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 | } | |
068535f1 | 233 | fl->fl_lmops = NULL; |
47831f35 TM |
234 | |
235 | } | |
a9e61e25 | 236 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 237 | |
1da177e4 | 238 | /* Free a lock which is not in use. */ |
05fa3135 | 239 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 240 | { |
5ce29646 MS |
241 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
242 | BUG_ON(!list_empty(&fl->fl_block)); | |
139ca04e | 243 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 244 | |
47831f35 | 245 | locks_release_private(fl); |
1da177e4 LT |
246 | kmem_cache_free(filelock_cache, fl); |
247 | } | |
05fa3135 | 248 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 | 249 | |
ed9814d8 JL |
250 | static void |
251 | locks_dispose_list(struct list_head *dispose) | |
252 | { | |
253 | struct file_lock *fl; | |
254 | ||
255 | while (!list_empty(dispose)) { | |
256 | fl = list_first_entry(dispose, struct file_lock, fl_block); | |
257 | list_del_init(&fl->fl_block); | |
258 | locks_free_lock(fl); | |
259 | } | |
260 | } | |
261 | ||
1da177e4 LT |
262 | void locks_init_lock(struct file_lock *fl) |
263 | { | |
ee19cc40 MS |
264 | memset(fl, 0, sizeof(struct file_lock)); |
265 | locks_init_lock_heads(fl); | |
1da177e4 LT |
266 | } |
267 | ||
268 | EXPORT_SYMBOL(locks_init_lock); | |
269 | ||
47831f35 TM |
270 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
271 | { | |
272 | if (fl->fl_ops) { | |
273 | if (fl->fl_ops->fl_copy_lock) | |
274 | fl->fl_ops->fl_copy_lock(new, fl); | |
275 | new->fl_ops = fl->fl_ops; | |
276 | } | |
bb8430a2 | 277 | if (fl->fl_lmops) |
47831f35 | 278 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
279 | } |
280 | ||
1da177e4 LT |
281 | /* |
282 | * Initialize a new lock from an existing file_lock structure. | |
283 | */ | |
1a747ee0 | 284 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
285 | { |
286 | new->fl_owner = fl->fl_owner; | |
287 | new->fl_pid = fl->fl_pid; | |
0996905f | 288 | new->fl_file = NULL; |
1da177e4 LT |
289 | new->fl_flags = fl->fl_flags; |
290 | new->fl_type = fl->fl_type; | |
291 | new->fl_start = fl->fl_start; | |
292 | new->fl_end = fl->fl_end; | |
0996905f TM |
293 | new->fl_ops = NULL; |
294 | new->fl_lmops = NULL; | |
295 | } | |
3dd7b71c | 296 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
297 | |
298 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
299 | { | |
566709bd JL |
300 | /* "new" must be a freshly-initialized lock */ |
301 | WARN_ON_ONCE(new->fl_ops); | |
0996905f TM |
302 | |
303 | __locks_copy_lock(new, fl); | |
304 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
305 | new->fl_ops = fl->fl_ops; |
306 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
307 | |
308 | locks_copy_private(new, fl); | |
1da177e4 LT |
309 | } |
310 | ||
311 | EXPORT_SYMBOL(locks_copy_lock); | |
312 | ||
313 | static inline int flock_translate_cmd(int cmd) { | |
314 | if (cmd & LOCK_MAND) | |
315 | return cmd & (LOCK_MAND | LOCK_RW); | |
316 | switch (cmd) { | |
317 | case LOCK_SH: | |
318 | return F_RDLCK; | |
319 | case LOCK_EX: | |
320 | return F_WRLCK; | |
321 | case LOCK_UN: | |
322 | return F_UNLCK; | |
323 | } | |
324 | return -EINVAL; | |
325 | } | |
326 | ||
327 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
328 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
329 | unsigned int cmd) | |
330 | { | |
331 | struct file_lock *fl; | |
332 | int type = flock_translate_cmd(cmd); | |
333 | if (type < 0) | |
334 | return type; | |
335 | ||
336 | fl = locks_alloc_lock(); | |
337 | if (fl == NULL) | |
338 | return -ENOMEM; | |
339 | ||
340 | fl->fl_file = filp; | |
73a8f5f7 | 341 | fl->fl_owner = filp; |
1da177e4 LT |
342 | fl->fl_pid = current->tgid; |
343 | fl->fl_flags = FL_FLOCK; | |
344 | fl->fl_type = type; | |
345 | fl->fl_end = OFFSET_MAX; | |
346 | ||
347 | *lock = fl; | |
348 | return 0; | |
349 | } | |
350 | ||
0ec4f431 | 351 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
352 | { |
353 | switch (type) { | |
354 | case F_RDLCK: | |
355 | case F_WRLCK: | |
356 | case F_UNLCK: | |
357 | fl->fl_type = type; | |
358 | break; | |
359 | default: | |
360 | return -EINVAL; | |
361 | } | |
362 | return 0; | |
363 | } | |
364 | ||
ef12e72a BF |
365 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
366 | struct flock64 *l) | |
1da177e4 | 367 | { |
1da177e4 | 368 | switch (l->l_whence) { |
f5579f8c | 369 | case SEEK_SET: |
ef12e72a | 370 | fl->fl_start = 0; |
1da177e4 | 371 | break; |
f5579f8c | 372 | case SEEK_CUR: |
ef12e72a | 373 | fl->fl_start = filp->f_pos; |
1da177e4 | 374 | break; |
f5579f8c | 375 | case SEEK_END: |
ef12e72a | 376 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
377 | break; |
378 | default: | |
379 | return -EINVAL; | |
380 | } | |
ef12e72a BF |
381 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
382 | return -EOVERFLOW; | |
383 | fl->fl_start += l->l_start; | |
384 | if (fl->fl_start < 0) | |
385 | return -EINVAL; | |
1da177e4 LT |
386 | |
387 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
388 | POSIX-2001 defines it. */ | |
4c780a46 | 389 | if (l->l_len > 0) { |
ef12e72a BF |
390 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
391 | return -EOVERFLOW; | |
392 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
393 | ||
4c780a46 | 394 | } else if (l->l_len < 0) { |
ef12e72a | 395 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 396 | return -EINVAL; |
ef12e72a BF |
397 | fl->fl_end = fl->fl_start - 1; |
398 | fl->fl_start += l->l_len; | |
399 | } else | |
400 | fl->fl_end = OFFSET_MAX; | |
401 | ||
1da177e4 LT |
402 | fl->fl_owner = current->files; |
403 | fl->fl_pid = current->tgid; | |
404 | fl->fl_file = filp; | |
405 | fl->fl_flags = FL_POSIX; | |
406 | fl->fl_ops = NULL; | |
407 | fl->fl_lmops = NULL; | |
408 | ||
409 | return assign_type(fl, l->l_type); | |
410 | } | |
411 | ||
ef12e72a BF |
412 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
413 | * style lock. | |
414 | */ | |
415 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
416 | struct flock *l) | |
1da177e4 | 417 | { |
ef12e72a BF |
418 | struct flock64 ll = { |
419 | .l_type = l->l_type, | |
420 | .l_whence = l->l_whence, | |
421 | .l_start = l->l_start, | |
422 | .l_len = l->l_len, | |
423 | }; | |
424 | ||
425 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 426 | } |
1da177e4 LT |
427 | |
428 | /* default lease lock manager operations */ | |
429 | static void lease_break_callback(struct file_lock *fl) | |
430 | { | |
431 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
432 | } | |
433 | ||
7b021967 | 434 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 435 | .lm_break = lease_break_callback, |
8fb47a4f | 436 | .lm_change = lease_modify, |
1da177e4 LT |
437 | }; |
438 | ||
439 | /* | |
440 | * Initialize a lease, use the default lock manager operations | |
441 | */ | |
0ec4f431 | 442 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 443 | { |
75dff55a TM |
444 | if (assign_type(fl, type) != 0) |
445 | return -EINVAL; | |
446 | ||
73a8f5f7 | 447 | fl->fl_owner = current->files; |
1da177e4 LT |
448 | fl->fl_pid = current->tgid; |
449 | ||
450 | fl->fl_file = filp; | |
451 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
452 | fl->fl_start = 0; |
453 | fl->fl_end = OFFSET_MAX; | |
454 | fl->fl_ops = NULL; | |
455 | fl->fl_lmops = &lease_manager_ops; | |
456 | return 0; | |
457 | } | |
458 | ||
459 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 460 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
461 | { |
462 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 463 | int error = -ENOMEM; |
1da177e4 LT |
464 | |
465 | if (fl == NULL) | |
e32b8ee2 | 466 | return ERR_PTR(error); |
1da177e4 LT |
467 | |
468 | error = lease_init(filp, type, fl); | |
75dff55a TM |
469 | if (error) { |
470 | locks_free_lock(fl); | |
e32b8ee2 | 471 | return ERR_PTR(error); |
75dff55a | 472 | } |
e32b8ee2 | 473 | return fl; |
1da177e4 LT |
474 | } |
475 | ||
476 | /* Check if two locks overlap each other. | |
477 | */ | |
478 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
479 | { | |
480 | return ((fl1->fl_end >= fl2->fl_start) && | |
481 | (fl2->fl_end >= fl1->fl_start)); | |
482 | } | |
483 | ||
484 | /* | |
485 | * Check whether two locks have the same owner. | |
486 | */ | |
33443c42 | 487 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 488 | { |
8fb47a4f | 489 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 490 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 491 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
492 | return fl1->fl_owner == fl2->fl_owner; |
493 | } | |
494 | ||
7012b02a | 495 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 496 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 497 | { |
7012b02a JL |
498 | lg_local_lock(&file_lock_lglock); |
499 | fl->fl_link_cpu = smp_processor_id(); | |
500 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
501 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
502 | } |
503 | ||
7012b02a | 504 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 505 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 506 | { |
7012b02a JL |
507 | /* |
508 | * Avoid taking lock if already unhashed. This is safe since this check | |
509 | * is done while holding the i_lock, and new insertions into the list | |
510 | * also require that it be held. | |
511 | */ | |
512 | if (hlist_unhashed(&fl->fl_link)) | |
513 | return; | |
514 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 515 | hlist_del_init(&fl->fl_link); |
7012b02a | 516 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
517 | } |
518 | ||
3999e493 JL |
519 | static unsigned long |
520 | posix_owner_key(struct file_lock *fl) | |
521 | { | |
522 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
523 | return fl->fl_lmops->lm_owner_key(fl); | |
524 | return (unsigned long)fl->fl_owner; | |
525 | } | |
526 | ||
6ca10ed8 | 527 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 528 | { |
3999e493 | 529 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
530 | } |
531 | ||
6ca10ed8 | 532 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 533 | { |
48f74186 | 534 | hash_del(&waiter->fl_link); |
88974691 JL |
535 | } |
536 | ||
1da177e4 LT |
537 | /* Remove waiter from blocker's block list. |
538 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 539 | * |
7b2296af | 540 | * Must be called with blocked_lock_lock held. |
1da177e4 | 541 | */ |
33443c42 | 542 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 543 | { |
88974691 | 544 | locks_delete_global_blocked(waiter); |
1da177e4 | 545 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
546 | waiter->fl_next = NULL; |
547 | } | |
548 | ||
1a9e64a7 | 549 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 550 | { |
7b2296af | 551 | spin_lock(&blocked_lock_lock); |
1da177e4 | 552 | __locks_delete_block(waiter); |
7b2296af | 553 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
554 | } |
555 | ||
556 | /* Insert waiter into blocker's block list. | |
557 | * We use a circular list so that processes can be easily woken up in | |
