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> |
1da177e4 | 129 | |
1da177e4 LT |
130 | #include <asm/uaccess.h> |
131 | ||
132 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
133 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
134 | #define IS_LEASE(fl) (fl->fl_flags & FL_LEASE) | |
135 | ||
ab83fa4b BF |
136 | static bool lease_breaking(struct file_lock *fl) |
137 | { | |
778fc546 BF |
138 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
139 | } | |
140 | ||
141 | static int target_leasetype(struct file_lock *fl) | |
142 | { | |
143 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
144 | return F_UNLCK; | |
145 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
146 | return F_RDLCK; | |
147 | return fl->fl_type; | |
ab83fa4b BF |
148 | } |
149 | ||
1da177e4 LT |
150 | int leases_enable = 1; |
151 | int lease_break_time = 45; | |
152 | ||
153 | #define for_each_lock(inode, lockp) \ | |
154 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
155 | ||
1c8c601a JL |
156 | /* |
157 | * The global file_lock_list is only used for displaying /proc/locks. Protected | |
158 | * by the file_lock_lock. | |
159 | */ | |
26bcbf96 | 160 | static LIST_HEAD(file_lock_list); |
88974691 | 161 | |
1c8c601a JL |
162 | /* |
163 | * The blocked_list is used to find POSIX lock loops for deadlock detection. | |
164 | * Protected by file_lock_lock. | |
165 | */ | |
1da177e4 | 166 | static LIST_HEAD(blocked_list); |
88974691 | 167 | |
1c8c601a JL |
168 | /* |
169 | * This lock protects the blocked_list, and the file_lock_list. Generally, if | |
170 | * you're accessing one of those lists, you want to be holding this lock. | |
171 | * | |
172 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
173 | * pointer for file_lock structures that are acting as lock requests (in | |
174 | * contrast to those that are acting as records of acquired locks). | |
175 | * | |
176 | * Note that when we acquire this lock in order to change the above fields, | |
177 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
178 | * protected by this lock, we can skip acquiring it iff we already hold the | |
179 | * i_lock. | |
180 | * | |
181 | * In particular, adding an entry to the fl_block list requires that you hold | |
182 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
183 | * an entry from the list however only requires the file_lock_lock. | |
184 | */ | |
72f98e72 | 185 | static DEFINE_SPINLOCK(file_lock_lock); |
1da177e4 | 186 | |
e18b890b | 187 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 188 | |
ee19cc40 | 189 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 190 | { |
ee19cc40 MS |
191 | INIT_LIST_HEAD(&fl->fl_link); |
192 | INIT_LIST_HEAD(&fl->fl_block); | |
193 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
194 | } |
195 | ||
1da177e4 | 196 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 197 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 198 | { |
ee19cc40 | 199 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
200 | |
201 | if (fl) | |
ee19cc40 | 202 | locks_init_lock_heads(fl); |
a51cb91d MS |
203 | |
204 | return fl; | |
1da177e4 | 205 | } |
c5b1f0d9 | 206 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 207 | |
a9e61e25 | 208 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
209 | { |
210 | if (fl->fl_ops) { | |
211 | if (fl->fl_ops->fl_release_private) | |
212 | fl->fl_ops->fl_release_private(fl); | |
213 | fl->fl_ops = NULL; | |
214 | } | |
068535f1 | 215 | fl->fl_lmops = NULL; |
47831f35 TM |
216 | |
217 | } | |
a9e61e25 | 218 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 219 | |
1da177e4 | 220 | /* Free a lock which is not in use. */ |
05fa3135 | 221 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 222 | { |
5ce29646 MS |
223 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
224 | BUG_ON(!list_empty(&fl->fl_block)); | |
225 | BUG_ON(!list_empty(&fl->fl_link)); | |
1da177e4 | 226 | |
47831f35 | 227 | locks_release_private(fl); |
1da177e4 LT |
228 | kmem_cache_free(filelock_cache, fl); |
229 | } | |
05fa3135 | 230 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
231 | |
232 | void locks_init_lock(struct file_lock *fl) | |
233 | { | |
ee19cc40 MS |
234 | memset(fl, 0, sizeof(struct file_lock)); |
235 | locks_init_lock_heads(fl); | |
1da177e4 LT |
236 | } |
237 | ||
238 | EXPORT_SYMBOL(locks_init_lock); | |
239 | ||
47831f35 TM |
240 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
241 | { | |
242 | if (fl->fl_ops) { | |
243 | if (fl->fl_ops->fl_copy_lock) | |
244 | fl->fl_ops->fl_copy_lock(new, fl); | |
245 | new->fl_ops = fl->fl_ops; | |
246 | } | |
bb8430a2 | 247 | if (fl->fl_lmops) |
47831f35 | 248 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
249 | } |
250 | ||
1da177e4 LT |
251 | /* |
252 | * Initialize a new lock from an existing file_lock structure. | |
253 | */ | |
1a747ee0 | 254 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
255 | { |
256 | new->fl_owner = fl->fl_owner; | |
257 | new->fl_pid = fl->fl_pid; | |
0996905f | 258 | new->fl_file = NULL; |
1da177e4 LT |
259 | new->fl_flags = fl->fl_flags; |
260 | new->fl_type = fl->fl_type; | |
261 | new->fl_start = fl->fl_start; | |
262 | new->fl_end = fl->fl_end; | |
0996905f TM |
263 | new->fl_ops = NULL; |
264 | new->fl_lmops = NULL; | |
265 | } | |
3dd7b71c | 266 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
267 | |
268 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
269 | { | |
270 | locks_release_private(new); | |
271 | ||
272 | __locks_copy_lock(new, fl); | |
273 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
274 | new->fl_ops = fl->fl_ops; |
275 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
276 | |
277 | locks_copy_private(new, fl); | |
1da177e4 LT |
278 | } |
279 | ||
280 | EXPORT_SYMBOL(locks_copy_lock); | |
281 | ||
282 | static inline int flock_translate_cmd(int cmd) { | |
283 | if (cmd & LOCK_MAND) | |
284 | return cmd & (LOCK_MAND | LOCK_RW); | |
285 | switch (cmd) { | |
286 | case LOCK_SH: | |
287 | return F_RDLCK; | |
288 | case LOCK_EX: | |
289 | return F_WRLCK; | |
290 | case LOCK_UN: | |
291 | return F_UNLCK; | |
292 | } | |
293 | return -EINVAL; | |
294 | } | |
295 | ||
296 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
297 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
298 | unsigned int cmd) | |
299 | { | |
300 | struct file_lock *fl; | |
301 | int type = flock_translate_cmd(cmd); | |
302 | if (type < 0) | |
303 | return type; | |
304 | ||
305 | fl = locks_alloc_lock(); | |
306 | if (fl == NULL) | |
307 | return -ENOMEM; | |
308 | ||
309 | fl->fl_file = filp; | |
310 | fl->fl_pid = current->tgid; | |
311 | fl->fl_flags = FL_FLOCK; | |
312 | fl->fl_type = type; | |
313 | fl->fl_end = OFFSET_MAX; | |
314 | ||
315 | *lock = fl; | |
316 | return 0; | |
317 | } | |
318 | ||
0ec4f431 | 319 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
320 | { |
321 | switch (type) { | |
322 | case F_RDLCK: | |
323 | case F_WRLCK: | |
324 | case F_UNLCK: | |
325 | fl->fl_type = type; | |
326 | break; | |
327 | default: | |
328 | return -EINVAL; | |
329 | } | |
330 | return 0; | |
331 | } | |
332 | ||
333 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX | |
334 | * style lock. | |
335 | */ | |
336 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
337 | struct flock *l) | |
338 | { | |
339 | off_t start, end; | |
340 | ||
341 | switch (l->l_whence) { | |
f5579f8c | 342 | case SEEK_SET: |
1da177e4 LT |
343 | start = 0; |
344 | break; | |
f5579f8c | 345 | case SEEK_CUR: |
1da177e4 LT |
346 | start = filp->f_pos; |
347 | break; | |
f5579f8c | 348 | case SEEK_END: |
496ad9aa | 349 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
350 | break; |
351 | default: | |
352 | return -EINVAL; | |
353 | } | |
354 | ||
355 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
356 | POSIX-2001 defines it. */ | |
357 | start += l->l_start; | |
4c780a46 TM |
358 | if (start < 0) |
359 | return -EINVAL; | |
360 | fl->fl_end = OFFSET_MAX; | |
361 | if (l->l_len > 0) { | |
362 | end = start + l->l_len - 1; | |
363 | fl->fl_end = end; | |
364 | } else if (l->l_len < 0) { | |
1da177e4 | 365 | end = start - 1; |
4c780a46 | 366 | fl->fl_end = end; |
1da177e4 | 367 | start += l->l_len; |
4c780a46 TM |
368 | if (start < 0) |
369 | return -EINVAL; | |
1da177e4 | 370 | } |
1da177e4 | 371 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
372 | if (fl->fl_end < fl->fl_start) |
373 | return -EOVERFLOW; | |
1da177e4 LT |
374 | |
375 | fl->fl_owner = current->files; | |
376 | fl->fl_pid = current->tgid; | |
377 | fl->fl_file = filp; | |
378 | fl->fl_flags = FL_POSIX; | |
379 | fl->fl_ops = NULL; | |
380 | fl->fl_lmops = NULL; | |
381 | ||
382 | return assign_type(fl, l->l_type); | |
383 | } | |
384 | ||
385 | #if BITS_PER_LONG == 32 | |
386 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, | |
387 | struct flock64 *l) | |
388 | { | |
389 | loff_t start; | |
390 | ||
391 | switch (l->l_whence) { | |
f5579f8c | 392 | case SEEK_SET: |
1da177e4 LT |
393 | start = 0; |
394 | break; | |
f5579f8c | 395 | case SEEK_CUR: |
1da177e4 LT |
396 | start = filp->f_pos; |
397 | break; | |
f5579f8c | 398 | case SEEK_END: |
496ad9aa | 399 | start = i_size_read(file_inode(filp)); |
1da177e4 LT |
400 | break; |
401 | default: | |
402 | return -EINVAL; | |
403 | } | |
404 | ||
4c780a46 TM |
405 | start += l->l_start; |
406 | if (start < 0) | |
1da177e4 | 407 | return -EINVAL; |
4c780a46 TM |
408 | fl->fl_end = OFFSET_MAX; |
409 | if (l->l_len > 0) { | |
410 | fl->fl_end = start + l->l_len - 1; | |
411 | } else if (l->l_len < 0) { | |
412 | fl->fl_end = start - 1; | |
413 | start += l->l_len; | |
414 | if (start < 0) | |
415 | return -EINVAL; | |
416 | } | |
1da177e4 | 417 | fl->fl_start = start; /* we record the absolute position */ |
4c780a46 TM |
418 | if (fl->fl_end < fl->fl_start) |
419 | return -EOVERFLOW; | |
1da177e4 LT |
420 | |
421 | fl->fl_owner = current->files; | |
422 | fl->fl_pid = current->tgid; | |
423 | fl->fl_file = filp; | |
424 | fl->fl_flags = FL_POSIX; | |
425 | fl->fl_ops = NULL; | |
426 | fl->fl_lmops = NULL; | |
427 | ||
f32cb532 | 428 | return assign_type(fl, l->l_type); |
1da177e4 LT |
429 | } |
430 | #endif | |
431 | ||
432 | /* default lease lock manager operations */ | |
433 | static void lease_break_callback(struct file_lock *fl) | |
434 | { | |
