fl || !IS_LEASE(fl)) | |
1491 | goto out; | |
1492 | ||
843c6b2f | 1493 | if (mode & O_NONBLOCK) { |
62af4f1f | 1494 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1495 | error = -EWOULDBLOCK; |
1496 | goto out; | |
1497 | } | |
1498 | ||
1499 | restart: | |
03d12ddf | 1500 | break_time = inode->i_flock->fl_break_time; |
f1c6bb2c | 1501 | if (break_time != 0) |
1da177e4 | 1502 | break_time -= jiffies; |
f1c6bb2c JL |
1503 | if (break_time == 0) |
1504 | break_time++; | |
03d12ddf | 1505 | locks_insert_block(inode->i_flock, new_fl); |
62af4f1f | 1506 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1507 | spin_unlock(&inode->i_lock); |
c45198ed | 1508 | locks_dispose_list(&dispose); |
4321e01e MW |
1509 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1510 | !new_fl->fl_next, break_time); | |
1c8c601a | 1511 | spin_lock(&inode->i_lock); |
62af4f1f | 1512 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1513 | locks_delete_block(new_fl); |
1da177e4 | 1514 | if (error >= 0) { |
778fc546 BF |
1515 | /* |
1516 | * Wait for the next conflicting lease that has not been | |
1517 | * broken yet | |
1518 | */ | |
03d12ddf JL |
1519 | if (error == 0) |
1520 | time_out_leases(inode, &dispose); | |
1521 | if (any_leases_conflict(inode, new_fl)) | |
1522 | goto restart; | |
1523 | ||
1da177e4 LT |
1524 | error = 0; |
1525 | } | |
1526 | ||
1527 | out: | |
1c8c601a | 1528 | spin_unlock(&inode->i_lock); |
c45198ed | 1529 | locks_dispose_list(&dispose); |
6d4b9e38 | 1530 | locks_free_lock(new_fl); |
1da177e4 LT |
1531 | return error; |
1532 | } | |
1533 | ||
1534 | EXPORT_SYMBOL(__break_lease); | |
1535 | ||
1536 | /** | |
a6b91919 | 1537 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1538 | * @inode: the inode |
1539 | * @time: pointer to a timespec which will contain the last modified time | |
1540 | * | |
1541 | * This is to force NFS clients to flush their caches for files with | |
1542 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1543 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1544 | */ |
1545 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1546 | { | |
bfe86024 JL |
1547 | bool has_lease = false; |
1548 | struct file_lock *flock; | |
1549 | ||
1550 | if (inode->i_flock) { | |
1551 | spin_lock(&inode->i_lock); | |
1552 | flock = inode->i_flock; | |
1553 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) | |
1554 | has_lease = true; | |
1555 | spin_unlock(&inode->i_lock); | |
1556 | } | |
1557 | ||
1558 | if (has_lease) | |
1da177e4 LT |
1559 | *time = current_fs_time(inode->i_sb); |
1560 | else | |
1561 | *time = inode->i_mtime; | |
1562 | } | |
1563 | ||
1564 | EXPORT_SYMBOL(lease_get_mtime); | |
1565 | ||
1566 | /** | |
1567 | * fcntl_getlease - Enquire what lease is currently active | |
1568 | * @filp: the file | |
1569 | * | |
1570 | * The value returned by this function will be one of | |
1571 | * (if no lease break is pending): | |
1572 | * | |
1573 | * %F_RDLCK to indicate a shared lease is held. | |
1574 | * | |
1575 | * %F_WRLCK to indicate an exclusive lease is held. | |
1576 | * | |
1577 | * %F_UNLCK to indicate no lease is held. | |
1578 | * | |
1579 | * (if a lease break is pending): | |
1580 | * | |
1581 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1582 | * changed to a shared lease (or removed). | |
1583 | * | |
1584 | * %F_UNLCK to indicate the lease needs to be removed. | |
1585 | * | |
1586 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1587 | * should be returned to userspace. | |
1588 | */ | |
1589 | int fcntl_getlease(struct file *filp) | |
1590 | { | |
1591 | struct file_lock *fl; | |
1c8c601a | 1592 | struct inode *inode = file_inode(filp); |
1da177e4 | 1593 | int type = F_UNLCK; |
c45198ed | 1594 | LIST_HEAD(dispose); |
1da177e4 | 1595 | |
1c8c601a | 1596 | spin_lock(&inode->i_lock); |
c45198ed | 1597 | time_out_leases(file_inode(filp), &dispose); |
496ad9aa | 1598 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); |
1da177e4 LT |
1599 | fl = fl->fl_next) { |
1600 | if (fl->fl_file == filp) { | |
778fc546 | 1601 | type = target_leasetype(fl); |
1da177e4 LT |
1602 | break; |
1603 | } | |
1604 | } | |
1c8c601a | 1605 | spin_unlock(&inode->i_lock); |
c45198ed | 1606 | locks_dispose_list(&dispose); |
1da177e4 LT |
1607 | return type; |
1608 | } | |
1609 | ||
24cbe784 JL |
1610 | /** |
1611 | * check_conflicting_open - see if the given dentry points to a file that has | |
1612 | * an existing open that would conflict with the | |
1613 | * desired lease. | |
1614 | * @dentry: dentry to check | |
1615 | * @arg: type of lease that we're trying to acquire | |
1616 | * | |
1617 | * Check to see if there's an existing open fd on this file that would | |
1618 | * conflict with the lease we're trying to set. | |
1619 | */ | |
1620 | static int | |
1621 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1622 | { | |
1623 | int ret = 0; | |
1624 | struct inode *inode = dentry->d_inode; | |
1625 | ||
1626 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1627 | return -EAGAIN; | |
1628 | ||
1629 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1630 | (atomic_read(&inode->i_count) > 1))) | |
1631 | ret = -EAGAIN; | |
1632 | ||
1633 | return ret; | |
1634 | } | |
1635 | ||
e6f5c789 JL |
1636 | static int |
1637 | generic_add_lease(struct file *filp, long arg, struct file_lock **flp, void **priv) | |
1da177e4 | 1638 | { |
7eaae282 | 1639 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1640 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1641 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1642 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1643 | int error; |