558 | * the order they blocked. The documentation doesn't require this but | |
559 | * it seems like the reasonable thing to do. | |
1c8c601a | 560 | * |
7b2296af | 561 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 562 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 563 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 564 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 565 | */ |
1c8c601a JL |
566 | static void __locks_insert_block(struct file_lock *blocker, |
567 | struct file_lock *waiter) | |
1da177e4 | 568 | { |
6dc0fe8f | 569 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 570 | waiter->fl_next = blocker; |
88974691 | 571 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 572 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
573 | locks_insert_global_blocked(waiter); |
574 | } | |
575 | ||
576 | /* Must be called with i_lock held. */ | |
577 | static void locks_insert_block(struct file_lock *blocker, | |
578 | struct file_lock *waiter) | |
579 | { | |
7b2296af | 580 | spin_lock(&blocked_lock_lock); |
1c8c601a | 581 | __locks_insert_block(blocker, waiter); |
7b2296af | 582 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
583 | } |
584 | ||
1cb36012 JL |
585 | /* |
586 | * Wake up processes blocked waiting for blocker. | |
587 | * | |
1c8c601a | 588 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
589 | */ |
590 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
591 | { | |
4e8c765d JL |
592 | /* |
593 | * Avoid taking global lock if list is empty. This is safe since new | |
594 | * blocked requests are only added to the list under the i_lock, and | |
595 | * the i_lock is always held here. Note that removal from the fl_block | |
596 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 597 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
598 | */ |
599 | if (list_empty(&blocker->fl_block)) | |
600 | return; | |
601 | ||
7b2296af | 602 | spin_lock(&blocked_lock_lock); |
1da177e4 | 603 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
604 | struct file_lock *waiter; |
605 | ||
606 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
607 | struct file_lock, fl_block); |
608 | __locks_delete_block(waiter); | |
8fb47a4f BF |
609 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
610 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
611 | else |
612 | wake_up(&waiter->fl_wait); | |
613 | } | |
7b2296af | 614 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
615 | } |
616 | ||
617 | /* Insert file lock fl into an inode's lock list at the position indicated | |
618 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
619 | * |
620 | * Must be called with the i_lock held! | |
1da177e4 LT |
621 | */ |
622 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
623 | { | |
ab1f1611 VG |
624 | fl->fl_nspid = get_pid(task_tgid(current)); |
625 | ||
1da177e4 LT |
626 | /* insert into file's list */ |
627 | fl->fl_next = *pos; | |
628 | *pos = fl; | |
88974691 JL |
629 | |
630 | locks_insert_global_locks(fl); | |
1da177e4 LT |
631 | } |
632 | ||
24cbe784 JL |
633 | /** |
634 | * locks_delete_lock - Delete a lock and then free it. | |
635 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
636 | * inode->i_flock list entry | |
637 | * | |
638 | * Unlink a lock from all lists and free the namespace reference, but don't | |
639 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
640 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
641 | * |
642 | * Must be called with the i_lock held! | |
1da177e4 | 643 | */ |
24cbe784 | 644 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
645 | { |
646 | struct file_lock *fl = *thisfl_p; | |
647 | ||
88974691 JL |
648 | locks_delete_global_locks(fl); |
649 | ||
1da177e4 LT |
650 | *thisfl_p = fl->fl_next; |
651 | fl->fl_next = NULL; | |
1da177e4 | 652 | |
ab1f1611 VG |
653 | if (fl->fl_nspid) { |
654 | put_pid(fl->fl_nspid); | |
655 | fl->fl_nspid = NULL; | |
656 | } | |
657 | ||
1da177e4 | 658 | locks_wake_up_blocks(fl); |
24cbe784 JL |
659 | } |
660 | ||
661 | /* | |
662 | * Unlink a lock from all lists and free it. | |
663 | * | |
664 | * Must be called with i_lock held! | |
665 | */ | |
ed9814d8 JL |
666 | static void locks_delete_lock(struct file_lock **thisfl_p, |
667 | struct list_head *dispose) | |
24cbe784 JL |
668 | { |
669 | struct file_lock *fl = *thisfl_p; | |
670 | ||
671 | locks_unlink_lock(thisfl_p); | |
ed9814d8 JL |
672 | if (dispose) |
673 | list_add(&fl->fl_block, dispose); | |
674 | else | |
675 | locks_free_lock(fl); | |
1da177e4 LT |
676 | } |
677 | ||
678 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
679 | * checks for shared/exclusive status of overlapping locks. | |
680 | */ | |
681 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
682 | { | |
683 | if (sys_fl->fl_type == F_WRLCK) | |
684 | return 1; | |
685 | if (caller_fl->fl_type == F_WRLCK) | |
686 | return 1; | |
687 | return 0; | |
688 | } | |
689 | ||
690 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
691 | * checking before calling the locks_conflict(). | |
692 | */ | |
693 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
694 | { | |
695 | /* POSIX locks owned by the same process do not conflict with | |
696 | * each other. | |
697 | */ | |
698 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
699 | return (0); | |
700 | ||
701 | /* Check whether they overlap */ | |
702 | if (!locks_overlap(caller_fl, sys_fl)) | |
703 | return 0; | |
704 | ||
705 | return (locks_conflict(caller_fl, sys_fl)); | |
706 | } | |
707 | ||
708 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
709 | * checking before calling the locks_conflict(). | |
710 | */ | |
711 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
712 | { | |
713 | /* FLOCK locks referring to the same filp do not conflict with | |
714 | * each other. | |
715 | */ | |
716 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
717 | return (0); | |
718 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
719 | return 0; | |
720 | ||
721 | return (locks_conflict(caller_fl, sys_fl)); | |
722 | } | |
723 | ||
6d34ac19 | 724 | void |
9d6a8c5c | 725 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
726 | { |
727 | struct file_lock *cfl; | |
1c8c601a | 728 | struct inode *inode = file_inode(filp); |
1da177e4 | 729 | |
1c8c601a | 730 | spin_lock(&inode->i_lock); |
496ad9aa | 731 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
732 | if (!IS_POSIX(cfl)) |
733 | continue; | |
b842e240 | 734 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
735 | break; |
736 | } | |
ab1f1611 | 737 | if (cfl) { |
9d6a8c5c | 738 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 739 | if (cfl->fl_nspid) |
6c5f3e7b | 740 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 741 | } else |
129a84de | 742 | fl->fl_type = F_UNLCK; |
1c8c601a | 743 | spin_unlock(&inode->i_lock); |
6d34ac19 | 744 | return; |
1da177e4 | 745 | } |
1da177e4 LT |
746 | EXPORT_SYMBOL(posix_test_lock); |
747 | ||
b533184f BF |
748 | /* |
749 | * Deadlock detection: | |
750 | * | |
751 | * We attempt to detect deadlocks that are due purely to posix file | |
752 | * locks. | |
1da177e4 | 753 | * |
b533184f BF |
754 | * We assume that a task can be waiting for at most one lock at a time. |
755 | * So for any acquired lock, the process holding that lock may be | |
756 | * waiting on at most one other lock. That lock in turns may be held by | |
757 | * someone waiting for at most one other lock. Given a requested lock | |
758 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
759 | * follow this chain of waiters to ensure we are not about to create a | |
760 | * cycle. | |
1da177e4 | 761 | * |
b533184f BF |
762 | * Since we do this before we ever put a process to sleep on a lock, we |
763 | * are ensured that there is never a cycle; that is what guarantees that | |
764 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 765 | * |
b533184f BF |
766 | * Note: the above assumption may not be true when handling lock |
767 | * requests from a broken NFS client. It may also fail in the presence | |
768 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 769 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 770 | * |
cff2fce5 | 771 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
772 | * Because the owner is not even nominally tied to a thread of |
773 | * execution, the deadlock detection below can't reasonably work well. Just | |
774 | * skip it for those. | |
775 | * | |
cff2fce5 | 776 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
777 | * locks that just checks for the case where two tasks are attempting to |
778 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 779 | */ |
97855b49 BF |
780 | |
781 | #define MAX_DEADLK_ITERATIONS 10 | |
782 | ||
b533184f BF |
783 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
784 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
785 | { | |
786 | struct file_lock *fl; | |
787 | ||
3999e493 | 788 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
789 | if (posix_same_owner(fl, block_fl)) |
790 | return fl->fl_next; | |
791 | } | |
792 | return NULL; | |
793 | } | |
794 | ||
7b2296af | 795 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 796 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
797 | struct file_lock *block_fl) |
798 | { | |
97855b49 | 799 | int i = 0; |
1da177e4 | 800 | |
57b65325 JL |
801 | /* |
802 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 803 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 804 | */ |
cff2fce5 | 805 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
806 | return 0; |
807 | ||
b533184f BF |
808 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
809 | if (i++ > MAX_DEADLK_ITERATIONS) | |
810 | return 0; | |
811 | if (posix_same_owner(caller_fl, block_fl)) | |
812 | return 1; | |
1da177e4 LT |
813 | } |
814 | return 0; | |
815 | } | |
816 | ||
1da177e4 | 817 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 818 | * after any leases, but before any posix locks. |
f475ae95 TM |
819 | * |
820 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
821 | * whether or not a lock was successfully freed by testing the return | |
822 | * value for -ENOENT. | |
1da177e4 | 823 | */ |
993dfa87 | 824 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 825 | { |
993dfa87 | 826 | struct file_lock *new_fl = NULL; |
1da177e4 | 827 | struct file_lock **before; |
496ad9aa | 828 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
829 | int error = 0; |
830 | int found = 0; | |
ed9814d8 | 831 | LIST_HEAD(dispose); |
1da177e4 | 832 | |
b89f4321 | 833 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 834 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
835 | if (!new_fl) |
836 | return -ENOMEM; | |
84d535ad PE |
837 | } |
838 | ||
1c8c601a | 839 | spin_lock(&inode->i_lock); |
b89f4321 AB |
840 | if (request->fl_flags & FL_ACCESS) |
841 | goto find_conflict; | |
842 | ||
1da177e4 LT |
843 | for_each_lock(inode, before) { |
844 | struct file_lock *fl = *before; | |
845 | if (IS_POSIX(fl)) | |
846 | break; | |
847 | if (IS_LEASE(fl)) | |
848 | continue; | |
849 | if (filp != fl->fl_file) | |
