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 | |
e18b890b | 205 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 206 | |
ee19cc40 | 207 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 208 | { |
139ca04e | 209 | INIT_HLIST_NODE(&fl->fl_link); |
ee19cc40 MS |
210 | INIT_LIST_HEAD(&fl->fl_block); |
211 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
212 | } |
213 | ||
1da177e4 | 214 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 215 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 216 | { |
ee19cc40 | 217 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
218 | |
219 | if (fl) | |
ee19cc40 | 220 | locks_init_lock_heads(fl); |
a51cb91d MS |
221 | |
222 | return fl; | |
1da177e4 | 223 | } |
c5b1f0d9 | 224 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 225 | |
a9e61e25 | 226 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
227 | { |
228 | if (fl->fl_ops) { | |
229 | if (fl->fl_ops->fl_release_private) | |
230 | fl->fl_ops->fl_release_private(fl); | |
231 | fl->fl_ops = NULL; | |
232 | } | |
47831f35 | 233 | |
5c97d7b1 KM |
234 | if (fl->fl_lmops) { |
235 | if (fl->fl_lmops->lm_put_owner) | |
236 | fl->fl_lmops->lm_put_owner(fl); | |
237 | fl->fl_lmops = NULL; | |
238 | } | |
47831f35 | 239 | } |
a9e61e25 | 240 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 241 | |
1da177e4 | 242 | /* Free a lock which is not in use. */ |
05fa3135 | 243 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 244 | { |
5ce29646 MS |
245 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
246 | BUG_ON(!list_empty(&fl->fl_block)); | |
139ca04e | 247 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 248 | |
47831f35 | 249 | locks_release_private(fl); |
1da177e4 LT |
250 | kmem_cache_free(filelock_cache, fl); |
251 | } | |
05fa3135 | 252 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 253 | |
ed9814d8 JL |
254 | static void |
255 | locks_dispose_list(struct list_head *dispose) | |
256 | { | |
257 | struct file_lock *fl; | |
258 | ||
259 | while (!list_empty(dispose)) { | |
260 | fl = list_first_entry(dispose, struct file_lock, fl_block); | |
261 | list_del_init(&fl->fl_block); | |
262 | locks_free_lock(fl); | |
263 | } | |
264 | } | |
265 | ||
1da177e4 LT |
266 | void locks_init_lock(struct file_lock *fl) |
267 | { | |
ee19cc40 MS |
268 | memset(fl, 0, sizeof(struct file_lock)); |
269 | locks_init_lock_heads(fl); | |
1da177e4 LT |
270 | } |
271 | ||
272 | EXPORT_SYMBOL(locks_init_lock); | |
273 | ||
47831f35 TM |
274 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
275 | { | |
276 | if (fl->fl_ops) { | |
277 | if (fl->fl_ops->fl_copy_lock) | |
278 | fl->fl_ops->fl_copy_lock(new, fl); | |
279 | new->fl_ops = fl->fl_ops; | |
280 | } | |
5c97d7b1 KM |
281 | |
282 | if (fl->fl_lmops) { | |
283 | if (fl->fl_lmops->lm_get_owner) | |
284 | fl->fl_lmops->lm_get_owner(new, fl); | |
47831f35 | 285 | new->fl_lmops = fl->fl_lmops; |
5c97d7b1 | 286 | } |
47831f35 TM |
287 | } |
288 | ||
1da177e4 LT |
289 | /* |
290 | * Initialize a new lock from an existing file_lock structure. | |
291 | */ | |
3fe0fff1 | 292 | void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) |
1da177e4 LT |
293 | { |
294 | new->fl_owner = fl->fl_owner; | |
295 | new->fl_pid = fl->fl_pid; | |
0996905f | 296 | new->fl_file = NULL; |
1da177e4 LT |
297 | new->fl_flags = fl->fl_flags; |
298 | new->fl_type = fl->fl_type; | |
299 | new->fl_start = fl->fl_start; | |
300 | new->fl_end = fl->fl_end; | |
0996905f TM |
301 | new->fl_ops = NULL; |
302 | new->fl_lmops = NULL; | |
303 | } | |
3fe0fff1 | 304 | EXPORT_SYMBOL(locks_copy_conflock); |
0996905f TM |
305 | |
306 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
307 | { | |
566709bd JL |
308 | /* "new" must be a freshly-initialized lock */ |
309 | WARN_ON_ONCE(new->fl_ops); | |
0996905f | 310 | |
3fe0fff1 | 311 | locks_copy_conflock(new, fl); |
0996905f | 312 | new->fl_file = fl->fl_file; |
1da177e4 LT |
313 | new->fl_ops = fl->fl_ops; |
314 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
315 | |
316 | locks_copy_private(new, fl); | |
1da177e4 LT |
317 | } |
318 | ||
319 | EXPORT_SYMBOL(locks_copy_lock); | |
320 | ||
321 | static inline int flock_translate_cmd(int cmd) { | |
322 | if (cmd & LOCK_MAND) | |
323 | return cmd & (LOCK_MAND | LOCK_RW); | |
324 | switch (cmd) { | |
325 | case LOCK_SH: | |
326 | return F_RDLCK; | |
327 | case LOCK_EX: | |
328 | return F_WRLCK; | |
329 | case LOCK_UN: | |
330 | return F_UNLCK; | |
331 | } | |
332 | return -EINVAL; | |
333 | } | |
334 | ||
335 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
336 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
337 | unsigned int cmd) | |
338 | { | |
339 | struct file_lock *fl; | |
340 | int type = flock_translate_cmd(cmd); | |
341 | if (type < 0) | |
342 | return type; | |
343 | ||
344 | fl = locks_alloc_lock(); | |
345 | if (fl == NULL) | |
346 | return -ENOMEM; | |
347 | ||
348 | fl->fl_file = filp; | |
73a8f5f7 | 349 | fl->fl_owner = filp; |
1da177e4 LT |
350 | fl->fl_pid = current->tgid; |
351 | fl->fl_flags = FL_FLOCK; | |
352 | fl->fl_type = type; | |
353 | fl->fl_end = OFFSET_MAX; | |
354 | ||
355 | *lock = fl; | |
356 | return 0; | |
357 | } | |
358 | ||
0ec4f431 | 359 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
360 | { |
361 | switch (type) { | |
362 | case F_RDLCK: | |
363 | case F_WRLCK: | |
364 | case F_UNLCK: | |
365 | fl->fl_type = type; | |
366 | break; | |
367 | default: | |
368 | return -EINVAL; | |
369 | } | |
370 | return 0; | |
371 | } | |
372 | ||
ef12e72a BF |
373 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
374 | struct flock64 *l) | |
1da177e4 | 375 | { |
1da177e4 | 376 | switch (l->l_whence) { |
f5579f8c | 377 | case SEEK_SET: |
ef12e72a | 378 | fl->fl_start = 0; |
1da177e4 | 379 | break; |
f5579f8c | 380 | case SEEK_CUR: |
ef12e72a | 381 | fl->fl_start = filp->f_pos; |
1da177e4 | 382 | break; |
f5579f8c | 383 | case SEEK_END: |
ef12e72a | 384 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
385 | break; |
386 | default: | |
387 | return -EINVAL; | |
388 | } | |
ef12e72a BF |
389 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
390 | return -EOVERFLOW; | |
391 | fl->fl_start += l->l_start; | |
392 | if (fl->fl_start < 0) | |
393 | return -EINVAL; | |
1da177e4 LT |
394 | |
395 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
396 | POSIX-2001 defines it. */ | |
4c780a46 | 397 | if (l->l_len > 0) { |
ef12e72a BF |
398 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
399 | return -EOVERFLOW; | |
400 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
401 | ||
4c780a46 | 402 | } else if (l->l_len < 0) { |
ef12e72a | 403 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 404 | return -EINVAL; |
ef12e72a BF |
405 | fl->fl_end = fl->fl_start - 1; |
406 | fl->fl_start += l->l_len; | |
407 | } else | |
408 | fl->fl_end = OFFSET_MAX; | |
409 | ||
1da177e4 LT |
410 | fl->fl_owner = current->files; |
411 | fl->fl_pid = current->tgid; | |
412 | fl->fl_file = filp; | |
413 | fl->fl_flags = FL_POSIX; | |
414 | fl->fl_ops = NULL; | |
415 | fl->fl_lmops = NULL; | |
416 | ||
417 | return assign_type(fl, l->l_type); | |
418 | } | |
419 | ||
ef12e72a BF |
420 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
421 | * style lock. | |
422 | */ | |
423 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
424 | struct flock *l) | |
1da177e4 | 425 | { |
ef12e72a BF |
426 | struct flock64 ll = { |
427 | .l_type = l->l_type, | |
428 | .l_whence = l->l_whence, | |
429 | .l_start = l->l_start, | |
430 | .l_len = l->l_len, | |
431 | }; | |
432 | ||
433 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 434 | } |
1da177e4 LT |
435 | |
436 | /* default lease lock manager operations */ | |
437 | static void lease_break_callback(struct file_lock *fl) | |
438 | { | |
439 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
440 | } | |
441 | ||
7b021967 | 442 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 443 | .lm_break = lease_break_callback, |
8fb47a4f | 444 | .lm_change = lease_modify, |
1da177e4 LT |
445 | }; |
446 | ||
447 | /* | |
448 | * Initialize a lease, use the default lock manager operations | |
449 | */ | |
0ec4f431 | 450 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 451 | { |
75dff55a TM |
452 | if (assign_type(fl, type) != 0) |
453 | return -EINVAL; | |
454 | ||
73a8f5f7 | 455 | fl->fl_owner = current->files; |
1da177e4 LT |
456 | fl->fl_pid = current->tgid; |
457 | ||
458 | fl->fl_file = filp; | |
459 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
460 | fl->fl_start = 0; |
461 | fl->fl_end = OFFSET_MAX; | |
462 | fl->fl_ops = NULL; | |
463 | fl->fl_lmops = &lease_manager_ops; | |
464 | return 0; | |
465 | } | |
466 | ||
467 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 468 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
469 | { |
470 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 471 | int error = -ENOMEM; |
1da177e4 LT |
472 | |
473 | if (fl == NULL) | |
e32b8ee2 | 474 | return ERR_PTR(error); |
1da177e4 LT |
475 | |
476 | error = lease_init(filp, type, fl); | |
75dff55a TM |
477 | if (error) { |
478 | locks_free_lock(fl); | |
e32b8ee2 | 479 | return ERR_PTR(error); |
75dff55a | 480 | } |
e32b8ee2 | 481 | return fl; |
1da177e4 LT |
482 | } |
483 | ||
484 | /* Check if two locks overlap each other. | |
485 | */ | |
486 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
487 | { | |
488 | return ((fl1->fl_end >= fl2->fl_start) && | |
489 | (fl2->fl_end >= fl1->fl_start)); | |
490 | } | |
491 | ||
492 | /* | |
493 | * Check whether two locks have the same owner. | |
494 | */ | |
33443c42 | 495 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 496 | { |
8fb47a4f | 497 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 498 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 499 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
500 | return fl1->fl_owner == fl2->fl_owner; |
501 | } | |
502 | ||
7012b02a | 503 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 504 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 505 | { |
7012b02a JL |
506 | lg_local_lock(&file_lock_lglock); |
507 | fl->fl_link_cpu = smp_processor_id(); | |
508 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
509 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
510 | } |
511 | ||
7012b02a | 512 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 513 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 514 | { |
7012b02a JL |
515 | /* |
516 | * Avoid taking lock if already unhashed. This is safe since this check | |
517 | * is done while holding the i_lock, and new insertions into the list | |
518 | * also require that it be held. | |
519 | */ | |
520 | if (hlist_unhashed(&fl->fl_link)) | |
521 | return; | |
522 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 523 | hlist_del_init(&fl->fl_link); |
7012b02a | 524 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
525 | } |
526 | ||
3999e493 JL |
527 | static unsigned long |
528 | posix_owner_key(struct file_lock *fl) | |
529 | { | |
530 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
531 | return fl->fl_lmops->lm_owner_key(fl); | |
532 | return (unsigned long)fl->fl_owner; | |
533 | } | |
534 | ||
6ca10ed8 | 535 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 536 | { |