c45198ed | 1644 | LIST_HEAD(dispose); |
1da177e4 | 1645 | |
096657b6 | 1646 | lease = *flp; |
62af4f1f JL |
1647 | trace_generic_add_lease(inode, lease); |
1648 | ||
df4e8d2c BF |
1649 | /* |
1650 | * In the delegation case we need mutual exclusion with | |
1651 | * a number of operations that take the i_mutex. We trylock | |
1652 | * because delegations are an optional optimization, and if | |
1653 | * there's some chance of a conflict--we'd rather not | |
1654 | * bother, maybe that's a sign this just isn't a good file to | |
1655 | * hand out a delegation on. | |
1656 | */ | |
1657 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1658 | return -EAGAIN; | |
1659 | ||
1660 | if (is_deleg && arg == F_WRLCK) { | |
1661 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1662 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1663 | WARN_ON_ONCE(1); |
1664 | return -EINVAL; | |
1665 | } | |
096657b6 | 1666 | |
f82b4b67 | 1667 | spin_lock(&inode->i_lock); |
c45198ed | 1668 | time_out_leases(inode, &dispose); |
24cbe784 JL |
1669 | error = check_conflicting_open(dentry, arg); |
1670 | if (error) | |
096657b6 | 1671 | goto out; |
6d5e8b05 | 1672 | |
1da177e4 LT |
1673 | /* |
1674 | * At this point, we know that if there is an exclusive | |
1675 | * lease on this file, then we hold it on this filp | |
1676 | * (otherwise our open of this file would have blocked). | |
1677 | * And if we are trying to acquire an exclusive lease, | |
1678 | * then the file is not open by anyone (including us) | |
1679 | * except for this filp. | |
1680 | */ | |
c1f24ef4 | 1681 | error = -EAGAIN; |
1da177e4 LT |
1682 | for (before = &inode->i_flock; |
1683 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1684 | before = &fl->fl_next) { | |
c1f24ef4 | 1685 | if (fl->fl_file == filp) { |
1da177e4 | 1686 | my_before = before; |
c1f24ef4 BF |
1687 | continue; |
1688 | } | |
1689 | /* | |
1690 | * No exclusive leases if someone else has a lease on | |
1691 | * this file: | |
1692 | */ | |
1693 | if (arg == F_WRLCK) | |
1694 | goto out; | |
1695 | /* | |
1696 | * Modifying our existing lease is OK, but no getting a | |
1697 | * new lease if someone else is opening for write: | |
1698 | */ | |
1699 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1700 | goto out; | |
1da177e4 LT |
1701 | } |
1702 | ||
1da177e4 | 1703 | if (my_before != NULL) { |
1c7dd2ff | 1704 | lease = *my_before; |
c45198ed | 1705 | error = lease->fl_lmops->lm_change(my_before, arg, &dispose); |
1c7dd2ff JL |
1706 | if (error) |
1707 | goto out; | |
1708 | goto out_setup; | |
1da177e4 LT |
1709 | } |
1710 | ||
1da177e4 LT |
1711 | error = -EINVAL; |
1712 | if (!leases_enable) | |
1713 | goto out; | |
1714 | ||
c5b1f0d9 | 1715 | locks_insert_lock(before, lease); |
24cbe784 JL |
1716 | /* |
1717 | * The check in break_lease() is lockless. It's possible for another | |
1718 | * open to race in after we did the earlier check for a conflicting | |
1719 | * open but before the lease was inserted. Check again for a | |
1720 | * conflicting open and cancel the lease if there is one. | |
1721 | * | |
1722 | * We also add a barrier here to ensure that the insertion of the lock | |
1723 | * precedes these checks. | |
1724 | */ | |
1725 | smp_mb(); | |
1726 | error = check_conflicting_open(dentry, arg); | |
1727 | if (error) | |
e6f5c789 | 1728 | goto out_unlink; |
1c7dd2ff JL |
1729 | |
1730 | out_setup: | |
1731 | if (lease->fl_lmops->lm_setup) | |
1732 | lease->fl_lmops->lm_setup(lease, priv); | |
1da177e4 | 1733 | out: |
f82b4b67 | 1734 | spin_unlock(&inode->i_lock); |
c45198ed | 1735 | locks_dispose_list(&dispose); |
df4e8d2c BF |
1736 | if (is_deleg) |
1737 | mutex_unlock(&inode->i_mutex); | |
1c7dd2ff JL |
1738 | if (!error && !my_before) |
1739 | *flp = NULL; | |
1da177e4 | 1740 | return error; |
e6f5c789 JL |
1741 | out_unlink: |
1742 | locks_unlink_lock(before); | |
1743 | goto out; | |
1da177e4 | 1744 | } |
8335ebd9 | 1745 | |
0efaa7e8 | 1746 | static int generic_delete_lease(struct file *filp) |
8335ebd9 | 1747 | { |
0efaa7e8 | 1748 | int error = -EAGAIN; |
8335ebd9 BF |
1749 | struct file_lock *fl, **before; |
1750 | struct dentry *dentry = filp->f_path.dentry; | |
1751 | struct inode *inode = dentry->d_inode; | |
c45198ed | 1752 | LIST_HEAD(dispose); |
8335ebd9 | 1753 | |
f82b4b67 | 1754 | spin_lock(&inode->i_lock); |
c45198ed | 1755 | time_out_leases(inode, &dispose); |
8335ebd9 BF |
1756 | for (before = &inode->i_flock; |
1757 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1758 | before = &fl->fl_next) { | |
0efaa7e8 JL |
1759 | if (fl->fl_file == filp) |
1760 | break; | |
8335ebd9 | 1761 | } |
0efaa7e8 | 1762 | trace_generic_delete_lease(inode, fl); |
52d304eb | 1763 | if (fl && IS_LEASE(fl)) |
c45198ed | 1764 | error = fl->fl_lmops->lm_change(before, F_UNLCK, &dispose); |
f82b4b67 | 1765 | spin_unlock(&inode->i_lock); |
c45198ed | 1766 | locks_dispose_list(&dispose); |
0efaa7e8 | 1767 | return error; |
8335ebd9 BF |
1768 | } |
1769 | ||
1770 | /** | |
1771 | * generic_setlease - sets a lease on an open file | |
1c7dd2ff JL |
1772 | * @filp: file pointer |
1773 | * @arg: type of lease to obtain | |
1774 | * @flp: input - file_lock to use, output - file_lock inserted | |
1775 | * @priv: private data for lm_setup (may be NULL if lm_setup | |
1776 | * doesn't require it) | |
8335ebd9 BF |
1777 | * |
1778 | * The (input) flp->fl_lmops->lm_break function is required | |
1779 | * by break_lease(). | |
8335ebd9 | 1780 | */ |
e6f5c789 JL |
1781 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp, |
1782 | void **priv) | |