435 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
436 | } | |
437 | ||
7b021967 | 438 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 439 | .lm_break = lease_break_callback, |
8fb47a4f | 440 | .lm_change = lease_modify, |
1da177e4 LT |
441 | }; |
442 | ||
443 | /* | |
444 | * Initialize a lease, use the default lock manager operations | |
445 | */ | |
0ec4f431 | 446 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 447 | { |
75dff55a TM |
448 | if (assign_type(fl, type) != 0) |
449 | return -EINVAL; | |
450 | ||
1da177e4 LT |
451 | fl->fl_owner = current->files; |
452 | fl->fl_pid = current->tgid; | |
453 | ||
454 | fl->fl_file = filp; | |
455 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
456 | fl->fl_start = 0; |
457 | fl->fl_end = OFFSET_MAX; | |
458 | fl->fl_ops = NULL; | |
459 | fl->fl_lmops = &lease_manager_ops; | |
460 | return 0; | |
461 | } | |
462 | ||
463 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 464 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
465 | { |
466 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 467 | int error = -ENOMEM; |
1da177e4 LT |
468 | |
469 | if (fl == NULL) | |
e32b8ee2 | 470 | return ERR_PTR(error); |
1da177e4 LT |
471 | |
472 | error = lease_init(filp, type, fl); | |
75dff55a TM |
473 | if (error) { |
474 | locks_free_lock(fl); | |
e32b8ee2 | 475 | return ERR_PTR(error); |
75dff55a | 476 | } |
e32b8ee2 | 477 | return fl; |
1da177e4 LT |
478 | } |
479 | ||
480 | /* Check if two locks overlap each other. | |
481 | */ | |
482 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
483 | { | |
484 | return ((fl1->fl_end >= fl2->fl_start) && | |
485 | (fl2->fl_end >= fl1->fl_start)); | |
486 | } | |
487 | ||
488 | /* | |
489 | * Check whether two locks have the same owner. | |
490 | */ | |
33443c42 | 491 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 492 | { |
8fb47a4f | 493 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 494 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 495 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
496 | return fl1->fl_owner == fl2->fl_owner; |
497 | } | |
498 | ||
88974691 JL |
499 | static inline void |
500 | locks_insert_global_locks(struct file_lock *fl) | |
501 | { | |
1c8c601a | 502 | spin_lock(&file_lock_lock); |
88974691 | 503 | list_add_tail(&fl->fl_link, &file_lock_list); |
1c8c601a | 504 | spin_unlock(&file_lock_lock); |
88974691 JL |
505 | } |
506 | ||
507 | static inline void | |
508 | locks_delete_global_locks(struct file_lock *fl) | |
509 | { | |
1c8c601a | 510 | spin_lock(&file_lock_lock); |
88974691 | 511 | list_del_init(&fl->fl_link); |
1c8c601a | 512 | spin_unlock(&file_lock_lock); |
88974691 JL |
513 | } |
514 | ||
515 | static inline void | |
516 | locks_insert_global_blocked(struct file_lock *waiter) | |
517 | { | |
518 | list_add(&waiter->fl_link, &blocked_list); | |
519 | } | |
520 | ||
521 | static inline void | |
522 | locks_delete_global_blocked(struct file_lock *waiter) | |
523 | { | |
524 | list_del_init(&waiter->fl_link); | |
525 | } | |
526 | ||
1da177e4 LT |
527 | /* Remove waiter from blocker's block list. |
528 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a JL |
529 | * |
530 | * Must be called with file_lock_lock held. | |
1da177e4 | 531 | */ |
33443c42 | 532 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 533 | { |
88974691 | 534 | locks_delete_global_blocked(waiter); |
1da177e4 | 535 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
536 | waiter->fl_next = NULL; |
537 | } | |
538 | ||
1a9e64a7 | 539 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 540 | { |
1c8c601a | 541 | spin_lock(&file_lock_lock); |
1da177e4 | 542 | __locks_delete_block(waiter); |
1c8c601a | 543 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
544 | } |
545 | ||
546 | /* Insert waiter into blocker's block list. | |
547 | * We use a circular list so that processes can be easily woken up in | |
548 | * the order they blocked. The documentation doesn't require this but | |
549 | * it seems like the reasonable thing to do. | |
1c8c601a | 550 | * |
4e8c765d JL |
551 | * Must be called with both the i_lock and file_lock_lock held. The fl_block |
552 | * list itself is protected by the file_lock_list, but by ensuring that the | |
553 | * i_lock is also held on insertions we can avoid taking the file_lock_lock | |
554 | * in some cases when we see that the fl_block list is empty. | |
1da177e4 | 555 | */ |
1c8c601a JL |
556 | static void __locks_insert_block(struct file_lock *blocker, |
557 | struct file_lock *waiter) | |
1da177e4 | 558 | { |
6dc0fe8f | 559 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 560 | waiter->fl_next = blocker; |
88974691 | 561 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
1da177e4 | 562 | if (IS_POSIX(blocker)) |
1c8c601a JL |
563 | locks_insert_global_blocked(waiter); |
564 | } | |
565 | ||
566 | /* Must be called with i_lock held. */ | |
567 | static void locks_insert_block(struct file_lock *blocker, | |
568 | struct file_lock *waiter) | |
569 | { | |
570 | spin_lock(&file_lock_lock); | |
571 | __locks_insert_block(blocker, waiter); | |
572 | spin_unlock(&file_lock_lock); | |
1da177e4 LT |
573 | } |
574 | ||
1cb36012 JL |
575 | /* |
576 | * Wake up processes blocked waiting for blocker. | |
577 | * | |
1c8c601a | 578 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
579 | */ |
580 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
581 | { | |
4e8c765d JL |
582 | /* |
583 | * Avoid taking global lock if list is empty. This is safe since new | |
584 | * blocked requests are only added to the list under the i_lock, and | |
585 | * the i_lock is always held here. Note that removal from the fl_block | |
586 | * list does not require the i_lock, so we must recheck list_empty() | |
587 | * after acquiring the file_lock_lock. | |
588 | */ | |
589 | if (list_empty(&blocker->fl_block)) | |
590 | return; | |
591 | ||
1c8c601a | 592 | spin_lock(&file_lock_lock); |
1da177e4 | 593 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
594 | struct file_lock *waiter; |
595 | ||
596 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
597 | struct file_lock, fl_block); |
598 | __locks_delete_block(waiter); | |
8fb47a4f BF |
599 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
600 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
601 | else |
602 | wake_up(&waiter->fl_wait); | |
603 | } | |
1c8c601a | 604 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
605 | } |
606 | ||
607 | /* Insert file lock fl into an inode's lock list at the position indicated | |
608 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
609 | * |
610 | * Must be called with the i_lock held! | |
1da177e4 LT |
611 | */ |
612 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
613 | { | |
ab1f1611 VG |
614 | fl->fl_nspid = get_pid(task_tgid(current)); |
615 | ||
1da177e4 LT |
616 | /* insert into file's list */ |
617 | fl->fl_next = *pos; | |
618 | *pos = fl; | |
88974691 JL |
619 | |
620 | locks_insert_global_locks(fl); | |
1da177e4 LT |
621 | } |
622 | ||
623 | /* | |
624 | * Delete a lock and then free it. | |
625 | * Wake up processes that are blocked waiting for this lock, | |
626 | * notify the FS that the lock has been cleared and | |
627 | * finally free the lock. | |
1c8c601a JL |
628 | * |
629 | * Must be called with the i_lock held! | |
1da177e4 LT |
630 | */ |
631 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
632 | { | |
633 | struct file_lock *fl = *thisfl_p; | |
634 | ||
88974691 JL |
635 | locks_delete_global_locks(fl); |
636 | ||
1da177e4 LT |
637 | *thisfl_p = fl->fl_next; |
638 | fl->fl_next = NULL; | |
1da177e4 | 639 | |
ab1f1611 VG |
640 | if (fl->fl_nspid) { |
641 | put_pid(fl->fl_nspid); | |
642 | fl->fl_nspid = NULL; | |
643 | } | |
644 | ||
1da177e4 LT |
645 | locks_wake_up_blocks(fl); |
646 | locks_free_lock(fl); | |
647 | } | |
648 | ||
649 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
650 | * checks for shared/exclusive status of overlapping locks. | |
651 | */ | |
652 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
653 | { | |
654 | if (sys_fl->fl_type == F_WRLCK) | |
655 | return 1; | |
656 | if (caller_fl->fl_type == F_WRLCK) | |
657 | return 1; | |
658 | return 0; | |
659 | } | |
660 | ||
661 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
662 | * checking before calling the locks_conflict(). | |
663 | */ | |
664 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
665 | { | |
666 | /* POSIX locks owned by the same process do not conflict with | |
667 | * each other. | |
668 | */ | |
669 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
670 | return (0); | |
671 | ||
672 | /* Check whether they overlap */ | |
673 | if (!locks_overlap(caller_fl, sys_fl)) | |
674 | return 0; | |
675 | ||
676 | return (locks_conflict(caller_fl, sys_fl)); | |
677 | } | |
678 | ||
679 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
680 | * checking before calling the locks_conflict(). | |
681 | */ | |
682 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
683 | { | |
684 | /* FLOCK locks referring to the same filp do not conflict with | |
685 | * each other. | |
686 | */ | |
687 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
688 | return (0); | |
689 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
690 | return 0; | |
691 | ||
692 | return (locks_conflict(caller_fl, sys_fl)); | |
693 | } | |
694 | ||
6d34ac19 | 695 | void |
9d6a8c5c | 696 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
697 | { |
698 | struct file_lock *cfl; | |
1c8c601a | 699 | struct inode *inode = file_inode(filp); |
1da177e4 | 700 | |
1c8c601a | 701 | spin_lock(&inode->i_lock); |
496ad9aa | 702 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
703 | if (!IS_POSIX(cfl)) |
704 | continue; | |
b842e240 | 705 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
706 | break; |
707 | } | |
ab1f1611 | 708 | if (cfl) { |
9d6a8c5c | 709 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 710 | if (cfl->fl_nspid) |
6c5f3e7b | 711 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 712 | } else |
129a84de | 713 | fl->fl_type = F_UNLCK; |
1c8c601a | 714 | spin_unlock(&inode->i_lock); |
6d34ac19 | 715 | return; |
1da177e4 | 716 | } |
1da177e4 LT |
717 | EXPORT_SYMBOL(posix_test_lock); |
718 | ||
b533184f BF |
719 | /* |
