]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/locks.c | |
3 | * | |
4 | * Provide support for fcntl()'s F_GETLK, F_SETLK, and F_SETLKW calls. | |
5 | * Doug Evans (dje@spiff.uucp), August 07, 1992 | |
6 | * | |
7 | * Deadlock detection added. | |
8 | * FIXME: one thing isn't handled yet: | |
9 | * - mandatory locks (requires lots of changes elsewhere) | |
10 | * Kelly Carmichael (kelly@[142.24.8.65]), September 17, 1994. | |
11 | * | |
12 | * Miscellaneous edits, and a total rewrite of posix_lock_file() code. | |
13 | * Kai Petzke (wpp@marie.physik.tu-berlin.de), 1994 | |
14 | * | |
15 | * Converted file_lock_table to a linked list from an array, which eliminates | |
16 | * the limits on how many active file locks are open. | |
17 | * Chad Page (pageone@netcom.com), November 27, 1994 | |
18 | * | |
19 | * Removed dependency on file descriptors. dup()'ed file descriptors now | |
20 | * get the same locks as the original file descriptors, and a close() on | |
21 | * any file descriptor removes ALL the locks on the file for the current | |
22 | * process. Since locks still depend on the process id, locks are inherited | |
23 | * after an exec() but not after a fork(). This agrees with POSIX, and both | |
24 | * BSD and SVR4 practice. | |
25 | * Andy Walker (andy@lysaker.kvaerner.no), February 14, 1995 | |
26 | * | |
27 | * Scrapped free list which is redundant now that we allocate locks | |
28 | * dynamically with kmalloc()/kfree(). | |
29 | * Andy Walker (andy@lysaker.kvaerner.no), February 21, 1995 | |
30 | * | |
31 | * Implemented two lock personalities - FL_FLOCK and FL_POSIX. | |
32 | * | |
33 | * FL_POSIX locks are created with calls to fcntl() and lockf() through the | |
34 | * fcntl() system call. They have the semantics described above. | |
35 | * | |
36 | * FL_FLOCK locks are created with calls to flock(), through the flock() | |
37 | * system call, which is new. Old C libraries implement flock() via fcntl() | |
38 | * and will continue to use the old, broken implementation. | |
39 | * | |
40 | * FL_FLOCK locks follow the 4.4 BSD flock() semantics. They are associated | |
41 | * with a file pointer (filp). As a result they can be shared by a parent | |
42 | * process and its children after a fork(). They are removed when the last | |
43 | * file descriptor referring to the file pointer is closed (unless explicitly | |
44 | * unlocked). | |
45 | * | |
46 | * FL_FLOCK locks never deadlock, an existing lock is always removed before | |
47 | * upgrading from shared to exclusive (or vice versa). When this happens | |
48 | * any processes blocked by the current lock are woken up and allowed to | |
49 | * run before the new lock is applied. | |
50 | * Andy Walker (andy@lysaker.kvaerner.no), June 09, 1995 | |
51 | * | |
52 | * Removed some race conditions in flock_lock_file(), marked other possible | |
53 | * races. Just grep for FIXME to see them. | |
54 | * Dmitry Gorodchanin (pgmdsg@ibi.com), February 09, 1996. | |
55 | * | |
56 | * Addressed Dmitry's concerns. Deadlock checking no longer recursive. | |
57 | * Lock allocation changed to GFP_ATOMIC as we can't afford to sleep | |
58 | * once we've checked for blocking and deadlocking. | |
59 | * Andy Walker (andy@lysaker.kvaerner.no), April 03, 1996. | |
60 | * | |
61 | * Initial implementation of mandatory locks. SunOS turned out to be | |
62 | * a rotten model, so I implemented the "obvious" semantics. | |
395cf969 | 63 | * See 'Documentation/filesystems/mandatory-locking.txt' for details. |
1da177e4 LT |
64 | * Andy Walker (andy@lysaker.kvaerner.no), April 06, 1996. |
65 | * | |
66 | * Don't allow mandatory locks on mmap()'ed files. Added simple functions to | |
67 | * check if a file has mandatory locks, used by mmap(), open() and creat() to | |
68 | * see if system call should be rejected. Ref. HP-UX/SunOS/Solaris Reference | |
69 | * Manual, Section 2. | |
70 | * Andy Walker (andy@lysaker.kvaerner.no), April 09, 1996. | |
71 | * | |
72 | * Tidied up block list handling. Added '/proc/locks' interface. | |
73 | * Andy Walker (andy@lysaker.kvaerner.no), April 24, 1996. | |
74 | * | |
75 | * Fixed deadlock condition for pathological code that mixes calls to | |
76 | * flock() and fcntl(). | |
77 | * Andy Walker (andy@lysaker.kvaerner.no), April 29, 1996. | |
78 | * | |
79 | * Allow only one type of locking scheme (FL_POSIX or FL_FLOCK) to be in use | |
80 | * for a given file at a time. Changed the CONFIG_LOCK_MANDATORY scheme to | |
81 | * guarantee sensible behaviour in the case where file system modules might | |
82 | * be compiled with different options than the kernel itself. | |
83 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
84 | * | |
85 | * Added a couple of missing wake_up() calls. Thanks to Thomas Meckel | |
86 | * (Thomas.Meckel@mni.fh-giessen.de) for spotting this. | |
87 | * Andy Walker (andy@lysaker.kvaerner.no), May 15, 1996. | |
88 | * | |
89 | * Changed FL_POSIX locks to use the block list in the same way as FL_FLOCK | |
90 | * locks. Changed process synchronisation to avoid dereferencing locks that | |
91 | * have already been freed. | |
92 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 21, 1996. | |
93 | * | |
94 | * Made the block list a circular list to minimise searching in the list. | |
95 | * Andy Walker (andy@lysaker.kvaerner.no), Sep 25, 1996. | |
96 | * | |
97 | * Made mandatory locking a mount option. Default is not to allow mandatory | |
98 | * locking. | |
99 | * Andy Walker (andy@lysaker.kvaerner.no), Oct 04, 1996. | |
100 | * | |
101 | * Some adaptations for NFS support. | |
102 | * Olaf Kirch (okir@monad.swb.de), Dec 1996, | |
103 | * | |
104 | * Fixed /proc/locks interface so that we can't overrun the buffer we are handed. | |
105 | * Andy Walker (andy@lysaker.kvaerner.no), May 12, 1997. | |
106 | * | |
107 | * Use slab allocator instead of kmalloc/kfree. | |
108 | * Use generic list implementation from <linux/list.h>. | |
109 | * Sped up posix_locks_deadlock by only considering blocked locks. | |