8335ebd9 BF |
1783 | { |
1784 | struct dentry *dentry = filp->f_path.dentry; | |
1785 | struct inode *inode = dentry->d_inode; | |
1786 | int error; | |
1787 | ||
8e96e3b7 | 1788 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1789 | return -EACCES; |
1790 | if (!S_ISREG(inode->i_mode)) | |
1791 | return -EINVAL; | |
1792 | error = security_file_lock(filp, arg); | |
1793 | if (error) | |
1794 | return error; | |
1795 | ||
8335ebd9 BF |
1796 | switch (arg) { |
1797 | case F_UNLCK: | |
0efaa7e8 | 1798 | return generic_delete_lease(filp); |
8335ebd9 BF |
1799 | case F_RDLCK: |
1800 | case F_WRLCK: | |
0efaa7e8 JL |
1801 | if (!(*flp)->fl_lmops->lm_break) { |
1802 | WARN_ON_ONCE(1); | |
1803 | return -ENOLCK; | |
1804 | } | |
e6f5c789 | 1805 | return generic_add_lease(filp, arg, flp, priv); |
8335ebd9 | 1806 | default: |
8d657eb3 | 1807 | return -EINVAL; |
8335ebd9 BF |
1808 | } |
1809 | } | |
0af1a450 | 1810 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1811 | |
b89f4321 | 1812 | /** |
e51673aa | 1813 | * vfs_setlease - sets a lease on an open file |
1c7dd2ff JL |
1814 | * @filp: file pointer |
1815 | * @arg: type of lease to obtain | |
1816 | * @lease: file_lock to use when adding a lease | |
1817 | * @priv: private info for lm_setup when adding a lease (may be | |
1818 | * NULL if lm_setup doesn't require it) | |
e51673aa JL |
1819 | * |
1820 | * Call this to establish a lease on the file. The "lease" argument is not | |
1821 | * used for F_UNLCK requests and may be NULL. For commands that set or alter | |
1822 | * an existing lease, the (*lease)->fl_lmops->lm_break operation must be set; | |
1823 | * if not, this function will return -ENOLCK (and generate a scary-looking | |
1824 | * stack trace). | |
1c7dd2ff JL |
1825 | * |
1826 | * The "priv" pointer is passed directly to the lm_setup function as-is. It | |
1827 | * may be NULL if the lm_setup operation doesn't require it. | |
1da177e4 | 1828 | */ |
e6f5c789 JL |
1829 | int |
1830 | vfs_setlease(struct file *filp, long arg, struct file_lock **lease, void **priv) | |
1da177e4 | 1831 | { |
1c7dd2ff | 1832 | if (filp->f_op->setlease) |
f82b4b67 | 1833 | return filp->f_op->setlease(filp, arg, lease, priv); |
1c7dd2ff | 1834 | else |
f82b4b67 | 1835 | return generic_setlease(filp, arg, lease, priv); |
1da177e4 | 1836 | } |
a9933cea | 1837 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1838 | |
0ceaf6c7 | 1839 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) |
1da177e4 | 1840 | { |
1c7dd2ff | 1841 | struct file_lock *fl; |
f7347ce4 | 1842 | struct fasync_struct *new; |
1da177e4 LT |
1843 | int error; |
1844 | ||
c5b1f0d9 AB |
1845 | fl = lease_alloc(filp, arg); |
1846 | if (IS_ERR(fl)) | |
1847 | return PTR_ERR(fl); | |
1da177e4 | 1848 | |
f7347ce4 LT |
1849 | new = fasync_alloc(); |
1850 | if (!new) { | |
1851 | locks_free_lock(fl); | |
1852 | return -ENOMEM; | |
1853 | } | |
1c7dd2ff | 1854 | new->fa_fd = fd; |
f7347ce4 | 1855 | |
1c7dd2ff | 1856 | error = vfs_setlease(filp, arg, &fl, (void **)&new); |
2dfb928f JL |
1857 | if (fl) |
1858 | locks_free_lock(fl); | |
f7347ce4 LT |
1859 | if (new) |
1860 | fasync_free(new); | |
1da177e4 LT |
1861 | return error; |
1862 | } | |
1863 | ||
0ceaf6c7 BF |
1864 | /** |
1865 | * fcntl_setlease - sets a lease on an open file | |
1866 | * @fd: open file descriptor | |
1867 | * @filp: file pointer | |
1868 | * @arg: type of lease to obtain | |
1869 | * | |
1870 | * Call this fcntl to establish a lease on the file. | |
1871 | * Note that you also need to call %F_SETSIG to | |
1872 | * receive a signal when the lease is broken. | |
1873 | */ | |
1874 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1875 | { | |
1876 | if (arg == F_UNLCK) | |
e6f5c789 | 1877 | return vfs_setlease(filp, F_UNLCK, NULL, NULL); |
0ceaf6c7 BF |
1878 | return do_fcntl_add_lease(fd, filp, arg); |
1879 | } | |
1880 | ||
1da177e4 LT |
1881 | /** |
1882 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1883 | * @filp: The file to apply the lock to | |
1884 | * @fl: The lock to be applied | |
1885 | * | |
1886 | * Add a FLOCK style lock to a file. | |
1887 | */ | |
1888 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1889 | { | |
1890 | int error; | |
1891 | might_sleep(); | |
1892 | for (;;) { | |
1893 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1894 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1895 | break; |
1896 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1897 | if (!error) | |
1898 | continue; | |
1899 | ||
1900 | locks_delete_block(fl); | |
1901 | break; | |
1902 | } | |
1903 | return error; | |
1904 | } | |
1905 | ||
1906 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1907 | ||
1908 | /** | |
1909 | * sys_flock: - flock() system call. | |
1910 | * @fd: the file descriptor to lock. | |
1911 | * @cmd: the type of lock to apply. | |
1912 | * | |
1913 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1914 | * The @cmd can be one of | |
1915 | * | |
1916 | * %LOCK_SH -- a shared lock. | |
1917 | * | |
1918 | * %LOCK_EX -- an exclusive lock. | |
1919 | * | |
1920 | * %LOCK_UN -- remove an existing lock. | |
1921 | * | |
1922 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1923 | * | |
1924 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1925 | * processes read and write access respectively. | |
1926 | */ | |
002c8976 | 1927 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1928 | { |
2903ff01 | 1929 | struct fd f = fdget(fd); |
1da177e4 LT |
1930 | struct file_lock *lock; |
1931 | int can_sleep, unlock; | |
1932 | int error; | |
1933 | ||
1934 | error = -EBADF; | |