3999e493 | 537 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
538 | } |
539 | ||
6ca10ed8 | 540 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 541 | { |
48f74186 | 542 | hash_del(&waiter->fl_link); |
88974691 JL |
543 | } |
544 | ||
1da177e4 LT |
545 | /* Remove waiter from blocker's block list. |
546 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 547 | * |
7b2296af | 548 | * Must be called with blocked_lock_lock held. |
1da177e4 | 549 | */ |
33443c42 | 550 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 551 | { |
88974691 | 552 | locks_delete_global_blocked(waiter); |
1da177e4 | 553 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
554 | waiter->fl_next = NULL; |
555 | } | |
556 | ||
1a9e64a7 | 557 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 558 | { |
7b2296af | 559 | spin_lock(&blocked_lock_lock); |
1da177e4 | 560 | __locks_delete_block(waiter); |
7b2296af | 561 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
562 | } |
563 | ||
564 | /* Insert waiter into blocker's block list. | |
565 | * We use a circular list so that processes can be easily woken up in | |
566 | * the order they blocked. The documentation doesn't require this but | |
567 | * it seems like the reasonable thing to do. | |
1c8c601a | 568 | * |
7b2296af | 569 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 570 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 571 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 572 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 573 | */ |
1c8c601a JL |
574 | static void __locks_insert_block(struct file_lock *blocker, |
575 | struct file_lock *waiter) | |
1da177e4 | 576 | { |
6dc0fe8f | 577 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 578 | waiter->fl_next = blocker; |
88974691 | 579 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 580 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
581 | locks_insert_global_blocked(waiter); |
582 | } | |
583 | ||
584 | /* Must be called with i_lock held. */ | |
585 | static void locks_insert_block(struct file_lock *blocker, | |
586 | struct file_lock *waiter) | |
587 | { | |
7b2296af | 588 | spin_lock(&blocked_lock_lock); |
1c8c601a | 589 | __locks_insert_block(blocker, waiter); |
7b2296af | 590 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
591 | } |
592 | ||
1cb36012 JL |
593 | /* |
594 | * Wake up processes blocked waiting for blocker. | |
595 | * | |
1c8c601a | 596 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
597 | */ |
598 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
599 | { | |
4e8c765d JL |
600 | /* |
601 | * Avoid taking global lock if list is empty. This is safe since new | |
602 | * blocked requests are only added to the list under the i_lock, and | |
603 | * the i_lock is always held here. Note that removal from the fl_block | |
604 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 605 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
606 | */ |
607 | if (list_empty(&blocker->fl_block)) | |
608 | return; | |
609 | ||
7b2296af | 610 | spin_lock(&blocked_lock_lock); |
1da177e4 | 611 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
612 | struct file_lock *waiter; |
613 | ||
614 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
615 | struct file_lock, fl_block); |
616 | __locks_delete_block(waiter); | |
8fb47a4f BF |
617 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
618 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
619 | else |
620 | wake_up(&waiter->fl_wait); | |
621 | } | |
7b2296af | 622 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
623 | } |
624 | ||
625 | /* Insert file lock fl into an inode's lock list at the position indicated | |
626 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
627 | * |
628 | * Must be called with the i_lock held! | |
1da177e4 LT |
629 | */ |
630 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
631 | { | |
ab1f1611 VG |
632 | fl->fl_nspid = get_pid(task_tgid(current)); |
633 | ||
1da177e4 LT |
634 | /* insert into file's list */ |
635 | fl->fl_next = *pos; | |
636 | *pos = fl; | |
88974691 JL |
637 | |
638 | locks_insert_global_locks(fl); | |
1da177e4 LT |
639 | } |
640 | ||
24cbe784 JL |
641 | /** |
642 | * locks_delete_lock - Delete a lock and then free it. | |
643 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
644 | * inode->i_flock list entry | |
645 | * | |
646 | * Unlink a lock from all lists and free the namespace reference, but don't | |
647 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
648 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
649 | * |
650 | * Must be called with the i_lock held! | |
1da177e4 | 651 | */ |
24cbe784 | 652 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
653 | { |
654 | struct file_lock *fl = *thisfl_p; | |
655 | ||
88974691 JL |
656 | locks_delete_global_locks(fl); |
657 | ||
1da177e4 LT |
658 | *thisfl_p = fl->fl_next; |
659 | fl->fl_next = NULL; | |
1da177e4 | 660 | |
ab1f1611 VG |
661 | if (fl->fl_nspid) { |
662 | put_pid(fl->fl_nspid); | |
663 | fl->fl_nspid = NULL; | |
664 | } | |
665 | ||
1da177e4 | 666 | locks_wake_up_blocks(fl); |
24cbe784 JL |
667 | } |
668 | ||
669 | /* | |
670 | * Unlink a lock from all lists and free it. | |
671 | * | |
672 | * Must be called with i_lock held! | |
673 | */ | |
ed9814d8 JL |
674 | static void locks_delete_lock(struct file_lock **thisfl_p, |
675 | struct list_head *dispose) | |
24cbe784 JL |
676 | { |
677 | struct file_lock *fl = *thisfl_p; | |
678 | ||
679 | locks_unlink_lock(thisfl_p); | |
ed9814d8 JL |
680 | if (dispose) |
681 | list_add(&fl->fl_block, dispose); | |
682 | else | |
683 | locks_free_lock(fl); | |
1da177e4 LT |
684 | } |
685 | ||
686 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
687 | * checks for shared/exclusive status of overlapping locks. | |
688 | */ | |
689 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
690 | { | |
691 | if (sys_fl->fl_type == F_WRLCK) | |
692 | return 1; | |
693 | if (caller_fl->fl_type == F_WRLCK) | |
694 | return 1; | |
695 | return 0; | |
696 | } | |
697 | ||
698 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
699 | * checking before calling the locks_conflict(). | |
700 | */ | |
701 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
702 | { | |
703 | /* POSIX locks owned by the same process do not conflict with | |
704 | * each other. | |
705 | */ | |
706 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
707 | return (0); | |
708 | ||
709 | /* Check whether they overlap */ | |
710 | if (!locks_overlap(caller_fl, sys_fl)) | |
711 | return 0; | |
712 | ||
713 | return (locks_conflict(caller_fl, sys_fl)); | |
714 | } | |
715 | ||
716 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
717 | * checking before calling the locks_conflict(). | |
718 | */ | |
719 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
720 | { | |
721 | /* FLOCK locks referring to the same filp do not conflict with | |
722 | * each other. | |
723 | */ | |
724 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
725 | return (0); | |
726 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
727 | return 0; | |
728 | ||
729 | return (locks_conflict(caller_fl, sys_fl)); | |
730 | } | |
731 | ||
6d34ac19 | 732 | void |
9d6a8c5c | 733 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
734 | { |
735 | struct file_lock *cfl; | |
1c8c601a | 736 | struct inode *inode = file_inode(filp); |
1da177e4 | 737 | |
1c8c601a | 738 | spin_lock(&inode->i_lock); |
496ad9aa | 739 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
740 | if (!IS_POSIX(cfl)) |
741 | continue; | |
b842e240 | 742 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
743 | break; |
744 | } | |
ab1f1611 | 745 | if (cfl) { |
3fe0fff1 | 746 | locks_copy_conflock(fl, cfl); |
ab1f1611 | 747 | if (cfl->fl_nspid) |
6c5f3e7b | 748 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 749 | } else |
129a84de | 750 | fl->fl_type = F_UNLCK; |
1c8c601a | 751 | spin_unlock(&inode->i_lock); |
6d34ac19 | 752 | return; |
1da177e4 | 753 | } |
1da177e4 LT |
754 | EXPORT_SYMBOL(posix_test_lock); |
755 | ||
b533184f BF |
756 | /* |
757 | * Deadlock detection: | |
758 | * | |
759 | * We attempt to detect deadlocks that are due purely to posix file | |
760 | * locks. | |
1da177e4 | 761 | * |
b533184f BF |
762 | * We assume that a task can be waiting for at most one lock at a time. |
763 | * So for any acquired lock, the process holding that lock may be | |
764 | * waiting on at most one other lock. That lock in turns may be held by | |
765 | * someone waiting for at most one other lock. Given a requested lock | |
766 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
767 | * follow this chain of waiters to ensure we are not about to create a | |
768 | * cycle. | |
1da177e4 | 769 | * |
b533184f BF |
770 | * Since we do this before we ever put a process to sleep on a lock, we |
771 | * are ensured that there is never a cycle; that is what guarantees that | |
772 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 773 | * |
b533184f BF |
774 | * Note: the above assumption may not be true when handling lock |
775 | * requests from a broken NFS client. It may also fail in the presence | |
776 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 777 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 778 | * |
cff2fce5 | 779 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
780 | * Because the owner is not even nominally tied to a thread of |
781 | * execution, the deadlock detection below can't reasonably work well. Just | |
782 | * skip it for those. | |
783 | * | |
cff2fce5 | 784 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
785 | * locks that just checks for the case where two tasks are attempting to |
786 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 787 | */ |
97855b49 BF |
788 | |
789 | #define MAX_DEADLK_ITERATIONS 10 | |
790 | ||
b533184f BF |
791 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
792 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
793 | { | |
794 | struct file_lock *fl; | |
795 | ||
3999e493 | 796 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
797 | if (posix_same_owner(fl, block_fl)) |
798 | return fl->fl_next; | |
799 | } | |
800 | return NULL; | |
801 | } | |
802 | ||
7b2296af | 803 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 804 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
805 | struct file_lock *block_fl) |
806 | { | |
97855b49 | 807 | int i = 0; |
1da177e4 | 808 | |
57b65325 JL |
809 | /* |
810 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 811 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 812 | */ |
cff2fce5 | 813 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
814 | return 0; |
815 | ||
b533184f BF |
816 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
817 | if (i++ > MAX_DEADLK_ITERATIONS) | |
818 | return 0; | |
819 | if (posix_same_owner(caller_fl, block_fl)) | |
820 | return 1; | |
1da177e4 LT |
821 | } |
822 | return 0; | |
823 | } | |
824 | ||
1da177e4 | 825 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 826 | * after any leases, but before any posix locks. |
f475ae95 TM |
827 | * |
828 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
829 | * whether or not a lock was successfully freed by testing the return | |
830 | * value for -ENOENT. | |