110 | * Matthew Wilcox <willy@debian.org>, March, 2000. | |
111 | * | |
112 | * Leases and LOCK_MAND | |
113 | * Matthew Wilcox <willy@debian.org>, June, 2000. | |
114 | * Stephen Rothwell <sfr@canb.auug.org.au>, June, 2000. | |
115 | */ | |
116 | ||
117 | #include <linux/capability.h> | |
118 | #include <linux/file.h> | |
9f3acc31 | 119 | #include <linux/fdtable.h> |
1da177e4 LT |
120 | #include <linux/fs.h> |
121 | #include <linux/init.h> | |
122 | #include <linux/module.h> | |
123 | #include <linux/security.h> | |
124 | #include <linux/slab.h> | |
1da177e4 LT |
125 | #include <linux/syscalls.h> |
126 | #include <linux/time.h> | |
4fb3a538 | 127 | #include <linux/rcupdate.h> |
ab1f1611 | 128 | #include <linux/pid_namespace.h> |
48f74186 | 129 | #include <linux/hashtable.h> |
7012b02a JL |
130 | #include <linux/percpu.h> |
131 | #include <linux/lglock.h> | |
1da177e4 | 132 | |
62af4f1f JL |
133 | #define CREATE_TRACE_POINTS |
134 | #include <trace/events/filelock.h> | |
135 | ||
1da177e4 LT |
136 | #include <asm/uaccess.h> |
137 | ||
138 | #define IS_POSIX(fl) (fl->fl_flags & FL_POSIX) | |
139 | #define IS_FLOCK(fl) (fl->fl_flags & FL_FLOCK) | |
617588d5 | 140 | #define IS_LEASE(fl) (fl->fl_flags & (FL_LEASE|FL_DELEG)) |
cff2fce5 | 141 | #define IS_OFDLCK(fl) (fl->fl_flags & FL_OFDLCK) |
1da177e4 | 142 | |
ab83fa4b BF |
143 | static bool lease_breaking(struct file_lock *fl) |
144 | { | |
778fc546 BF |
145 | return fl->fl_flags & (FL_UNLOCK_PENDING | FL_DOWNGRADE_PENDING); |
146 | } | |
147 | ||
148 | static int target_leasetype(struct file_lock *fl) | |
149 | { | |
150 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
151 | return F_UNLCK; | |
152 | if (fl->fl_flags & FL_DOWNGRADE_PENDING) | |
153 | return F_RDLCK; | |
154 | return fl->fl_type; | |
ab83fa4b BF |
155 | } |
156 | ||
1da177e4 LT |
157 | int leases_enable = 1; |
158 | int lease_break_time = 45; | |
159 | ||
160 | #define for_each_lock(inode, lockp) \ | |
161 | for (lockp = &inode->i_flock; *lockp != NULL; lockp = &(*lockp)->fl_next) | |
162 | ||
1c8c601a | 163 | /* |
7012b02a JL |
164 | * The global file_lock_list is only used for displaying /proc/locks, so we |
165 | * keep a list on each CPU, with each list protected by its own spinlock via | |
166 | * the file_lock_lglock. Note that alterations to the list also require that | |
167 | * the relevant i_lock is held. | |
1c8c601a | 168 | */ |
7012b02a JL |
169 | DEFINE_STATIC_LGLOCK(file_lock_lglock); |
170 | static DEFINE_PER_CPU(struct hlist_head, file_lock_list); | |
88974691 | 171 | |
1c8c601a | 172 | /* |
48f74186 | 173 | * The blocked_hash is used to find POSIX lock loops for deadlock detection. |
7b2296af | 174 | * It is protected by blocked_lock_lock. |
48f74186 JL |
175 | * |
176 | * We hash locks by lockowner in order to optimize searching for the lock a | |
177 | * particular lockowner is waiting on. | |
178 | * | |
179 | * FIXME: make this value scale via some heuristic? We generally will want more | |
180 | * buckets when we have more lockowners holding locks, but that's a little | |
181 | * difficult to determine without knowing what the workload will look like. | |
1c8c601a | 182 | */ |
48f74186 JL |
183 | #define BLOCKED_HASH_BITS 7 |
184 | static DEFINE_HASHTABLE(blocked_hash, BLOCKED_HASH_BITS); | |
88974691 | 185 | |
1c8c601a | 186 | /* |
7b2296af JL |
187 | * This lock protects the blocked_hash. Generally, if you're accessing it, you |
188 | * want to be holding this lock. | |
1c8c601a JL |
189 | * |
190 | * In addition, it also protects the fl->fl_block list, and the fl->fl_next | |
191 | * pointer for file_lock structures that are acting as lock requests (in | |
192 | * contrast to those that are acting as records of acquired locks). | |
193 | * | |
194 | * Note that when we acquire this lock in order to change the above fields, | |
195 | * we often hold the i_lock as well. In certain cases, when reading the fields | |
196 | * protected by this lock, we can skip acquiring it iff we already hold the | |
197 | * i_lock. | |
198 | * | |
199 | * In particular, adding an entry to the fl_block list requires that you hold | |
200 | * both the i_lock and the blocked_lock_lock (acquired in that order). Deleting | |
201 | * an entry from the list however only requires the file_lock_lock. | |
202 | */ | |
7b2296af | 203 | static DEFINE_SPINLOCK(blocked_lock_lock); |
1da177e4 | 204 | |
e18b890b | 205 | static struct kmem_cache *filelock_cache __read_mostly; |
1da177e4 | 206 | |
ee19cc40 | 207 | static void locks_init_lock_heads(struct file_lock *fl) |
a51cb91d | 208 | { |
139ca04e | 209 | INIT_HLIST_NODE(&fl->fl_link); |
ee19cc40 MS |
210 | INIT_LIST_HEAD(&fl->fl_block); |
211 | init_waitqueue_head(&fl->fl_wait); | |
a51cb91d MS |
212 | } |
213 | ||
1da177e4 | 214 | /* Allocate an empty lock structure. */ |
c5b1f0d9 | 215 | struct file_lock *locks_alloc_lock(void) |
1da177e4 | 216 | { |
ee19cc40 | 217 | struct file_lock *fl = kmem_cache_zalloc(filelock_cache, GFP_KERNEL); |
a51cb91d MS |
218 | |
219 | if (fl) | |
ee19cc40 | 220 | locks_init_lock_heads(fl); |
a51cb91d MS |
221 | |
222 | return fl; | |
1da177e4 | 223 | } |
c5b1f0d9 | 224 | EXPORT_SYMBOL_GPL(locks_alloc_lock); |
1da177e4 | 225 | |
a9e61e25 | 226 | void locks_release_private(struct file_lock *fl) |
47831f35 TM |
227 | { |
228 | if (fl->fl_ops) { | |
229 | if (fl->fl_ops->fl_release_private) | |
230 | fl->fl_ops->fl_release_private(fl); | |
231 | fl->fl_ops = NULL; | |
232 | } | |
068535f1 | 233 | fl->fl_lmops = NULL; |
47831f35 TM |
234 | |
235 | } | |
a9e61e25 | 236 | EXPORT_SYMBOL_GPL(locks_release_private); |
47831f35 | 237 | |
1da177e4 | 238 | /* Free a lock which is not in use. */ |
05fa3135 | 239 | void locks_free_lock(struct file_lock *fl) |
1da177e4 | 240 | { |
5ce29646 MS |
241 | BUG_ON(waitqueue_active(&fl->fl_wait)); |
242 | BUG_ON(!list_empty(&fl->fl_block)); | |
139ca04e | 243 | BUG_ON(!hlist_unhashed(&fl->fl_link)); |
1da177e4 | 244 | |
47831f35 | 245 | locks_release_private(fl); |