850 | continue; | |
993dfa87 | 851 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
852 | goto out; |
853 | found = 1; | |
ed9814d8 | 854 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
855 | break; |
856 | } | |
1da177e4 | 857 | |
f475ae95 TM |
858 | if (request->fl_type == F_UNLCK) { |
859 | if ((request->fl_flags & FL_EXISTS) && !found) | |
860 | error = -ENOENT; | |
993dfa87 | 861 | goto out; |
f475ae95 | 862 | } |
1da177e4 LT |
863 | |
864 | /* | |
865 | * If a higher-priority process was blocked on the old file lock, | |
866 | * give it the opportunity to lock the file. | |
867 | */ | |
b89f4321 | 868 | if (found) { |
1c8c601a | 869 | spin_unlock(&inode->i_lock); |
def01bc5 | 870 | cond_resched(); |
1c8c601a | 871 | spin_lock(&inode->i_lock); |
b89f4321 | 872 | } |
1da177e4 | 873 | |
f07f18dd | 874 | find_conflict: |
1da177e4 LT |
875 | for_each_lock(inode, before) { |
876 | struct file_lock *fl = *before; | |
877 | if (IS_POSIX(fl)) | |
878 | break; | |
879 | if (IS_LEASE(fl)) | |
880 | continue; | |
993dfa87 | 881 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
882 | continue; |
883 | error = -EAGAIN; | |
bde74e4b MS |
884 | if (!(request->fl_flags & FL_SLEEP)) |
885 | goto out; | |
886 | error = FILE_LOCK_DEFERRED; | |
887 | locks_insert_block(fl, request); | |
1da177e4 LT |
888 | goto out; |
889 | } | |
f07f18dd TM |
890 | if (request->fl_flags & FL_ACCESS) |
891 | goto out; | |
993dfa87 | 892 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 893 | locks_insert_lock(before, new_fl); |
993dfa87 | 894 | new_fl = NULL; |
9cedc194 | 895 | error = 0; |
1da177e4 LT |
896 | |
897 | out: | |
1c8c601a | 898 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
899 | if (new_fl) |
900 | locks_free_lock(new_fl); | |
ed9814d8 | 901 | locks_dispose_list(&dispose); |
1da177e4 LT |
902 | return error; |
903 | } | |
904 | ||
150b3934 | 905 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
906 | { |
907 | struct file_lock *fl; | |
39005d02 MS |
908 | struct file_lock *new_fl = NULL; |
909 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
910 | struct file_lock *left = NULL; |
911 | struct file_lock *right = NULL; | |
912 | struct file_lock **before; | |
b9746ef8 JL |
913 | int error; |
914 | bool added = false; | |
ed9814d8 | 915 | LIST_HEAD(dispose); |
1da177e4 LT |
916 | |
917 | /* | |
918 | * We may need two file_lock structures for this operation, | |
919 | * so we get them in advance to avoid races. | |
39005d02 MS |
920 | * |
921 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 922 | */ |
39005d02 MS |
923 | if (!(request->fl_flags & FL_ACCESS) && |
924 | (request->fl_type != F_UNLCK || | |
925 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
926 | new_fl = locks_alloc_lock(); | |
927 | new_fl2 = locks_alloc_lock(); | |
928 | } | |
1da177e4 | 929 | |
1c8c601a | 930 | spin_lock(&inode->i_lock); |
1cb36012 JL |
931 | /* |
932 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
933 | * there are any, either return error or put the request on the | |
48f74186 | 934 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 935 | */ |
1da177e4 LT |
936 | if (request->fl_type != F_UNLCK) { |
937 | for_each_lock(inode, before) { | |
526985b9 | 938 | fl = *before; |
1da177e4 LT |
939 | if (!IS_POSIX(fl)) |
940 | continue; | |
941 | if (!posix_locks_conflict(request, fl)) | |
942 | continue; | |
5842add2 | 943 | if (conflock) |
1a747ee0 | 944 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
945 | error = -EAGAIN; |
946 | if (!(request->fl_flags & FL_SLEEP)) | |
947 | goto out; | |
1c8c601a JL |
948 | /* |
949 | * Deadlock detection and insertion into the blocked | |
950 | * locks list must be done while holding the same lock! | |
951 | */ | |
1da177e4 | 952 | error = -EDEADLK; |
7b2296af | 953 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
954 | if (likely(!posix_locks_deadlock(request, fl))) { |
955 | error = FILE_LOCK_DEFERRED; | |
956 | __locks_insert_block(fl, request); | |
957 | } | |
7b2296af | 958 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
959 | goto out; |
960 | } | |
961 | } | |
962 | ||
963 | /* If we're just looking for a conflict, we're done. */ | |
964 | error = 0; | |
965 | if (request->fl_flags & FL_ACCESS) | |
966 | goto out; | |
967 | ||
1da177e4 | 968 | /* |
1da177e4 LT |
969 | * Find the first old lock with the same owner as the new lock. |
970 | */ | |
971 | ||
972 | before = &inode->i_flock; | |
973 | ||
974 | /* First skip locks owned by other processes. */ | |
975 | while ((fl = *before) && (!IS_POSIX(fl) || | |
976 | !posix_same_owner(request, fl))) { | |
977 | before = &fl->fl_next; | |
978 | } | |
979 | ||
1cb36012 | 980 | /* Process locks with this owner. */ |
1da177e4 LT |
981 | while ((fl = *before) && posix_same_owner(request, fl)) { |
982 | /* Detect adjacent or overlapping regions (if same lock type) | |
983 | */ | |
984 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
985 | /* In all comparisons of start vs end, use |
986 | * "start - 1" rather than "end + 1". If end | |
987 | * is OFFSET_MAX, end + 1 will become negative. | |
988 | */ | |
1da177e4 LT |
989 | if (fl->fl_end < request->fl_start - 1) |
990 | goto next_lock; | |
991 | /* If the next lock in the list has entirely bigger | |
992 | * addresses than the new one, insert the lock here. | |
993 | */ | |
449231d6 | 994 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
995 | break; |
996 | ||
997 | /* If we come here, the new and old lock are of the | |
998 | * same type and adjacent or overlapping. Make one | |
999 | * lock yielding from the lower start address of both | |
1000 | * locks to the higher end address. | |
1001 | */ | |
1002 | if (fl->fl_start > request->fl_start) | |
1003 | fl->fl_start = request->fl_start; | |
1004 | else | |
1005 | request->fl_start = fl->fl_start; | |
1006 | if (fl->fl_end < request->fl_end) | |
1007 | fl->fl_end = request->fl_end; | |
1008 | else | |
1009 | request->fl_end = fl->fl_end; | |
1010 | if (added) { | |
ed9814d8 | 1011 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1012 | continue; |
1013 | } | |
1014 | request = fl; | |
b9746ef8 | 1015 | added = true; |
1da177e4 LT |
1016 | } |
1017 | else { | |
1018 | /* Processing for different lock types is a bit | |
1019 | * more complex. | |
1020 | */ | |
1021 | if (fl->fl_end < request->fl_start) | |
1022 | goto next_lock; | |
1023 | if (fl->fl_start > request->fl_end) | |
1024 | break; | |
1025 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1026 | added = true; |
1da177e4 LT |
1027 | if (fl->fl_start < request->fl_start) |
1028 | left = fl; | |
1029 | /* If the next lock in the list has a higher end | |
1030 | * address than the new one, insert the new one here. | |
1031 | */ | |
1032 | if (fl->fl_end > request->fl_end) { | |
1033 | right = fl; | |
1034 | break; | |
1035 | } | |
1036 | if (fl->fl_start >= request->fl_start) { | |
1037 | /* The new lock completely replaces an old | |
1038 | * one (This may happen several times). | |
1039 | */ | |
1040 | if (added) { | |
ed9814d8 | 1041 | locks_delete_lock(before, &dispose); |
1da177e4 LT |
1042 | continue; |
1043 | } | |
b84d49f9 JL |
1044 | /* |
1045 | * Replace the old lock with new_fl, and | |
1046 | * remove the old one. It's safe to do the | |
1047 | * insert here since we know that we won't be | |
1048 | * using new_fl later, and that the lock is | |
1049 | * just replacing an existing lock. | |
1da177e4 | 1050 | */ |
b84d49f9 JL |
1051 | error = -ENOLCK; |
1052 | if (!new_fl) | |
1053 | goto out; | |
1054 | locks_copy_lock(new_fl, request); | |
1055 | request = new_fl; | |
1056 | new_fl = NULL; | |
ed9814d8 | 1057 | locks_delete_lock(before, &dispose); |
b84d49f9 | 1058 | locks_insert_lock(before, request); |
b9746ef8 | 1059 | added = true; |
1da177e4 LT |
1060 | } |
1061 | } | |
1062 | /* Go on to next lock. | |
1063 | */ | |
1064 | next_lock: | |
1065 | before = &fl->fl_next; | |
1066 | } | |
1067 | ||
0d9a490a | 1068 | /* |
1cb36012 JL |
1069 | * The above code only modifies existing locks in case of merging or |
1070 | * replacing. If new lock(s) need to be inserted all modifications are | |
1071 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1072 | */ |
1073 | error = -ENOLCK; /* "no luck" */ | |
1074 | if (right && left == right && !new_fl2) | |
1075 | goto out; | |
1076 | ||
1da177e4 LT |
1077 | error = 0; |
1078 | if (!added) { | |
f475ae95 TM |
1079 | if (request->fl_type == F_UNLCK) { |
1080 | if (request->fl_flags & FL_EXISTS) | |
1081 | error = -ENOENT; | |
1da177e4 | 1082 | goto out; |
f475ae95 | 1083 | } |
0d9a490a MS |
1084 | |
1085 | if (!new_fl) { | |
1086 | error = -ENOLCK; | |
1087 | goto out; | |
1088 | } | |
1da177e4 LT |
1089 | locks_copy_lock(new_fl, request); |
1090 | locks_insert_lock(before, new_fl); | |
1091 | new_fl = NULL; | |
1092 | } | |
1093 | if (right) { | |
1094 | if (left == right) { | |
1095 | /* The new lock breaks the old one in two pieces, | |
1096 | * so we have to use the second new lock. | |
1097 | */ | |
1098 | left = new_fl2; | |
1099 | new_fl2 = NULL; | |
1100 | locks_copy_lock(left, right); | |
1101 | locks_insert_lock(before, left); | |
1102 | } | |
1103 | right->fl_start = request->fl_end + 1; | |
1104 | locks_wake_up_blocks(right); | |
1105 | } | |
1106 | if (left) { | |
1107 | left->fl_end = request->fl_start - 1; | |
1108 | locks_wake_up_blocks(left); | |
1109 | } | |
1110 | out: | |
1c8c601a | 1111 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1112 | /* |
1113 | * Free any unused locks. | |
1114 | */ | |
1115 | if (new_fl) | |
1116 | locks_free_lock(new_fl); | |
1117 | if (new_fl2) | |
1118 | locks_free_lock(new_fl2); | |
ed9814d8 | 1119 | locks_dispose_list(&dispose); |
1da177e4 LT |
1120 | return error; |
1121 | } | |
1122 | ||
1123 | /** | |
1124 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1125 | * @filp: The file to apply the lock to | |
1126 | * @fl: The lock to be applied | |
150b3934 | 1127 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1128 | * |
1129 | * Add a POSIX style lock to a file. | |
1130 | * We merge adjacent & overlapping locks whenever possible. | |
1131 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1132 | * |
1133 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1134 | * whether or not a lock was successfully freed by testing the return | |
1135 | * value for -ENOENT. | |
1da177e4 | 1136 | */ |
150b3934 | 1137 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1138 | struct file_lock *conflock) |
1139 | { | |
496ad9aa | 1140 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1141 | } |
150b3934 | 1142 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1143 | |
1144 | /** | |
1145 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1146 | * @filp: The file to apply the lock to | |
1147 | * @fl: The lock to be applied | |
1148 | * | |
1149 | * Add a POSIX style lock to a file. | |
1150 | * We merge adjacent & overlapping locks whenever possible. | |
1151 | * POSIX locks are sorted by owner task, then by starting address | |
1152 | */ | |
1153 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1154 | { | |