720 | * Deadlock detection: | |
721 | * | |
722 | * We attempt to detect deadlocks that are due purely to posix file | |
723 | * locks. | |
1da177e4 | 724 | * |
b533184f BF |
725 | * We assume that a task can be waiting for at most one lock at a time. |
726 | * So for any acquired lock, the process holding that lock may be | |
727 | * waiting on at most one other lock. That lock in turns may be held by | |
728 | * someone waiting for at most one other lock. Given a requested lock | |
729 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
730 | * follow this chain of waiters to ensure we are not about to create a | |
731 | * cycle. | |
1da177e4 | 732 | * |
b533184f BF |
733 | * Since we do this before we ever put a process to sleep on a lock, we |
734 | * are ensured that there is never a cycle; that is what guarantees that | |
735 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 736 | * |
b533184f BF |
737 | * Note: the above assumption may not be true when handling lock |
738 | * requests from a broken NFS client. It may also fail in the presence | |
739 | * of tasks (such as posix threads) sharing the same open file table. | |
740 | * | |
741 | * To handle those cases, we just bail out after a few iterations. | |
1da177e4 | 742 | */ |
97855b49 BF |
743 | |
744 | #define MAX_DEADLK_ITERATIONS 10 | |
745 | ||
b533184f BF |
746 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
747 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
748 | { | |
749 | struct file_lock *fl; | |
750 | ||
751 | list_for_each_entry(fl, &blocked_list, fl_link) { | |
752 | if (posix_same_owner(fl, block_fl)) | |
753 | return fl->fl_next; | |
754 | } | |
755 | return NULL; | |
756 | } | |
757 | ||
1c8c601a | 758 | /* Must be called with the file_lock_lock held! */ |
b0904e14 | 759 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
760 | struct file_lock *block_fl) |
761 | { | |
97855b49 | 762 | int i = 0; |
1da177e4 | 763 | |
b533184f BF |
764 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
765 | if (i++ > MAX_DEADLK_ITERATIONS) | |
766 | return 0; | |
767 | if (posix_same_owner(caller_fl, block_fl)) | |
768 | return 1; | |
1da177e4 LT |
769 | } |
770 | return 0; | |
771 | } | |
772 | ||
1da177e4 | 773 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 774 | * after any leases, but before any posix locks. |
f475ae95 TM |
775 | * |
776 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
777 | * whether or not a lock was successfully freed by testing the return | |
778 | * value for -ENOENT. | |
1da177e4 | 779 | */ |
993dfa87 | 780 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 781 | { |
993dfa87 | 782 | struct file_lock *new_fl = NULL; |
1da177e4 | 783 | struct file_lock **before; |
496ad9aa | 784 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
785 | int error = 0; |
786 | int found = 0; | |
787 | ||
b89f4321 | 788 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 789 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
790 | if (!new_fl) |
791 | return -ENOMEM; | |
84d535ad PE |
792 | } |
793 | ||
1c8c601a | 794 | spin_lock(&inode->i_lock); |
b89f4321 AB |
795 | if (request->fl_flags & FL_ACCESS) |
796 | goto find_conflict; | |
797 | ||
1da177e4 LT |
798 | for_each_lock(inode, before) { |
799 | struct file_lock *fl = *before; | |
800 | if (IS_POSIX(fl)) | |
801 | break; | |
802 | if (IS_LEASE(fl)) | |
803 | continue; | |
804 | if (filp != fl->fl_file) | |
805 | continue; | |
993dfa87 | 806 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
807 | goto out; |
808 | found = 1; | |
809 | locks_delete_lock(before); | |
810 | break; | |
811 | } | |
1da177e4 | 812 | |
f475ae95 TM |
813 | if (request->fl_type == F_UNLCK) { |
814 | if ((request->fl_flags & FL_EXISTS) && !found) | |
815 | error = -ENOENT; | |
993dfa87 | 816 | goto out; |
f475ae95 | 817 | } |
1da177e4 LT |
818 | |
819 | /* | |
820 | * If a higher-priority process was blocked on the old file lock, | |
821 | * give it the opportunity to lock the file. | |
822 | */ | |
b89f4321 | 823 | if (found) { |
1c8c601a | 824 | spin_unlock(&inode->i_lock); |
def01bc5 | 825 | cond_resched(); |
1c8c601a | 826 | spin_lock(&inode->i_lock); |
b89f4321 | 827 | } |
1da177e4 | 828 | |
f07f18dd | 829 | find_conflict: |
1da177e4 LT |
830 | for_each_lock(inode, before) { |
831 | struct file_lock *fl = *before; | |
832 | if (IS_POSIX(fl)) | |
833 | break; | |
834 | if (IS_LEASE(fl)) | |
835 | continue; | |
993dfa87 | 836 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
837 | continue; |
838 | error = -EAGAIN; | |
bde74e4b MS |
839 | if (!(request->fl_flags & FL_SLEEP)) |
840 | goto out; | |
841 | error = FILE_LOCK_DEFERRED; | |
842 | locks_insert_block(fl, request); | |
1da177e4 LT |
843 | goto out; |
844 | } | |
f07f18dd TM |
845 | if (request->fl_flags & FL_ACCESS) |
846 | goto out; | |
993dfa87 | 847 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 848 | locks_insert_lock(before, new_fl); |
993dfa87 | 849 | new_fl = NULL; |
9cedc194 | 850 | error = 0; |
1da177e4 LT |
851 | |
852 | out: | |
1c8c601a | 853 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
854 | if (new_fl) |
855 | locks_free_lock(new_fl); | |
1da177e4 LT |
856 | return error; |
857 | } | |
858 | ||
150b3934 | 859 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
860 | { |
861 | struct file_lock *fl; | |
39005d02 MS |
862 | struct file_lock *new_fl = NULL; |
863 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
864 | struct file_lock *left = NULL; |
865 | struct file_lock *right = NULL; | |
866 | struct file_lock **before; | |
b9746ef8 JL |
867 | int error; |
868 | bool added = false; | |
1da177e4 LT |
869 | |
870 | /* | |
871 | * We may need two file_lock structures for this operation, | |
872 | * so we get them in advance to avoid races. | |
39005d02 MS |
873 | * |
874 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 875 | */ |
39005d02 MS |
876 | if (!(request->fl_flags & FL_ACCESS) && |
877 | (request->fl_type != F_UNLCK || | |
878 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
879 | new_fl = locks_alloc_lock(); | |
880 | new_fl2 = locks_alloc_lock(); | |
881 | } | |
1da177e4 | 882 | |
1c8c601a | 883 | spin_lock(&inode->i_lock); |
1cb36012 JL |
884 | /* |
885 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
886 | * there are any, either return error or put the request on the | |
887 | * blocker's list of waiters and the global blocked_list. | |
888 | */ | |
1da177e4 LT |
889 | if (request->fl_type != F_UNLCK) { |
890 | for_each_lock(inode, before) { | |
526985b9 | 891 | fl = *before; |
1da177e4 LT |
892 | if (!IS_POSIX(fl)) |
893 | continue; | |
894 | if (!posix_locks_conflict(request, fl)) | |
895 | continue; | |
5842add2 | 896 | if (conflock) |
1a747ee0 | 897 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
898 | error = -EAGAIN; |
899 | if (!(request->fl_flags & FL_SLEEP)) | |
900 | goto out; | |
1c8c601a JL |
901 | /* |
902 | * Deadlock detection and insertion into the blocked | |
903 | * locks list must be done while holding the same lock! | |
904 | */ | |
1da177e4 | 905 | error = -EDEADLK; |
1c8c601a JL |
906 | spin_lock(&file_lock_lock); |
907 | if (likely(!posix_locks_deadlock(request, fl))) { | |
908 | error = FILE_LOCK_DEFERRED; | |
909 | __locks_insert_block(fl, request); | |
910 | } | |
911 | spin_unlock(&file_lock_lock); | |
1da177e4 LT |
912 | goto out; |
913 | } | |
914 | } | |
915 | ||
916 | /* If we're just looking for a conflict, we're done. */ | |
917 | error = 0; | |
918 | if (request->fl_flags & FL_ACCESS) | |
919 | goto out; | |
920 | ||
1da177e4 | 921 | /* |
1da177e4 LT |
922 | * Find the first old lock with the same owner as the new lock. |
923 | */ | |
924 | ||
925 | before = &inode->i_flock; | |
926 | ||
927 | /* First skip locks owned by other processes. */ | |
928 | while ((fl = *before) && (!IS_POSIX(fl) || | |
929 | !posix_same_owner(request, fl))) { | |
930 | before = &fl->fl_next; | |
931 | } | |
932 | ||
1cb36012 | 933 | /* Process locks with this owner. */ |
1da177e4 LT |
934 | while ((fl = *before) && posix_same_owner(request, fl)) { |
935 | /* Detect adjacent or overlapping regions (if same lock type) | |
936 | */ | |
937 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
938 | /* In all comparisons of start vs end, use |
939 | * "start - 1" rather than "end + 1". If end | |
940 | * is OFFSET_MAX, end + 1 will become negative. | |
941 | */ | |
1da177e4 LT |
942 | if (fl->fl_end < request->fl_start - 1) |
943 | goto next_lock; | |
944 | /* If the next lock in the list has entirely bigger | |
945 | * addresses than the new one, insert the lock here. | |
946 | */ | |
449231d6 | 947 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
948 | break; |
949 | ||
950 | /* If we come here, the new and old lock are of the | |
951 | * same type and adjacent or overlapping. Make one | |
952 | * lock yielding from the lower start address of both | |
953 | * locks to the higher end address. | |
954 | */ | |
955 | if (fl->fl_start > request->fl_start) | |
956 | fl->fl_start = request->fl_start; | |
957 | else | |
958 | request->fl_start = fl->fl_start; | |
959 | if (fl->fl_end < request->fl_end) | |
960 | fl->fl_end = request->fl_end; | |
961 | else | |
962 | request->fl_end = fl->fl_end; | |
963 | if (added) { | |
964 | locks_delete_lock(before); | |
965 | continue; | |
966 | } | |
967 | request = fl; | |
b9746ef8 | 968 | added = true; |
1da177e4 LT |
969 | } |
970 | else { | |
971 | /* Processing for different lock types is a bit | |
972 | * more complex. | |
973 | */ | |
974 | if (fl->fl_end < request->fl_start) | |
975 | goto next_lock; | |
976 | if (fl->fl_start > request->fl_end) | |
977 | break; | |
978 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 979 | added = true; |
1da177e4 LT |
980 | if (fl->fl_start < request->fl_start) |
981 | left = fl; | |
982 | /* If the next lock in the list has a higher end | |
983 | * address than the new one, insert the new one here. | |
984 | */ | |
985 | if (fl->fl_end > request->fl_end) { | |
986 | right = fl; | |
987 | break; | |
988 | } | |
989 | if (fl->fl_start >= request->fl_start) { | |
990 | /* The new lock completely replaces an old | |
991 | * one (This may happen several times). | |
992 | */ | |
993 | if (added) { | |
994 | locks_delete_lock(before); | |
995 | continue; | |
996 | } | |