2903ff01 | 1935 | if (!f.file) |
1da177e4 LT |
1936 | goto out; |
1937 | ||
1938 | can_sleep = !(cmd & LOCK_NB); | |
1939 | cmd &= ~LOCK_NB; | |
1940 | unlock = (cmd == LOCK_UN); | |
1941 | ||
aeb5d727 | 1942 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1943 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1944 | goto out_putf; |
1945 | ||
6e129d00 JL |
1946 | lock = flock_make_lock(f.file, cmd); |
1947 | if (IS_ERR(lock)) { | |
1948 | error = PTR_ERR(lock); | |
1da177e4 | 1949 | goto out_putf; |
6e129d00 JL |
1950 | } |
1951 | ||
1da177e4 LT |
1952 | if (can_sleep) |
1953 | lock->fl_flags |= FL_SLEEP; | |
1954 | ||
2903ff01 | 1955 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1956 | if (error) |
1957 | goto out_free; | |
1958 | ||
72c2d531 | 1959 | if (f.file->f_op->flock) |
2903ff01 | 1960 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1961 | (can_sleep) ? F_SETLKW : F_SETLK, |
1962 | lock); | |
1963 | else | |
2903ff01 | 1964 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1965 | |
1966 | out_free: | |
993dfa87 | 1967 | locks_free_lock(lock); |
1da177e4 LT |
1968 | |
1969 | out_putf: | |
2903ff01 | 1970 | fdput(f); |
1da177e4 LT |
1971 | out: |
1972 | return error; | |
1973 | } | |
1974 | ||
3ee17abd BF |
1975 | /** |
1976 | * vfs_test_lock - test file byte range lock | |
1977 | * @filp: The file to test lock for | |
6924c554 | 1978 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1979 | * |
1980 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1981 | * setting conf->fl_type to something other than F_UNLCK. | |
1982 | */ | |
1983 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1984 | { | |
72c2d531 | 1985 | if (filp->f_op->lock) |
3ee17abd BF |
1986 | return filp->f_op->lock(filp, F_GETLK, fl); |
1987 | posix_test_lock(filp, fl); | |
1988 | return 0; | |
1989 | } | |
1990 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1991 | ||
c2fa1b8a BF |
1992 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1993 | { | |
cff2fce5 | 1994 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1995 | #if BITS_PER_LONG == 32 |
1996 | /* | |
1997 | * Make sure we can represent the posix lock via | |
1998 | * legacy 32bit flock. | |
1999 | */ | |
2000 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
2001 | return -EOVERFLOW; | |
2002 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
2003 | return -EOVERFLOW; | |
2004 | #endif | |
2005 | flock->l_start = fl->fl_start; | |
2006 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2007 | fl->fl_end - fl->fl_start + 1; | |
2008 | flock->l_whence = 0; | |
129a84de | 2009 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
2010 | return 0; |
2011 | } | |
2012 | ||
2013 | #if BITS_PER_LONG == 32 | |
2014 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
2015 | { | |
cff2fce5 | 2016 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
2017 | flock->l_start = fl->fl_start; |
2018 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
2019 | fl->fl_end - fl->fl_start + 1; | |
2020 | flock->l_whence = 0; | |
2021 | flock->l_type = fl->fl_type; | |
2022 | } | |
2023 | #endif | |
2024 | ||
1da177e4 LT |
2025 | /* Report the first existing lock that would conflict with l. |
2026 | * This implements the F_GETLK command of fcntl(). | |
2027 | */ | |
c1e62b8f | 2028 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 2029 | { |
9d6a8c5c | 2030 | struct file_lock file_lock; |
1da177e4 LT |
2031 | struct flock flock; |
2032 | int error; | |
2033 | ||
2034 | error = -EFAULT; | |
2035 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2036 | goto out; | |
2037 | error = -EINVAL; | |
2038 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2039 | goto out; | |
2040 | ||
2041 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
2042 | if (error) | |
2043 | goto out; | |
2044 | ||
0d3f7a2d | 2045 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2046 | error = -EINVAL; |
2047 | if (flock.l_pid != 0) | |
2048 | goto out; | |
2049 | ||
5d50ffd7 | 2050 | cmd = F_GETLK; |
cff2fce5 | 2051 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2052 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2053 | } |
2054 | ||
3ee17abd BF |
2055 | error = vfs_test_lock(filp, &file_lock); |
2056 | if (error) | |
2057 | goto out; | |
1da177e4 | 2058 | |
9d6a8c5c ME |
2059 | flock.l_type = file_lock.fl_type; |
2060 | if (file_lock.fl_type != F_UNLCK) { | |
2061 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 2062 | if (error) |
f328296e | 2063 | goto rel_priv; |
1da177e4 LT |
2064 | } |
2065 | error = -EFAULT; | |
2066 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2067 | error = 0; | |
f328296e KM |
2068 | rel_priv: |
2069 | locks_release_private(&file_lock); | |
1da177e4 LT |
2070 | out: |
2071 | return error; | |
2072 | } | |
2073 | ||
7723ec97 ME |
2074 | /** |
2075 | * vfs_lock_file - file byte range lock | |
2076 | * @filp: The file to apply the lock to | |
2077 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2078 | * @fl: The lock to be applied | |
150b3934 ME |
2079 | * @conf: Place to return a copy of the conflicting lock, if found. |
2080 | * | |
2081 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2082 | * as the final argument. | |
2083 | * | |
2084 | * If the filesystem defines a private ->lock() method, then @conf will | |
2085 | * be left unchanged; so a caller that cares should initialize it to | |
2086 | * some acceptable default. | |
2beb6614 ME |
2087 | * |