1da177e4 | 831 | */ |
993dfa87 | 832 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 833 | { |
993dfa87 | 834 | struct file_lock *new_fl = NULL; |
1da177e4 | 835 | struct file_lock **before; |
496ad9aa | 836 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
837 | int error = 0; |
838 | int found = 0; | |
ed9814d8 | 839 | LIST_HEAD(dispose); |
1da177e4 | 840 | |
b89f4321 | 841 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 842 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
843 | if (!new_fl) |
844 | return -ENOMEM; | |
84d535ad PE |
845 | } |
846 | ||
1c8c601a | 847 | spin_lock(&inode->i_lock); |
b89f4321 AB |
848 | if (request->fl_flags & FL_ACCESS) |
849 | goto find_conflict; | |
850 | ||
1da177e4 LT |
851 | for_each_lock(inode, before) { |
852 | struct file_lock *fl = *before; | |
853 | if (IS_POSIX(fl)) | |
854 | break; | |
855 | if (IS_LEASE(fl)) | |
856 | continue; | |
857 | if (filp != fl->fl_file) | |
858 | continue; | |
993dfa87 | 859 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
860 | goto out; |
861 | found = 1; | |
ed9814d8 | 862 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
863 | break; |
864 | } | |
1da177e4 | 865 | |
f475ae95 TM |
866 | if (request->fl_type == F_UNLCK) { |
867 | if ((request->fl_flags & FL_EXISTS) && !found) | |
868 | error = -ENOENT; | |
993dfa87 | 869 | goto out; |
f475ae95 | 870 | } |
1da177e4 LT |
871 | |
872 | /* | |
873 | * If a higher-priority process was blocked on the old file lock, | |
874 | * give it the opportunity to lock the file. | |
875 | */ | |
b89f4321 | 876 | if (found) { |
1c8c601a | 877 | spin_unlock(&inode->i_lock); |
def01bc5 | 878 | cond_resched(); |
1c8c601a | 879 | spin_lock(&inode->i_lock); |
b89f4321 | 880 | } |
1da177e4 | 881 | |
f07f18dd | 882 | find_conflict: |
1da177e4 LT |
883 | for_each_lock(inode, before) { |
884 | struct file_lock *fl = *before; | |
885 | if (IS_POSIX(fl)) | |
886 | break; | |
887 | if (IS_LEASE(fl)) | |
888 | continue; | |
993dfa87 | 889 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
890 | continue; |
891 | error = -EAGAIN; | |
bde74e4b MS |
892 | if (!(request->fl_flags & FL_SLEEP)) |
893 | goto out; | |
894 | error = FILE_LOCK_DEFERRED; | |
895 | locks_insert_block(fl, request); | |
1da177e4 LT |
896 | goto out; |
897 | } | |
f07f18dd TM |
898 | if (request->fl_flags & FL_ACCESS) |
899 | goto out; | |
993dfa87 | 900 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 901 | locks_insert_lock(before, new_fl); |
993dfa87 | 902 | new_fl = NULL; |
9cedc194 | 903 | error = 0; |
1da177e4 LT |
904 | |
905 | out: | |
1c8c601a | 906 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
907 | if (new_fl) |
908 | locks_free_lock(new_fl); | |
ed9814d8 | 909 | locks_dispose_list(&dispose); |
1da177e4 LT |
910 | return error; |
911 | } | |
912 | ||
150b3934 | 913 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
914 | { |
915 | struct file_lock *fl; | |
39005d02 MS |
916 | struct file_lock *new_fl = NULL; |
917 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
918 | struct file_lock *left = NULL; |
919 | struct file_lock *right = NULL; | |
920 | struct file_lock **before; | |
b9746ef8 JL |
921 | int error; |
922 | bool added = false; | |
ed9814d8 | 923 | LIST_HEAD(dispose); |
1da177e4 LT |
924 | |
925 | /* | |
926 | * We may need two file_lock structures for this operation, | |
927 | * so we get them in advance to avoid races. | |
39005d02 MS |
928 | * |
929 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 930 | */ |
39005d02 MS |
931 | if (!(request->fl_flags & FL_ACCESS) && |
932 | (request->fl_type != F_UNLCK || | |
933 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
934 | new_fl = locks_alloc_lock(); | |
935 | new_fl2 = locks_alloc_lock(); | |
936 | } | |
1da177e4 | 937 | |
1c8c601a | 938 | spin_lock(&inode->i_lock); |
1cb36012 JL |
939 | /* |
940 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
941 | * there are any, either return error or put the request on the | |
48f74186 | 942 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 943 | */ |
1da177e4 LT |
944 | if (request->fl_type != F_UNLCK) { |
945 | for_each_lock(inode, before) { | |
526985b9 | 946 | fl = *before; |
1da177e4 LT |
947 | if (!IS_POSIX(fl)) |
948 | continue; | |
949 | if (!posix_locks_conflict(request, fl)) | |
950 | continue; | |
5842add2 | 951 | if (conflock) |
3fe0fff1 | 952 | locks_copy_conflock(conflock, fl); |
1da177e4 LT |
953 | error = -EAGAIN; |
954 | if (!(request->fl_flags & FL_SLEEP)) | |
955 | goto out; | |
1c8c601a JL |
956 | /* |
957 | * Deadlock detection and insertion into the blocked | |
958 | * locks list must be done while holding the same lock! | |
959 | */ | |
1da177e4 | 960 | error = -EDEADLK; |
7b2296af | 961 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
962 | if (likely(!posix_locks_deadlock(request, fl))) { |
963 | error = FILE_LOCK_DEFERRED; | |
964 | __locks_insert_block(fl, request); | |
965 | } | |
7b2296af | 966 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
967 | goto out; |
968 | } | |
969 | } | |
970 | ||
971 | /* If we're just looking for a conflict, we're done. */ | |
972 | error = 0; | |
973 | if (request->fl_flags & FL_ACCESS) | |
974 | goto out; | |
975 | ||
1da177e4 | 976 | /* |
1da177e4 LT |
977 | * Find the first old lock with the same owner as the new lock. |
978 | */ | |
979 | ||
980 | before = &inode->i_flock; | |
981 | ||
982 | /* First skip locks owned by other processes. */ | |
983 | while ((fl = *before) && (!IS_POSIX(fl) || | |
984 | !posix_same_owner(request, fl))) { | |
985 | before = &fl->fl_next; | |
986 | } | |
987 | ||
1cb36012 | 988 | /* Process locks with this owner. */ |
1da177e4 LT |
989 | while ((fl = *before) && posix_same_owner(request, fl)) { |
990 | /* Detect adjacent or overlapping regions (if same lock type) | |
991 | */ | |
992 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
993 | /* In all comparisons of start vs end, use |
994 | * "start - 1" rather than "end + 1". If end | |
995 | * is OFFSET_MAX, end + 1 will become negative. | |
996 | */ | |
1da177e4 LT |
997 | if (fl->fl_end < request->fl_start - 1) |
998 | goto next_lock; | |
999 | /* If the next lock in the list has entirely bigger | |
1000 | * addresses than the new one, insert the lock here. | |
1001 | */ | |
449231d6 | 1002 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
1003 | break; |
1004 | ||
1005 | /* If we come here, the new and old lock are of the | |
1006 | * same type and adjacent or overlapping. Make one | |
1007 | * lock yielding from the lower start address of both | |
1008 | * locks to the higher end address. | |
1009 | */ | |
1010 | if (fl->fl_start > request->fl_start) | |
1011 | fl->fl_start = request->fl_start; | |
1012 | else | |
1013 | request->fl_start = fl->fl_start; | |
1014 | if (fl->fl_end < request->fl_end) | |
1015 | fl->fl_end = request->fl_end; | |
1016 | else | |
1017 | request->fl_end = fl->fl_end; | |
1018 | if (added) { | |
ed9814d8 | 1019 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1020 | continue; |
1021 | } | |
1022 | request = fl; | |
b9746ef8 | 1023 | added = true; |
1da177e4 LT |
1024 | } |
1025 | else { | |
1026 | /* Processing for different lock types is a bit | |
1027 | * more complex. | |
1028 | */ | |
1029 | if (fl->fl_end < request->fl_start) | |
1030 | goto next_lock; | |
1031 | if (fl->fl_start > request->fl_end) | |
1032 | break; | |
1033 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1034 | added = true; |
1da177e4 LT |
1035 | if (fl->fl_start < request->fl_start) |
1036 | left = fl; | |
1037 | /* If the next lock in the list has a higher end | |
1038 | * address than the new one, insert the new one here. | |
1039 | */ | |
1040 | if (fl->fl_end > request->fl_end) { | |
1041 | right = fl; | |
1042 | break; | |
1043 | } | |
1044 | if (fl->fl_start >= request->fl_start) { | |
1045 | /* The new lock completely replaces an old | |
1046 | * one (This may happen several times). | |
1047 | */ | |
1048 | if (added) { | |
ed9814d8 | 1049 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1050 | continue; |
1051 | } | |
b84d49f9 JL |
1052 | /* |
1053 | * Replace the old lock with new_fl, and | |
1054 | * remove the old one. It's safe to do the | |
1055 | * insert here since we know that we won't be | |
1056 | * using new_fl later, and that the lock is | |
1057 | * just replacing an existing lock. | |
1da177e4 | 1058 | */ |
b84d49f9 JL |
1059 | error = -ENOLCK; |
1060 | if (!new_fl) | |
1061 | goto out; | |
1062 | locks_copy_lock(new_fl, request); | |
1063 | request = new_fl; | |
1064 | new_fl = NULL; | |
ed9814d8 | 1065 | locks_delete_lock(before, &dispose); |
b84d49f9 | 1066 | locks_insert_lock(before, request); |
b9746ef8 | 1067 | added = true; |
1da177e4 LT |
1068 | } |
1069 | } | |
1070 | /* Go on to next lock. | |
1071 | */ | |
1072 | next_lock: | |
1073 | before = &fl->fl_next; | |
1074 | } | |
1075 | ||
0d9a490a | 1076 | /* |
1cb36012 JL |
1077 | * The above code only modifies existing locks in case of merging or |
1078 | * replacing. If new lock(s) need to be inserted all modifications are | |
1079 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1080 | */ |
1081 | error = -ENOLCK; /* "no luck" */ | |
1082 | if (right && left == right && !new_fl2) | |
1083 | goto out; | |
1084 | ||
1da177e4 LT |
1085 | error = 0; |
1086 | if (!added) { | |
f475ae95 TM |
1087 | if (request->fl_type == F_UNLCK) { |
1088 | if (request->fl_flags & FL_EXISTS) | |
1089 | error = -ENOENT; | |
1da177e4 | 1090 | goto out; |
f475ae95 | 1091 | } |
0d9a490a MS |
1092 | |
1093 | if (!new_fl) { | |
1094 | error = -ENOLCK; | |
1095 | goto out; | |
1096 | } | |
1da177e4 LT |
1097 | locks_copy_lock(new_fl, request); |
1098 | locks_insert_lock(before, new_fl); | |
1099 | new_fl = NULL; | |
1100 | } | |
1101 | if (right) { | |
1102 | if (left == right) { | |
1103 | /* The new lock breaks the old one in two pieces, | |
1104 | * so we have to use the second new lock. | |
1105 | */ | |
1106 | left = new_fl2; | |
1107 | new_fl2 = NULL; | |
1108 | locks_copy_lock(left, right); | |
1109 | locks_insert_lock(before, left); | |
1110 | } | |
1111 | right->fl_start = request->fl_end + 1; | |
1112 | locks_wake_up_blocks(right); | |
1113 | } | |
1114 | if (left) { | |
1115 | left->fl_end = request->fl_start - 1; | |
1116 | locks_wake_up_blocks(left); | |
1117 | } | |
1118 | out: | |
1c8c601a | 1119 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1120 | /* |
1121 | * Free any unused locks. | |
1122 | */ | |
1123 | if (new_fl) | |
1124 | locks_free_lock(new_fl); | |
1125 | if (new_fl2) | |
1126 | locks_free_lock(new_fl2); | |
ed9814d8 | 1127 | locks_dispose_list(&dispose); |
1da177e4 LT |
1128 | return error; |
1129 | } | |
1130 | ||
1131 | /** | |
1132 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1133 | * @filp: The file to apply the lock to | |
1134 | * @fl: The lock to be applied | |
150b3934 | 1135 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1136 | * |
1137 | * Add a POSIX style lock to a file. | |
1138 | * We merge adjacent & overlapping locks whenever possible. | |
1139 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1140 | * |
1141 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1142 | * whether or not a lock was successfully freed by testing the return | |