1155 | int error; | |
1156 | might_sleep (); | |
1157 | for (;;) { | |
150b3934 | 1158 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1159 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1160 | break; |
1161 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1162 | if (!error) | |
1163 | continue; | |
1164 | ||
1165 | locks_delete_block(fl); | |
1166 | break; | |
1167 | } | |
1168 | return error; | |
1169 | } | |
1170 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1171 | ||
1172 | /** | |
1173 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1174 | * @file: the file to check |
1da177e4 LT |
1175 | * |
1176 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1177 | * This function is called from locks_verify_locked() only. | |
1178 | */ | |
d7a06983 | 1179 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1180 | { |
d7a06983 | 1181 | struct inode *inode = file_inode(file); |
1da177e4 LT |
1182 | struct file_lock *fl; |
1183 | ||
1184 | /* | |
1185 | * Search the lock list for this inode for any POSIX locks. | |
1186 | */ | |
1c8c601a | 1187 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1188 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1189 | if (!IS_POSIX(fl)) | |
1190 | continue; | |
73a8f5f7 CH |
1191 | if (fl->fl_owner != current->files && |
1192 | fl->fl_owner != file) | |
1da177e4 LT |
1193 | break; |
1194 | } | |
1c8c601a | 1195 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1196 | return fl ? -EAGAIN : 0; |
1197 | } | |
1198 | ||
1199 | /** | |
1200 | * locks_mandatory_area - Check for a conflicting lock | |
1201 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1202 | * for shared | |
1203 | * @inode: the file to check | |
1204 | * @filp: how the file was opened (if it was) | |
1205 | * @offset: start of area to check | |
1206 | * @count: length of area to check | |
1207 | * | |
1208 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1209 | * This function is called from rw_verify_area() and | |
1210 | * locks_verify_truncate(). | |
1211 | */ | |
1212 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1213 | struct file *filp, loff_t offset, | |
1214 | size_t count) | |
1215 | { | |
1216 | struct file_lock fl; | |
1217 | int error; | |
29723ade | 1218 | bool sleep = false; |
1da177e4 LT |
1219 | |
1220 | locks_init_lock(&fl); | |
1da177e4 LT |
1221 | fl.fl_pid = current->tgid; |
1222 | fl.fl_file = filp; | |
1223 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1224 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1225 | sleep = true; |
1da177e4 LT |
1226 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1227 | fl.fl_start = offset; | |
1228 | fl.fl_end = offset + count - 1; | |
1229 | ||
1230 | for (;;) { | |
29723ade | 1231 | if (filp) { |
73a8f5f7 | 1232 | fl.fl_owner = filp; |
29723ade JL |
1233 | fl.fl_flags &= ~FL_SLEEP; |
1234 | error = __posix_lock_file(inode, &fl, NULL); | |
1235 | if (!error) | |
1236 | break; | |
1237 | } | |
1238 | ||
1239 | if (sleep) | |
1240 | fl.fl_flags |= FL_SLEEP; | |
1241 | fl.fl_owner = current->files; | |
150b3934 | 1242 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1243 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1244 | break; |
1245 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1246 | if (!error) { | |
1247 | /* | |
1248 | * If we've been sleeping someone might have | |
1249 | * changed the permissions behind our back. | |
1250 | */ | |
a16877ca | 1251 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1252 | continue; |
1253 | } | |
1254 | ||
1255 | locks_delete_block(&fl); | |
1256 | break; | |
1257 | } | |
1258 | ||
1259 | return error; | |
1260 | } | |
1261 | ||
1262 | EXPORT_SYMBOL(locks_mandatory_area); | |
1263 | ||
778fc546 BF |
1264 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1265 | { | |
1266 | switch (arg) { | |
1267 | case F_UNLCK: | |
1268 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1269 | /* fall through: */ | |
1270 | case F_RDLCK: | |
1271 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1272 | } | |
1273 | } | |
1274 | ||
1da177e4 LT |
1275 | /* We already had a lease on this file; just change its type */ |
1276 | int lease_modify(struct file_lock **before, int arg) | |
1277 | { | |
1278 | struct file_lock *fl = *before; | |
1279 | int error = assign_type(fl, arg); | |
1280 | ||
1281 | if (error) | |
1282 | return error; | |
778fc546 | 1283 | lease_clear_pending(fl, arg); |
1da177e4 | 1284 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1285 | if (arg == F_UNLCK) { |
1286 | struct file *filp = fl->fl_file; | |
1287 | ||
1288 | f_delown(filp); | |
1289 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1290 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1291 | if (fl->fl_fasync != NULL) { | |
1292 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1293 | fl->fl_fasync = NULL; | |
1294 | } | |
ed9814d8 | 1295 | locks_delete_lock(before, NULL); |
3b6e2723 | 1296 | } |
1da177e4 LT |
1297 | return 0; |
1298 | } | |
1299 | ||
1300 | EXPORT_SYMBOL(lease_modify); | |
1301 | ||
778fc546 BF |
1302 | static bool past_time(unsigned long then) |
1303 | { | |
1304 | if (!then) | |
1305 | /* 0 is a special value meaning "this never expires": */ | |
1306 | return false; | |
1307 | return time_after(jiffies, then); | |
1308 | } | |
1309 | ||
1da177e4 LT |
1310 | static void time_out_leases(struct inode *inode) |
1311 | { | |
1312 | struct file_lock **before; | |
1313 | struct file_lock *fl; | |
1314 | ||
1315 | before = &inode->i_flock; | |
ab83fa4b | 1316 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1317 | trace_time_out_leases(inode, fl); |
778fc546 BF |
1318 | if (past_time(fl->fl_downgrade_time)) |
1319 | lease_modify(before, F_RDLCK); | |
1320 | if (past_time(fl->fl_break_time)) | |
1321 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1322 | if (fl == *before) /* lease_modify may have freed fl */ |
1323 | before = &fl->fl_next; | |
1324 | } | |
1325 | } | |
1326 | ||
df4e8d2c BF |
1327 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1328 | { | |
1329 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1330 | return false; | |
1331 | return locks_conflict(breaker, lease); | |
1332 | } | |
1333 | ||
1da177e4 LT |
1334 | /** |
1335 | * __break_lease - revoke all outstanding leases on file | |
1336 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1337 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1338 | * break all leases | |
1339 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1340 | * only delegations | |
1da177e4 | 1341 | * |
87250dd2 | 1342 | * break_lease (inlined for speed) has checked there already is at least |
1343 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1344 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1345 | * specified %O_NONBLOCK to your open(). |
1346 | */ | |
df4e8d2c | 1347 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1348 | { |
778fc546 | 1349 | int error = 0; |
1da177e4 LT |
1350 | struct file_lock *new_fl, *flock; |
1351 | struct file_lock *fl; | |
1da177e4 LT |
1352 | unsigned long break_time; |
1353 | int i_have_this_lease = 0; | |
df4e8d2c | 1354 | bool lease_conflict = false; |
8737c930 | 1355 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1356 | |
8737c930 | 1357 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1358 | if (IS_ERR(new_fl)) |
1359 | return PTR_ERR(new_fl); | |
df4e8d2c | 1360 | new_fl->fl_flags = type; |
1da177e4 | 1361 | |
1c8c601a | 1362 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1363 | |
1364 | time_out_leases(inode); | |
1365 | ||
1366 | flock = inode->i_flock; | |
1367 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1368 | goto out; | |
1369 | ||
df4e8d2c BF |
1370 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1371 | if (leases_conflict(fl, new_fl)) { | |
1372 | lease_conflict = true; | |
1373 | if (fl->fl_owner == current->files) | |
1374 | i_have_this_lease = 1; | |
1375 | } | |
1376 | } | |
1377 | if (!lease_conflict) | |
778fc546 BF |
1378 | goto out; |
1379 | ||
1da177e4 LT |
1380 | break_time = 0; |
1381 | if (lease_break_time > 0) { | |
1382 | break_time = jiffies + lease_break_time * HZ; | |
1383 | if (break_time == 0) | |
1384 | break_time++; /* so that 0 means no break time */ | |
1385 | } | |
1386 | ||
1387 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
df4e8d2c BF |
1388 | if (!leases_conflict(fl, new_fl)) |
1389 | continue; | |
778fc546 BF |
1390 | if (want_write) { |
1391 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1392 | continue; | |
1393 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1394 | fl->fl_break_time = break_time; |
778fc546 BF |
1395 | } else { |
1396 | if (lease_breaking(flock)) | |
1397 | continue; | |
1398 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1399 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1400 | } |
778fc546 | 1401 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1402 | } |
1403 | ||
1404 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
62af4f1f | 1405 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1406 | error = -EWOULDBLOCK; |
1407 | goto out; | |
1408 | } | |
1409 | ||
1410 | restart: | |
1411 | break_time = flock->fl_break_time; | |
f1c6bb2c | 1412 | if (break_time != 0) |
1da177e4 | 1413 | break_time -= jiffies; |
f1c6bb2c JL |
1414 | if (break_time == 0) |
1415 | break_time++; | |
4321e01e | 1416 | locks_insert_block(flock, new_fl); |
62af4f1f | 1417 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1418 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1419 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1420 | !new_fl->fl_next, break_time); | |
1c8c601a | 1421 | spin_lock(&inode->i_lock); |
62af4f1f | 1422 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1423 | locks_delete_block(new_fl); |
1da177e4 LT |
1424 | if (error >= 0) { |
1425 | if (error == 0) | |
1426 | time_out_leases(inode); | |
778fc546 BF |
1427 | /* |
1428 | * Wait for the next conflicting lease that has not been | |
1429 | * broken yet | |
1430 | */ | |
1da177e4 LT |
1431 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1432 | flock = flock->fl_next) { | |
df4e8d2c | 1433 | if (leases_conflict(new_fl, flock)) |
1da177e4 LT |
1434 | goto restart; |
1435 | } | |
1436 | error = 0; | |
1437 | } | |
1438 | ||
1439 | out: | |
1c8c601a | 1440 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1441 | locks_free_lock(new_fl); |
1da177e4 LT |
1442 | return error; |
1443 | } | |
1444 | ||
1445 | EXPORT_SYMBOL(__break_lease); | |
1446 | ||
1447 | /** | |
a6b91919 | 1448 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1449 | * @inode: the inode |
1450 | * @time: pointer to a timespec which will contain the last modified time | |
1451 | * | |
1452 | * This is to force NFS clients to flush their caches for files with | |
1453 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1454 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1455 | */ |
1456 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1457 | { | |
1458 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1459 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1460 | *time = current_fs_time(inode->i_sb); |