1da177e4 LT |
246 | kmem_cache_free(filelock_cache, fl); |
247 | } | |
05fa3135 | 248 | EXPORT_SYMBOL(locks_free_lock); |
1da177e4 LT |
249 | |
250 | void locks_init_lock(struct file_lock *fl) | |
251 | { | |
ee19cc40 MS |
252 | memset(fl, 0, sizeof(struct file_lock)); |
253 | locks_init_lock_heads(fl); | |
1da177e4 LT |
254 | } |
255 | ||
256 | EXPORT_SYMBOL(locks_init_lock); | |
257 | ||
47831f35 TM |
258 | static void locks_copy_private(struct file_lock *new, struct file_lock *fl) |
259 | { | |
260 | if (fl->fl_ops) { | |
261 | if (fl->fl_ops->fl_copy_lock) | |
262 | fl->fl_ops->fl_copy_lock(new, fl); | |
263 | new->fl_ops = fl->fl_ops; | |
264 | } | |
bb8430a2 | 265 | if (fl->fl_lmops) |
47831f35 | 266 | new->fl_lmops = fl->fl_lmops; |
47831f35 TM |
267 | } |
268 | ||
1da177e4 LT |
269 | /* |
270 | * Initialize a new lock from an existing file_lock structure. | |
271 | */ | |
1a747ee0 | 272 | void __locks_copy_lock(struct file_lock *new, const struct file_lock *fl) |
1da177e4 LT |
273 | { |
274 | new->fl_owner = fl->fl_owner; | |
275 | new->fl_pid = fl->fl_pid; | |
0996905f | 276 | new->fl_file = NULL; |
1da177e4 LT |
277 | new->fl_flags = fl->fl_flags; |
278 | new->fl_type = fl->fl_type; | |
279 | new->fl_start = fl->fl_start; | |
280 | new->fl_end = fl->fl_end; | |
0996905f TM |
281 | new->fl_ops = NULL; |
282 | new->fl_lmops = NULL; | |
283 | } | |
3dd7b71c | 284 | EXPORT_SYMBOL(__locks_copy_lock); |
0996905f TM |
285 | |
286 | void locks_copy_lock(struct file_lock *new, struct file_lock *fl) | |
287 | { | |
566709bd JL |
288 | /* "new" must be a freshly-initialized lock */ |
289 | WARN_ON_ONCE(new->fl_ops); | |
0996905f TM |
290 | |
291 | __locks_copy_lock(new, fl); | |
292 | new->fl_file = fl->fl_file; | |
1da177e4 LT |
293 | new->fl_ops = fl->fl_ops; |
294 | new->fl_lmops = fl->fl_lmops; | |
47831f35 TM |
295 | |
296 | locks_copy_private(new, fl); | |
1da177e4 LT |
297 | } |
298 | ||
299 | EXPORT_SYMBOL(locks_copy_lock); | |
300 | ||
301 | static inline int flock_translate_cmd(int cmd) { | |
302 | if (cmd & LOCK_MAND) | |
303 | return cmd & (LOCK_MAND | LOCK_RW); | |
304 | switch (cmd) { | |
305 | case LOCK_SH: | |
306 | return F_RDLCK; | |
307 | case LOCK_EX: | |
308 | return F_WRLCK; | |
309 | case LOCK_UN: | |
310 | return F_UNLCK; | |
311 | } | |
312 | return -EINVAL; | |
313 | } | |
314 | ||
315 | /* Fill in a file_lock structure with an appropriate FLOCK lock. */ | |
316 | static int flock_make_lock(struct file *filp, struct file_lock **lock, | |
317 | unsigned int cmd) | |
318 | { | |
319 | struct file_lock *fl; | |
320 | int type = flock_translate_cmd(cmd); | |
321 | if (type < 0) | |
322 | return type; | |
323 | ||
324 | fl = locks_alloc_lock(); | |
325 | if (fl == NULL) | |
326 | return -ENOMEM; | |
327 | ||
328 | fl->fl_file = filp; | |
73a8f5f7 | 329 | fl->fl_owner = filp; |
1da177e4 LT |
330 | fl->fl_pid = current->tgid; |
331 | fl->fl_flags = FL_FLOCK; | |
332 | fl->fl_type = type; | |
333 | fl->fl_end = OFFSET_MAX; | |
334 | ||
335 | *lock = fl; | |
336 | return 0; | |
337 | } | |
338 | ||
0ec4f431 | 339 | static int assign_type(struct file_lock *fl, long type) |
1da177e4 LT |
340 | { |
341 | switch (type) { | |
342 | case F_RDLCK: | |
343 | case F_WRLCK: | |
344 | case F_UNLCK: | |
345 | fl->fl_type = type; | |
346 | break; | |
347 | default: | |
348 | return -EINVAL; | |
349 | } | |
350 | return 0; | |
351 | } | |
352 | ||
ef12e72a BF |
353 | static int flock64_to_posix_lock(struct file *filp, struct file_lock *fl, |
354 | struct flock64 *l) | |
1da177e4 | 355 | { |
1da177e4 | 356 | switch (l->l_whence) { |
f5579f8c | 357 | case SEEK_SET: |
ef12e72a | 358 | fl->fl_start = 0; |
1da177e4 | 359 | break; |
f5579f8c | 360 | case SEEK_CUR: |
ef12e72a | 361 | fl->fl_start = filp->f_pos; |
1da177e4 | 362 | break; |
f5579f8c | 363 | case SEEK_END: |
ef12e72a | 364 | fl->fl_start = i_size_read(file_inode(filp)); |
1da177e4 LT |
365 | break; |
366 | default: | |
367 | return -EINVAL; | |
368 | } | |
ef12e72a BF |
369 | if (l->l_start > OFFSET_MAX - fl->fl_start) |
370 | return -EOVERFLOW; | |
371 | fl->fl_start += l->l_start; | |
372 | if (fl->fl_start < 0) | |
373 | return -EINVAL; | |
1da177e4 LT |
374 | |
375 | /* POSIX-1996 leaves the case l->l_len < 0 undefined; | |
376 | POSIX-2001 defines it. */ | |
4c780a46 | 377 | if (l->l_len > 0) { |
ef12e72a BF |
378 | if (l->l_len - 1 > OFFSET_MAX - fl->fl_start) |
379 | return -EOVERFLOW; | |
380 | fl->fl_end = fl->fl_start + l->l_len - 1; | |
381 | ||
4c780a46 | 382 | } else if (l->l_len < 0) { |
ef12e72a | 383 | if (fl->fl_start + l->l_len < 0) |
4c780a46 | 384 | return -EINVAL; |
ef12e72a BF |
385 | fl->fl_end = fl->fl_start - 1; |
386 | fl->fl_start += l->l_len; | |
387 | } else | |
388 | fl->fl_end = OFFSET_MAX; | |
389 | ||
1da177e4 LT |
390 | fl->fl_owner = current->files; |
391 | fl->fl_pid = current->tgid; | |
392 | fl->fl_file = filp; | |
393 | fl->fl_flags = FL_POSIX; | |
394 | fl->fl_ops = NULL; | |
395 | fl->fl_lmops = NULL; | |
396 | ||
397 | return assign_type(fl, l->l_type); | |
398 | } | |
399 | ||
ef12e72a BF |
400 | /* Verify a "struct flock" and copy it to a "struct file_lock" as a POSIX |
401 | * style lock. | |
402 | */ | |
403 | static int flock_to_posix_lock(struct file *filp, struct file_lock *fl, | |
404 | struct flock *l) | |
1da177e4 | 405 | { |
ef12e72a BF |
406 | struct flock64 ll = { |
407 | .l_type = l->l_type, | |
408 | .l_whence = l->l_whence, | |
409 | .l_start = l->l_start, | |
410 | .l_len = l->l_len, | |
411 | }; | |
412 | ||
413 | return flock64_to_posix_lock(filp, fl, &ll); | |
1da177e4 | 414 | } |
1da177e4 LT |
415 | |
416 | /* default lease lock manager operations */ | |
417 | static void lease_break_callback(struct file_lock *fl) | |
418 | { | |
419 | kill_fasync(&fl->fl_fasync, SIGIO, POLL_MSG); | |
420 | } | |
421 | ||
7b021967 | 422 | static const struct lock_manager_operations lease_manager_ops = { |
8fb47a4f | 423 | .lm_break = lease_break_callback, |
8fb47a4f | 424 | .lm_change = lease_modify, |