1143 | * value for -ENOENT. | |
1da177e4 | 1144 | */ |
150b3934 | 1145 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1146 | struct file_lock *conflock) |
1147 | { | |
496ad9aa | 1148 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1149 | } |
150b3934 | 1150 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1151 | |
1152 | /** | |
1153 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1154 | * @filp: The file to apply the lock to | |
1155 | * @fl: The lock to be applied | |
1156 | * | |
1157 | * Add a POSIX style lock to a file. | |
1158 | * We merge adjacent & overlapping locks whenever possible. | |
1159 | * POSIX locks are sorted by owner task, then by starting address | |
1160 | */ | |
1161 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1162 | { | |
1163 | int error; | |
1164 | might_sleep (); | |
1165 | for (;;) { | |
150b3934 | 1166 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1167 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1168 | break; |
1169 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1170 | if (!error) | |
1171 | continue; | |
1172 | ||
1173 | locks_delete_block(fl); | |
1174 | break; | |
1175 | } | |
1176 | return error; | |
1177 | } | |
1178 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1179 | ||
1180 | /** | |
1181 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1182 | * @file: the file to check |
1da177e4 LT |
1183 | * |
1184 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1185 | * This function is called from locks_verify_locked() only. | |
1186 | */ | |
d7a06983 | 1187 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1188 | { |
d7a06983 | 1189 | struct inode *inode = file_inode(file); |
1da177e4 LT |
1190 | struct file_lock *fl; |
1191 | ||
1192 | /* | |
1193 | * Search the lock list for this inode for any POSIX locks. | |
1194 | */ | |
1c8c601a | 1195 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1196 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1197 | if (!IS_POSIX(fl)) | |
1198 | continue; | |
73a8f5f7 CH |
1199 | if (fl->fl_owner != current->files && |
1200 | fl->fl_owner != file) | |
1da177e4 LT |
1201 | break; |
1202 | } | |
1c8c601a | 1203 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1204 | return fl ? -EAGAIN : 0; |
1205 | } | |
1206 | ||
1207 | /** | |
1208 | * locks_mandatory_area - Check for a conflicting lock | |
1209 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1210 | * for shared | |
1211 | * @inode: the file to check | |
1212 | * @filp: how the file was opened (if it was) | |
1213 | * @offset: start of area to check | |
1214 | * @count: length of area to check | |
1215 | * | |
1216 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1217 | * This function is called from rw_verify_area() and | |
1218 | * locks_verify_truncate(). | |
1219 | */ | |
1220 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1221 | struct file *filp, loff_t offset, | |
1222 | size_t count) | |
1223 | { | |
1224 | struct file_lock fl; | |
1225 | int error; | |
29723ade | 1226 | bool sleep = false; |
1da177e4 LT |
1227 | |
1228 | locks_init_lock(&fl); | |
1da177e4 LT |
1229 | fl.fl_pid = current->tgid; |
1230 | fl.fl_file = filp; | |
1231 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1232 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1233 | sleep = true; |
1da177e4 LT |
1234 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1235 | fl.fl_start = offset; | |
1236 | fl.fl_end = offset + count - 1; | |
1237 | ||
1238 | for (;;) { | |
29723ade | 1239 | if (filp) { |
73a8f5f7 | 1240 | fl.fl_owner = filp; |
29723ade JL |
1241 | fl.fl_flags &= ~FL_SLEEP; |
1242 | error = __posix_lock_file(inode, &fl, NULL); | |
1243 | if (!error) | |
1244 | break; | |
1245 | } | |
1246 | ||
1247 | if (sleep) | |
1248 | fl.fl_flags |= FL_SLEEP; | |
1249 | fl.fl_owner = current->files; | |
150b3934 | 1250 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1251 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1252 | break; |
1253 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1254 | if (!error) { | |
1255 | /* | |
1256 | * If we've been sleeping someone might have | |
1257 | * changed the permissions behind our back. | |
1258 | */ | |
a16877ca | 1259 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1260 | continue; |
1261 | } | |
1262 | ||
1263 | locks_delete_block(&fl); | |
1264 | break; | |
1265 | } | |
1266 | ||
1267 | return error; | |
1268 | } | |
1269 | ||
1270 | EXPORT_SYMBOL(locks_mandatory_area); | |
1271 | ||
778fc546 BF |
1272 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1273 | { | |
1274 | switch (arg) { | |
1275 | case F_UNLCK: | |
1276 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1277 | /* fall through: */ | |
1278 | case F_RDLCK: | |
1279 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1280 | } | |
1281 | } | |
1282 | ||
1da177e4 LT |
1283 | /* We already had a lease on this file; just change its type */ |
1284 | int lease_modify(struct file_lock **before, int arg) | |
1285 | { | |
1286 | struct file_lock *fl = *before; | |
1287 | int error = assign_type(fl, arg); | |
1288 | ||
1289 | if (error) | |
1290 | return error; | |
778fc546 | 1291 | lease_clear_pending(fl, arg); |
1da177e4 | 1292 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1293 | if (arg == F_UNLCK) { |
1294 | struct file *filp = fl->fl_file; | |
1295 | ||
1296 | f_delown(filp); | |
1297 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1298 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1299 | if (fl->fl_fasync != NULL) { | |
1300 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1301 | fl->fl_fasync = NULL; | |
1302 | } | |
ed9814d8 | 1303 | locks_delete_lock(before, NULL); |
3b6e2723 | 1304 | } |
1da177e4 LT |
1305 | return 0; |
1306 | } | |
1307 | ||
1308 | EXPORT_SYMBOL(lease_modify); | |
1309 | ||
778fc546 BF |
1310 | static bool past_time(unsigned long then) |
1311 | { | |
1312 | if (!then) | |
1313 | /* 0 is a special value meaning "this never expires": */ | |
1314 | return false; | |
1315 | return time_after(jiffies, then); | |
1316 | } | |
1317 | ||
1da177e4 LT |
1318 | static void time_out_leases(struct inode *inode) |
1319 | { | |
1320 | struct file_lock **before; | |
1321 | struct file_lock *fl; | |
1322 | ||
1323 | before = &inode->i_flock; | |
ab83fa4b | 1324 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1325 | trace_time_out_leases(inode, fl); |
778fc546 BF |
1326 | if (past_time(fl->fl_downgrade_time)) |
1327 | lease_modify(before, F_RDLCK); | |
1328 | if (past_time(fl->fl_break_time)) | |
1329 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1330 | if (fl == *before) /* lease_modify may have freed fl */ |
1331 | before = &fl->fl_next; | |
1332 | } | |
1333 | } | |
1334 | ||
df4e8d2c BF |
1335 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1336 | { | |
1337 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1338 | return false; | |
1339 | return locks_conflict(breaker, lease); | |
1340 | } | |
1341 | ||
1da177e4 LT |
1342 | /** |
1343 | * __break_lease - revoke all outstanding leases on file | |
1344 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1345 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1346 | * break all leases | |
1347 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1348 | * only delegations | |
1da177e4 | 1349 | * |
87250dd2 | 1350 | * break_lease (inlined for speed) has checked there already is at least |
1351 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1352 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1353 | * specified %O_NONBLOCK to your open(). |
1354 | */ | |
df4e8d2c | 1355 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1356 | { |
778fc546 | 1357 | int error = 0; |
1da177e4 LT |
1358 | struct file_lock *new_fl, *flock; |
1359 | struct file_lock *fl; | |
1da177e4 LT |
1360 | unsigned long break_time; |
1361 | int i_have_this_lease = 0; | |
df4e8d2c | 1362 | bool lease_conflict = false; |
8737c930 | 1363 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1364 | |
8737c930 | 1365 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1366 | if (IS_ERR(new_fl)) |
1367 | return PTR_ERR(new_fl); | |
df4e8d2c | 1368 | new_fl->fl_flags = type; |
1da177e4 | 1369 | |
1c8c601a | 1370 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1371 | |
1372 | time_out_leases(inode); | |
1373 | ||
1374 | flock = inode->i_flock; | |
1375 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1376 | goto out; | |
1377 | ||
df4e8d2c BF |
1378 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1379 | if (leases_conflict(fl, new_fl)) { | |
1380 | lease_conflict = true; | |
1381 | if (fl->fl_owner == current->files) | |
1382 | i_have_this_lease = 1; | |
1383 | } | |
1384 | } | |
1385 | if (!lease_conflict) | |
778fc546 BF |
1386 | goto out; |
1387 | ||
1da177e4 LT |
1388 | break_time = 0; |
1389 | if (lease_break_time > 0) { | |
1390 | break_time = jiffies + lease_break_time * HZ; | |
1391 | if (break_time == 0) | |
1392 | break_time++; /* so that 0 means no break time */ | |
1393 | } | |
1394 | ||
1395 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
df4e8d2c BF |
1396 | if (!leases_conflict(fl, new_fl)) |
1397 | continue; | |
778fc546 BF |
1398 | if (want_write) { |
1399 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1400 | continue; | |
1401 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1402 | fl->fl_break_time = break_time; |
778fc546 BF |
1403 | } else { |
1404 | if (lease_breaking(flock)) | |
1405 | continue; | |
1406 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1407 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1408 | } |
778fc546 | 1409 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1410 | } |
1411 | ||
1412 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
62af4f1f | 1413 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1414 | error = -EWOULDBLOCK; |
1415 | goto out; | |
1416 | } | |
1417 | ||
1418 | restart: | |
1419 | break_time = flock->fl_break_time; | |
f1c6bb2c | 1420 | if (break_time != 0) |
1da177e4 | 1421 | break_time -= jiffies; |
f1c6bb2c JL |
1422 | if (break_time == 0) |
1423 | break_time++; | |
4321e01e | 1424 | locks_insert_block(flock, new_fl); |
62af4f1f | 1425 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1426 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1427 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1428 | !new_fl->fl_next, break_time); | |
1c8c601a | 1429 | spin_lock(&inode->i_lock); |
62af4f1f | 1430 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1431 | locks_delete_block(new_fl); |
1da177e4 LT |
1432 | if (error >= 0) { |
1433 | if (error == 0) | |
1434 | time_out_leases(inode); | |
778fc546 BF |
1435 | /* |
1436 | * Wait for the next conflicting lease that has not been | |
1437 | * broken yet | |
1438 | */ | |
1da177e4 LT |
1439 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1440 | flock = flock->fl_next) { | |
df4e8d2c | 1441 | if (leases_conflict(new_fl, flock)) |
1da177e4 LT |
1442 | goto restart; |
1443 | } | |
1444 | error = 0; | |
1445 | } | |
1446 | ||
1447 | out: | |
1c8c601a | 1448 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1449 | locks_free_lock(new_fl); |
1da177e4 LT |
1450 | return error; |
1451 | } | |
1452 | ||
1453 | EXPORT_SYMBOL(__break_lease); | |