997 | /* Replace the old lock with the new one. | |
998 | * Wake up anybody waiting for the old one, | |
999 | * as the change in lock type might satisfy | |
1000 | * their needs. | |
1001 | */ | |
1002 | locks_wake_up_blocks(fl); | |
1003 | fl->fl_start = request->fl_start; | |
1004 | fl->fl_end = request->fl_end; | |
1005 | fl->fl_type = request->fl_type; | |
47831f35 TM |
1006 | locks_release_private(fl); |
1007 | locks_copy_private(fl, request); | |
1da177e4 | 1008 | request = fl; |
b9746ef8 | 1009 | added = true; |
1da177e4 LT |
1010 | } |
1011 | } | |
1012 | /* Go on to next lock. | |
1013 | */ | |
1014 | next_lock: | |
1015 | before = &fl->fl_next; | |
1016 | } | |
1017 | ||
0d9a490a | 1018 | /* |
1cb36012 JL |
1019 | * The above code only modifies existing locks in case of merging or |
1020 | * replacing. If new lock(s) need to be inserted all modifications are | |
1021 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1022 | */ |
1023 | error = -ENOLCK; /* "no luck" */ | |
1024 | if (right && left == right && !new_fl2) | |
1025 | goto out; | |
1026 | ||
1da177e4 LT |
1027 | error = 0; |
1028 | if (!added) { | |
f475ae95 TM |
1029 | if (request->fl_type == F_UNLCK) { |
1030 | if (request->fl_flags & FL_EXISTS) | |
1031 | error = -ENOENT; | |
1da177e4 | 1032 | goto out; |
f475ae95 | 1033 | } |
0d9a490a MS |
1034 | |
1035 | if (!new_fl) { | |
1036 | error = -ENOLCK; | |
1037 | goto out; | |
1038 | } | |
1da177e4 LT |
1039 | locks_copy_lock(new_fl, request); |
1040 | locks_insert_lock(before, new_fl); | |
1041 | new_fl = NULL; | |
1042 | } | |
1043 | if (right) { | |
1044 | if (left == right) { | |
1045 | /* The new lock breaks the old one in two pieces, | |
1046 | * so we have to use the second new lock. | |
1047 | */ | |
1048 | left = new_fl2; | |
1049 | new_fl2 = NULL; | |
1050 | locks_copy_lock(left, right); | |
1051 | locks_insert_lock(before, left); | |
1052 | } | |
1053 | right->fl_start = request->fl_end + 1; | |
1054 | locks_wake_up_blocks(right); | |
1055 | } | |
1056 | if (left) { | |
1057 | left->fl_end = request->fl_start - 1; | |
1058 | locks_wake_up_blocks(left); | |
1059 | } | |
1060 | out: | |
1c8c601a | 1061 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1062 | /* |
1063 | * Free any unused locks. | |
1064 | */ | |
1065 | if (new_fl) | |
1066 | locks_free_lock(new_fl); | |
1067 | if (new_fl2) | |
1068 | locks_free_lock(new_fl2); | |
1069 | return error; | |
1070 | } | |
1071 | ||
1072 | /** | |
1073 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1074 | * @filp: The file to apply the lock to | |
1075 | * @fl: The lock to be applied | |
150b3934 | 1076 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1077 | * |
1078 | * Add a POSIX style lock to a file. | |
1079 | * We merge adjacent & overlapping locks whenever possible. | |
1080 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1081 | * |
1082 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1083 | * whether or not a lock was successfully freed by testing the return | |
1084 | * value for -ENOENT. | |
1da177e4 | 1085 | */ |
150b3934 | 1086 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1087 | struct file_lock *conflock) |
1088 | { | |
496ad9aa | 1089 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1090 | } |
150b3934 | 1091 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1092 | |
1093 | /** | |
1094 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1095 | * @filp: The file to apply the lock to | |
1096 | * @fl: The lock to be applied | |
1097 | * | |
1098 | * Add a POSIX style lock to a file. | |
1099 | * We merge adjacent & overlapping locks whenever possible. | |
1100 | * POSIX locks are sorted by owner task, then by starting address | |
1101 | */ | |
1102 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1103 | { | |
1104 | int error; | |
1105 | might_sleep (); | |
1106 | for (;;) { | |
150b3934 | 1107 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1108 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1109 | break; |
1110 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1111 | if (!error) | |
1112 | continue; | |
1113 | ||
1114 | locks_delete_block(fl); | |
1115 | break; | |
1116 | } | |
1117 | return error; | |
1118 | } | |
1119 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1120 | ||
1121 | /** | |
1122 | * locks_mandatory_locked - Check for an active lock | |
1123 | * @inode: the file to check | |
1124 | * | |
1125 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1126 | * This function is called from locks_verify_locked() only. | |
1127 | */ | |
1128 | int locks_mandatory_locked(struct inode *inode) | |
1129 | { | |
1130 | fl_owner_t owner = current->files; | |
1131 | struct file_lock *fl; | |
1132 | ||
1133 | /* | |
1134 | * Search the lock list for this inode for any POSIX locks. | |
1135 | */ | |
1c8c601a | 1136 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1137 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1138 | if (!IS_POSIX(fl)) | |
1139 | continue; | |
1140 | if (fl->fl_owner != owner) | |
1141 | break; | |
1142 | } | |
1c8c601a | 1143 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1144 | return fl ? -EAGAIN : 0; |
1145 | } | |
1146 | ||
1147 | /** | |
1148 | * locks_mandatory_area - Check for a conflicting lock | |
1149 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1150 | * for shared | |
1151 | * @inode: the file to check | |
1152 | * @filp: how the file was opened (if it was) | |
1153 | * @offset: start of area to check | |
1154 | * @count: length of area to check | |
1155 | * | |
1156 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1157 | * This function is called from rw_verify_area() and | |
1158 | * locks_verify_truncate(). | |
1159 | */ | |
1160 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1161 | struct file *filp, loff_t offset, | |
1162 | size_t count) | |
1163 | { | |
1164 | struct file_lock fl; | |
1165 | int error; | |
1166 | ||
1167 | locks_init_lock(&fl); | |
1168 | fl.fl_owner = current->files; | |
1169 | fl.fl_pid = current->tgid; | |
1170 | fl.fl_file = filp; | |
1171 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1172 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
1173 | fl.fl_flags |= FL_SLEEP; | |
1174 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; | |
1175 | fl.fl_start = offset; | |
1176 | fl.fl_end = offset + count - 1; | |
1177 | ||
1178 | for (;;) { | |
150b3934 | 1179 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1180 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1181 | break; |
1182 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1183 | if (!error) { | |
1184 | /* | |
1185 | * If we've been sleeping someone might have | |
1186 | * changed the permissions behind our back. | |
1187 | */ | |
a16877ca | 1188 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1189 | continue; |
1190 | } | |
1191 | ||
1192 | locks_delete_block(&fl); | |
1193 | break; | |
1194 | } | |
1195 | ||
1196 | return error; | |
1197 | } | |
1198 | ||
1199 | EXPORT_SYMBOL(locks_mandatory_area); | |
1200 | ||
778fc546 BF |
1201 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1202 | { | |
1203 | switch (arg) { | |
1204 | case F_UNLCK: | |
1205 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1206 | /* fall through: */ | |
1207 | case F_RDLCK: | |
1208 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1209 | } | |
1210 | } | |
1211 | ||
1da177e4 LT |
1212 | /* We already had a lease on this file; just change its type */ |
1213 | int lease_modify(struct file_lock **before, int arg) | |
1214 | { | |
1215 | struct file_lock *fl = *before; | |
1216 | int error = assign_type(fl, arg); | |
1217 | ||
1218 | if (error) | |
1219 | return error; | |
778fc546 | 1220 | lease_clear_pending(fl, arg); |
1da177e4 | 1221 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1222 | if (arg == F_UNLCK) { |
1223 | struct file *filp = fl->fl_file; | |
1224 | ||
1225 | f_delown(filp); | |
1226 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1227 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1228 | if (fl->fl_fasync != NULL) { | |
1229 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1230 | fl->fl_fasync = NULL; | |
1231 | } | |
1da177e4 | 1232 | locks_delete_lock(before); |
3b6e2723 | 1233 | } |
1da177e4 LT |
1234 | return 0; |
1235 | } | |
1236 | ||
1237 | EXPORT_SYMBOL(lease_modify); | |
1238 | ||
778fc546 BF |
1239 | static bool past_time(unsigned long then) |
1240 | { | |
1241 | if (!then) | |
1242 | /* 0 is a special value meaning "this never expires": */ | |
1243 | return false; | |
1244 | return time_after(jiffies, then); | |
1245 | } | |
1246 | ||
1da177e4 LT |
1247 | static void time_out_leases(struct inode *inode) |
1248 | { | |
1249 | struct file_lock **before; | |
1250 | struct file_lock *fl; | |
1251 | ||
1252 | before = &inode->i_flock; | |
ab83fa4b | 1253 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
778fc546 BF |
1254 | if (past_time(fl->fl_downgrade_time)) |
1255 | lease_modify(before, F_RDLCK); | |
1256 | if (past_time(fl->fl_break_time)) | |
1257 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1258 | if (fl == *before) /* lease_modify may have freed fl */ |
1259 | before = &fl->fl_next; | |
1260 | } | |
1261 | } | |
1262 | ||
1263 | /** | |
1264 | * __break_lease - revoke all outstanding leases on file | |
1265 | * @inode: the inode of the file to return | |
1266 | * @mode: the open mode (read or write) | |
1267 | * | |
87250dd2 | 1268 | * break_lease (inlined for speed) has checked there already is at least |
1269 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1270 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1271 | * specified %O_NONBLOCK to your open(). |
1272 | */ | |
1273 | int __break_lease(struct inode *inode, unsigned int mode) | |
1274 | { | |
778fc546 | 1275 | int error = 0; |
1da177e4 LT |
1276 | struct file_lock *new_fl, *flock; |
1277 | struct file_lock *fl; | |
1da177e4 LT |
1278 | unsigned long break_time; |
1279 | int i_have_this_lease = 0; | |
8737c930 | 1280 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1281 | |
8737c930 | 1282 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1283 | if (IS_ERR(new_fl)) |
1284 | return PTR_ERR(new_fl); | |
1da177e4 | 1285 | |
1c8c601a | 1286 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1287 | |
1288 | time_out_leases(inode); | |