1da177e4 LT |
425 | }; |
426 | ||
427 | /* | |
428 | * Initialize a lease, use the default lock manager operations | |
429 | */ | |
0ec4f431 | 430 | static int lease_init(struct file *filp, long type, struct file_lock *fl) |
1da177e4 | 431 | { |
75dff55a TM |
432 | if (assign_type(fl, type) != 0) |
433 | return -EINVAL; | |
434 | ||
73a8f5f7 | 435 | fl->fl_owner = current->files; |
1da177e4 LT |
436 | fl->fl_pid = current->tgid; |
437 | ||
438 | fl->fl_file = filp; | |
439 | fl->fl_flags = FL_LEASE; | |
1da177e4 LT |
440 | fl->fl_start = 0; |
441 | fl->fl_end = OFFSET_MAX; | |
442 | fl->fl_ops = NULL; | |
443 | fl->fl_lmops = &lease_manager_ops; | |
444 | return 0; | |
445 | } | |
446 | ||
447 | /* Allocate a file_lock initialised to this type of lease */ | |
0ec4f431 | 448 | static struct file_lock *lease_alloc(struct file *filp, long type) |
1da177e4 LT |
449 | { |
450 | struct file_lock *fl = locks_alloc_lock(); | |
75dff55a | 451 | int error = -ENOMEM; |
1da177e4 LT |
452 | |
453 | if (fl == NULL) | |
e32b8ee2 | 454 | return ERR_PTR(error); |
1da177e4 LT |
455 | |
456 | error = lease_init(filp, type, fl); | |
75dff55a TM |
457 | if (error) { |
458 | locks_free_lock(fl); | |
e32b8ee2 | 459 | return ERR_PTR(error); |
75dff55a | 460 | } |
e32b8ee2 | 461 | return fl; |
1da177e4 LT |
462 | } |
463 | ||
464 | /* Check if two locks overlap each other. | |
465 | */ | |
466 | static inline int locks_overlap(struct file_lock *fl1, struct file_lock *fl2) | |
467 | { | |
468 | return ((fl1->fl_end >= fl2->fl_start) && | |
469 | (fl2->fl_end >= fl1->fl_start)); | |
470 | } | |
471 | ||
472 | /* | |
473 | * Check whether two locks have the same owner. | |
474 | */ | |
33443c42 | 475 | static int posix_same_owner(struct file_lock *fl1, struct file_lock *fl2) |
1da177e4 | 476 | { |
8fb47a4f | 477 | if (fl1->fl_lmops && fl1->fl_lmops->lm_compare_owner) |
1da177e4 | 478 | return fl2->fl_lmops == fl1->fl_lmops && |
8fb47a4f | 479 | fl1->fl_lmops->lm_compare_owner(fl1, fl2); |
1da177e4 LT |
480 | return fl1->fl_owner == fl2->fl_owner; |
481 | } | |
482 | ||
7012b02a | 483 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 484 | static void locks_insert_global_locks(struct file_lock *fl) |
88974691 | 485 | { |
7012b02a JL |
486 | lg_local_lock(&file_lock_lglock); |
487 | fl->fl_link_cpu = smp_processor_id(); | |
488 | hlist_add_head(&fl->fl_link, this_cpu_ptr(&file_lock_list)); | |
489 | lg_local_unlock(&file_lock_lglock); | |
88974691 JL |
490 | } |
491 | ||
7012b02a | 492 | /* Must be called with the i_lock held! */ |
6ca10ed8 | 493 | static void locks_delete_global_locks(struct file_lock *fl) |
88974691 | 494 | { |
7012b02a JL |
495 | /* |
496 | * Avoid taking lock if already unhashed. This is safe since this check | |
497 | * is done while holding the i_lock, and new insertions into the list | |
498 | * also require that it be held. | |
499 | */ | |
500 | if (hlist_unhashed(&fl->fl_link)) | |
501 | return; | |
502 | lg_local_lock_cpu(&file_lock_lglock, fl->fl_link_cpu); | |
139ca04e | 503 | hlist_del_init(&fl->fl_link); |
7012b02a | 504 | lg_local_unlock_cpu(&file_lock_lglock, fl->fl_link_cpu); |
88974691 JL |
505 | } |
506 | ||
3999e493 JL |
507 | static unsigned long |
508 | posix_owner_key(struct file_lock *fl) | |
509 | { | |
510 | if (fl->fl_lmops && fl->fl_lmops->lm_owner_key) | |
511 | return fl->fl_lmops->lm_owner_key(fl); | |
512 | return (unsigned long)fl->fl_owner; | |
513 | } | |
514 | ||
6ca10ed8 | 515 | static void locks_insert_global_blocked(struct file_lock *waiter) |
88974691 | 516 | { |
3999e493 | 517 | hash_add(blocked_hash, &waiter->fl_link, posix_owner_key(waiter)); |
88974691 JL |
518 | } |
519 | ||
6ca10ed8 | 520 | static void locks_delete_global_blocked(struct file_lock *waiter) |
88974691 | 521 | { |
48f74186 | 522 | hash_del(&waiter->fl_link); |
88974691 JL |
523 | } |
524 | ||
1da177e4 LT |
525 | /* Remove waiter from blocker's block list. |
526 | * When blocker ends up pointing to itself then the list is empty. | |
1c8c601a | 527 | * |
7b2296af | 528 | * Must be called with blocked_lock_lock held. |
1da177e4 | 529 | */ |
33443c42 | 530 | static void __locks_delete_block(struct file_lock *waiter) |
1da177e4 | 531 | { |
88974691 | 532 | locks_delete_global_blocked(waiter); |
1da177e4 | 533 | list_del_init(&waiter->fl_block); |
1da177e4 LT |
534 | waiter->fl_next = NULL; |
535 | } | |
536 | ||
1a9e64a7 | 537 | static void locks_delete_block(struct file_lock *waiter) |
1da177e4 | 538 | { |
7b2296af | 539 | spin_lock(&blocked_lock_lock); |
1da177e4 | 540 | __locks_delete_block(waiter); |
7b2296af | 541 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
542 | } |
543 | ||
544 | /* Insert waiter into blocker's block list. | |
545 | * We use a circular list so that processes can be easily woken up in | |
546 | * the order they blocked. The documentation doesn't require this but | |
547 | * it seems like the reasonable thing to do. | |
1c8c601a | 548 | * |
7b2296af | 549 | * Must be called with both the i_lock and blocked_lock_lock held. The fl_block |
46dad760 | 550 | * list itself is protected by the blocked_lock_lock, but by ensuring that the |
7b2296af | 551 | * i_lock is also held on insertions we can avoid taking the blocked_lock_lock |
4e8c765d | 552 | * in some cases when we see that the fl_block list is empty. |
1da177e4 | 553 | */ |
1c8c601a JL |
554 | static void __locks_insert_block(struct file_lock *blocker, |
555 | struct file_lock *waiter) | |
1da177e4 | 556 | { |
6dc0fe8f | 557 | BUG_ON(!list_empty(&waiter->fl_block)); |
1da177e4 | 558 | waiter->fl_next = blocker; |
88974691 | 559 | list_add_tail(&waiter->fl_block, &blocker->fl_block); |
cff2fce5 | 560 | if (IS_POSIX(blocker) && !IS_OFDLCK(blocker)) |
1c8c601a JL |
561 | locks_insert_global_blocked(waiter); |
562 | } | |
563 | ||
564 | /* Must be called with i_lock held. */ | |
565 | static void locks_insert_block(struct file_lock *blocker, | |
566 | struct file_lock *waiter) | |
567 | { | |