1461 | else | |
1462 | *time = inode->i_mtime; | |
1463 | } | |
1464 | ||
1465 | EXPORT_SYMBOL(lease_get_mtime); | |
1466 | ||
1467 | /** | |
1468 | * fcntl_getlease - Enquire what lease is currently active | |
1469 | * @filp: the file | |
1470 | * | |
1471 | * The value returned by this function will be one of | |
1472 | * (if no lease break is pending): | |
1473 | * | |
1474 | * %F_RDLCK to indicate a shared lease is held. | |
1475 | * | |
1476 | * %F_WRLCK to indicate an exclusive lease is held. | |
1477 | * | |
1478 | * %F_UNLCK to indicate no lease is held. | |
1479 | * | |
1480 | * (if a lease break is pending): | |
1481 | * | |
1482 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1483 | * changed to a shared lease (or removed). | |
1484 | * | |
1485 | * %F_UNLCK to indicate the lease needs to be removed. | |
1486 | * | |
1487 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1488 | * should be returned to userspace. | |
1489 | */ | |
1490 | int fcntl_getlease(struct file *filp) | |
1491 | { | |
1492 | struct file_lock *fl; | |
1c8c601a | 1493 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1494 | int type = F_UNLCK; |
1495 | ||
1c8c601a | 1496 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1497 | time_out_leases(file_inode(filp)); |
1498 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1499 | fl = fl->fl_next) { |
1500 | if (fl->fl_file == filp) { | |
778fc546 | 1501 | type = target_leasetype(fl); |
1da177e4 LT |
1502 | break; |
1503 | } | |
1504 | } | |
1c8c601a | 1505 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1506 | return type; |
1507 | } | |
1508 | ||
24cbe784 JL |
1509 | /** |
1510 | * check_conflicting_open - see if the given dentry points to a file that has | |
1511 | * an existing open that would conflict with the | |
1512 | * desired lease. | |
1513 | * @dentry: dentry to check | |
1514 | * @arg: type of lease that we're trying to acquire | |
1515 | * | |
1516 | * Check to see if there's an existing open fd on this file that would | |
1517 | * conflict with the lease we're trying to set. | |
1518 | */ | |
1519 | static int | |
1520 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1521 | { | |
1522 | int ret = 0; | |
1523 | struct inode *inode = dentry->d_inode; | |
1524 | ||
1525 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1526 | return -EAGAIN; | |
1527 | ||
1528 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1529 | (atomic_read(&inode->i_count) > 1))) | |
1530 | ret = -EAGAIN; | |
1531 | ||
1532 | return ret; | |
1533 | } | |
1534 | ||
d4f22d19 | 1535 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1536 | { |
7eaae282 | 1537 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1538 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1539 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1540 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1541 | int error; |
1da177e4 | 1542 | |
096657b6 | 1543 | lease = *flp; |
62af4f1f JL |
1544 | trace_generic_add_lease(inode, lease); |
1545 | ||
df4e8d2c BF |
1546 | /* |
1547 | * In the delegation case we need mutual exclusion with | |
1548 | * a number of operations that take the i_mutex. We trylock | |
1549 | * because delegations are an optional optimization, and if | |
1550 | * there's some chance of a conflict--we'd rather not | |
1551 | * bother, maybe that's a sign this just isn't a good file to | |
1552 | * hand out a delegation on. | |
1553 | */ | |
1554 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1555 | return -EAGAIN; | |
1556 | ||
1557 | if (is_deleg && arg == F_WRLCK) { | |
1558 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1559 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1560 | WARN_ON_ONCE(1); |
1561 | return -EINVAL; | |
1562 | } | |
096657b6 | 1563 | |
24cbe784 JL |
1564 | error = check_conflicting_open(dentry, arg); |
1565 | if (error) | |
096657b6 | 1566 | goto out; |
6d5e8b05 | 1567 | |
1da177e4 LT |
1568 | /* |
1569 | * At this point, we know that if there is an exclusive | |
1570 | * lease on this file, then we hold it on this filp | |
1571 | * (otherwise our open of this file would have blocked). | |
1572 | * And if we are trying to acquire an exclusive lease, | |
1573 | * then the file is not open by anyone (including us) | |
1574 | * except for this filp. | |
1575 | */ | |
c1f24ef4 | 1576 | error = -EAGAIN; |
1da177e4 LT |
1577 | for (before = &inode->i_flock; |
1578 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1579 | before = &fl->fl_next) { | |
c1f24ef4 | 1580 | if (fl->fl_file == filp) { |
1da177e4 | 1581 | my_before = before; |
c1f24ef4 BF |
1582 | continue; |
1583 | } | |
1584 | /* | |
1585 | * No exclusive leases if someone else has a lease on | |
1586 | * this file: | |
1587 | */ | |
1588 | if (arg == F_WRLCK) | |
1589 | goto out; | |
1590 | /* | |
1591 | * Modifying our existing lease is OK, but no getting a | |
1592 | * new lease if someone else is opening for write: | |
1593 | */ | |
1594 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1595 | goto out; | |
1da177e4 LT |
1596 | } |
1597 | ||
1da177e4 | 1598 | if (my_before != NULL) { |
8fb47a4f | 1599 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1600 | if (!error) |
1601 | *flp = *my_before; | |
1da177e4 LT |
1602 | goto out; |
1603 | } | |
1604 | ||
1da177e4 LT |
1605 | error = -EINVAL; |
1606 | if (!leases_enable) | |
1607 | goto out; | |
1608 | ||
c5b1f0d9 | 1609 | locks_insert_lock(before, lease); |
24cbe784 JL |
1610 | /* |
1611 | * The check in break_lease() is lockless. It's possible for another | |
1612 | * open to race in after we did the earlier check for a conflicting | |
1613 | * open but before the lease was inserted. Check again for a | |
1614 | * conflicting open and cancel the lease if there is one. | |
1615 | * | |
1616 | * We also add a barrier here to ensure that the insertion of the lock | |
1617 | * precedes these checks. | |
1618 | */ | |
1619 | smp_mb(); | |
1620 | error = check_conflicting_open(dentry, arg); | |
1621 | if (error) | |
1622 | locks_unlink_lock(flp); | |
1da177e4 | 1623 | out: |
df4e8d2c BF |
1624 | if (is_deleg) |
1625 | mutex_unlock(&inode->i_mutex); | |
1da177e4 LT |
1626 | return error; |
1627 | } | |
8335ebd9 | 1628 | |
d4f22d19 | 1629 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1630 | { |
1631 | struct file_lock *fl, **before; | |
1632 | struct dentry *dentry = filp->f_path.dentry; | |
1633 | struct inode *inode = dentry->d_inode; | |
1634 | ||
62af4f1f JL |
1635 | trace_generic_delete_lease(inode, *flp); |
1636 | ||
8335ebd9 BF |
1637 | for (before = &inode->i_flock; |
1638 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1639 | before = &fl->fl_next) { | |
1640 | if (fl->fl_file != filp) | |
1641 | continue; | |
1642 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1643 | } | |
1644 | return -EAGAIN; | |
1645 | } | |
1646 | ||
1647 | /** | |
1648 | * generic_setlease - sets a lease on an open file | |
1649 | * @filp: file pointer | |
1650 | * @arg: type of lease to obtain | |
1651 | * @flp: input - file_lock to use, output - file_lock inserted | |
1652 | * | |
1653 | * The (input) flp->fl_lmops->lm_break function is required | |
1654 | * by break_lease(). | |
1655 | * | |
1c8c601a | 1656 | * Called with inode->i_lock held. |
8335ebd9 BF |
1657 | */ |
1658 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1659 | { | |
1660 | struct dentry *dentry = filp->f_path.dentry; | |
1661 | struct inode *inode = dentry->d_inode; | |
1662 | int error; | |
1663 | ||
8e96e3b7 | 1664 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1665 | return -EACCES; |
1666 | if (!S_ISREG(inode->i_mode)) | |
1667 | return -EINVAL; | |
1668 | error = security_file_lock(filp, arg); | |
1669 | if (error) | |
1670 | return error; | |
1671 | ||
1672 | time_out_leases(inode); | |
1673 | ||
1674 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1675 | ||
1676 | switch (arg) { | |
1677 | case F_UNLCK: | |
1678 | return generic_delete_lease(filp, flp); | |
1679 | case F_RDLCK: | |
1680 | case F_WRLCK: | |
1681 | return generic_add_lease(filp, arg, flp); | |
1682 | default: | |
8d657eb3 | 1683 | return -EINVAL; |
8335ebd9 BF |
1684 | } |
1685 | } | |
0af1a450 | 1686 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1687 | |
b89f4321 AB |
1688 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1689 | { | |
72c2d531 | 1690 | if (filp->f_op->setlease) |
b89f4321 AB |
1691 | return filp->f_op->setlease(filp, arg, lease); |
1692 | else | |
1693 | return generic_setlease(filp, arg, lease); | |
1694 | } | |
1695 | ||
1696 | /** | |
a9933cea | 1697 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1698 | * @filp: file pointer |
1699 | * @arg: type of lease to obtain | |
1700 | * @lease: file_lock to use | |
1701 | * | |
1702 | * Call this to establish a lease on the file. | |
8fb47a4f | 1703 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1704 | * break_lease will oops! |
1705 | * | |
1706 | * This will call the filesystem's setlease file method, if | |
1707 | * defined. Note that there is no getlease method; instead, the | |
1708 | * filesystem setlease method should call back to setlease() to | |
1709 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1710 | * find it. Since fcntl_getlease() only reports whether the current | |
1711 | * task holds a lease, a cluster filesystem need only do this for | |
1712 | * leases held by processes on this node. | |
1713 | * | |
1714 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1715 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1716 | * filesystem's open, create, and (on truncate) setattr methods. |
1717 | * | |
1718 | * Warning: the only current setlease methods exist only to disable | |
1719 | * leases in certain cases. More vfs changes may be required to | |
1720 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1721 | */ |
1722 | ||
a9933cea | 1723 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1724 | { |
1c8c601a | 1725 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1726 | int error; |
1727 | ||
1c8c601a | 1728 | spin_lock(&inode->i_lock); |
b89f4321 | 1729 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1730 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1731 | |
1732 | return error; | |
1733 | } | |
a9933cea | 1734 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1735 | |
0ceaf6c7 BF |
1736 | static int do_fcntl_delete_lease(struct file *filp) |
1737 | { | |
1738 | struct file_lock fl, *flp = &fl; | |
1739 | ||
1740 | lease_init(filp, F_UNLCK, flp); | |
1741 | ||
1742 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1743 | } | |
1744 | ||
1745 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1746 | { |
3df057ac | 1747 | struct file_lock *fl, *ret; |
1c8c601a | 1748 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1749 | struct fasync_struct *new; |
1da177e4 LT |
1750 | int error; |
1751 | ||
c5b1f0d9 AB |
1752 | fl = lease_alloc(filp, arg); |
1753 | if (IS_ERR(fl)) | |
1754 | return PTR_ERR(fl); | |
1da177e4 | 1755 | |
f7347ce4 LT |
1756 | new = fasync_alloc(); |
1757 | if (!new) { | |
1758 | locks_free_lock(fl); | |