2088 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2089 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2090 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2091 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2092 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2093 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2094 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2095 | * request completes. |
2096 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2097 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2098 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2099 | * nonzero return code and the file system should release the lock. The file |
2100 | * system is also responsible to keep a corresponding posix lock when it | |
2101 | * grants a lock so the VFS can find out which locks are locally held and do | |
2102 | * the correct lock cleanup when required. | |
2103 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2104 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2105 | * return code. |
7723ec97 | 2106 | */ |
150b3934 | 2107 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2108 | { |
72c2d531 | 2109 | if (filp->f_op->lock) |
7723ec97 ME |
2110 | return filp->f_op->lock(filp, cmd, fl); |
2111 | else | |
150b3934 | 2112 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2113 | } |
2114 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2115 | ||
b648a6de MS |
2116 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2117 | struct file_lock *fl) | |
2118 | { | |
2119 | int error; | |
2120 | ||
2121 | error = security_file_lock(filp, fl->fl_type); | |
2122 | if (error) | |
2123 | return error; | |
2124 | ||
764c76b3 MS |
2125 | for (;;) { |
2126 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2127 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2128 | break; |
764c76b3 MS |
2129 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2130 | if (!error) | |
2131 | continue; | |
2132 | ||
2133 | locks_delete_block(fl); | |
2134 | break; | |
b648a6de MS |
2135 | } |
2136 | ||
2137 | return error; | |
2138 | } | |
2139 | ||
cf01f4ee JL |
2140 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2141 | static int | |
2142 | check_fmode_for_setlk(struct file_lock *fl) | |
2143 | { | |
2144 | switch (fl->fl_type) { | |
2145 | case F_RDLCK: | |
2146 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2147 | return -EBADF; | |
2148 | break; | |
2149 | case F_WRLCK: | |
2150 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2151 | return -EBADF; | |
2152 | } | |
2153 | return 0; | |
2154 | } | |
2155 | ||
1da177e4 LT |
2156 | /* Apply the lock described by l to an open file descriptor. |
2157 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2158 | */ | |
c293621b PS |
2159 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2160 | struct flock __user *l) | |
1da177e4 LT |
2161 | { |
2162 | struct file_lock *file_lock = locks_alloc_lock(); | |
2163 | struct flock flock; | |
2164 | struct inode *inode; | |
0b2bac2f | 2165 | struct file *f; |
1da177e4 LT |
2166 | int error; |
2167 | ||
2168 | if (file_lock == NULL) | |
2169 | return -ENOLCK; | |
2170 | ||
2171 | /* | |
2172 | * This might block, so we do it before checking the inode. | |
2173 | */ | |
2174 | error = -EFAULT; | |
2175 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2176 | goto out; | |
2177 | ||
496ad9aa | 2178 | inode = file_inode(filp); |
1da177e4 LT |
2179 | |
2180 | /* Don't allow mandatory locks on files that may be memory mapped | |
2181 | * and shared. | |
2182 | */ | |
a16877ca | 2183 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2184 | error = -EAGAIN; |
2185 | goto out; | |
2186 | } | |
2187 | ||
c293621b | 2188 | again: |
1da177e4 LT |
2189 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2190 | if (error) | |
2191 | goto out; | |
5d50ffd7 | 2192 | |
cf01f4ee JL |
2193 | error = check_fmode_for_setlk(file_lock); |
2194 | if (error) | |
2195 | goto out; | |
2196 | ||
5d50ffd7 JL |
2197 | /* |
2198 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2199 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2200 | */ |
2201 | switch (cmd) { | |
0d3f7a2d | 2202 | case F_OFD_SETLK: |
90478939 JL |
2203 | error = -EINVAL; |
2204 | if (flock.l_pid != 0) | |
2205 | goto out; | |
2206 | ||
5d50ffd7 | 2207 | cmd = F_SETLK; |
cff2fce5 | 2208 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2209 | file_lock->fl_owner = filp; |
5d50ffd7 | 2210 | break; |
0d3f7a2d | 2211 | case F_OFD_SETLKW: |
90478939 JL |
2212 | error = -EINVAL; |
2213 | if (flock.l_pid != 0) | |
2214 | goto out; | |
2215 | ||
5d50ffd7 | 2216 | cmd = F_SETLKW; |
cff2fce5 | 2217 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2218 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2219 | /* Fallthrough */ |
2220 | case F_SETLKW: | |
1da177e4 LT |
2221 | file_lock->fl_flags |= FL_SLEEP; |
2222 | } | |
5d50ffd7 | 2223 | |
b648a6de | 2224 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2225 | |
c293621b PS |
2226 | /* |
2227 | * Attempt to detect a close/fcntl race and recover by | |
2228 | * releasing the lock that was just acquired. | |
2229 | */ | |
0b2bac2f AV |
2230 | /* |
2231 | * we need that spin_lock here - it prevents reordering between | |
2232 | * update of inode->i_flock and check for it done in close(). | |
2233 | * rcu_read_lock() wouldn't do. | |
2234 | */ | |
2235 | spin_lock(¤t->files->file_lock); | |
2236 | f = fcheck(fd); | |
2237 | spin_unlock(¤t->files->file_lock); | |
2238 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2239 | flock.l_type = F_UNLCK; |
2240 | goto again; | |
1da177e4 LT |
2241 | } |
2242 | ||
c293621b | 2243 | out: |
1da177e4 LT |
2244 | locks_free_lock(file_lock); |
2245 | return error; | |
2246 | } | |
2247 | ||