7b2296af | 568 | spin_lock(&blocked_lock_lock); |
1c8c601a | 569 | __locks_insert_block(blocker, waiter); |
7b2296af | 570 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
571 | } |
572 | ||
1cb36012 JL |
573 | /* |
574 | * Wake up processes blocked waiting for blocker. | |
575 | * | |
1c8c601a | 576 | * Must be called with the inode->i_lock held! |
1da177e4 LT |
577 | */ |
578 | static void locks_wake_up_blocks(struct file_lock *blocker) | |
579 | { | |
4e8c765d JL |
580 | /* |
581 | * Avoid taking global lock if list is empty. This is safe since new | |
582 | * blocked requests are only added to the list under the i_lock, and | |
583 | * the i_lock is always held here. Note that removal from the fl_block | |
584 | * list does not require the i_lock, so we must recheck list_empty() | |
7b2296af | 585 | * after acquiring the blocked_lock_lock. |
4e8c765d JL |
586 | */ |
587 | if (list_empty(&blocker->fl_block)) | |
588 | return; | |
589 | ||
7b2296af | 590 | spin_lock(&blocked_lock_lock); |
1da177e4 | 591 | while (!list_empty(&blocker->fl_block)) { |
f0c1cd0e PE |
592 | struct file_lock *waiter; |
593 | ||
594 | waiter = list_first_entry(&blocker->fl_block, | |
1da177e4 LT |
595 | struct file_lock, fl_block); |
596 | __locks_delete_block(waiter); | |
8fb47a4f BF |
597 | if (waiter->fl_lmops && waiter->fl_lmops->lm_notify) |
598 | waiter->fl_lmops->lm_notify(waiter); | |
1da177e4 LT |
599 | else |
600 | wake_up(&waiter->fl_wait); | |
601 | } | |
7b2296af | 602 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
603 | } |
604 | ||
605 | /* Insert file lock fl into an inode's lock list at the position indicated | |
606 | * by pos. At the same time add the lock to the global file lock list. | |
1c8c601a JL |
607 | * |
608 | * Must be called with the i_lock held! | |
1da177e4 LT |
609 | */ |
610 | static void locks_insert_lock(struct file_lock **pos, struct file_lock *fl) | |
611 | { | |
ab1f1611 VG |
612 | fl->fl_nspid = get_pid(task_tgid(current)); |
613 | ||
1da177e4 LT |
614 | /* insert into file's list */ |
615 | fl->fl_next = *pos; | |
616 | *pos = fl; | |
88974691 JL |
617 | |
618 | locks_insert_global_locks(fl); | |
1da177e4 LT |
619 | } |
620 | ||
24cbe784 JL |
621 | /** |
622 | * locks_delete_lock - Delete a lock and then free it. | |
623 | * @thisfl_p: pointer that points to the fl_next field of the previous | |
624 | * inode->i_flock list entry | |
625 | * | |
626 | * Unlink a lock from all lists and free the namespace reference, but don't | |
627 | * free it yet. Wake up processes that are blocked waiting for this lock and | |
628 | * notify the FS that the lock has been cleared. | |
1c8c601a JL |
629 | * |
630 | * Must be called with the i_lock held! | |
1da177e4 | 631 | */ |
24cbe784 | 632 | static void locks_unlink_lock(struct file_lock **thisfl_p) |
1da177e4 LT |
633 | { |
634 | struct file_lock *fl = *thisfl_p; | |
635 | ||
88974691 JL |
636 | locks_delete_global_locks(fl); |
637 | ||
1da177e4 LT |
638 | *thisfl_p = fl->fl_next; |
639 | fl->fl_next = NULL; | |
1da177e4 | 640 | |
ab1f1611 VG |
641 | if (fl->fl_nspid) { |
642 | put_pid(fl->fl_nspid); | |
643 | fl->fl_nspid = NULL; | |
644 | } | |
645 | ||
1da177e4 | 646 | locks_wake_up_blocks(fl); |
24cbe784 JL |
647 | } |
648 | ||
649 | /* | |
650 | * Unlink a lock from all lists and free it. | |
651 | * | |
652 | * Must be called with i_lock held! | |
653 | */ | |
654 | static void locks_delete_lock(struct file_lock **thisfl_p) | |
655 | { | |
656 | struct file_lock *fl = *thisfl_p; | |
657 | ||
658 | locks_unlink_lock(thisfl_p); | |
1da177e4 LT |
659 | locks_free_lock(fl); |
660 | } | |
661 | ||
662 | /* Determine if lock sys_fl blocks lock caller_fl. Common functionality | |
663 | * checks for shared/exclusive status of overlapping locks. | |
664 | */ | |
665 | static int locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
666 | { | |
667 | if (sys_fl->fl_type == F_WRLCK) | |
668 | return 1; | |
669 | if (caller_fl->fl_type == F_WRLCK) | |
670 | return 1; | |
671 | return 0; | |
672 | } | |
673 | ||
674 | /* Determine if lock sys_fl blocks lock caller_fl. POSIX specific | |
675 | * checking before calling the locks_conflict(). | |
676 | */ | |
677 | static int posix_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
678 | { | |
679 | /* POSIX locks owned by the same process do not conflict with | |
680 | * each other. | |
681 | */ | |
682 | if (!IS_POSIX(sys_fl) || posix_same_owner(caller_fl, sys_fl)) | |
683 | return (0); | |
684 | ||
685 | /* Check whether they overlap */ | |
686 | if (!locks_overlap(caller_fl, sys_fl)) | |
687 | return 0; | |
688 | ||
689 | return (locks_conflict(caller_fl, sys_fl)); | |
690 | } | |
691 | ||
692 | /* Determine if lock sys_fl blocks lock caller_fl. FLOCK specific | |
693 | * checking before calling the locks_conflict(). | |
694 | */ | |
695 | static int flock_locks_conflict(struct file_lock *caller_fl, struct file_lock *sys_fl) | |
696 | { | |
697 | /* FLOCK locks referring to the same filp do not conflict with | |
698 | * each other. | |
699 | */ | |
700 | if (!IS_FLOCK(sys_fl) || (caller_fl->fl_file == sys_fl->fl_file)) | |
701 | return (0); | |
702 | if ((caller_fl->fl_type & LOCK_MAND) || (sys_fl->fl_type & LOCK_MAND)) | |
703 | return 0; | |
704 | ||
705 | return (locks_conflict(caller_fl, sys_fl)); | |
706 | } | |
707 | ||
6d34ac19 | 708 | void |
9d6a8c5c | 709 | posix_test_lock(struct file *filp, struct file_lock *fl) |
1da177e4 LT |
710 | { |
711 | struct file_lock *cfl; | |
1c8c601a | 712 | struct inode *inode = file_inode(filp); |
1da177e4 | 713 | |
1c8c601a | 714 | spin_lock(&inode->i_lock); |
496ad9aa | 715 | for (cfl = file_inode(filp)->i_flock; cfl; cfl = cfl->fl_next) { |
1da177e4 LT |
716 | if (!IS_POSIX(cfl)) |
717 | continue; | |
b842e240 | 718 | if (posix_locks_conflict(fl, cfl)) |
1da177e4 LT |
719 | break; |
720 | } | |
ab1f1611 | 721 | if (cfl) { |
9d6a8c5c | 722 | __locks_copy_lock(fl, cfl); |
ab1f1611 | 723 | if (cfl->fl_nspid) |
6c5f3e7b | 724 | fl->fl_pid = pid_vnr(cfl->fl_nspid); |