1454 | ||
1455 | /** | |
a6b91919 | 1456 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1457 | * @inode: the inode |
1458 | * @time: pointer to a timespec which will contain the last modified time | |
1459 | * | |
1460 | * This is to force NFS clients to flush their caches for files with | |
1461 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1462 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1463 | */ |
1464 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1465 | { | |
1466 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1467 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1468 | *time = current_fs_time(inode->i_sb); |
1469 | else | |
1470 | *time = inode->i_mtime; | |
1471 | } | |
1472 | ||
1473 | EXPORT_SYMBOL(lease_get_mtime); | |
1474 | ||
1475 | /** | |
1476 | * fcntl_getlease - Enquire what lease is currently active | |
1477 | * @filp: the file | |
1478 | * | |
1479 | * The value returned by this function will be one of | |
1480 | * (if no lease break is pending): | |
1481 | * | |
1482 | * %F_RDLCK to indicate a shared lease is held. | |
1483 | * | |
1484 | * %F_WRLCK to indicate an exclusive lease is held. | |
1485 | * | |
1486 | * %F_UNLCK to indicate no lease is held. | |
1487 | * | |
1488 | * (if a lease break is pending): | |
1489 | * | |
1490 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1491 | * changed to a shared lease (or removed). | |
1492 | * | |
1493 | * %F_UNLCK to indicate the lease needs to be removed. | |
1494 | * | |
1495 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1496 | * should be returned to userspace. | |
1497 | */ | |
1498 | int fcntl_getlease(struct file *filp) | |
1499 | { | |
1500 | struct file_lock *fl; | |
1c8c601a | 1501 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1502 | int type = F_UNLCK; |
1503 | ||
1c8c601a | 1504 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1505 | time_out_leases(file_inode(filp)); |
1506 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1507 | fl = fl->fl_next) { |
1508 | if (fl->fl_file == filp) { | |
778fc546 | 1509 | type = target_leasetype(fl); |
1da177e4 LT |
1510 | break; |
1511 | } | |
1512 | } | |
1c8c601a | 1513 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1514 | return type; |
1515 | } | |
1516 | ||
24cbe784 JL |
1517 | /** |
1518 | * check_conflicting_open - see if the given dentry points to a file that has | |
1519 | * an existing open that would conflict with the | |
1520 | * desired lease. | |
1521 | * @dentry: dentry to check | |
1522 | * @arg: type of lease that we're trying to acquire | |
1523 | * | |
1524 | * Check to see if there's an existing open fd on this file that would | |
1525 | * conflict with the lease we're trying to set. | |
1526 | */ | |
1527 | static int | |
1528 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1529 | { | |
1530 | int ret = 0; | |
1531 | struct inode *inode = dentry->d_inode; | |
1532 | ||
1533 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1534 | return -EAGAIN; | |
1535 | ||
1536 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1537 | (atomic_read(&inode->i_count) > 1))) | |
1538 | ret = -EAGAIN; | |
1539 | ||
1540 | return ret; | |
1541 | } | |
1542 | ||
d4f22d19 | 1543 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1544 | { |
7eaae282 | 1545 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1546 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1547 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1548 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1549 | int error; |
1da177e4 | 1550 | |
096657b6 | 1551 | lease = *flp; |
62af4f1f JL |
1552 | trace_generic_add_lease(inode, lease); |
1553 | ||
df4e8d2c BF |
1554 | /* |
1555 | * In the delegation case we need mutual exclusion with | |
1556 | * a number of operations that take the i_mutex. We trylock | |
1557 | * because delegations are an optional optimization, and if | |
1558 | * there's some chance of a conflict--we'd rather not | |
1559 | * bother, maybe that's a sign this just isn't a good file to | |
1560 | * hand out a delegation on. | |
1561 | */ | |
1562 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1563 | return -EAGAIN; | |
1564 | ||
1565 | if (is_deleg && arg == F_WRLCK) { | |
1566 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1567 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1568 | WARN_ON_ONCE(1); |
1569 | return -EINVAL; | |
1570 | } | |
096657b6 | 1571 | |
24cbe784 JL |
1572 | error = check_conflicting_open(dentry, arg); |
1573 | if (error) | |
096657b6 | 1574 | goto out; |
6d5e8b05 | 1575 | |
1da177e4 LT |
1576 | /* |
1577 | * At this point, we know that if there is an exclusive | |
1578 | * lease on this file, then we hold it on this filp | |
1579 | * (otherwise our open of this file would have blocked). | |
1580 | * And if we are trying to acquire an exclusive lease, | |
1581 | * then the file is not open by anyone (including us) | |
1582 | * except for this filp. | |
1583 | */ | |
c1f24ef4 | 1584 | error = -EAGAIN; |
1da177e4 LT |
1585 | for (before = &inode->i_flock; |
1586 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1587 | before = &fl->fl_next) { | |
c1f24ef4 | 1588 | if (fl->fl_file == filp) { |
1da177e4 | 1589 | my_before = before; |
c1f24ef4 BF |
1590 | continue; |
1591 | } | |
1592 | /* | |
1593 | * No exclusive leases if someone else has a lease on | |
1594 | * this file: | |
1595 | */ | |
1596 | if (arg == F_WRLCK) | |
1597 | goto out; | |
1598 | /* | |
1599 | * Modifying our existing lease is OK, but no getting a | |
1600 | * new lease if someone else is opening for write: | |
1601 | */ | |
1602 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1603 | goto out; | |
1da177e4 LT |
1604 | } |
1605 | ||
1da177e4 | 1606 | if (my_before != NULL) { |
8fb47a4f | 1607 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1608 | if (!error) |
1609 | *flp = *my_before; | |
1da177e4 LT |
1610 | goto out; |
1611 | } | |
1612 | ||
1da177e4 LT |
1613 | error = -EINVAL; |
1614 | if (!leases_enable) | |
1615 | goto out; | |
1616 | ||
c5b1f0d9 | 1617 | locks_insert_lock(before, lease); |
24cbe784 JL |
1618 | /* |
1619 | * The check in break_lease() is lockless. It's possible for another | |
1620 | * open to race in after we did the earlier check for a conflicting | |
1621 | * open but before the lease was inserted. Check again for a | |
1622 | * conflicting open and cancel the lease if there is one. | |
1623 | * | |
1624 | * We also add a barrier here to ensure that the insertion of the lock | |
1625 | * precedes these checks. | |
1626 | */ | |
1627 | smp_mb(); | |
1628 | error = check_conflicting_open(dentry, arg); | |
1629 | if (error) | |
f39b913c | 1630 | locks_unlink_lock(before); |
1da177e4 | 1631 | out: |
df4e8d2c BF |
1632 | if (is_deleg) |
1633 | mutex_unlock(&inode->i_mutex); | |
1da177e4 LT |
1634 | return error; |
1635 | } | |
8335ebd9 | 1636 | |
d4f22d19 | 1637 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1638 | { |
1639 | struct file_lock *fl, **before; | |
1640 | struct dentry *dentry = filp->f_path.dentry; | |
1641 | struct inode *inode = dentry->d_inode; | |
1642 | ||
62af4f1f JL |
1643 | trace_generic_delete_lease(inode, *flp); |
1644 | ||
8335ebd9 BF |
1645 | for (before = &inode->i_flock; |
1646 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1647 | before = &fl->fl_next) { | |
1648 | if (fl->fl_file != filp) | |
1649 | continue; | |
1650 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1651 | } | |
1652 | return -EAGAIN; | |
1653 | } | |
1654 | ||
1655 | /** | |
1656 | * generic_setlease - sets a lease on an open file | |
1657 | * @filp: file pointer | |
1658 | * @arg: type of lease to obtain | |
1659 | * @flp: input - file_lock to use, output - file_lock inserted | |
1660 | * | |
1661 | * The (input) flp->fl_lmops->lm_break function is required | |
1662 | * by break_lease(). | |
1663 | * | |
1c8c601a | 1664 | * Called with inode->i_lock held. |
8335ebd9 BF |
1665 | */ |
1666 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1667 | { | |
1668 | struct dentry *dentry = filp->f_path.dentry; | |
1669 | struct inode *inode = dentry->d_inode; | |
1670 | int error; | |
1671 | ||
8e96e3b7 | 1672 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1673 | return -EACCES; |
1674 | if (!S_ISREG(inode->i_mode)) | |
1675 | return -EINVAL; | |
1676 | error = security_file_lock(filp, arg); | |
1677 | if (error) | |
1678 | return error; | |
1679 | ||
1680 | time_out_leases(inode); | |
1681 | ||
1682 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1683 | ||
1684 | switch (arg) { | |
1685 | case F_UNLCK: | |
1686 | return generic_delete_lease(filp, flp); | |
1687 | case F_RDLCK: | |
1688 | case F_WRLCK: | |
1689 | return generic_add_lease(filp, arg, flp); | |
1690 | default: | |
8d657eb3 | 1691 | return -EINVAL; |
8335ebd9 BF |
1692 | } |
1693 | } | |
0af1a450 | 1694 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1695 | |
b89f4321 AB |
1696 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1697 | { | |
72c2d531 | 1698 | if (filp->f_op->setlease) |
b89f4321 AB |
1699 | return filp->f_op->setlease(filp, arg, lease); |
1700 | else | |
1701 | return generic_setlease(filp, arg, lease); | |
1702 | } | |
1703 | ||
1704 | /** | |
a9933cea | 1705 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1706 | * @filp: file pointer |
1707 | * @arg: type of lease to obtain | |
1708 | * @lease: file_lock to use | |
1709 | * | |
1710 | * Call this to establish a lease on the file. | |
8fb47a4f | 1711 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1712 | * break_lease will oops! |
1713 | * | |
1714 | * This will call the filesystem's setlease file method, if | |
1715 | * defined. Note that there is no getlease method; instead, the | |
1716 | * filesystem setlease method should call back to setlease() to | |
1717 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1718 | * find it. Since fcntl_getlease() only reports whether the current | |
1719 | * task holds a lease, a cluster filesystem need only do this for | |
1720 | * leases held by processes on this node. | |
1721 | * | |
1722 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1723 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1724 | * filesystem's open, create, and (on truncate) setattr methods. |
1725 | * | |
1726 | * Warning: the only current setlease methods exist only to disable | |
1727 | * leases in certain cases. More vfs changes may be required to | |
1728 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1729 | */ |
1730 | ||
a9933cea | 1731 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1732 | { |
1c8c601a | 1733 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1734 | int error; |
1735 | ||
1c8c601a | 1736 | spin_lock(&inode->i_lock); |
b89f4321 | 1737 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1738 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1739 | |
1740 | return error; | |
1741 | } | |
a9933cea | 1742 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1743 | |
0ceaf6c7 BF |
1744 | static int do_fcntl_delete_lease(struct file *filp) |
1745 | { | |
1746 | struct file_lock fl, *flp = &fl; | |
1747 | ||
1748 | lease_init(filp, F_UNLCK, flp); | |
1749 | ||
1750 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1751 | } | |