2248 | #if BITS_PER_LONG == 32 | |
2249 | /* Report the first existing lock that would conflict with l. | |
2250 | * This implements the F_GETLK command of fcntl(). | |
2251 | */ | |
c1e62b8f | 2252 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2253 | { |
9d6a8c5c | 2254 | struct file_lock file_lock; |
1da177e4 LT |
2255 | struct flock64 flock; |
2256 | int error; | |
2257 | ||
2258 | error = -EFAULT; | |
2259 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2260 | goto out; | |
2261 | error = -EINVAL; | |
2262 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2263 | goto out; | |
2264 | ||
2265 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2266 | if (error) | |
2267 | goto out; | |
2268 | ||
0d3f7a2d | 2269 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2270 | error = -EINVAL; |
2271 | if (flock.l_pid != 0) | |
2272 | goto out; | |
2273 | ||
5d50ffd7 | 2274 | cmd = F_GETLK64; |
cff2fce5 | 2275 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2276 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2277 | } |
2278 | ||
3ee17abd BF |
2279 | error = vfs_test_lock(filp, &file_lock); |
2280 | if (error) | |
2281 | goto out; | |
2282 | ||
9d6a8c5c ME |
2283 | flock.l_type = file_lock.fl_type; |
2284 | if (file_lock.fl_type != F_UNLCK) | |
2285 | posix_lock_to_flock64(&flock, &file_lock); | |
2286 | ||
1da177e4 LT |
2287 | error = -EFAULT; |
2288 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2289 | error = 0; | |
f328296e KM |
2290 | |
2291 | locks_release_private(&file_lock); | |
1da177e4 LT |
2292 | out: |
2293 | return error; | |
2294 | } | |
2295 | ||
2296 | /* Apply the lock described by l to an open file descriptor. | |
2297 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2298 | */ | |
c293621b PS |
2299 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2300 | struct flock64 __user *l) | |
1da177e4 LT |
2301 | { |
2302 | struct file_lock *file_lock = locks_alloc_lock(); | |
2303 | struct flock64 flock; | |
2304 | struct inode *inode; | |
0b2bac2f | 2305 | struct file *f; |
1da177e4 LT |
2306 | int error; |
2307 | ||
2308 | if (file_lock == NULL) | |
2309 | return -ENOLCK; | |
2310 | ||
2311 | /* | |
2312 | * This might block, so we do it before checking the inode. | |
2313 | */ | |
2314 | error = -EFAULT; | |
2315 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2316 | goto out; | |
2317 | ||
496ad9aa | 2318 | inode = file_inode(filp); |
1da177e4 LT |
2319 | |
2320 | /* Don't allow mandatory locks on files that may be memory mapped | |
2321 | * and shared. | |
2322 | */ | |
a16877ca | 2323 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2324 | error = -EAGAIN; |
2325 | goto out; | |
2326 | } | |
2327 | ||
c293621b | 2328 | again: |
1da177e4 LT |
2329 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2330 | if (error) | |
2331 | goto out; | |
5d50ffd7 | 2332 | |
cf01f4ee JL |
2333 | error = check_fmode_for_setlk(file_lock); |
2334 | if (error) | |
2335 | goto out; | |
2336 | ||
5d50ffd7 JL |
2337 | /* |
2338 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2339 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2340 | */ |
2341 | switch (cmd) { | |
0d3f7a2d | 2342 | case F_OFD_SETLK: |
90478939 JL |
2343 | error = -EINVAL; |
2344 | if (flock.l_pid != 0) | |
2345 | goto out; | |
2346 | ||
5d50ffd7 | 2347 | cmd = F_SETLK64; |
cff2fce5 | 2348 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2349 | file_lock->fl_owner = filp; |
5d50ffd7 | 2350 | break; |
0d3f7a2d | 2351 | case F_OFD_SETLKW: |
90478939 JL |
2352 | error = -EINVAL; |
2353 | if (flock.l_pid != 0) | |
2354 | goto out; | |
2355 | ||
5d50ffd7 | 2356 | cmd = F_SETLKW64; |
cff2fce5 | 2357 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2358 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2359 | /* Fallthrough */ |
2360 | case F_SETLKW64: | |
1da177e4 LT |
2361 | file_lock->fl_flags |= FL_SLEEP; |
2362 | } | |
5d50ffd7 | 2363 | |
b648a6de | 2364 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2365 | |
c293621b PS |
2366 | /* |
2367 | * Attempt to detect a close/fcntl race and recover by | |
2368 | * releasing the lock that was just acquired. | |
2369 | */ | |
0b2bac2f AV |
2370 | spin_lock(¤t->files->file_lock); |
2371 | f = fcheck(fd); | |
2372 | spin_unlock(¤t->files->file_lock); | |
2373 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2374 | flock.l_type = F_UNLCK; |
2375 | goto again; | |
1da177e4 LT |
2376 | } |
2377 | ||
2378 | out: | |
2379 | locks_free_lock(file_lock); | |
2380 | return error; | |
2381 | } | |
2382 | #endif /* BITS_PER_LONG == 32 */ | |
2383 | ||
2384 | /* | |
2385 | * This function is called when the file is being removed | |
2386 | * from the task's fd array. POSIX locks belonging to this task | |
2387 | * are deleted at this time. | |
2388 | */ | |
2389 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2390 | { | |
ff7b86b8 | 2391 | struct file_lock lock; |
1da177e4 LT |
2392 | |
2393 | /* | |
2394 | * If there are no locks held on this file, we don't need to call | |
2395 | * posix_lock_file(). Another process could be setting a lock on this | |
2396 | * file at the same time, but we wouldn't remove that lock anyway. | |
2397 | */ | |
496ad9aa | 2398 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2399 | return; |
2400 | ||
2401 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2402 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2403 | lock.fl_start = 0; |
2404 | lock.fl_end = OFFSET_MAX; | |
2405 | lock.fl_owner = owner; | |
2406 | lock.fl_pid = current->tgid; | |
2407 | lock.fl_file = filp; | |
2408 | lock.fl_ops = NULL; | |