ab1f1611 | 725 | } else |
129a84de | 726 | fl->fl_type = F_UNLCK; |
1c8c601a | 727 | spin_unlock(&inode->i_lock); |
6d34ac19 | 728 | return; |
1da177e4 | 729 | } |
1da177e4 LT |
730 | EXPORT_SYMBOL(posix_test_lock); |
731 | ||
b533184f BF |
732 | /* |
733 | * Deadlock detection: | |
734 | * | |
735 | * We attempt to detect deadlocks that are due purely to posix file | |
736 | * locks. | |
1da177e4 | 737 | * |
b533184f BF |
738 | * We assume that a task can be waiting for at most one lock at a time. |
739 | * So for any acquired lock, the process holding that lock may be | |
740 | * waiting on at most one other lock. That lock in turns may be held by | |
741 | * someone waiting for at most one other lock. Given a requested lock | |
742 | * caller_fl which is about to wait for a conflicting lock block_fl, we | |
743 | * follow this chain of waiters to ensure we are not about to create a | |
744 | * cycle. | |
1da177e4 | 745 | * |
b533184f BF |
746 | * Since we do this before we ever put a process to sleep on a lock, we |
747 | * are ensured that there is never a cycle; that is what guarantees that | |
748 | * the while() loop in posix_locks_deadlock() eventually completes. | |
97855b49 | 749 | * |
b533184f BF |
750 | * Note: the above assumption may not be true when handling lock |
751 | * requests from a broken NFS client. It may also fail in the presence | |
752 | * of tasks (such as posix threads) sharing the same open file table. | |
b533184f | 753 | * To handle those cases, we just bail out after a few iterations. |
57b65325 | 754 | * |
cff2fce5 | 755 | * For FL_OFDLCK locks, the owner is the filp, not the files_struct. |
57b65325 JL |
756 | * Because the owner is not even nominally tied to a thread of |
757 | * execution, the deadlock detection below can't reasonably work well. Just | |
758 | * skip it for those. | |
759 | * | |
cff2fce5 | 760 | * In principle, we could do a more limited deadlock detection on FL_OFDLCK |
57b65325 JL |
761 | * locks that just checks for the case where two tasks are attempting to |
762 | * upgrade from read to write locks on the same inode. | |
1da177e4 | 763 | */ |
97855b49 BF |
764 | |
765 | #define MAX_DEADLK_ITERATIONS 10 | |
766 | ||
b533184f BF |
767 | /* Find a lock that the owner of the given block_fl is blocking on. */ |
768 | static struct file_lock *what_owner_is_waiting_for(struct file_lock *block_fl) | |
769 | { | |
770 | struct file_lock *fl; | |
771 | ||
3999e493 | 772 | hash_for_each_possible(blocked_hash, fl, fl_link, posix_owner_key(block_fl)) { |
b533184f BF |
773 | if (posix_same_owner(fl, block_fl)) |
774 | return fl->fl_next; | |
775 | } | |
776 | return NULL; | |
777 | } | |
778 | ||
7b2296af | 779 | /* Must be called with the blocked_lock_lock held! */ |
b0904e14 | 780 | static int posix_locks_deadlock(struct file_lock *caller_fl, |
1da177e4 LT |
781 | struct file_lock *block_fl) |
782 | { | |
97855b49 | 783 | int i = 0; |
1da177e4 | 784 | |
57b65325 JL |
785 | /* |
786 | * This deadlock detector can't reasonably detect deadlocks with | |
cff2fce5 | 787 | * FL_OFDLCK locks, since they aren't owned by a process, per-se. |
57b65325 | 788 | */ |
cff2fce5 | 789 | if (IS_OFDLCK(caller_fl)) |
57b65325 JL |
790 | return 0; |
791 | ||
b533184f BF |
792 | while ((block_fl = what_owner_is_waiting_for(block_fl))) { |
793 | if (i++ > MAX_DEADLK_ITERATIONS) | |
794 | return 0; | |
795 | if (posix_same_owner(caller_fl, block_fl)) | |
796 | return 1; | |
1da177e4 LT |
797 | } |
798 | return 0; | |
799 | } | |
800 | ||
1da177e4 | 801 | /* Try to create a FLOCK lock on filp. We always insert new FLOCK locks |
02888f41 | 802 | * after any leases, but before any posix locks. |
f475ae95 TM |
803 | * |
804 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
805 | * whether or not a lock was successfully freed by testing the return | |
806 | * value for -ENOENT. | |
1da177e4 | 807 | */ |
993dfa87 | 808 | static int flock_lock_file(struct file *filp, struct file_lock *request) |
1da177e4 | 809 | { |
993dfa87 | 810 | struct file_lock *new_fl = NULL; |
1da177e4 | 811 | struct file_lock **before; |
496ad9aa | 812 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
813 | int error = 0; |
814 | int found = 0; | |
815 | ||
b89f4321 | 816 | if (!(request->fl_flags & FL_ACCESS) && (request->fl_type != F_UNLCK)) { |
84d535ad | 817 | new_fl = locks_alloc_lock(); |
b89f4321 AB |
818 | if (!new_fl) |
819 | return -ENOMEM; | |
84d535ad PE |
820 | } |
821 | ||
1c8c601a | 822 | spin_lock(&inode->i_lock); |
b89f4321 AB |
823 | if (request->fl_flags & FL_ACCESS) |
824 | goto find_conflict; | |
825 | ||
1da177e4 LT |
826 | for_each_lock(inode, before) { |
827 | struct file_lock *fl = *before; | |
828 | if (IS_POSIX(fl)) | |
829 | break; | |
830 | if (IS_LEASE(fl)) | |
831 | continue; | |
832 | if (filp != fl->fl_file) | |
833 | continue; | |
993dfa87 | 834 | if (request->fl_type == fl->fl_type) |
1da177e4 LT |
835 | goto out; |
836 | found = 1; | |
837 | locks_delete_lock(before); | |
838 | break; | |
839 | } | |
1da177e4 | 840 | |
f475ae95 TM |
841 | if (request->fl_type == F_UNLCK) { |
842 | if ((request->fl_flags & FL_EXISTS) && !found) | |
843 | error = -ENOENT; | |
993dfa87 | 844 | goto out; |
f475ae95 | 845 | } |
1da177e4 LT |
846 | |
847 | /* | |
848 | * If a higher-priority process was blocked on the old file lock, | |
849 | * give it the opportunity to lock the file. | |
850 | */ | |
b89f4321 | 851 | if (found) { |
1c8c601a | 852 | spin_unlock(&inode->i_lock); |
def01bc5 | 853 | cond_resched(); |
1c8c601a | 854 | spin_lock(&inode->i_lock); |
b89f4321 | 855 | } |
1da177e4 | 856 | |
f07f18dd | 857 | find_conflict: |
1da177e4 LT |
858 | for_each_lock(inode, before) { |
859 | struct file_lock *fl = *before; | |
860 | if (IS_POSIX(fl)) | |
861 | break; | |
862 | if (IS_LEASE(fl)) | |
863 | continue; | |
993dfa87 | 864 | if (!flock_locks_conflict(request, fl)) |
1da177e4 LT |
865 | continue; |
866 | error = -EAGAIN; | |
bde74e4b MS |