1289 | ||
1290 | flock = inode->i_flock; | |
1291 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1292 | goto out; | |
1293 | ||
778fc546 BF |
1294 | if (!locks_conflict(flock, new_fl)) |
1295 | goto out; | |
1296 | ||
1da177e4 LT |
1297 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) |
1298 | if (fl->fl_owner == current->files) | |
1299 | i_have_this_lease = 1; | |
1300 | ||
1da177e4 LT |
1301 | break_time = 0; |
1302 | if (lease_break_time > 0) { | |
1303 | break_time = jiffies + lease_break_time * HZ; | |
1304 | if (break_time == 0) | |
1305 | break_time++; /* so that 0 means no break time */ | |
1306 | } | |
1307 | ||
1308 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
778fc546 BF |
1309 | if (want_write) { |
1310 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1311 | continue; | |
1312 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1313 | fl->fl_break_time = break_time; |
778fc546 BF |
1314 | } else { |
1315 | if (lease_breaking(flock)) | |
1316 | continue; | |
1317 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1318 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1319 | } |
778fc546 | 1320 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1321 | } |
1322 | ||
1323 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
1324 | error = -EWOULDBLOCK; | |
1325 | goto out; | |
1326 | } | |
1327 | ||
1328 | restart: | |
1329 | break_time = flock->fl_break_time; | |
1330 | if (break_time != 0) { | |
1331 | break_time -= jiffies; | |
1332 | if (break_time == 0) | |
1333 | break_time++; | |
1334 | } | |
4321e01e | 1335 | locks_insert_block(flock, new_fl); |
1c8c601a | 1336 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1337 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1338 | !new_fl->fl_next, break_time); | |
1c8c601a JL |
1339 | spin_lock(&inode->i_lock); |
1340 | locks_delete_block(new_fl); | |
1da177e4 LT |
1341 | if (error >= 0) { |
1342 | if (error == 0) | |
1343 | time_out_leases(inode); | |
778fc546 BF |
1344 | /* |
1345 | * Wait for the next conflicting lease that has not been | |
1346 | * broken yet | |
1347 | */ | |
1da177e4 LT |
1348 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1349 | flock = flock->fl_next) { | |
778fc546 | 1350 | if (locks_conflict(new_fl, flock)) |
1da177e4 LT |
1351 | goto restart; |
1352 | } | |
1353 | error = 0; | |
1354 | } | |
1355 | ||
1356 | out: | |
1c8c601a | 1357 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1358 | locks_free_lock(new_fl); |
1da177e4 LT |
1359 | return error; |
1360 | } | |
1361 | ||
1362 | EXPORT_SYMBOL(__break_lease); | |
1363 | ||
1364 | /** | |
a6b91919 | 1365 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1366 | * @inode: the inode |
1367 | * @time: pointer to a timespec which will contain the last modified time | |
1368 | * | |
1369 | * This is to force NFS clients to flush their caches for files with | |
1370 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1371 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1372 | */ |
1373 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1374 | { | |
1375 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1376 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1377 | *time = current_fs_time(inode->i_sb); |
1378 | else | |
1379 | *time = inode->i_mtime; | |
1380 | } | |
1381 | ||
1382 | EXPORT_SYMBOL(lease_get_mtime); | |
1383 | ||
1384 | /** | |
1385 | * fcntl_getlease - Enquire what lease is currently active | |
1386 | * @filp: the file | |
1387 | * | |
1388 | * The value returned by this function will be one of | |
1389 | * (if no lease break is pending): | |
1390 | * | |
1391 | * %F_RDLCK to indicate a shared lease is held. | |
1392 | * | |
1393 | * %F_WRLCK to indicate an exclusive lease is held. | |
1394 | * | |
1395 | * %F_UNLCK to indicate no lease is held. | |
1396 | * | |
1397 | * (if a lease break is pending): | |
1398 | * | |
1399 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1400 | * changed to a shared lease (or removed). | |
1401 | * | |
1402 | * %F_UNLCK to indicate the lease needs to be removed. | |
1403 | * | |
1404 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1405 | * should be returned to userspace. | |
1406 | */ | |
1407 | int fcntl_getlease(struct file *filp) | |
1408 | { | |
1409 | struct file_lock *fl; | |
1c8c601a | 1410 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1411 | int type = F_UNLCK; |
1412 | ||
1c8c601a | 1413 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1414 | time_out_leases(file_inode(filp)); |
1415 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1416 | fl = fl->fl_next) { |
1417 | if (fl->fl_file == filp) { | |
778fc546 | 1418 | type = target_leasetype(fl); |
1da177e4 LT |
1419 | break; |
1420 | } | |
1421 | } | |
1c8c601a | 1422 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1423 | return type; |
1424 | } | |
1425 | ||
d4f22d19 | 1426 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1427 | { |
7eaae282 | 1428 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1429 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1430 | struct inode *inode = dentry->d_inode; |
c1f24ef4 | 1431 | int error; |
1da177e4 | 1432 | |
096657b6 BF |
1433 | lease = *flp; |
1434 | ||
8335ebd9 BF |
1435 | error = -EAGAIN; |
1436 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
096657b6 | 1437 | goto out; |
8335ebd9 BF |
1438 | if ((arg == F_WRLCK) |
1439 | && ((dentry->d_count > 1) | |
1440 | || (atomic_read(&inode->i_count) > 1))) | |
096657b6 | 1441 | goto out; |
6d5e8b05 | 1442 | |
1da177e4 LT |
1443 | /* |
1444 | * At this point, we know that if there is an exclusive | |
1445 | * lease on this file, then we hold it on this filp | |
1446 | * (otherwise our open of this file would have blocked). | |
1447 | * And if we are trying to acquire an exclusive lease, | |
1448 | * then the file is not open by anyone (including us) | |
1449 | * except for this filp. | |
1450 | */ | |
c1f24ef4 | 1451 | error = -EAGAIN; |
1da177e4 LT |
1452 | for (before = &inode->i_flock; |
1453 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1454 | before = &fl->fl_next) { | |
c1f24ef4 | 1455 | if (fl->fl_file == filp) { |
1da177e4 | 1456 | my_before = before; |
c1f24ef4 BF |
1457 | continue; |
1458 | } | |
1459 | /* | |
1460 | * No exclusive leases if someone else has a lease on | |
1461 | * this file: | |
1462 | */ | |
1463 | if (arg == F_WRLCK) | |
1464 | goto out; | |
1465 | /* | |
1466 | * Modifying our existing lease is OK, but no getting a | |
1467 | * new lease if someone else is opening for write: | |
1468 | */ | |
1469 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1470 | goto out; | |
1da177e4 LT |
1471 | } |
1472 | ||
1da177e4 | 1473 | if (my_before != NULL) { |
8fb47a4f | 1474 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1475 | if (!error) |
1476 | *flp = *my_before; | |
1da177e4 LT |
1477 | goto out; |
1478 | } | |
1479 | ||
1da177e4 LT |
1480 | error = -EINVAL; |
1481 | if (!leases_enable) | |
1482 | goto out; | |
1483 | ||
c5b1f0d9 | 1484 | locks_insert_lock(before, lease); |
85c59580 | 1485 | return 0; |
1da177e4 | 1486 | |
1da177e4 LT |
1487 | out: |
1488 | return error; | |
1489 | } | |
8335ebd9 | 1490 | |
d4f22d19 | 1491 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1492 | { |
1493 | struct file_lock *fl, **before; | |
1494 | struct dentry *dentry = filp->f_path.dentry; | |
1495 | struct inode *inode = dentry->d_inode; | |
1496 | ||
1497 | for (before = &inode->i_flock; | |
1498 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1499 | before = &fl->fl_next) { | |
1500 | if (fl->fl_file != filp) | |
1501 | continue; | |
1502 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1503 | } | |
1504 | return -EAGAIN; | |
1505 | } | |
1506 | ||
1507 | /** | |
1508 | * generic_setlease - sets a lease on an open file | |
1509 | * @filp: file pointer | |
1510 | * @arg: type of lease to obtain | |
1511 | * @flp: input - file_lock to use, output - file_lock inserted | |
1512 | * | |
1513 | * The (input) flp->fl_lmops->lm_break function is required | |
1514 | * by break_lease(). | |
1515 | * | |
1c8c601a | 1516 | * Called with inode->i_lock held. |
8335ebd9 BF |
1517 | */ |
1518 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1519 | { | |
1520 | struct dentry *dentry = filp->f_path.dentry; | |
1521 | struct inode *inode = dentry->d_inode; | |
1522 | int error; | |
1523 | ||
8e96e3b7 | 1524 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1525 | return -EACCES; |
1526 | if (!S_ISREG(inode->i_mode)) | |
1527 | return -EINVAL; | |
1528 | error = security_file_lock(filp, arg); | |
1529 | if (error) | |
1530 | return error; | |
1531 | ||
1532 | time_out_leases(inode); | |
1533 | ||
1534 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1535 | ||
1536 | switch (arg) { | |
1537 | case F_UNLCK: | |
1538 | return generic_delete_lease(filp, flp); | |
1539 | case F_RDLCK: | |
1540 | case F_WRLCK: | |
1541 | return generic_add_lease(filp, arg, flp); | |
1542 | default: | |
8d657eb3 | 1543 | return -EINVAL; |
8335ebd9 BF |
1544 | } |
1545 | } | |
0af1a450 | 1546 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1547 | |
b89f4321 AB |
1548 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1549 | { | |
1550 | if (filp->f_op && filp->f_op->setlease) | |
1551 | return filp->f_op->setlease(filp, arg, lease); | |
1552 | else | |
1553 | return generic_setlease(filp, arg, lease); | |
1554 | } | |
1555 | ||
1556 | /** | |
a9933cea | 1557 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1558 | * @filp: file pointer |
1559 | * @arg: type of lease to obtain | |
1560 | * @lease: file_lock to use | |
1561 | * | |
1562 | * Call this to establish a lease on the file. | |
8fb47a4f | 1563 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1564 | * break_lease will oops! |
1565 | * | |
1566 | * This will call the filesystem's setlease file method, if | |
1567 | * defined. Note that there is no getlease method; instead, the | |
1568 | * filesystem setlease method should call back to setlease() to | |
1569 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1570 | * find it. Since fcntl_getlease() only reports whether the current | |
1571 | * task holds a lease, a cluster filesystem need only do this for | |
1572 | * leases held by processes on this node. | |
1573 | * | |