1752 | ||
1753 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1754 | { |
3df057ac | 1755 | struct file_lock *fl, *ret; |
1c8c601a | 1756 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1757 | struct fasync_struct *new; |
1da177e4 LT |
1758 | int error; |
1759 | ||
c5b1f0d9 AB |
1760 | fl = lease_alloc(filp, arg); |
1761 | if (IS_ERR(fl)) | |
1762 | return PTR_ERR(fl); | |
1da177e4 | 1763 | |
f7347ce4 LT |
1764 | new = fasync_alloc(); |
1765 | if (!new) { | |
1766 | locks_free_lock(fl); | |
1767 | return -ENOMEM; | |
1768 | } | |
3df057ac | 1769 | ret = fl; |
1c8c601a | 1770 | spin_lock(&inode->i_lock); |
8896b93f | 1771 | error = __vfs_setlease(filp, arg, &ret); |
2dfb928f JL |
1772 | if (error) |
1773 | goto out_unlock; | |
1774 | if (ret == fl) | |
1775 | fl = NULL; | |
1da177e4 | 1776 | |
f7347ce4 LT |
1777 | /* |
1778 | * fasync_insert_entry() returns the old entry if any. | |
1779 | * If there was no old entry, then it used 'new' and | |
1780 | * inserted it into the fasync list. Clear new so that | |
1781 | * we don't release it here. | |
1782 | */ | |
3df057ac | 1783 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1784 | new = NULL; |
1785 | ||
8896b93f | 1786 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
2dfb928f | 1787 | out_unlock: |
1c8c601a | 1788 | spin_unlock(&inode->i_lock); |
2dfb928f JL |
1789 | if (fl) |
1790 | locks_free_lock(fl); | |
f7347ce4 LT |
1791 | if (new) |
1792 | fasync_free(new); | |
1da177e4 LT |
1793 | return error; |
1794 | } | |
1795 | ||
0ceaf6c7 BF |
1796 | /** |
1797 | * fcntl_setlease - sets a lease on an open file | |
1798 | * @fd: open file descriptor | |
1799 | * @filp: file pointer | |
1800 | * @arg: type of lease to obtain | |
1801 | * | |
1802 | * Call this fcntl to establish a lease on the file. | |
1803 | * Note that you also need to call %F_SETSIG to | |
1804 | * receive a signal when the lease is broken. | |
1805 | */ | |
1806 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1807 | { | |
1808 | if (arg == F_UNLCK) | |
1809 | return do_fcntl_delete_lease(filp); | |
1810 | return do_fcntl_add_lease(fd, filp, arg); | |
1811 | } | |
1812 | ||
1da177e4 LT |
1813 | /** |
1814 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1815 | * @filp: The file to apply the lock to | |
1816 | * @fl: The lock to be applied | |
1817 | * | |
1818 | * Add a FLOCK style lock to a file. | |
1819 | */ | |
1820 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1821 | { | |
1822 | int error; | |
1823 | might_sleep(); | |
1824 | for (;;) { | |
1825 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1826 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1827 | break; |
1828 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1829 | if (!error) | |
1830 | continue; | |
1831 | ||
1832 | locks_delete_block(fl); | |
1833 | break; | |
1834 | } | |
1835 | return error; | |
1836 | } | |
1837 | ||
1838 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1839 | ||
1840 | /** | |
1841 | * sys_flock: - flock() system call. | |
1842 | * @fd: the file descriptor to lock. | |
1843 | * @cmd: the type of lock to apply. | |
1844 | * | |
1845 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1846 | * The @cmd can be one of | |
1847 | * | |
1848 | * %LOCK_SH -- a shared lock. | |
1849 | * | |
1850 | * %LOCK_EX -- an exclusive lock. | |
1851 | * | |
1852 | * %LOCK_UN -- remove an existing lock. | |
1853 | * | |
1854 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1855 | * | |
1856 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1857 | * processes read and write access respectively. | |
1858 | */ | |
002c8976 | 1859 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1860 | { |
2903ff01 | 1861 | struct fd f = fdget(fd); |
1da177e4 LT |
1862 | struct file_lock *lock; |
1863 | int can_sleep, unlock; | |
1864 | int error; | |
1865 | ||
1866 | error = -EBADF; | |
2903ff01 | 1867 | if (!f.file) |
1da177e4 LT |
1868 | goto out; |
1869 | ||
1870 | can_sleep = !(cmd & LOCK_NB); | |
1871 | cmd &= ~LOCK_NB; | |
1872 | unlock = (cmd == LOCK_UN); | |
1873 | ||
aeb5d727 | 1874 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1875 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1876 | goto out_putf; |
1877 | ||
2903ff01 | 1878 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1879 | if (error) |
1880 | goto out_putf; | |
1881 | if (can_sleep) | |
1882 | lock->fl_flags |= FL_SLEEP; | |
1883 | ||
2903ff01 | 1884 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1885 | if (error) |
1886 | goto out_free; | |
1887 | ||
72c2d531 | 1888 | if (f.file->f_op->flock) |
2903ff01 | 1889 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1890 | (can_sleep) ? F_SETLKW : F_SETLK, |
1891 | lock); | |
1892 | else | |
2903ff01 | 1893 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1894 | |
1895 | out_free: | |
993dfa87 | 1896 | locks_free_lock(lock); |
1da177e4 LT |
1897 | |
1898 | out_putf: | |
2903ff01 | 1899 | fdput(f); |
1da177e4 LT |
1900 | out: |
1901 | return error; | |
1902 | } | |
1903 | ||
3ee17abd BF |
1904 | /** |
1905 | * vfs_test_lock - test file byte range lock | |
1906 | * @filp: The file to test lock for | |
6924c554 | 1907 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1908 | * |
1909 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1910 | * setting conf->fl_type to something other than F_UNLCK. | |
1911 | */ | |
1912 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1913 | { | |
72c2d531 | 1914 | if (filp->f_op->lock) |
3ee17abd BF |
1915 | return filp->f_op->lock(filp, F_GETLK, fl); |
1916 | posix_test_lock(filp, fl); | |
1917 | return 0; | |
1918 | } | |
1919 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1920 | ||
c2fa1b8a BF |
1921 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1922 | { | |
cff2fce5 | 1923 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1924 | #if BITS_PER_LONG == 32 |
1925 | /* | |
1926 | * Make sure we can represent the posix lock via | |
1927 | * legacy 32bit flock. | |
1928 | */ | |
1929 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1930 | return -EOVERFLOW; | |
1931 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1932 | return -EOVERFLOW; | |
1933 | #endif | |
1934 | flock->l_start = fl->fl_start; | |
1935 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1936 | fl->fl_end - fl->fl_start + 1; | |
1937 | flock->l_whence = 0; | |
129a84de | 1938 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1939 | return 0; |
1940 | } | |
1941 | ||
1942 | #if BITS_PER_LONG == 32 | |
1943 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1944 | { | |
cff2fce5 | 1945 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1946 | flock->l_start = fl->fl_start; |
1947 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1948 | fl->fl_end - fl->fl_start + 1; | |
1949 | flock->l_whence = 0; | |
1950 | flock->l_type = fl->fl_type; | |
1951 | } | |
1952 | #endif | |
1953 | ||
1da177e4 LT |
1954 | /* Report the first existing lock that would conflict with l. |
1955 | * This implements the F_GETLK command of fcntl(). | |
1956 | */ | |
c1e62b8f | 1957 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 1958 | { |
9d6a8c5c | 1959 | struct file_lock file_lock; |
1da177e4 LT |
1960 | struct flock flock; |
1961 | int error; | |
1962 | ||
1963 | error = -EFAULT; | |
1964 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1965 | goto out; | |
1966 | error = -EINVAL; | |
1967 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1968 | goto out; | |
1969 | ||
1970 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1971 | if (error) | |
1972 | goto out; | |
1973 | ||
0d3f7a2d | 1974 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
1975 | error = -EINVAL; |
1976 | if (flock.l_pid != 0) | |
1977 | goto out; | |
1978 | ||
5d50ffd7 | 1979 | cmd = F_GETLK; |
cff2fce5 | 1980 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 1981 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
1982 | } |
1983 | ||
3ee17abd BF |
1984 | error = vfs_test_lock(filp, &file_lock); |
1985 | if (error) | |
1986 | goto out; | |
1da177e4 | 1987 | |
9d6a8c5c ME |
1988 | flock.l_type = file_lock.fl_type; |
1989 | if (file_lock.fl_type != F_UNLCK) { | |
1990 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1991 | if (error) |
1da177e4 | 1992 | goto out; |
1da177e4 LT |
1993 | } |
1994 | error = -EFAULT; | |
1995 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1996 | error = 0; | |
1997 | out: | |
1998 | return error; | |
1999 | } | |
2000 | ||
7723ec97 ME |
2001 | /** |
2002 | * vfs_lock_file - file byte range lock | |
2003 | * @filp: The file to apply the lock to | |
2004 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
2005 | * @fl: The lock to be applied | |
150b3934 ME |
2006 | * @conf: Place to return a copy of the conflicting lock, if found. |
2007 | * | |
2008 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2009 | * as the final argument. | |
2010 | * | |
2011 | * If the filesystem defines a private ->lock() method, then @conf will | |
2012 | * be left unchanged; so a caller that cares should initialize it to | |
2013 | * some acceptable default. | |
2beb6614 ME |
2014 | * |
2015 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2016 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2017 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2018 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2019 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2020 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2021 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2022 | * request completes. |
2023 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2024 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2025 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2026 | * nonzero return code and the file system should release the lock. The file |
2027 | * system is also responsible to keep a corresponding posix lock when it | |
2028 | * grants a lock so the VFS can find out which locks are locally held and do | |
2029 | * the correct lock cleanup when required. | |
2030 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2031 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2032 | * return code. |
7723ec97 | 2033 | */ |
150b3934 | 2034 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2035 | { |
72c2d531 | 2036 | if (filp->f_op->lock) |
7723ec97 ME |
2037 | return filp->f_op->lock(filp, cmd, fl); |
2038 | else | |
150b3934 | 2039 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2040 | } |
2041 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2042 | ||
b648a6de MS |
2043 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2044 | struct file_lock *fl) | |
2045 | { | |
2046 | int error; | |
2047 | ||
2048 | error = security_file_lock(filp, fl->fl_type); | |
2049 | if (error) | |
2050 | return error; | |
2051 | ||
764c76b3 MS |
2052 | for (;;) { |
2053 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2054 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2055 | break; |
764c76b3 MS |
2056 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2057 | if (!error) | |
2058 | continue; | |
2059 | ||
2060 | locks_delete_block(fl); | |
2061 | break; | |
b648a6de MS |
2062 | } |
2063 | ||
2064 | return error; | |
2065 | } | |
2066 | ||
cf01f4ee JL |