1759 | return -ENOMEM; | |
1760 | } | |
3df057ac | 1761 | ret = fl; |
1c8c601a | 1762 | spin_lock(&inode->i_lock); |
8896b93f | 1763 | error = __vfs_setlease(filp, arg, &ret); |
2dfb928f JL |
1764 | if (error) |
1765 | goto out_unlock; | |
1766 | if (ret == fl) | |
1767 | fl = NULL; | |
1da177e4 | 1768 | |
f7347ce4 LT |
1769 | /* |
1770 | * fasync_insert_entry() returns the old entry if any. | |
1771 | * If there was no old entry, then it used 'new' and | |
1772 | * inserted it into the fasync list. Clear new so that | |
1773 | * we don't release it here. | |
1774 | */ | |
3df057ac | 1775 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1776 | new = NULL; |
1777 | ||
8896b93f | 1778 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
2dfb928f | 1779 | out_unlock: |
1c8c601a | 1780 | spin_unlock(&inode->i_lock); |
2dfb928f JL |
1781 | if (fl) |
1782 | locks_free_lock(fl); | |
f7347ce4 LT |
1783 | if (new) |
1784 | fasync_free(new); | |
1da177e4 LT |
1785 | return error; |
1786 | } | |
1787 | ||
0ceaf6c7 BF |
1788 | /** |
1789 | * fcntl_setlease - sets a lease on an open file | |
1790 | * @fd: open file descriptor | |
1791 | * @filp: file pointer | |
1792 | * @arg: type of lease to obtain | |
1793 | * | |
1794 | * Call this fcntl to establish a lease on the file. | |
1795 | * Note that you also need to call %F_SETSIG to | |
1796 | * receive a signal when the lease is broken. | |
1797 | */ | |
1798 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1799 | { | |
1800 | if (arg == F_UNLCK) | |
1801 | return do_fcntl_delete_lease(filp); | |
1802 | return do_fcntl_add_lease(fd, filp, arg); | |
1803 | } | |
1804 | ||
1da177e4 LT |
1805 | /** |
1806 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1807 | * @filp: The file to apply the lock to | |
1808 | * @fl: The lock to be applied | |
1809 | * | |
1810 | * Add a FLOCK style lock to a file. | |
1811 | */ | |
1812 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1813 | { | |
1814 | int error; | |
1815 | might_sleep(); | |
1816 | for (;;) { | |
1817 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1818 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1819 | break; |
1820 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1821 | if (!error) | |
1822 | continue; | |
1823 | ||
1824 | locks_delete_block(fl); | |
1825 | break; | |
1826 | } | |
1827 | return error; | |
1828 | } | |
1829 | ||
1830 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1831 | ||
1832 | /** | |
1833 | * sys_flock: - flock() system call. | |
1834 | * @fd: the file descriptor to lock. | |
1835 | * @cmd: the type of lock to apply. | |
1836 | * | |
1837 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1838 | * The @cmd can be one of | |
1839 | * | |
1840 | * %LOCK_SH -- a shared lock. | |
1841 | * | |
1842 | * %LOCK_EX -- an exclusive lock. | |
1843 | * | |
1844 | * %LOCK_UN -- remove an existing lock. | |
1845 | * | |
1846 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1847 | * | |
1848 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1849 | * processes read and write access respectively. | |
1850 | */ | |
002c8976 | 1851 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1852 | { |
2903ff01 | 1853 | struct fd f = fdget(fd); |
1da177e4 LT |
1854 | struct file_lock *lock; |
1855 | int can_sleep, unlock; | |
1856 | int error; | |
1857 | ||
1858 | error = -EBADF; | |
2903ff01 | 1859 | if (!f.file) |
1da177e4 LT |
1860 | goto out; |
1861 | ||
1862 | can_sleep = !(cmd & LOCK_NB); | |
1863 | cmd &= ~LOCK_NB; | |
1864 | unlock = (cmd == LOCK_UN); | |
1865 | ||
aeb5d727 | 1866 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1867 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1868 | goto out_putf; |
1869 | ||
2903ff01 | 1870 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1871 | if (error) |
1872 | goto out_putf; | |
1873 | if (can_sleep) | |
1874 | lock->fl_flags |= FL_SLEEP; | |
1875 | ||
2903ff01 | 1876 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1877 | if (error) |
1878 | goto out_free; | |
1879 | ||
72c2d531 | 1880 | if (f.file->f_op->flock) |
2903ff01 | 1881 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1882 | (can_sleep) ? F_SETLKW : F_SETLK, |
1883 | lock); | |
1884 | else | |
2903ff01 | 1885 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1886 | |
1887 | out_free: | |
993dfa87 | 1888 | locks_free_lock(lock); |
1da177e4 LT |
1889 | |
1890 | out_putf: | |
2903ff01 | 1891 | fdput(f); |
1da177e4 LT |
1892 | out: |
1893 | return error; | |
1894 | } | |
1895 | ||
3ee17abd BF |
1896 | /** |
1897 | * vfs_test_lock - test file byte range lock | |
1898 | * @filp: The file to test lock for | |
6924c554 | 1899 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1900 | * |
1901 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1902 | * setting conf->fl_type to something other than F_UNLCK. | |
1903 | */ | |
1904 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1905 | { | |
72c2d531 | 1906 | if (filp->f_op->lock) |
3ee17abd BF |
1907 | return filp->f_op->lock(filp, F_GETLK, fl); |
1908 | posix_test_lock(filp, fl); | |
1909 | return 0; | |
1910 | } | |
1911 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1912 | ||
c2fa1b8a BF |
1913 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1914 | { | |
cff2fce5 | 1915 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1916 | #if BITS_PER_LONG == 32 |
1917 | /* | |
1918 | * Make sure we can represent the posix lock via | |
1919 | * legacy 32bit flock. | |
1920 | */ | |
1921 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1922 | return -EOVERFLOW; | |
1923 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1924 | return -EOVERFLOW; | |
1925 | #endif | |
1926 | flock->l_start = fl->fl_start; | |
1927 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1928 | fl->fl_end - fl->fl_start + 1; | |
1929 | flock->l_whence = 0; | |
129a84de | 1930 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1931 | return 0; |
1932 | } | |
1933 | ||
1934 | #if BITS_PER_LONG == 32 | |
1935 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1936 | { | |
cff2fce5 | 1937 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1938 | flock->l_start = fl->fl_start; |
1939 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1940 | fl->fl_end - fl->fl_start + 1; | |
1941 | flock->l_whence = 0; | |
1942 | flock->l_type = fl->fl_type; | |
1943 | } | |
1944 | #endif | |
1945 | ||
1da177e4 LT |
1946 | /* Report the first existing lock that would conflict with l. |
1947 | * This implements the F_GETLK command of fcntl(). | |
1948 | */ | |
c1e62b8f | 1949 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 1950 | { |
9d6a8c5c | 1951 | struct file_lock file_lock; |
1da177e4 LT |
1952 | struct flock flock; |
1953 | int error; | |
1954 | ||
1955 | error = -EFAULT; | |
1956 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1957 | goto out; | |
1958 | error = -EINVAL; | |
1959 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1960 | goto out; | |
1961 | ||
1962 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1963 | if (error) | |
1964 | goto out; | |
1965 | ||
0d3f7a2d | 1966 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
1967 | error = -EINVAL; |
1968 | if (flock.l_pid != 0) | |
1969 | goto out; | |
1970 | ||
5d50ffd7 | 1971 | cmd = F_GETLK; |
cff2fce5 | 1972 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 1973 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
1974 | } |
1975 | ||
3ee17abd BF |
1976 | error = vfs_test_lock(filp, &file_lock); |
1977 | if (error) | |
1978 | goto out; | |
1da177e4 | 1979 | |
9d6a8c5c ME |
1980 | flock.l_type = file_lock.fl_type; |
1981 | if (file_lock.fl_type != F_UNLCK) { | |
1982 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1983 | if (error) |
1da177e4 | 1984 | goto out; |
1da177e4 LT |
1985 | } |
1986 | error = -EFAULT; | |
1987 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1988 | error = 0; | |
1989 | out: | |
1990 | return error; | |
1991 | } | |
1992 | ||
7723ec97 ME |
1993 | /** |
1994 | * vfs_lock_file - file byte range lock | |
1995 | * @filp: The file to apply the lock to | |
1996 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1997 | * @fl: The lock to be applied | |
150b3934 ME |
1998 | * @conf: Place to return a copy of the conflicting lock, if found. |
1999 | * | |
2000 | * A caller that doesn't care about the conflicting lock may pass NULL | |
2001 | * as the final argument. | |
2002 | * | |
2003 | * If the filesystem defines a private ->lock() method, then @conf will | |
2004 | * be left unchanged; so a caller that cares should initialize it to | |
2005 | * some acceptable default. | |
2beb6614 ME |
2006 | * |
2007 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
2008 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
2009 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 2010 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
2011 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
2012 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 2013 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
2014 | * request completes. |
2015 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
2016 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
2017 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2018 | * nonzero return code and the file system should release the lock. The file |
2019 | * system is also responsible to keep a corresponding posix lock when it | |
2020 | * grants a lock so the VFS can find out which locks are locally held and do | |
2021 | * the correct lock cleanup when required. | |
2022 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2023 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2024 | * return code. |
7723ec97 | 2025 | */ |
150b3934 | 2026 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2027 | { |
72c2d531 | 2028 | if (filp->f_op->lock) |
7723ec97 ME |
2029 | return filp->f_op->lock(filp, cmd, fl); |
2030 | else | |
150b3934 | 2031 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2032 | } |
2033 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2034 | ||
b648a6de MS |
2035 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2036 | struct file_lock *fl) | |
2037 | { | |
2038 | int error; | |
2039 | ||
2040 | error = security_file_lock(filp, fl->fl_type); | |
2041 | if (error) | |
2042 | return error; | |
2043 | ||
764c76b3 MS |
2044 | for (;;) { |
2045 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2046 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2047 | break; |
764c76b3 MS |
2048 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2049 | if (!error) | |
2050 | continue; | |
2051 | ||
2052 | locks_delete_block(fl); | |
2053 | break; | |
b648a6de MS |
2054 | } |
2055 | ||
2056 | return error; | |
2057 | } | |
2058 | ||
cf01f4ee JL |
2059 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2060 | static int | |
2061 | check_fmode_for_setlk(struct file_lock *fl) | |
2062 | { | |
2063 | switch (fl->fl_type) { | |
2064 | case F_RDLCK: | |
2065 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2066 | return -EBADF; | |
2067 | break; | |