867 | if (!(request->fl_flags & FL_SLEEP)) |
868 | goto out; | |
869 | error = FILE_LOCK_DEFERRED; | |
870 | locks_insert_block(fl, request); | |
1da177e4 LT |
871 | goto out; |
872 | } | |
f07f18dd TM |
873 | if (request->fl_flags & FL_ACCESS) |
874 | goto out; | |
993dfa87 | 875 | locks_copy_lock(new_fl, request); |
0e2f6db8 | 876 | locks_insert_lock(before, new_fl); |
993dfa87 | 877 | new_fl = NULL; |
9cedc194 | 878 | error = 0; |
1da177e4 LT |
879 | |
880 | out: | |
1c8c601a | 881 | spin_unlock(&inode->i_lock); |
993dfa87 TM |
882 | if (new_fl) |
883 | locks_free_lock(new_fl); | |
1da177e4 LT |
884 | return error; |
885 | } | |
886 | ||
150b3934 | 887 | static int __posix_lock_file(struct inode *inode, struct file_lock *request, struct file_lock *conflock) |
1da177e4 LT |
888 | { |
889 | struct file_lock *fl; | |
39005d02 MS |
890 | struct file_lock *new_fl = NULL; |
891 | struct file_lock *new_fl2 = NULL; | |
1da177e4 LT |
892 | struct file_lock *left = NULL; |
893 | struct file_lock *right = NULL; | |
894 | struct file_lock **before; | |
b9746ef8 JL |
895 | int error; |
896 | bool added = false; | |
1da177e4 LT |
897 | |
898 | /* | |
899 | * We may need two file_lock structures for this operation, | |
900 | * so we get them in advance to avoid races. | |
39005d02 MS |
901 | * |
902 | * In some cases we can be sure, that no new locks will be needed | |
1da177e4 | 903 | */ |
39005d02 MS |
904 | if (!(request->fl_flags & FL_ACCESS) && |
905 | (request->fl_type != F_UNLCK || | |
906 | request->fl_start != 0 || request->fl_end != OFFSET_MAX)) { | |
907 | new_fl = locks_alloc_lock(); | |
908 | new_fl2 = locks_alloc_lock(); | |
909 | } | |
1da177e4 | 910 | |
1c8c601a | 911 | spin_lock(&inode->i_lock); |
1cb36012 JL |
912 | /* |
913 | * New lock request. Walk all POSIX locks and look for conflicts. If | |
914 | * there are any, either return error or put the request on the | |
48f74186 | 915 | * blocker's list of waiters and the global blocked_hash. |
1cb36012 | 916 | */ |
1da177e4 LT |
917 | if (request->fl_type != F_UNLCK) { |
918 | for_each_lock(inode, before) { | |
526985b9 | 919 | fl = *before; |
1da177e4 LT |
920 | if (!IS_POSIX(fl)) |
921 | continue; | |
922 | if (!posix_locks_conflict(request, fl)) | |
923 | continue; | |
5842add2 | 924 | if (conflock) |
1a747ee0 | 925 | __locks_copy_lock(conflock, fl); |
1da177e4 LT |
926 | error = -EAGAIN; |
927 | if (!(request->fl_flags & FL_SLEEP)) | |
928 | goto out; | |
1c8c601a JL |
929 | /* |
930 | * Deadlock detection and insertion into the blocked | |
931 | * locks list must be done while holding the same lock! | |
932 | */ | |
1da177e4 | 933 | error = -EDEADLK; |
7b2296af | 934 | spin_lock(&blocked_lock_lock); |
1c8c601a JL |
935 | if (likely(!posix_locks_deadlock(request, fl))) { |
936 | error = FILE_LOCK_DEFERRED; | |
937 | __locks_insert_block(fl, request); | |
938 | } | |
7b2296af | 939 | spin_unlock(&blocked_lock_lock); |
1da177e4 LT |
940 | goto out; |
941 | } | |
942 | } | |
943 | ||
944 | /* If we're just looking for a conflict, we're done. */ | |
945 | error = 0; | |
946 | if (request->fl_flags & FL_ACCESS) | |
947 | goto out; | |
948 | ||
1da177e4 | 949 | /* |
1da177e4 LT |
950 | * Find the first old lock with the same owner as the new lock. |
951 | */ | |
952 | ||
953 | before = &inode->i_flock; | |
954 | ||
955 | /* First skip locks owned by other processes. */ | |
956 | while ((fl = *before) && (!IS_POSIX(fl) || | |
957 | !posix_same_owner(request, fl))) { | |
958 | before = &fl->fl_next; | |
959 | } | |
960 | ||
1cb36012 | 961 | /* Process locks with this owner. */ |
1da177e4 LT |
962 | while ((fl = *before) && posix_same_owner(request, fl)) { |
963 | /* Detect adjacent or overlapping regions (if same lock type) | |
964 | */ | |
965 | if (request->fl_type == fl->fl_type) { | |
449231d6 OK |
966 | /* In all comparisons of start vs end, use |
967 | * "start - 1" rather than "end + 1". If end | |
968 | * is OFFSET_MAX, end + 1 will become negative. | |
969 | */ | |
1da177e4 LT |
970 | if (fl->fl_end < request->fl_start - 1) |
971 | goto next_lock; | |
972 | /* If the next lock in the list has entirely bigger | |
973 | * addresses than the new one, insert the lock here. | |
974 | */ | |
449231d6 | 975 | if (fl->fl_start - 1 > request->fl_end) |
1da177e4 LT |
976 | break; |
977 | ||
978 | /* If we come here, the new and old lock are of the | |
979 | * same type and adjacent or overlapping. Make one | |
980 | * lock yielding from the lower start address of both | |
981 | * locks to the higher end address. | |
982 | */ | |
983 | if (fl->fl_start > request->fl_start) | |
984 | fl->fl_start = request->fl_start; | |
985 | else | |
986 | request->fl_start = fl->fl_start; | |
987 | if (fl->fl_end < request->fl_end) | |
988 | fl->fl_end = request->fl_end; | |
989 | else | |
990 | request->fl_end = fl->fl_end; | |
991 | if (added) { | |
992 | locks_delete_lock(before); | |
993 | continue; | |
994 | } | |
995 | request = fl; | |
b9746ef8 | 996 | added = true; |
1da177e4 LT |
997 | } |
998 | else { | |
999 | /* Processing for different lock types is a bit | |
1000 | * more complex. | |
1001 | */ | |
1002 | if (fl->fl_end < request->fl_start) | |
1003 | goto next_lock; | |
1004 | if (fl->fl_start > request->fl_end) | |
1005 | break; | |
1006 | if (request->fl_type == F_UNLCK) | |
b9746ef8 | 1007 | added = true; |
1da177e4 LT |
1008 | if (fl->fl_start < request->fl_start) |
1009 | left = fl; | |
1010 | /* If the next lock in the list has a higher end | |
1011 | * address than the new one, insert the new one here. | |
1012 | */ | |
1013 | if (fl->fl_end > request->fl_end) { | |
1014 | right = fl; | |
1015 | break; | |
1016 | } | |
1017 | if (fl->fl_start >= request->fl_start) { | |
1018 | /* The new lock completely replaces an old | |
1019 | * one (This may happen several times). | |
1020 | */ | |
1021 | if (added) { | |
1022 | locks_delete_lock(before); | |
1023 | continue; | |
1024 | } | |
b84d49f9 JL |
1025 | /* |
1026 | * Replace the old lock with new_fl, and | |