1574 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1575 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1576 | * filesystem's open, create, and (on truncate) setattr methods. |
1577 | * | |
1578 | * Warning: the only current setlease methods exist only to disable | |
1579 | * leases in certain cases. More vfs changes may be required to | |
1580 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1581 | */ |
1582 | ||
a9933cea | 1583 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1584 | { |
1c8c601a | 1585 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1586 | int error; |
1587 | ||
1c8c601a | 1588 | spin_lock(&inode->i_lock); |
b89f4321 | 1589 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1590 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1591 | |
1592 | return error; | |
1593 | } | |
a9933cea | 1594 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1595 | |
0ceaf6c7 BF |
1596 | static int do_fcntl_delete_lease(struct file *filp) |
1597 | { | |
1598 | struct file_lock fl, *flp = &fl; | |
1599 | ||
1600 | lease_init(filp, F_UNLCK, flp); | |
1601 | ||
1602 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1603 | } | |
1604 | ||
1605 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1606 | { |
3df057ac | 1607 | struct file_lock *fl, *ret; |
1c8c601a | 1608 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1609 | struct fasync_struct *new; |
1da177e4 LT |
1610 | int error; |
1611 | ||
c5b1f0d9 AB |
1612 | fl = lease_alloc(filp, arg); |
1613 | if (IS_ERR(fl)) | |
1614 | return PTR_ERR(fl); | |
1da177e4 | 1615 | |
f7347ce4 LT |
1616 | new = fasync_alloc(); |
1617 | if (!new) { | |
1618 | locks_free_lock(fl); | |
1619 | return -ENOMEM; | |
1620 | } | |
3df057ac | 1621 | ret = fl; |
1c8c601a | 1622 | spin_lock(&inode->i_lock); |
8896b93f | 1623 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 | 1624 | if (error) { |
1c8c601a | 1625 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1626 | locks_free_lock(fl); |
1627 | goto out_free_fasync; | |
1628 | } | |
3df057ac BF |
1629 | if (ret != fl) |
1630 | locks_free_lock(fl); | |
1da177e4 | 1631 | |
f7347ce4 LT |
1632 | /* |
1633 | * fasync_insert_entry() returns the old entry if any. | |
1634 | * If there was no old entry, then it used 'new' and | |
1635 | * inserted it into the fasync list. Clear new so that | |
1636 | * we don't release it here. | |
1637 | */ | |
3df057ac | 1638 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1639 | new = NULL; |
1640 | ||
8896b93f | 1641 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1c8c601a | 1642 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1643 | |
1644 | out_free_fasync: | |
f7347ce4 LT |
1645 | if (new) |
1646 | fasync_free(new); | |
1da177e4 LT |
1647 | return error; |
1648 | } | |
1649 | ||
0ceaf6c7 BF |
1650 | /** |
1651 | * fcntl_setlease - sets a lease on an open file | |
1652 | * @fd: open file descriptor | |
1653 | * @filp: file pointer | |
1654 | * @arg: type of lease to obtain | |
1655 | * | |
1656 | * Call this fcntl to establish a lease on the file. | |
1657 | * Note that you also need to call %F_SETSIG to | |
1658 | * receive a signal when the lease is broken. | |
1659 | */ | |
1660 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1661 | { | |
1662 | if (arg == F_UNLCK) | |
1663 | return do_fcntl_delete_lease(filp); | |
1664 | return do_fcntl_add_lease(fd, filp, arg); | |
1665 | } | |
1666 | ||
1da177e4 LT |
1667 | /** |
1668 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1669 | * @filp: The file to apply the lock to | |
1670 | * @fl: The lock to be applied | |
1671 | * | |
1672 | * Add a FLOCK style lock to a file. | |
1673 | */ | |
1674 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1675 | { | |
1676 | int error; | |
1677 | might_sleep(); | |
1678 | for (;;) { | |
1679 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1680 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1681 | break; |
1682 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1683 | if (!error) | |
1684 | continue; | |
1685 | ||
1686 | locks_delete_block(fl); | |
1687 | break; | |
1688 | } | |
1689 | return error; | |
1690 | } | |
1691 | ||
1692 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1693 | ||
1694 | /** | |
1695 | * sys_flock: - flock() system call. | |
1696 | * @fd: the file descriptor to lock. | |
1697 | * @cmd: the type of lock to apply. | |
1698 | * | |
1699 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1700 | * The @cmd can be one of | |
1701 | * | |
1702 | * %LOCK_SH -- a shared lock. | |
1703 | * | |
1704 | * %LOCK_EX -- an exclusive lock. | |
1705 | * | |
1706 | * %LOCK_UN -- remove an existing lock. | |
1707 | * | |
1708 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1709 | * | |
1710 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1711 | * processes read and write access respectively. | |
1712 | */ | |
002c8976 | 1713 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1714 | { |
2903ff01 | 1715 | struct fd f = fdget(fd); |
1da177e4 LT |
1716 | struct file_lock *lock; |
1717 | int can_sleep, unlock; | |
1718 | int error; | |
1719 | ||
1720 | error = -EBADF; | |
2903ff01 | 1721 | if (!f.file) |
1da177e4 LT |
1722 | goto out; |
1723 | ||
1724 | can_sleep = !(cmd & LOCK_NB); | |
1725 | cmd &= ~LOCK_NB; | |
1726 | unlock = (cmd == LOCK_UN); | |
1727 | ||
aeb5d727 | 1728 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1729 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1730 | goto out_putf; |
1731 | ||
2903ff01 | 1732 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1733 | if (error) |
1734 | goto out_putf; | |
1735 | if (can_sleep) | |
1736 | lock->fl_flags |= FL_SLEEP; | |
1737 | ||
2903ff01 | 1738 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1739 | if (error) |
1740 | goto out_free; | |
1741 | ||
2903ff01 AV |
1742 | if (f.file->f_op && f.file->f_op->flock) |
1743 | error = f.file->f_op->flock(f.file, | |
1da177e4 LT |
1744 | (can_sleep) ? F_SETLKW : F_SETLK, |
1745 | lock); | |
1746 | else | |
2903ff01 | 1747 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1748 | |
1749 | out_free: | |
993dfa87 | 1750 | locks_free_lock(lock); |
1da177e4 LT |
1751 | |
1752 | out_putf: | |
2903ff01 | 1753 | fdput(f); |
1da177e4 LT |
1754 | out: |
1755 | return error; | |
1756 | } | |
1757 | ||
3ee17abd BF |
1758 | /** |
1759 | * vfs_test_lock - test file byte range lock | |
1760 | * @filp: The file to test lock for | |
6924c554 | 1761 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1762 | * |
1763 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1764 | * setting conf->fl_type to something other than F_UNLCK. | |
1765 | */ | |
1766 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1767 | { | |
1768 | if (filp->f_op && filp->f_op->lock) | |
1769 | return filp->f_op->lock(filp, F_GETLK, fl); | |
1770 | posix_test_lock(filp, fl); | |
1771 | return 0; | |
1772 | } | |
1773 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1774 | ||
c2fa1b8a BF |
1775 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1776 | { | |
1777 | flock->l_pid = fl->fl_pid; | |
1778 | #if BITS_PER_LONG == 32 | |
1779 | /* | |
1780 | * Make sure we can represent the posix lock via | |
1781 | * legacy 32bit flock. | |
1782 | */ | |
1783 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1784 | return -EOVERFLOW; | |
1785 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1786 | return -EOVERFLOW; | |
1787 | #endif | |
1788 | flock->l_start = fl->fl_start; | |
1789 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1790 | fl->fl_end - fl->fl_start + 1; | |
1791 | flock->l_whence = 0; | |
129a84de | 1792 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1793 | return 0; |
1794 | } | |
1795 | ||
1796 | #if BITS_PER_LONG == 32 | |
1797 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1798 | { | |
1799 | flock->l_pid = fl->fl_pid; | |
1800 | flock->l_start = fl->fl_start; | |
1801 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1802 | fl->fl_end - fl->fl_start + 1; | |
1803 | flock->l_whence = 0; | |
1804 | flock->l_type = fl->fl_type; | |
1805 | } | |
1806 | #endif | |
1807 | ||
1da177e4 LT |
1808 | /* Report the first existing lock that would conflict with l. |
1809 | * This implements the F_GETLK command of fcntl(). | |
1810 | */ | |
1811 | int fcntl_getlk(struct file *filp, struct flock __user *l) | |
1812 | { | |
9d6a8c5c | 1813 | struct file_lock file_lock; |
1da177e4 LT |
1814 | struct flock flock; |
1815 | int error; | |
1816 | ||
1817 | error = -EFAULT; | |
1818 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1819 | goto out; | |
1820 | error = -EINVAL; | |
1821 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1822 | goto out; | |
1823 | ||
1824 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1825 | if (error) | |
1826 | goto out; | |
1827 | ||
3ee17abd BF |
1828 | error = vfs_test_lock(filp, &file_lock); |
1829 | if (error) | |
1830 | goto out; | |
1da177e4 | 1831 | |
9d6a8c5c ME |
1832 | flock.l_type = file_lock.fl_type; |
1833 | if (file_lock.fl_type != F_UNLCK) { | |
1834 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1835 | if (error) |
1da177e4 | 1836 | goto out; |
1da177e4 LT |
1837 | } |
1838 | error = -EFAULT; | |
1839 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1840 | error = 0; | |
1841 | out: | |
1842 | return error; | |
1843 | } | |
1844 | ||
7723ec97 ME |
1845 | /** |
1846 | * vfs_lock_file - file byte range lock | |
1847 | * @filp: The file to apply the lock to | |
1848 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1849 | * @fl: The lock to be applied | |
150b3934 ME |
1850 | * @conf: Place to return a copy of the conflicting lock, if found. |
1851 | * | |
1852 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1853 | * as the final argument. | |
1854 | * | |
1855 | * If the filesystem defines a private ->lock() method, then @conf will | |
1856 | * be left unchanged; so a caller that cares should initialize it to | |
1857 | * some acceptable default. | |
2beb6614 ME |
1858 | * |
1859 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1860 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1861 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1862 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1863 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1864 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1865 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1866 | * request completes. |