2068 | case F_WRLCK: | |
2069 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2070 | return -EBADF; | |
2071 | } | |
2072 | return 0; | |
2073 | } | |
2074 | ||
1da177e4 LT |
2075 | /* Apply the lock described by l to an open file descriptor. |
2076 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2077 | */ | |
c293621b PS |
2078 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2079 | struct flock __user *l) | |
1da177e4 LT |
2080 | { |
2081 | struct file_lock *file_lock = locks_alloc_lock(); | |
2082 | struct flock flock; | |
2083 | struct inode *inode; | |
0b2bac2f | 2084 | struct file *f; |
1da177e4 LT |
2085 | int error; |
2086 | ||
2087 | if (file_lock == NULL) | |
2088 | return -ENOLCK; | |
2089 | ||
2090 | /* | |
2091 | * This might block, so we do it before checking the inode. | |
2092 | */ | |
2093 | error = -EFAULT; | |
2094 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2095 | goto out; | |
2096 | ||
496ad9aa | 2097 | inode = file_inode(filp); |
1da177e4 LT |
2098 | |
2099 | /* Don't allow mandatory locks on files that may be memory mapped | |
2100 | * and shared. | |
2101 | */ | |
a16877ca | 2102 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2103 | error = -EAGAIN; |
2104 | goto out; | |
2105 | } | |
2106 | ||
c293621b | 2107 | again: |
1da177e4 LT |
2108 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2109 | if (error) | |
2110 | goto out; | |
5d50ffd7 | 2111 | |
cf01f4ee JL |
2112 | error = check_fmode_for_setlk(file_lock); |
2113 | if (error) | |
2114 | goto out; | |
2115 | ||
5d50ffd7 JL |
2116 | /* |
2117 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2118 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2119 | */ |
2120 | switch (cmd) { | |
0d3f7a2d | 2121 | case F_OFD_SETLK: |
90478939 JL |
2122 | error = -EINVAL; |
2123 | if (flock.l_pid != 0) | |
2124 | goto out; | |
2125 | ||
5d50ffd7 | 2126 | cmd = F_SETLK; |
cff2fce5 | 2127 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2128 | file_lock->fl_owner = filp; |
5d50ffd7 | 2129 | break; |
0d3f7a2d | 2130 | case F_OFD_SETLKW: |
90478939 JL |
2131 | error = -EINVAL; |
2132 | if (flock.l_pid != 0) | |
2133 | goto out; | |
2134 | ||
5d50ffd7 | 2135 | cmd = F_SETLKW; |
cff2fce5 | 2136 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2137 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2138 | /* Fallthrough */ |
2139 | case F_SETLKW: | |
1da177e4 LT |
2140 | file_lock->fl_flags |= FL_SLEEP; |
2141 | } | |
5d50ffd7 | 2142 | |
b648a6de | 2143 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2144 | |
c293621b PS |
2145 | /* |
2146 | * Attempt to detect a close/fcntl race and recover by | |
2147 | * releasing the lock that was just acquired. | |
2148 | */ | |
0b2bac2f AV |
2149 | /* |
2150 | * we need that spin_lock here - it prevents reordering between | |
2151 | * update of inode->i_flock and check for it done in close(). | |
2152 | * rcu_read_lock() wouldn't do. | |
2153 | */ | |
2154 | spin_lock(¤t->files->file_lock); | |
2155 | f = fcheck(fd); | |
2156 | spin_unlock(¤t->files->file_lock); | |
2157 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2158 | flock.l_type = F_UNLCK; |
2159 | goto again; | |
1da177e4 LT |
2160 | } |
2161 | ||
c293621b | 2162 | out: |
1da177e4 LT |
2163 | locks_free_lock(file_lock); |
2164 | return error; | |
2165 | } | |
2166 | ||
2167 | #if BITS_PER_LONG == 32 | |
2168 | /* Report the first existing lock that would conflict with l. | |
2169 | * This implements the F_GETLK command of fcntl(). | |
2170 | */ | |
c1e62b8f | 2171 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2172 | { |
9d6a8c5c | 2173 | struct file_lock file_lock; |
1da177e4 LT |
2174 | struct flock64 flock; |
2175 | int error; | |
2176 | ||
2177 | error = -EFAULT; | |
2178 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2179 | goto out; | |
2180 | error = -EINVAL; | |
2181 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2182 | goto out; | |
2183 | ||
2184 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2185 | if (error) | |
2186 | goto out; | |
2187 | ||
0d3f7a2d | 2188 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2189 | error = -EINVAL; |
2190 | if (flock.l_pid != 0) | |
2191 | goto out; | |
2192 | ||
5d50ffd7 | 2193 | cmd = F_GETLK64; |
cff2fce5 | 2194 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2195 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2196 | } |
2197 | ||
3ee17abd BF |
2198 | error = vfs_test_lock(filp, &file_lock); |
2199 | if (error) | |
2200 | goto out; | |
2201 | ||
9d6a8c5c ME |
2202 | flock.l_type = file_lock.fl_type; |
2203 | if (file_lock.fl_type != F_UNLCK) | |
2204 | posix_lock_to_flock64(&flock, &file_lock); | |
2205 | ||
1da177e4 LT |
2206 | error = -EFAULT; |
2207 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2208 | error = 0; | |
2209 | ||
2210 | out: | |
2211 | return error; | |
2212 | } | |
2213 | ||
2214 | /* Apply the lock described by l to an open file descriptor. | |
2215 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2216 | */ | |
c293621b PS |
2217 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2218 | struct flock64 __user *l) | |
1da177e4 LT |
2219 | { |
2220 | struct file_lock *file_lock = locks_alloc_lock(); | |
2221 | struct flock64 flock; | |
2222 | struct inode *inode; | |
0b2bac2f | 2223 | struct file *f; |
1da177e4 LT |
2224 | int error; |
2225 | ||
2226 | if (file_lock == NULL) | |
2227 | return -ENOLCK; | |
2228 | ||
2229 | /* | |
2230 | * This might block, so we do it before checking the inode. | |
2231 | */ | |
2232 | error = -EFAULT; | |
2233 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2234 | goto out; | |
2235 | ||
496ad9aa | 2236 | inode = file_inode(filp); |
1da177e4 LT |
2237 | |
2238 | /* Don't allow mandatory locks on files that may be memory mapped | |
2239 | * and shared. | |
2240 | */ | |
a16877ca | 2241 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2242 | error = -EAGAIN; |
2243 | goto out; | |
2244 | } | |
2245 | ||
c293621b | 2246 | again: |
1da177e4 LT |
2247 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2248 | if (error) | |
2249 | goto out; | |
5d50ffd7 | 2250 | |
cf01f4ee JL |
2251 | error = check_fmode_for_setlk(file_lock); |
2252 | if (error) | |
2253 | goto out; | |
2254 | ||
5d50ffd7 JL |
2255 | /* |
2256 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2257 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2258 | */ |
2259 | switch (cmd) { | |
0d3f7a2d | 2260 | case F_OFD_SETLK: |
90478939 JL |
2261 | error = -EINVAL; |
2262 | if (flock.l_pid != 0) | |
2263 | goto out; | |
2264 | ||
5d50ffd7 | 2265 | cmd = F_SETLK64; |
cff2fce5 | 2266 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2267 | file_lock->fl_owner = filp; |
5d50ffd7 | 2268 | break; |
0d3f7a2d | 2269 | case F_OFD_SETLKW: |
90478939 JL |
2270 | error = -EINVAL; |
2271 | if (flock.l_pid != 0) | |
2272 | goto out; | |
2273 | ||
5d50ffd7 | 2274 | cmd = F_SETLKW64; |
cff2fce5 | 2275 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2276 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2277 | /* Fallthrough */ |
2278 | case F_SETLKW64: | |
1da177e4 LT |
2279 | file_lock->fl_flags |= FL_SLEEP; |
2280 | } | |
5d50ffd7 | 2281 | |
b648a6de | 2282 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2283 | |
c293621b PS |
2284 | /* |
2285 | * Attempt to detect a close/fcntl race and recover by | |
2286 | * releasing the lock that was just acquired. | |
2287 | */ | |
0b2bac2f AV |
2288 | spin_lock(¤t->files->file_lock); |
2289 | f = fcheck(fd); | |
2290 | spin_unlock(¤t->files->file_lock); | |
2291 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2292 | flock.l_type = F_UNLCK; |
2293 | goto again; | |
1da177e4 LT |
2294 | } |
2295 | ||
2296 | out: | |
2297 | locks_free_lock(file_lock); | |
2298 | return error; | |
2299 | } | |
2300 | #endif /* BITS_PER_LONG == 32 */ | |
2301 | ||
2302 | /* | |
2303 | * This function is called when the file is being removed | |
2304 | * from the task's fd array. POSIX locks belonging to this task | |
2305 | * are deleted at this time. | |
2306 | */ | |
2307 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2308 | { | |
ff7b86b8 | 2309 | struct file_lock lock; |
1da177e4 LT |
2310 | |
2311 | /* | |
2312 | * If there are no locks held on this file, we don't need to call | |
2313 | * posix_lock_file(). Another process could be setting a lock on this | |
2314 | * file at the same time, but we wouldn't remove that lock anyway. | |
2315 | */ | |
496ad9aa | 2316 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2317 | return; |
2318 | ||
2319 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2320 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2321 | lock.fl_start = 0; |
2322 | lock.fl_end = OFFSET_MAX; | |
2323 | lock.fl_owner = owner; | |
2324 | lock.fl_pid = current->tgid; | |
2325 | lock.fl_file = filp; | |
2326 | lock.fl_ops = NULL; | |
2327 | lock.fl_lmops = NULL; | |
2328 | ||
150b3934 | 2329 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2330 | |
1da177e4 LT |
2331 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2332 | lock.fl_ops->fl_release_private(&lock); | |
2333 | } | |
2334 | ||
2335 | EXPORT_SYMBOL(locks_remove_posix); | |
2336 | ||
2337 | /* | |
2338 | * This function is called on the last close of an open file. | |
2339 | */ | |
78ed8a13 | 2340 | void locks_remove_file(struct file *filp) |
1da177e4 | 2341 | { |
496ad9aa | 2342 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2343 | struct file_lock *fl; |
2344 | struct file_lock **before; | |
ed9814d8 | 2345 | LIST_HEAD(dispose); |
1da177e4 LT |
2346 | |
2347 | if (!inode->i_flock) | |
2348 | return; | |
2349 | ||
73a8f5f7 | 2350 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2351 | |
72c2d531 | 2352 | if (filp->f_op->flock) { |
1da177e4 | 2353 | struct file_lock fl = { |
73a8f5f7 | 2354 | .fl_owner = filp, |
1da177e4 LT |
2355 | .fl_pid = current->tgid, |
2356 | .fl_file = filp, | |
2357 | .fl_flags = FL_FLOCK, | |
2358 | .fl_type = F_UNLCK, | |
2359 | .fl_end = OFFSET_MAX, | |
2360 | }; | |
2361 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2362 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2363 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2364 | } |
2365 | ||
1c8c601a | 2366 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2367 | before = &inode->i_flock; |
2368 | ||
2369 | while ((fl = *before) != NULL) { | |
2370 | if (fl->fl_file == filp) { | |
1da177e4 LT |
2371 | if (IS_LEASE(fl)) { |
2372 | lease_modify(before, F_UNLCK); | |
2373 | continue; | |
2374 | } | |
8c3cac5e JL |
2375 | |
2376 | /* | |
2377 | * There's a leftover lock on the list of a type that | |
2378 | * we didn't expect to see. Most likely a classic | |
2379 | * POSIX lock that ended up not getting released | |
2380 | * properly, or that raced onto the list somehow. Log | |
2381 | * some info about it and then just remove it from | |
2382 | * the list. | |
2383 | */ | |
2384 | WARN(!IS_FLOCK(fl), | |
2385 | "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", | |
2386 | MAJOR(inode->i_sb->s_dev), | |
2387 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2388 | fl->fl_type, fl->fl_flags, | |