2067 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2068 | static int | |
2069 | check_fmode_for_setlk(struct file_lock *fl) | |
2070 | { | |
2071 | switch (fl->fl_type) { | |
2072 | case F_RDLCK: | |
2073 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2074 | return -EBADF; | |
2075 | break; | |
2076 | case F_WRLCK: | |
2077 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2078 | return -EBADF; | |
2079 | } | |
2080 | return 0; | |
2081 | } | |
2082 | ||
1da177e4 LT |
2083 | /* Apply the lock described by l to an open file descriptor. |
2084 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2085 | */ | |
c293621b PS |
2086 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2087 | struct flock __user *l) | |
1da177e4 LT |
2088 | { |
2089 | struct file_lock *file_lock = locks_alloc_lock(); | |
2090 | struct flock flock; | |
2091 | struct inode *inode; | |
0b2bac2f | 2092 | struct file *f; |
1da177e4 LT |
2093 | int error; |
2094 | ||
2095 | if (file_lock == NULL) | |
2096 | return -ENOLCK; | |
2097 | ||
2098 | /* | |
2099 | * This might block, so we do it before checking the inode. | |
2100 | */ | |
2101 | error = -EFAULT; | |
2102 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2103 | goto out; | |
2104 | ||
496ad9aa | 2105 | inode = file_inode(filp); |
1da177e4 LT |
2106 | |
2107 | /* Don't allow mandatory locks on files that may be memory mapped | |
2108 | * and shared. | |
2109 | */ | |
a16877ca | 2110 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2111 | error = -EAGAIN; |
2112 | goto out; | |
2113 | } | |
2114 | ||
c293621b | 2115 | again: |
1da177e4 LT |
2116 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2117 | if (error) | |
2118 | goto out; | |
5d50ffd7 | 2119 | |
cf01f4ee JL |
2120 | error = check_fmode_for_setlk(file_lock); |
2121 | if (error) | |
2122 | goto out; | |
2123 | ||
5d50ffd7 JL |
2124 | /* |
2125 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2126 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2127 | */ |
2128 | switch (cmd) { | |
0d3f7a2d | 2129 | case F_OFD_SETLK: |
90478939 JL |
2130 | error = -EINVAL; |
2131 | if (flock.l_pid != 0) | |
2132 | goto out; | |
2133 | ||
5d50ffd7 | 2134 | cmd = F_SETLK; |
cff2fce5 | 2135 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2136 | file_lock->fl_owner = filp; |
5d50ffd7 | 2137 | break; |
0d3f7a2d | 2138 | case F_OFD_SETLKW: |
90478939 JL |
2139 | error = -EINVAL; |
2140 | if (flock.l_pid != 0) | |
2141 | goto out; | |
2142 | ||
5d50ffd7 | 2143 | cmd = F_SETLKW; |
cff2fce5 | 2144 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2145 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2146 | /* Fallthrough */ |
2147 | case F_SETLKW: | |
1da177e4 LT |
2148 | file_lock->fl_flags |= FL_SLEEP; |
2149 | } | |
5d50ffd7 | 2150 | |
b648a6de | 2151 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2152 | |
c293621b PS |
2153 | /* |
2154 | * Attempt to detect a close/fcntl race and recover by | |
2155 | * releasing the lock that was just acquired. | |
2156 | */ | |
0b2bac2f AV |
2157 | /* |
2158 | * we need that spin_lock here - it prevents reordering between | |
2159 | * update of inode->i_flock and check for it done in close(). | |
2160 | * rcu_read_lock() wouldn't do. | |
2161 | */ | |
2162 | spin_lock(¤t->files->file_lock); | |
2163 | f = fcheck(fd); | |
2164 | spin_unlock(¤t->files->file_lock); | |
2165 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2166 | flock.l_type = F_UNLCK; |
2167 | goto again; | |
1da177e4 LT |
2168 | } |
2169 | ||
c293621b | 2170 | out: |
1da177e4 LT |
2171 | locks_free_lock(file_lock); |
2172 | return error; | |
2173 | } | |
2174 | ||
2175 | #if BITS_PER_LONG == 32 | |
2176 | /* Report the first existing lock that would conflict with l. | |
2177 | * This implements the F_GETLK command of fcntl(). | |
2178 | */ | |
c1e62b8f | 2179 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2180 | { |
9d6a8c5c | 2181 | struct file_lock file_lock; |
1da177e4 LT |
2182 | struct flock64 flock; |
2183 | int error; | |
2184 | ||
2185 | error = -EFAULT; | |
2186 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2187 | goto out; | |
2188 | error = -EINVAL; | |
2189 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2190 | goto out; | |
2191 | ||
2192 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2193 | if (error) | |
2194 | goto out; | |
2195 | ||
0d3f7a2d | 2196 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2197 | error = -EINVAL; |
2198 | if (flock.l_pid != 0) | |
2199 | goto out; | |
2200 | ||
5d50ffd7 | 2201 | cmd = F_GETLK64; |
cff2fce5 | 2202 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2203 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2204 | } |
2205 | ||
3ee17abd BF |
2206 | error = vfs_test_lock(filp, &file_lock); |
2207 | if (error) | |
2208 | goto out; | |
2209 | ||
9d6a8c5c ME |
2210 | flock.l_type = file_lock.fl_type; |
2211 | if (file_lock.fl_type != F_UNLCK) | |
2212 | posix_lock_to_flock64(&flock, &file_lock); | |
2213 | ||
1da177e4 LT |
2214 | error = -EFAULT; |
2215 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2216 | error = 0; | |
2217 | ||
2218 | out: | |
2219 | return error; | |
2220 | } | |
2221 | ||
2222 | /* Apply the lock described by l to an open file descriptor. | |
2223 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2224 | */ | |
c293621b PS |
2225 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2226 | struct flock64 __user *l) | |
1da177e4 LT |
2227 | { |
2228 | struct file_lock *file_lock = locks_alloc_lock(); | |
2229 | struct flock64 flock; | |
2230 | struct inode *inode; | |
0b2bac2f | 2231 | struct file *f; |
1da177e4 LT |
2232 | int error; |
2233 | ||
2234 | if (file_lock == NULL) | |
2235 | return -ENOLCK; | |
2236 | ||
2237 | /* | |
2238 | * This might block, so we do it before checking the inode. | |
2239 | */ | |
2240 | error = -EFAULT; | |
2241 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2242 | goto out; | |
2243 | ||
496ad9aa | 2244 | inode = file_inode(filp); |
1da177e4 LT |
2245 | |
2246 | /* Don't allow mandatory locks on files that may be memory mapped | |
2247 | * and shared. | |
2248 | */ | |
a16877ca | 2249 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2250 | error = -EAGAIN; |
2251 | goto out; | |
2252 | } | |
2253 | ||
c293621b | 2254 | again: |
1da177e4 LT |
2255 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2256 | if (error) | |
2257 | goto out; | |
5d50ffd7 | 2258 | |
cf01f4ee JL |
2259 | error = check_fmode_for_setlk(file_lock); |
2260 | if (error) | |
2261 | goto out; | |
2262 | ||
5d50ffd7 JL |
2263 | /* |
2264 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2265 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2266 | */ |
2267 | switch (cmd) { | |
0d3f7a2d | 2268 | case F_OFD_SETLK: |
90478939 JL |
2269 | error = -EINVAL; |
2270 | if (flock.l_pid != 0) | |
2271 | goto out; | |
2272 | ||
5d50ffd7 | 2273 | cmd = F_SETLK64; |
cff2fce5 | 2274 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2275 | file_lock->fl_owner = filp; |
5d50ffd7 | 2276 | break; |
0d3f7a2d | 2277 | case F_OFD_SETLKW: |
90478939 JL |
2278 | error = -EINVAL; |
2279 | if (flock.l_pid != 0) | |
2280 | goto out; | |
2281 | ||
5d50ffd7 | 2282 | cmd = F_SETLKW64; |
cff2fce5 | 2283 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2284 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2285 | /* Fallthrough */ |
2286 | case F_SETLKW64: | |
1da177e4 LT |
2287 | file_lock->fl_flags |= FL_SLEEP; |
2288 | } | |
5d50ffd7 | 2289 | |
b648a6de | 2290 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2291 | |
c293621b PS |
2292 | /* |
2293 | * Attempt to detect a close/fcntl race and recover by | |
2294 | * releasing the lock that was just acquired. | |
2295 | */ | |
0b2bac2f AV |
2296 | spin_lock(¤t->files->file_lock); |
2297 | f = fcheck(fd); | |
2298 | spin_unlock(¤t->files->file_lock); | |
2299 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2300 | flock.l_type = F_UNLCK; |
2301 | goto again; | |
1da177e4 LT |
2302 | } |
2303 | ||
2304 | out: | |
2305 | locks_free_lock(file_lock); | |
2306 | return error; | |
2307 | } | |
2308 | #endif /* BITS_PER_LONG == 32 */ | |
2309 | ||
2310 | /* | |
2311 | * This function is called when the file is being removed | |
2312 | * from the task's fd array. POSIX locks belonging to this task | |
2313 | * are deleted at this time. | |
2314 | */ | |
2315 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2316 | { | |
ff7b86b8 | 2317 | struct file_lock lock; |
1da177e4 LT |
2318 | |
2319 | /* | |
2320 | * If there are no locks held on this file, we don't need to call | |
2321 | * posix_lock_file(). Another process could be setting a lock on this | |
2322 | * file at the same time, but we wouldn't remove that lock anyway. | |
2323 | */ | |
496ad9aa | 2324 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2325 | return; |
2326 | ||
2327 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2328 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2329 | lock.fl_start = 0; |
2330 | lock.fl_end = OFFSET_MAX; | |
2331 | lock.fl_owner = owner; | |
2332 | lock.fl_pid = current->tgid; | |
2333 | lock.fl_file = filp; | |
2334 | lock.fl_ops = NULL; | |
2335 | lock.fl_lmops = NULL; | |
2336 | ||
150b3934 | 2337 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2338 | |
1da177e4 LT |
2339 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2340 | lock.fl_ops->fl_release_private(&lock); | |
2341 | } | |
2342 | ||
2343 | EXPORT_SYMBOL(locks_remove_posix); | |
2344 | ||
2345 | /* | |
2346 | * This function is called on the last close of an open file. | |
2347 | */ | |
78ed8a13 | 2348 | void locks_remove_file(struct file *filp) |
1da177e4 | 2349 | { |
496ad9aa | 2350 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2351 | struct file_lock *fl; |
2352 | struct file_lock **before; | |
ed9814d8 | 2353 | LIST_HEAD(dispose); |
1da177e4 LT |
2354 | |
2355 | if (!inode->i_flock) | |
2356 | return; | |
2357 | ||
73a8f5f7 | 2358 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2359 | |
72c2d531 | 2360 | if (filp->f_op->flock) { |
1da177e4 | 2361 | struct file_lock fl = { |
73a8f5f7 | 2362 | .fl_owner = filp, |
1da177e4 LT |
2363 | .fl_pid = current->tgid, |
2364 | .fl_file = filp, | |
2365 | .fl_flags = FL_FLOCK, | |
2366 | .fl_type = F_UNLCK, | |
2367 | .fl_end = OFFSET_MAX, | |
2368 | }; | |
2369 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2370 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2371 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2372 | } |
2373 | ||
1c8c601a | 2374 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2375 | before = &inode->i_flock; |
2376 | ||
2377 | while ((fl = *before) != NULL) { | |
2378 | if (fl->fl_file == filp) { | |
1da177e4 LT |
2379 | if (IS_LEASE(fl)) { |
2380 | lease_modify(before, F_UNLCK); | |
2381 | continue; | |
2382 | } | |
8c3cac5e JL |
2383 | |
2384 | /* | |
2385 | * There's a leftover lock on the list of a type that | |
2386 | * we didn't expect to see. Most likely a classic | |
2387 | * POSIX lock that ended up not getting released | |
2388 | * properly, or that raced onto the list somehow. Log | |
2389 | * some info about it and then just remove it from | |
2390 | * the list. | |
2391 | */ | |
2392 | WARN(!IS_FLOCK(fl), | |