2409 | lock.fl_lmops = NULL; | |
2410 | ||
150b3934 | 2411 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2412 | |
1da177e4 LT |
2413 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2414 | lock.fl_ops->fl_release_private(&lock); | |
2415 | } | |
2416 | ||
2417 | EXPORT_SYMBOL(locks_remove_posix); | |
2418 | ||
dd459bb1 JL |
2419 | static void |
2420 | locks_remove_flock(struct file *filp) | |
2421 | { | |
2422 | struct file_lock fl = { | |
2423 | .fl_owner = filp, | |
2424 | .fl_pid = current->tgid, | |
2425 | .fl_file = filp, | |
2426 | .fl_flags = FL_FLOCK, | |
2427 | .fl_type = F_UNLCK, | |
2428 | .fl_end = OFFSET_MAX, | |
2429 | }; | |
5263e31e | 2430 | struct file_lock_context *flctx = file_inode(filp)->i_flctx; |
dd459bb1 | 2431 | |
5263e31e | 2432 | if (!flctx || list_empty(&flctx->flc_flock)) |
dd459bb1 JL |
2433 | return; |
2434 | ||
2435 | if (filp->f_op->flock) | |
2436 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
2437 | else | |
2438 | flock_lock_file(filp, &fl); | |
2439 | ||
2440 | if (fl.fl_ops && fl.fl_ops->fl_release_private) | |
2441 | fl.fl_ops->fl_release_private(&fl); | |
2442 | } | |
2443 | ||
1da177e4 LT |
2444 | /* |
2445 | * This function is called on the last close of an open file. | |
2446 | */ | |
78ed8a13 | 2447 | void locks_remove_file(struct file *filp) |
1da177e4 | 2448 | { |
496ad9aa | 2449 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2450 | struct file_lock *fl; |
2451 | struct file_lock **before; | |
ed9814d8 | 2452 | LIST_HEAD(dispose); |
1da177e4 | 2453 | |
dd459bb1 | 2454 | /* remove any OFD locks */ |
73a8f5f7 | 2455 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2456 | |
dd459bb1 JL |
2457 | /* remove flock locks */ |
2458 | locks_remove_flock(filp); | |
2459 | ||
2460 | if (!inode->i_flock) | |
2461 | return; | |
1da177e4 | 2462 | |
1c8c601a | 2463 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2464 | before = &inode->i_flock; |
2465 | ||
2466 | while ((fl = *before) != NULL) { | |
2467 | if (fl->fl_file == filp) { | |
1da177e4 | 2468 | if (IS_LEASE(fl)) { |
c45198ed | 2469 | lease_modify(before, F_UNLCK, &dispose); |
1da177e4 LT |
2470 | continue; |
2471 | } | |
8c3cac5e JL |
2472 | |
2473 | /* | |
2474 | * There's a leftover lock on the list of a type that | |
2475 | * we didn't expect to see. Most likely a classic | |
2476 | * POSIX lock that ended up not getting released | |
2477 | * properly, or that raced onto the list somehow. Log | |
2478 | * some info about it and then just remove it from | |
2479 | * the list. | |
2480 | */ | |
dd459bb1 | 2481 | WARN(1, "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", |
8c3cac5e JL |
2482 | MAJOR(inode->i_sb->s_dev), |
2483 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2484 | fl->fl_type, fl->fl_flags, | |
2485 | fl->fl_start, fl->fl_end); | |
2486 | ||
ed9814d8 | 2487 | locks_delete_lock(before, &dispose); |
8c3cac5e | 2488 | continue; |
1da177e4 LT |
2489 | } |
2490 | before = &fl->fl_next; | |
2491 | } | |
1c8c601a | 2492 | spin_unlock(&inode->i_lock); |
ed9814d8 | 2493 | locks_dispose_list(&dispose); |
1da177e4 LT |
2494 | } |
2495 | ||
1da177e4 LT |
2496 | /** |
2497 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2498 | * @waiter: the lock which was waiting |
2499 | * | |
2500 | * lockd needs to block waiting for locks. | |
2501 | */ | |
64a318ee | 2502 | int |
f891a29f | 2503 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2504 | { |
64a318ee BF |
2505 | int status = 0; |
2506 | ||
7b2296af | 2507 | spin_lock(&blocked_lock_lock); |
5996a298 | 2508 | if (waiter->fl_next) |
1da177e4 | 2509 | __locks_delete_block(waiter); |
64a318ee BF |
2510 | else |
2511 | status = -ENOENT; | |
7b2296af | 2512 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2513 | return status; |
1da177e4 | 2514 | } |
1da177e4 LT |
2515 | EXPORT_SYMBOL(posix_unblock_lock); |
2516 | ||
9b9d2ab4 ME |
2517 | /** |
2518 | * vfs_cancel_lock - file byte range unblock lock | |
2519 | * @filp: The file to apply the unblock to | |
2520 | * @fl: The lock to be unblocked | |
2521 | * | |
2522 | * Used by lock managers to cancel blocked requests | |
2523 | */ | |
2524 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2525 | { | |
72c2d531 | 2526 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2527 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2528 | return 0; | |
2529 | } | |
2530 | ||
2531 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2532 | ||
7f8ada98 | 2533 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2534 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2535 | #include <linux/seq_file.h> |
2536 | ||
7012b02a JL |
2537 | struct locks_iterator { |
2538 | int li_cpu; | |
2539 | loff_t li_pos; | |
2540 | }; | |
2541 | ||
7f8ada98 | 2542 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2543 | loff_t id, char *pfx) |
1da177e4 LT |
2544 | { |
2545 | struct inode *inode = NULL; | |
ab1f1611 VG |
2546 | unsigned int fl_pid; |
2547 | ||
2548 | if (fl->fl_nspid) | |
6c5f3e7b | 2549 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2550 | else |
2551 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2552 | |
2553 | if (fl->fl_file != NULL) | |
496ad9aa | 2554 | inode = file_inode(fl->fl_file); |
1da177e4 | 2555 | |
99dc8292 | 2556 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2557 | if (IS_POSIX(fl)) { |
c918d42a | 2558 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2559 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2560 | else if (IS_OFDLCK(fl)) |
5315c26a | 2561 | seq_puts(f, "OFDLCK"); |
c918d42a | 2562 | else |
5315c26a | 2563 | seq_puts(f, "POSIX "); |