1867 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1868 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1869 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
1870 | * nonzero return code and the file system should release the lock. The file |
1871 | * system is also responsible to keep a corresponding posix lock when it | |
1872 | * grants a lock so the VFS can find out which locks are locally held and do | |
1873 | * the correct lock cleanup when required. | |
1874 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 1875 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 1876 | * return code. |
7723ec97 | 1877 | */ |
150b3934 | 1878 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 ME |
1879 | { |
1880 | if (filp->f_op && filp->f_op->lock) | |
1881 | return filp->f_op->lock(filp, cmd, fl); | |
1882 | else | |
150b3934 | 1883 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
1884 | } |
1885 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
1886 | ||
b648a6de MS |
1887 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
1888 | struct file_lock *fl) | |
1889 | { | |
1890 | int error; | |
1891 | ||
1892 | error = security_file_lock(filp, fl->fl_type); | |
1893 | if (error) | |
1894 | return error; | |
1895 | ||
764c76b3 MS |
1896 | for (;;) { |
1897 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
1898 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 1899 | break; |
764c76b3 MS |
1900 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
1901 | if (!error) | |
1902 | continue; | |
1903 | ||
1904 | locks_delete_block(fl); | |
1905 | break; | |
b648a6de MS |
1906 | } |
1907 | ||
1908 | return error; | |
1909 | } | |
1910 | ||
1da177e4 LT |
1911 | /* Apply the lock described by l to an open file descriptor. |
1912 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
1913 | */ | |
c293621b PS |
1914 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
1915 | struct flock __user *l) | |
1da177e4 LT |
1916 | { |
1917 | struct file_lock *file_lock = locks_alloc_lock(); | |
1918 | struct flock flock; | |
1919 | struct inode *inode; | |
0b2bac2f | 1920 | struct file *f; |
1da177e4 LT |
1921 | int error; |
1922 | ||
1923 | if (file_lock == NULL) | |
1924 | return -ENOLCK; | |
1925 | ||
1926 | /* | |
1927 | * This might block, so we do it before checking the inode. | |
1928 | */ | |
1929 | error = -EFAULT; | |
1930 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1931 | goto out; | |
1932 | ||
496ad9aa | 1933 | inode = file_inode(filp); |
1da177e4 LT |
1934 | |
1935 | /* Don't allow mandatory locks on files that may be memory mapped | |
1936 | * and shared. | |
1937 | */ | |
a16877ca | 1938 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
1939 | error = -EAGAIN; |
1940 | goto out; | |
1941 | } | |
1942 | ||
c293621b | 1943 | again: |
1da177e4 LT |
1944 | error = flock_to_posix_lock(filp, file_lock, &flock); |
1945 | if (error) | |
1946 | goto out; | |
1947 | if (cmd == F_SETLKW) { | |
1948 | file_lock->fl_flags |= FL_SLEEP; | |
1949 | } | |
1950 | ||
1951 | error = -EBADF; | |
1952 | switch (flock.l_type) { | |
1953 | case F_RDLCK: | |
1954 | if (!(filp->f_mode & FMODE_READ)) | |
1955 | goto out; | |
1956 | break; | |
1957 | case F_WRLCK: | |
1958 | if (!(filp->f_mode & FMODE_WRITE)) | |
1959 | goto out; | |
1960 | break; | |
1961 | case F_UNLCK: | |
1962 | break; | |
1963 | default: | |
1964 | error = -EINVAL; | |
1965 | goto out; | |
1966 | } | |
1967 | ||
b648a6de | 1968 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 1969 | |
c293621b PS |
1970 | /* |
1971 | * Attempt to detect a close/fcntl race and recover by | |
1972 | * releasing the lock that was just acquired. | |
1973 | */ | |
0b2bac2f AV |
1974 | /* |
1975 | * we need that spin_lock here - it prevents reordering between | |
1976 | * update of inode->i_flock and check for it done in close(). | |
1977 | * rcu_read_lock() wouldn't do. | |
1978 | */ | |
1979 | spin_lock(¤t->files->file_lock); | |
1980 | f = fcheck(fd); | |
1981 | spin_unlock(¤t->files->file_lock); | |
1982 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
1983 | flock.l_type = F_UNLCK; |
1984 | goto again; | |
1da177e4 LT |
1985 | } |
1986 | ||
c293621b | 1987 | out: |
1da177e4 LT |
1988 | locks_free_lock(file_lock); |
1989 | return error; | |
1990 | } | |
1991 | ||
1992 | #if BITS_PER_LONG == 32 | |
1993 | /* Report the first existing lock that would conflict with l. | |
1994 | * This implements the F_GETLK command of fcntl(). | |
1995 | */ | |
1996 | int fcntl_getlk64(struct file *filp, struct flock64 __user *l) | |
1997 | { | |
9d6a8c5c | 1998 | struct file_lock file_lock; |
1da177e4 LT |
1999 | struct flock64 flock; |
2000 | int error; | |
2001 | ||
2002 | error = -EFAULT; | |
2003 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2004 | goto out; | |
2005 | error = -EINVAL; | |
2006 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2007 | goto out; | |
2008 | ||
2009 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2010 | if (error) | |
2011 | goto out; | |
2012 | ||
3ee17abd BF |
2013 | error = vfs_test_lock(filp, &file_lock); |
2014 | if (error) | |
2015 | goto out; | |
2016 | ||
9d6a8c5c ME |
2017 | flock.l_type = file_lock.fl_type; |
2018 | if (file_lock.fl_type != F_UNLCK) | |
2019 | posix_lock_to_flock64(&flock, &file_lock); | |
2020 | ||
1da177e4 LT |
2021 | error = -EFAULT; |
2022 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2023 | error = 0; | |
2024 | ||
2025 | out: | |
2026 | return error; | |
2027 | } | |
2028 | ||
2029 | /* Apply the lock described by l to an open file descriptor. | |
2030 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2031 | */ | |
c293621b PS |
2032 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2033 | struct flock64 __user *l) | |
1da177e4 LT |
2034 | { |
2035 | struct file_lock *file_lock = locks_alloc_lock(); | |
2036 | struct flock64 flock; | |
2037 | struct inode *inode; | |
0b2bac2f | 2038 | struct file *f; |
1da177e4 LT |
2039 | int error; |
2040 | ||
2041 | if (file_lock == NULL) | |
2042 | return -ENOLCK; | |
2043 | ||
2044 | /* | |
2045 | * This might block, so we do it before checking the inode. | |
2046 | */ | |
2047 | error = -EFAULT; | |
2048 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2049 | goto out; | |
2050 | ||
496ad9aa | 2051 | inode = file_inode(filp); |
1da177e4 LT |
2052 | |
2053 | /* Don't allow mandatory locks on files that may be memory mapped | |
2054 | * and shared. | |
2055 | */ | |
a16877ca | 2056 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2057 | error = -EAGAIN; |
2058 | goto out; | |
2059 | } | |
2060 | ||
c293621b | 2061 | again: |
1da177e4 LT |
2062 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2063 | if (error) | |
2064 | goto out; | |
2065 | if (cmd == F_SETLKW64) { | |
2066 | file_lock->fl_flags |= FL_SLEEP; | |
2067 | } | |
2068 | ||
2069 | error = -EBADF; | |
2070 | switch (flock.l_type) { | |
2071 | case F_RDLCK: | |
2072 | if (!(filp->f_mode & FMODE_READ)) | |
2073 | goto out; | |
2074 | break; | |
2075 | case F_WRLCK: | |
2076 | if (!(filp->f_mode & FMODE_WRITE)) | |
2077 | goto out; | |
2078 | break; | |
2079 | case F_UNLCK: | |
2080 | break; | |
2081 | default: | |
2082 | error = -EINVAL; | |
2083 | goto out; | |
2084 | } | |
2085 | ||
b648a6de | 2086 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2087 | |
c293621b PS |
2088 | /* |
2089 | * Attempt to detect a close/fcntl race and recover by | |
2090 | * releasing the lock that was just acquired. | |
2091 | */ | |
0b2bac2f AV |
2092 | spin_lock(¤t->files->file_lock); |
2093 | f = fcheck(fd); | |
2094 | spin_unlock(¤t->files->file_lock); | |
2095 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2096 | flock.l_type = F_UNLCK; |
2097 | goto again; | |
1da177e4 LT |
2098 | } |
2099 | ||
2100 | out: | |
2101 | locks_free_lock(file_lock); | |
2102 | return error; | |
2103 | } | |
2104 | #endif /* BITS_PER_LONG == 32 */ | |
2105 | ||
2106 | /* | |
2107 | * This function is called when the file is being removed | |
2108 | * from the task's fd array. POSIX locks belonging to this task | |
2109 | * are deleted at this time. | |
2110 | */ | |
2111 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2112 | { | |
ff7b86b8 | 2113 | struct file_lock lock; |
1da177e4 LT |
2114 | |
2115 | /* | |
2116 | * If there are no locks held on this file, we don't need to call | |
2117 | * posix_lock_file(). Another process could be setting a lock on this | |
2118 | * file at the same time, but we wouldn't remove that lock anyway. | |
2119 | */ | |
496ad9aa | 2120 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2121 | return; |
2122 | ||
2123 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2124 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2125 | lock.fl_start = 0; |
2126 | lock.fl_end = OFFSET_MAX; | |
2127 | lock.fl_owner = owner; | |
2128 | lock.fl_pid = current->tgid; | |
2129 | lock.fl_file = filp; | |
2130 | lock.fl_ops = NULL; | |
2131 | lock.fl_lmops = NULL; | |
2132 | ||
150b3934 | 2133 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2134 | |
1da177e4 LT |
2135 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2136 | lock.fl_ops->fl_release_private(&lock); | |
2137 | } | |
2138 | ||
2139 | EXPORT_SYMBOL(locks_remove_posix); | |
2140 | ||
2141 | /* | |
2142 | * This function is called on the last close of an open file. | |
2143 | */ | |
2144 | void locks_remove_flock(struct file *filp) | |
2145 | { | |
496ad9aa | 2146 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2147 | struct file_lock *fl; |
2148 | struct file_lock **before; | |
2149 | ||
2150 | if (!inode->i_flock) | |
2151 | return; | |
2152 | ||
2153 | if (filp->f_op && filp->f_op->flock) { | |
2154 | struct file_lock fl = { | |
2155 | .fl_pid = current->tgid, | |
2156 | .fl_file = filp, | |
2157 | .fl_flags = FL_FLOCK, | |
2158 | .fl_type = F_UNLCK, | |
2159 | .fl_end = OFFSET_MAX, | |
2160 | }; | |
2161 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2162 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2163 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2164 | } |
2165 | ||
1c8c601a | 2166 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2167 | before = &inode->i_flock; |