2389 | fl->fl_start, fl->fl_end); | |
2390 | ||
ed9814d8 | 2391 | locks_delete_lock(before, &dispose); |
8c3cac5e | 2392 | continue; |
1da177e4 LT |
2393 | } |
2394 | before = &fl->fl_next; | |
2395 | } | |
1c8c601a | 2396 | spin_unlock(&inode->i_lock); |
ed9814d8 | 2397 | locks_dispose_list(&dispose); |
1da177e4 LT |
2398 | } |
2399 | ||
1da177e4 LT |
2400 | /** |
2401 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2402 | * @waiter: the lock which was waiting |
2403 | * | |
2404 | * lockd needs to block waiting for locks. | |
2405 | */ | |
64a318ee | 2406 | int |
f891a29f | 2407 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2408 | { |
64a318ee BF |
2409 | int status = 0; |
2410 | ||
7b2296af | 2411 | spin_lock(&blocked_lock_lock); |
5996a298 | 2412 | if (waiter->fl_next) |
1da177e4 | 2413 | __locks_delete_block(waiter); |
64a318ee BF |
2414 | else |
2415 | status = -ENOENT; | |
7b2296af | 2416 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2417 | return status; |
1da177e4 | 2418 | } |
1da177e4 LT |
2419 | EXPORT_SYMBOL(posix_unblock_lock); |
2420 | ||
9b9d2ab4 ME |
2421 | /** |
2422 | * vfs_cancel_lock - file byte range unblock lock | |
2423 | * @filp: The file to apply the unblock to | |
2424 | * @fl: The lock to be unblocked | |
2425 | * | |
2426 | * Used by lock managers to cancel blocked requests | |
2427 | */ | |
2428 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2429 | { | |
72c2d531 | 2430 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2431 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2432 | return 0; | |
2433 | } | |
2434 | ||
2435 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2436 | ||
7f8ada98 | 2437 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2438 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2439 | #include <linux/seq_file.h> |
2440 | ||
7012b02a JL |
2441 | struct locks_iterator { |
2442 | int li_cpu; | |
2443 | loff_t li_pos; | |
2444 | }; | |
2445 | ||
7f8ada98 | 2446 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2447 | loff_t id, char *pfx) |
1da177e4 LT |
2448 | { |
2449 | struct inode *inode = NULL; | |
ab1f1611 VG |
2450 | unsigned int fl_pid; |
2451 | ||
2452 | if (fl->fl_nspid) | |
6c5f3e7b | 2453 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2454 | else |
2455 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2456 | |
2457 | if (fl->fl_file != NULL) | |
496ad9aa | 2458 | inode = file_inode(fl->fl_file); |
1da177e4 | 2459 | |
99dc8292 | 2460 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2461 | if (IS_POSIX(fl)) { |
c918d42a | 2462 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2463 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2464 | else if (IS_OFDLCK(fl)) |
5315c26a | 2465 | seq_puts(f, "OFDLCK"); |
c918d42a | 2466 | else |
5315c26a | 2467 | seq_puts(f, "POSIX "); |
c918d42a JL |
2468 | |
2469 | seq_printf(f, " %s ", | |
1da177e4 | 2470 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2471 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2472 | } else if (IS_FLOCK(fl)) { |
2473 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2474 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2475 | } else { |
5315c26a | 2476 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2477 | } |
2478 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2479 | if (fl->fl_flags & FL_DELEG) |
2480 | seq_puts(f, "DELEG "); | |
2481 | else | |
2482 | seq_puts(f, "LEASE "); | |
2483 | ||
ab83fa4b | 2484 | if (lease_breaking(fl)) |
5315c26a | 2485 | seq_puts(f, "BREAKING "); |
1da177e4 | 2486 | else if (fl->fl_file) |
5315c26a | 2487 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2488 | else |
5315c26a | 2489 | seq_puts(f, "BREAKER "); |
1da177e4 | 2490 | } else { |
5315c26a | 2491 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2492 | } |
2493 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2494 | seq_printf(f, "%s ", |
1da177e4 LT |
2495 | (fl->fl_type & LOCK_READ) |
2496 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2497 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2498 | } else { | |
7f8ada98 | 2499 | seq_printf(f, "%s ", |
ab83fa4b | 2500 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2501 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2502 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2503 | } |
2504 | if (inode) { | |
2505 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2506 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2507 | inode->i_sb->s_id, inode->i_ino); |
2508 | #else | |
2509 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2510 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2511 | MAJOR(inode->i_sb->s_dev), |
2512 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2513 | #endif | |
2514 | } else { | |
ab1f1611 | 2515 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2516 | } |
2517 | if (IS_POSIX(fl)) { | |
2518 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2519 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2520 | else |
7f8ada98 | 2521 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2522 | } else { |
5315c26a | 2523 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2524 | } |
2525 | } | |
2526 | ||
7f8ada98 | 2527 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2528 | { |
7012b02a | 2529 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2530 | struct file_lock *fl, *bfl; |
1da177e4 | 2531 | |
139ca04e | 2532 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2533 | |
7012b02a | 2534 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2535 | |
7f8ada98 | 2536 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2537 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2538 | |
7f8ada98 PE |
2539 | return 0; |
2540 | } | |
1da177e4 | 2541 | |
7f8ada98 | 2542 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2543 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2544 | { |
7012b02a | 2545 | struct locks_iterator *iter = f->private; |
99dc8292 | 2546 | |
7012b02a JL |
2547 | iter->li_pos = *pos + 1; |
2548 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2549 | spin_lock(&blocked_lock_lock); |
7012b02a | 2550 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2551 | } |
1da177e4 | 2552 | |
7f8ada98 PE |
2553 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2554 | { | |
7012b02a JL |
2555 | struct locks_iterator *iter = f->private; |
2556 | ||
2557 | ++iter->li_pos; | |
2558 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2559 | } |
1da177e4 | 2560 | |
7f8ada98 | 2561 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2562 | __releases(&blocked_lock_lock) |
7f8ada98 | 2563 | { |
7b2296af | 2564 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2565 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2566 | } |
2567 | ||
d8ba7a36 | 2568 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2569 | .start = locks_start, |
2570 | .next = locks_next, | |
2571 | .stop = locks_stop, | |
2572 | .show = locks_show, | |
2573 | }; | |
d8ba7a36 AD |
2574 | |
2575 | static int locks_open(struct inode *inode, struct file *filp) | |
2576 | { | |
7012b02a JL |
2577 | return seq_open_private(filp, &locks_seq_operations, |
2578 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2579 | } |
2580 | ||
2581 | static const struct file_operations proc_locks_operations = { | |
2582 | .open = locks_open, | |
2583 | .read = seq_read, | |
2584 | .llseek = seq_lseek, | |
99dc8292 | 2585 | .release = seq_release_private, |
d8ba7a36 AD |
2586 | }; |
2587 | ||
2588 | static int __init proc_locks_init(void) | |
2589 | { | |
2590 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2591 | return 0; | |
2592 | } | |
2593 | module_init(proc_locks_init); | |
7f8ada98 PE |
2594 | #endif |
2595 | ||
1da177e4 LT |
2596 | /** |
2597 | * lock_may_read - checks that the region is free of locks | |
2598 | * @inode: the inode that is being read | |
2599 | * @start: the first byte to read | |
2600 | * @len: the number of bytes to read | |
2601 | * | |
2602 | * Emulates Windows locking requirements. Whole-file | |
2603 | * mandatory locks (share modes) can prohibit a read and | |
2604 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2605 | * | |
2606 | * N.B. this function is only ever called | |
2607 | * from knfsd and ownership of locks is never checked. | |
2608 | */ | |
2609 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2610 | { | |
2611 | struct file_lock *fl; | |
2612 | int result = 1; | |
1c8c601a JL |
2613 | |
2614 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2615 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2616 | if (IS_POSIX(fl)) { | |
2617 | if (fl->fl_type == F_RDLCK) | |
2618 | continue; | |
2619 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2620 | continue; | |
2621 | } else if (IS_FLOCK(fl)) { | |
2622 | if (!(fl->fl_type & LOCK_MAND)) | |
2623 | continue; | |
2624 | if (fl->fl_type & LOCK_READ) | |
2625 | continue; | |
2626 | } else | |
2627 | continue; | |
2628 | result = 0; | |
2629 | break; | |
2630 | } | |
1c8c601a | 2631 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2632 | return result; |
2633 | } | |
2634 | ||
2635 | EXPORT_SYMBOL(lock_may_read); | |
2636 | ||
2637 | /** | |
2638 | * lock_may_write - checks that the region is free of locks | |
2639 | * @inode: the inode that is being written | |
2640 | * @start: the first byte to write | |
2641 | * @len: the number of bytes to write | |
2642 | * | |
2643 | * Emulates Windows locking requirements. Whole-file | |
2644 | * mandatory locks (share modes) can prohibit a write and | |
2645 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2646 | * | |
2647 | * N.B. this function is only ever called | |
2648 | * from knfsd and ownership of locks is never checked. | |
2649 | */ | |
2650 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2651 | { | |
2652 | struct file_lock *fl; | |
2653 | int result = 1; | |
1c8c601a JL |
2654 | |
2655 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2656 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2657 | if (IS_POSIX(fl)) { | |
2658 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2659 | continue; | |
2660 | } else if (IS_FLOCK(fl)) { | |
2661 | if (!(fl->fl_type & LOCK_MAND)) | |
2662 | continue; | |
2663 | if (fl->fl_type & LOCK_WRITE) | |
2664 | continue; | |
2665 | } else | |
2666 | continue; | |
2667 | result = 0; | |
2668 | break; | |
2669 | } | |
1c8c601a | 2670 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2671 | return result; |
2672 | } | |
2673 | ||
2674 | EXPORT_SYMBOL(lock_may_write); | |
2675 | ||
1da177e4 LT |
2676 | static int __init filelock_init(void) |
2677 | { | |
7012b02a JL |
2678 | int i; |
2679 | ||
1da177e4 | 2680 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2681 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2682 | ||
7012b02a JL |
2683 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2684 | ||
2685 | for_each_possible_cpu(i) | |
2686 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2687 | ||
1da177e4 LT |
2688 | return 0; |
2689 | } | |
2690 | ||
2691 | core_initcall(filelock_init); |