c918d42a JL |
2564 | |
2565 | seq_printf(f, " %s ", | |
1da177e4 | 2566 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2567 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2568 | } else if (IS_FLOCK(fl)) { |
2569 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2570 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2571 | } else { |
5315c26a | 2572 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2573 | } |
2574 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2575 | if (fl->fl_flags & FL_DELEG) |
2576 | seq_puts(f, "DELEG "); | |
2577 | else | |
2578 | seq_puts(f, "LEASE "); | |
2579 | ||
ab83fa4b | 2580 | if (lease_breaking(fl)) |
5315c26a | 2581 | seq_puts(f, "BREAKING "); |
1da177e4 | 2582 | else if (fl->fl_file) |
5315c26a | 2583 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2584 | else |
5315c26a | 2585 | seq_puts(f, "BREAKER "); |
1da177e4 | 2586 | } else { |
5315c26a | 2587 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2588 | } |
2589 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2590 | seq_printf(f, "%s ", |
1da177e4 LT |
2591 | (fl->fl_type & LOCK_READ) |
2592 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2593 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2594 | } else { | |
7f8ada98 | 2595 | seq_printf(f, "%s ", |
ab83fa4b | 2596 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2597 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2598 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2599 | } |
2600 | if (inode) { | |
2601 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2602 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2603 | inode->i_sb->s_id, inode->i_ino); |
2604 | #else | |
2605 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2606 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2607 | MAJOR(inode->i_sb->s_dev), |
2608 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2609 | #endif | |
2610 | } else { | |
ab1f1611 | 2611 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2612 | } |
2613 | if (IS_POSIX(fl)) { | |
2614 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2615 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2616 | else |
7f8ada98 | 2617 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2618 | } else { |
5315c26a | 2619 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2620 | } |
2621 | } | |
2622 | ||
7f8ada98 | 2623 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2624 | { |
7012b02a | 2625 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2626 | struct file_lock *fl, *bfl; |
1da177e4 | 2627 | |
139ca04e | 2628 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2629 | |
7012b02a | 2630 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2631 | |
7f8ada98 | 2632 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2633 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2634 | |
7f8ada98 PE |
2635 | return 0; |
2636 | } | |
1da177e4 | 2637 | |
7f8ada98 | 2638 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2639 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2640 | { |
7012b02a | 2641 | struct locks_iterator *iter = f->private; |
99dc8292 | 2642 | |
7012b02a JL |
2643 | iter->li_pos = *pos + 1; |
2644 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2645 | spin_lock(&blocked_lock_lock); |
7012b02a | 2646 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2647 | } |
1da177e4 | 2648 | |
7f8ada98 PE |
2649 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2650 | { | |
7012b02a JL |
2651 | struct locks_iterator *iter = f->private; |
2652 | ||
2653 | ++iter->li_pos; | |
2654 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2655 | } |
1da177e4 | 2656 | |
7f8ada98 | 2657 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2658 | __releases(&blocked_lock_lock) |
7f8ada98 | 2659 | { |
7b2296af | 2660 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2661 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2662 | } |
2663 | ||
d8ba7a36 | 2664 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2665 | .start = locks_start, |
2666 | .next = locks_next, | |
2667 | .stop = locks_stop, | |
2668 | .show = locks_show, | |
2669 | }; | |
d8ba7a36 AD |
2670 | |
2671 | static int locks_open(struct inode *inode, struct file *filp) | |
2672 | { | |
7012b02a JL |
2673 | return seq_open_private(filp, &locks_seq_operations, |
2674 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2675 | } |
2676 | ||
2677 | static const struct file_operations proc_locks_operations = { | |
2678 | .open = locks_open, | |
2679 | .read = seq_read, | |
2680 | .llseek = seq_lseek, | |
99dc8292 | 2681 | .release = seq_release_private, |
d8ba7a36 AD |
2682 | }; |
2683 | ||
2684 | static int __init proc_locks_init(void) | |
2685 | { | |
2686 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2687 | return 0; | |
2688 | } | |
2689 | module_init(proc_locks_init); | |
7f8ada98 PE |
2690 | #endif |
2691 | ||
1da177e4 LT |
2692 | static int __init filelock_init(void) |
2693 | { | |
7012b02a JL |
2694 | int i; |
2695 | ||
4a075e39 JL |
2696 | flctx_cache = kmem_cache_create("file_lock_ctx", |
2697 | sizeof(struct file_lock_context), 0, SLAB_PANIC, NULL); | |
2698 | ||
1da177e4 | 2699 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2700 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2701 | ||
7012b02a JL |
2702 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2703 | ||
2704 | for_each_possible_cpu(i) | |
2705 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2706 | ||
1da177e4 LT |
2707 | return 0; |
2708 | } | |
2709 | ||
2710 | core_initcall(filelock_init); |