2393 | "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", | |
2394 | MAJOR(inode->i_sb->s_dev), | |
2395 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2396 | fl->fl_type, fl->fl_flags, | |
2397 | fl->fl_start, fl->fl_end); | |
2398 | ||
ed9814d8 | 2399 | locks_delete_lock(before, &dispose); |
8c3cac5e | 2400 | continue; |
1da177e4 LT |
2401 | } |
2402 | before = &fl->fl_next; | |
2403 | } | |
1c8c601a | 2404 | spin_unlock(&inode->i_lock); |
ed9814d8 | 2405 | locks_dispose_list(&dispose); |
1da177e4 LT |
2406 | } |
2407 | ||
1da177e4 LT |
2408 | /** |
2409 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2410 | * @waiter: the lock which was waiting |
2411 | * | |
2412 | * lockd needs to block waiting for locks. | |
2413 | */ | |
64a318ee | 2414 | int |
f891a29f | 2415 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2416 | { |
64a318ee BF |
2417 | int status = 0; |
2418 | ||
7b2296af | 2419 | spin_lock(&blocked_lock_lock); |
5996a298 | 2420 | if (waiter->fl_next) |
1da177e4 | 2421 | __locks_delete_block(waiter); |
64a318ee BF |
2422 | else |
2423 | status = -ENOENT; | |
7b2296af | 2424 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2425 | return status; |
1da177e4 | 2426 | } |
1da177e4 LT |
2427 | EXPORT_SYMBOL(posix_unblock_lock); |
2428 | ||
9b9d2ab4 ME |
2429 | /** |
2430 | * vfs_cancel_lock - file byte range unblock lock | |
2431 | * @filp: The file to apply the unblock to | |
2432 | * @fl: The lock to be unblocked | |
2433 | * | |
2434 | * Used by lock managers to cancel blocked requests | |
2435 | */ | |
2436 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2437 | { | |
72c2d531 | 2438 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2439 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2440 | return 0; | |
2441 | } | |
2442 | ||
2443 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2444 | ||
7f8ada98 | 2445 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2446 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2447 | #include <linux/seq_file.h> |
2448 | ||
7012b02a JL |
2449 | struct locks_iterator { |
2450 | int li_cpu; | |
2451 | loff_t li_pos; | |
2452 | }; | |
2453 | ||
7f8ada98 | 2454 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2455 | loff_t id, char *pfx) |
1da177e4 LT |
2456 | { |
2457 | struct inode *inode = NULL; | |
ab1f1611 VG |
2458 | unsigned int fl_pid; |
2459 | ||
2460 | if (fl->fl_nspid) | |
6c5f3e7b | 2461 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2462 | else |
2463 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2464 | |
2465 | if (fl->fl_file != NULL) | |
496ad9aa | 2466 | inode = file_inode(fl->fl_file); |
1da177e4 | 2467 | |
99dc8292 | 2468 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2469 | if (IS_POSIX(fl)) { |
c918d42a | 2470 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2471 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2472 | else if (IS_OFDLCK(fl)) |
5315c26a | 2473 | seq_puts(f, "OFDLCK"); |
c918d42a | 2474 | else |
5315c26a | 2475 | seq_puts(f, "POSIX "); |
c918d42a JL |
2476 | |
2477 | seq_printf(f, " %s ", | |
1da177e4 | 2478 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2479 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2480 | } else if (IS_FLOCK(fl)) { |
2481 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2482 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2483 | } else { |
5315c26a | 2484 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2485 | } |
2486 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2487 | if (fl->fl_flags & FL_DELEG) |
2488 | seq_puts(f, "DELEG "); | |
2489 | else | |
2490 | seq_puts(f, "LEASE "); | |
2491 | ||
ab83fa4b | 2492 | if (lease_breaking(fl)) |
5315c26a | 2493 | seq_puts(f, "BREAKING "); |
1da177e4 | 2494 | else if (fl->fl_file) |
5315c26a | 2495 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2496 | else |
5315c26a | 2497 | seq_puts(f, "BREAKER "); |
1da177e4 | 2498 | } else { |
5315c26a | 2499 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2500 | } |
2501 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2502 | seq_printf(f, "%s ", |
1da177e4 LT |
2503 | (fl->fl_type & LOCK_READ) |
2504 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2505 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2506 | } else { | |
7f8ada98 | 2507 | seq_printf(f, "%s ", |
ab83fa4b | 2508 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2509 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2510 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2511 | } |
2512 | if (inode) { | |
2513 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2514 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2515 | inode->i_sb->s_id, inode->i_ino); |
2516 | #else | |
2517 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2518 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2519 | MAJOR(inode->i_sb->s_dev), |
2520 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2521 | #endif | |
2522 | } else { | |
ab1f1611 | 2523 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2524 | } |
2525 | if (IS_POSIX(fl)) { | |
2526 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2527 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2528 | else |
7f8ada98 | 2529 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2530 | } else { |
5315c26a | 2531 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2532 | } |
2533 | } | |
2534 | ||
7f8ada98 | 2535 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2536 | { |
7012b02a | 2537 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2538 | struct file_lock *fl, *bfl; |
1da177e4 | 2539 | |
139ca04e | 2540 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2541 | |
7012b02a | 2542 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2543 | |
7f8ada98 | 2544 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2545 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2546 | |
7f8ada98 PE |
2547 | return 0; |
2548 | } | |
1da177e4 | 2549 | |
7f8ada98 | 2550 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2551 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2552 | { |
7012b02a | 2553 | struct locks_iterator *iter = f->private; |
99dc8292 | 2554 | |
7012b02a JL |
2555 | iter->li_pos = *pos + 1; |
2556 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2557 | spin_lock(&blocked_lock_lock); |
7012b02a | 2558 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2559 | } |
1da177e4 | 2560 | |
7f8ada98 PE |
2561 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2562 | { | |
7012b02a JL |
2563 | struct locks_iterator *iter = f->private; |
2564 | ||
2565 | ++iter->li_pos; | |
2566 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2567 | } |
1da177e4 | 2568 | |
7f8ada98 | 2569 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2570 | __releases(&blocked_lock_lock) |
7f8ada98 | 2571 | { |
7b2296af | 2572 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2573 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2574 | } |
2575 | ||
d8ba7a36 | 2576 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2577 | .start = locks_start, |
2578 | .next = locks_next, | |
2579 | .stop = locks_stop, | |
2580 | .show = locks_show, | |
2581 | }; | |
d8ba7a36 AD |
2582 | |
2583 | static int locks_open(struct inode *inode, struct file *filp) | |
2584 | { | |
7012b02a JL |
2585 | return seq_open_private(filp, &locks_seq_operations, |
2586 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2587 | } |
2588 | ||
2589 | static const struct file_operations proc_locks_operations = { | |
2590 | .open = locks_open, | |
2591 | .read = seq_read, | |
2592 | .llseek = seq_lseek, | |
99dc8292 | 2593 | .release = seq_release_private, |
d8ba7a36 AD |
2594 | }; |
2595 | ||
2596 | static int __init proc_locks_init(void) | |
2597 | { | |
2598 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2599 | return 0; | |
2600 | } | |
2601 | module_init(proc_locks_init); | |
7f8ada98 PE |
2602 | #endif |
2603 | ||
1da177e4 LT |
2604 | /** |
2605 | * lock_may_read - checks that the region is free of locks | |
2606 | * @inode: the inode that is being read | |
2607 | * @start: the first byte to read | |
2608 | * @len: the number of bytes to read | |
2609 | * | |
2610 | * Emulates Windows locking requirements. Whole-file | |
2611 | * mandatory locks (share modes) can prohibit a read and | |
2612 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2613 | * | |
2614 | * N.B. this function is only ever called | |
2615 | * from knfsd and ownership of locks is never checked. | |
2616 | */ | |
2617 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2618 | { | |
2619 | struct file_lock *fl; | |
2620 | int result = 1; | |
1c8c601a JL |
2621 | |
2622 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2623 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2624 | if (IS_POSIX(fl)) { | |
2625 | if (fl->fl_type == F_RDLCK) | |
2626 | continue; | |
2627 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2628 | continue; | |
2629 | } else if (IS_FLOCK(fl)) { | |
2630 | if (!(fl->fl_type & LOCK_MAND)) | |
2631 | continue; | |
2632 | if (fl->fl_type & LOCK_READ) | |
2633 | continue; | |
2634 | } else | |
2635 | continue; | |
2636 | result = 0; | |
2637 | break; | |
2638 | } | |
1c8c601a | 2639 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2640 | return result; |
2641 | } | |
2642 | ||
2643 | EXPORT_SYMBOL(lock_may_read); | |
2644 | ||
2645 | /** | |
2646 | * lock_may_write - checks that the region is free of locks | |
2647 | * @inode: the inode that is being written | |
2648 | * @start: the first byte to write | |
2649 | * @len: the number of bytes to write | |
2650 | * | |
2651 | * Emulates Windows locking requirements. Whole-file | |
2652 | * mandatory locks (share modes) can prohibit a write and | |
2653 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2654 | * | |
2655 | * N.B. this function is only ever called | |
2656 | * from knfsd and ownership of locks is never checked. | |
2657 | */ | |
2658 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2659 | { | |
2660 | struct file_lock *fl; | |
2661 | int result = 1; | |
1c8c601a JL |
2662 | |
2663 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2664 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2665 | if (IS_POSIX(fl)) { | |
2666 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2667 | continue; | |
2668 | } else if (IS_FLOCK(fl)) { | |
2669 | if (!(fl->fl_type & LOCK_MAND)) | |
2670 | continue; | |
2671 | if (fl->fl_type & LOCK_WRITE) | |
2672 | continue; | |
2673 | } else | |
2674 | continue; | |
2675 | result = 0; | |
2676 | break; | |
2677 | } | |
1c8c601a | 2678 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2679 | return result; |
2680 | } | |
2681 | ||
2682 | EXPORT_SYMBOL(lock_may_write); | |
2683 | ||
1da177e4 LT |
2684 | static int __init filelock_init(void) |
2685 | { | |
7012b02a JL |
2686 | int i; |
2687 | ||
1da177e4 | 2688 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2689 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2690 | ||
7012b02a JL |
2691 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2692 | ||
2693 | for_each_possible_cpu(i) | |
2694 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2695 | ||
1da177e4 LT |
2696 | return 0; |
2697 | } | |
2698 | ||
2699 | core_initcall(filelock_init); |