2168 | ||
2169 | while ((fl = *before) != NULL) { | |
2170 | if (fl->fl_file == filp) { | |
c293621b | 2171 | if (IS_FLOCK(fl)) { |
1da177e4 LT |
2172 | locks_delete_lock(before); |
2173 | continue; | |
2174 | } | |
2175 | if (IS_LEASE(fl)) { | |
2176 | lease_modify(before, F_UNLCK); | |
2177 | continue; | |
2178 | } | |
2179 | /* What? */ | |
2180 | BUG(); | |
2181 | } | |
2182 | before = &fl->fl_next; | |
2183 | } | |
1c8c601a | 2184 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2185 | } |
2186 | ||
1da177e4 LT |
2187 | /** |
2188 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2189 | * @waiter: the lock which was waiting |
2190 | * | |
2191 | * lockd needs to block waiting for locks. | |
2192 | */ | |
64a318ee | 2193 | int |
f891a29f | 2194 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2195 | { |
64a318ee BF |
2196 | int status = 0; |
2197 | ||
1c8c601a | 2198 | spin_lock(&file_lock_lock); |
5996a298 | 2199 | if (waiter->fl_next) |
1da177e4 | 2200 | __locks_delete_block(waiter); |
64a318ee BF |
2201 | else |
2202 | status = -ENOENT; | |
1c8c601a | 2203 | spin_unlock(&file_lock_lock); |
64a318ee | 2204 | return status; |
1da177e4 | 2205 | } |
1da177e4 LT |
2206 | EXPORT_SYMBOL(posix_unblock_lock); |
2207 | ||
9b9d2ab4 ME |
2208 | /** |
2209 | * vfs_cancel_lock - file byte range unblock lock | |
2210 | * @filp: The file to apply the unblock to | |
2211 | * @fl: The lock to be unblocked | |
2212 | * | |
2213 | * Used by lock managers to cancel blocked requests | |
2214 | */ | |
2215 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2216 | { | |
2217 | if (filp->f_op && filp->f_op->lock) | |
2218 | return filp->f_op->lock(filp, F_CANCELLK, fl); | |
2219 | return 0; | |
2220 | } | |
2221 | ||
2222 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2223 | ||
7f8ada98 | 2224 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2225 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2226 | #include <linux/seq_file.h> |
2227 | ||
2228 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, | |
99dc8292 | 2229 | loff_t id, char *pfx) |
1da177e4 LT |
2230 | { |
2231 | struct inode *inode = NULL; | |
ab1f1611 VG |
2232 | unsigned int fl_pid; |
2233 | ||
2234 | if (fl->fl_nspid) | |
6c5f3e7b | 2235 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2236 | else |
2237 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2238 | |
2239 | if (fl->fl_file != NULL) | |
496ad9aa | 2240 | inode = file_inode(fl->fl_file); |
1da177e4 | 2241 | |
99dc8292 | 2242 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2243 | if (IS_POSIX(fl)) { |
7f8ada98 | 2244 | seq_printf(f, "%6s %s ", |
1da177e4 LT |
2245 | (fl->fl_flags & FL_ACCESS) ? "ACCESS" : "POSIX ", |
2246 | (inode == NULL) ? "*NOINODE*" : | |
a16877ca | 2247 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2248 | } else if (IS_FLOCK(fl)) { |
2249 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2250 | seq_printf(f, "FLOCK MSNFS "); |
1da177e4 | 2251 | } else { |
7f8ada98 | 2252 | seq_printf(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2253 | } |
2254 | } else if (IS_LEASE(fl)) { | |
7f8ada98 | 2255 | seq_printf(f, "LEASE "); |
ab83fa4b | 2256 | if (lease_breaking(fl)) |
7f8ada98 | 2257 | seq_printf(f, "BREAKING "); |
1da177e4 | 2258 | else if (fl->fl_file) |
7f8ada98 | 2259 | seq_printf(f, "ACTIVE "); |
1da177e4 | 2260 | else |
7f8ada98 | 2261 | seq_printf(f, "BREAKER "); |
1da177e4 | 2262 | } else { |
7f8ada98 | 2263 | seq_printf(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2264 | } |
2265 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2266 | seq_printf(f, "%s ", |
1da177e4 LT |
2267 | (fl->fl_type & LOCK_READ) |
2268 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2269 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2270 | } else { | |
7f8ada98 | 2271 | seq_printf(f, "%s ", |
ab83fa4b | 2272 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2273 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2274 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2275 | } |
2276 | if (inode) { | |
2277 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2278 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2279 | inode->i_sb->s_id, inode->i_ino); |
2280 | #else | |
2281 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2282 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2283 | MAJOR(inode->i_sb->s_dev), |
2284 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2285 | #endif | |
2286 | } else { | |
ab1f1611 | 2287 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2288 | } |
2289 | if (IS_POSIX(fl)) { | |
2290 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2291 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2292 | else |
7f8ada98 | 2293 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2294 | } else { |
7f8ada98 | 2295 | seq_printf(f, "0 EOF\n"); |
1da177e4 LT |
2296 | } |
2297 | } | |
2298 | ||
7f8ada98 | 2299 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2300 | { |
7f8ada98 | 2301 | struct file_lock *fl, *bfl; |
1da177e4 | 2302 | |
7f8ada98 | 2303 | fl = list_entry(v, struct file_lock, fl_link); |
1da177e4 | 2304 | |
99dc8292 | 2305 | lock_get_status(f, fl, *((loff_t *)f->private), ""); |
1da177e4 | 2306 | |
7f8ada98 | 2307 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
99dc8292 | 2308 | lock_get_status(f, bfl, *((loff_t *)f->private), " ->"); |
094f2825 | 2309 | |
7f8ada98 PE |
2310 | return 0; |
2311 | } | |
1da177e4 | 2312 | |
7f8ada98 PE |
2313 | static void *locks_start(struct seq_file *f, loff_t *pos) |
2314 | { | |
99dc8292 JM |
2315 | loff_t *p = f->private; |
2316 | ||
1c8c601a | 2317 | spin_lock(&file_lock_lock); |
99dc8292 | 2318 | *p = (*pos + 1); |
7f8ada98 PE |
2319 | return seq_list_start(&file_lock_list, *pos); |
2320 | } | |
1da177e4 | 2321 | |
7f8ada98 PE |
2322 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2323 | { | |
99dc8292 JM |
2324 | loff_t *p = f->private; |
2325 | ++*p; | |
7f8ada98 PE |
2326 | return seq_list_next(v, &file_lock_list, pos); |
2327 | } | |
1da177e4 | 2328 | |
7f8ada98 PE |
2329 | static void locks_stop(struct seq_file *f, void *v) |
2330 | { | |
1c8c601a | 2331 | spin_unlock(&file_lock_lock); |
1da177e4 LT |
2332 | } |
2333 | ||
d8ba7a36 | 2334 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2335 | .start = locks_start, |
2336 | .next = locks_next, | |
2337 | .stop = locks_stop, | |
2338 | .show = locks_show, | |
2339 | }; | |
d8ba7a36 AD |
2340 | |
2341 | static int locks_open(struct inode *inode, struct file *filp) | |
2342 | { | |
99dc8292 | 2343 | return seq_open_private(filp, &locks_seq_operations, sizeof(loff_t)); |
d8ba7a36 AD |
2344 | } |
2345 | ||
2346 | static const struct file_operations proc_locks_operations = { | |
2347 | .open = locks_open, | |
2348 | .read = seq_read, | |
2349 | .llseek = seq_lseek, | |
99dc8292 | 2350 | .release = seq_release_private, |
d8ba7a36 AD |
2351 | }; |
2352 | ||
2353 | static int __init proc_locks_init(void) | |
2354 | { | |
2355 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2356 | return 0; | |
2357 | } | |
2358 | module_init(proc_locks_init); | |
7f8ada98 PE |
2359 | #endif |
2360 | ||
1da177e4 LT |
2361 | /** |
2362 | * lock_may_read - checks that the region is free of locks | |
2363 | * @inode: the inode that is being read | |
2364 | * @start: the first byte to read | |
2365 | * @len: the number of bytes to read | |
2366 | * | |
2367 | * Emulates Windows locking requirements. Whole-file | |
2368 | * mandatory locks (share modes) can prohibit a read and | |
2369 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2370 | * | |
2371 | * N.B. this function is only ever called | |
2372 | * from knfsd and ownership of locks is never checked. | |
2373 | */ | |
2374 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2375 | { | |
2376 | struct file_lock *fl; | |
2377 | int result = 1; | |
1c8c601a JL |
2378 | |
2379 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2380 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2381 | if (IS_POSIX(fl)) { | |
2382 | if (fl->fl_type == F_RDLCK) | |
2383 | continue; | |
2384 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2385 | continue; | |
2386 | } else if (IS_FLOCK(fl)) { | |
2387 | if (!(fl->fl_type & LOCK_MAND)) | |
2388 | continue; | |
2389 | if (fl->fl_type & LOCK_READ) | |
2390 | continue; | |
2391 | } else | |
2392 | continue; | |
2393 | result = 0; | |
2394 | break; | |
2395 | } | |
1c8c601a | 2396 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2397 | return result; |
2398 | } | |
2399 | ||
2400 | EXPORT_SYMBOL(lock_may_read); | |
2401 | ||
2402 | /** | |
2403 | * lock_may_write - checks that the region is free of locks | |
2404 | * @inode: the inode that is being written | |
2405 | * @start: the first byte to write | |
2406 | * @len: the number of bytes to write | |
2407 | * | |
2408 | * Emulates Windows locking requirements. Whole-file | |
2409 | * mandatory locks (share modes) can prohibit a write and | |
2410 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2411 | * | |
2412 | * N.B. this function is only ever called | |
2413 | * from knfsd and ownership of locks is never checked. | |
2414 | */ | |
2415 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2416 | { | |
2417 | struct file_lock *fl; | |
2418 | int result = 1; | |
1c8c601a JL |
2419 | |
2420 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2421 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2422 | if (IS_POSIX(fl)) { | |
2423 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2424 | continue; | |
2425 | } else if (IS_FLOCK(fl)) { | |
2426 | if (!(fl->fl_type & LOCK_MAND)) | |
2427 | continue; | |
2428 | if (fl->fl_type & LOCK_WRITE) | |
2429 | continue; | |
2430 | } else | |
2431 | continue; | |
2432 | result = 0; | |
2433 | break; | |
2434 | } | |
1c8c601a | 2435 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2436 | return result; |
2437 | } | |
2438 | ||
2439 | EXPORT_SYMBOL(lock_may_write); | |
2440 | ||
1da177e4 LT |
2441 | static int __init filelock_init(void) |
2442 | { | |
2443 | filelock_cache = kmem_cache_create("file_lock_cache", | |
ee19cc40 MS |
2444 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2445 | ||
1da177e4 LT |
2446 | return 0; |
2447 | } | |
2448 | ||
2449 | core_initcall(filelock_init); |