1027 | * remove the old one. It's safe to do the | |
1028 | * insert here since we know that we won't be | |
1029 | * using new_fl later, and that the lock is | |
1030 | * just replacing an existing lock. | |
1da177e4 | 1031 | */ |
b84d49f9 JL |
1032 | error = -ENOLCK; |
1033 | if (!new_fl) | |
1034 | goto out; | |
1035 | locks_copy_lock(new_fl, request); | |
1036 | request = new_fl; | |
1037 | new_fl = NULL; | |
1038 | locks_delete_lock(before); | |
1039 | locks_insert_lock(before, request); | |
b9746ef8 | 1040 | added = true; |
1da177e4 LT |
1041 | } |
1042 | } | |
1043 | /* Go on to next lock. | |
1044 | */ | |
1045 | next_lock: | |
1046 | before = &fl->fl_next; | |
1047 | } | |
1048 | ||
0d9a490a | 1049 | /* |
1cb36012 JL |
1050 | * The above code only modifies existing locks in case of merging or |
1051 | * replacing. If new lock(s) need to be inserted all modifications are | |
1052 | * done below this, so it's safe yet to bail out. | |
0d9a490a MS |
1053 | */ |
1054 | error = -ENOLCK; /* "no luck" */ | |
1055 | if (right && left == right && !new_fl2) | |
1056 | goto out; | |
1057 | ||
1da177e4 LT |
1058 | error = 0; |
1059 | if (!added) { | |
f475ae95 TM |
1060 | if (request->fl_type == F_UNLCK) { |
1061 | if (request->fl_flags & FL_EXISTS) | |
1062 | error = -ENOENT; | |
1da177e4 | 1063 | goto out; |
f475ae95 | 1064 | } |
0d9a490a MS |
1065 | |
1066 | if (!new_fl) { | |
1067 | error = -ENOLCK; | |
1068 | goto out; | |
1069 | } | |
1da177e4 LT |
1070 | locks_copy_lock(new_fl, request); |
1071 | locks_insert_lock(before, new_fl); | |
1072 | new_fl = NULL; | |
1073 | } | |
1074 | if (right) { | |
1075 | if (left == right) { | |
1076 | /* The new lock breaks the old one in two pieces, | |
1077 | * so we have to use the second new lock. | |
1078 | */ | |
1079 | left = new_fl2; | |
1080 | new_fl2 = NULL; | |
1081 | locks_copy_lock(left, right); | |
1082 | locks_insert_lock(before, left); | |
1083 | } | |
1084 | right->fl_start = request->fl_end + 1; | |
1085 | locks_wake_up_blocks(right); | |
1086 | } | |
1087 | if (left) { | |
1088 | left->fl_end = request->fl_start - 1; | |
1089 | locks_wake_up_blocks(left); | |
1090 | } | |
1091 | out: | |
1c8c601a | 1092 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1093 | /* |
1094 | * Free any unused locks. | |
1095 | */ | |
1096 | if (new_fl) | |
1097 | locks_free_lock(new_fl); | |
1098 | if (new_fl2) | |
1099 | locks_free_lock(new_fl2); | |
1100 | return error; | |
1101 | } | |
1102 | ||
1103 | /** | |
1104 | * posix_lock_file - Apply a POSIX-style lock to a file | |
1105 | * @filp: The file to apply the lock to | |
1106 | * @fl: The lock to be applied | |
150b3934 | 1107 | * @conflock: Place to return a copy of the conflicting lock, if found. |
1da177e4 LT |
1108 | * |
1109 | * Add a POSIX style lock to a file. | |
1110 | * We merge adjacent & overlapping locks whenever possible. | |
1111 | * POSIX locks are sorted by owner task, then by starting address | |
f475ae95 TM |
1112 | * |
1113 | * Note that if called with an FL_EXISTS argument, the caller may determine | |
1114 | * whether or not a lock was successfully freed by testing the return | |
1115 | * value for -ENOENT. | |
1da177e4 | 1116 | */ |
150b3934 | 1117 | int posix_lock_file(struct file *filp, struct file_lock *fl, |
5842add2 AA |
1118 | struct file_lock *conflock) |
1119 | { | |
496ad9aa | 1120 | return __posix_lock_file(file_inode(filp), fl, conflock); |
1da177e4 | 1121 | } |
150b3934 | 1122 | EXPORT_SYMBOL(posix_lock_file); |
1da177e4 LT |
1123 | |
1124 | /** | |
1125 | * posix_lock_file_wait - Apply a POSIX-style lock to a file | |
1126 | * @filp: The file to apply the lock to | |
1127 | * @fl: The lock to be applied | |
1128 | * | |
1129 | * Add a POSIX style lock to a file. | |
1130 | * We merge adjacent & overlapping locks whenever possible. | |
1131 | * POSIX locks are sorted by owner task, then by starting address | |
1132 | */ | |
1133 | int posix_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1134 | { | |
1135 | int error; | |
1136 | might_sleep (); | |
1137 | for (;;) { | |
150b3934 | 1138 | error = posix_lock_file(filp, fl, NULL); |
bde74e4b | 1139 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1140 | break; |
1141 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1142 | if (!error) | |
1143 | continue; | |
1144 | ||
1145 | locks_delete_block(fl); | |
1146 | break; | |
1147 | } | |
1148 | return error; | |
1149 | } | |
1150 | EXPORT_SYMBOL(posix_lock_file_wait); | |
1151 | ||
1152 | /** | |
1153 | * locks_mandatory_locked - Check for an active lock | |
d7a06983 | 1154 | * @file: the file to check |
1da177e4 LT |
1155 | * |
1156 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1157 | * This function is called from locks_verify_locked() only. | |
1158 | */ | |
d7a06983 | 1159 | int locks_mandatory_locked(struct file *file) |
1da177e4 | 1160 | { |
d7a06983 | 1161 | struct inode *inode = file_inode(file); |
1da177e4 LT |
1162 | struct file_lock *fl; |
1163 | ||
1164 | /* | |
1165 | * Search the lock list for this inode for any POSIX locks. | |
1166 | */ | |
1c8c601a | 1167 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1168 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
1169 | if (!IS_POSIX(fl)) | |
1170 | continue; | |
73a8f5f7 CH |
1171 | if (fl->fl_owner != current->files && |
1172 | fl->fl_owner != file) | |
1da177e4 LT |
1173 | break; |
1174 | } | |
1c8c601a | 1175 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1176 | return fl ? -EAGAIN : 0; |
1177 | } | |
1178 | ||
1179 | /** | |
1180 | * locks_mandatory_area - Check for a conflicting lock | |
1181 | * @read_write: %FLOCK_VERIFY_WRITE for exclusive access, %FLOCK_VERIFY_READ | |
1182 | * for shared | |
1183 | * @inode: the file to check | |
1184 | * @filp: how the file was opened (if it was) | |
1185 | * @offset: start of area to check | |
1186 | * @count: length of area to check | |
1187 | * | |
1188 | * Searches the inode's list of locks to find any POSIX locks which conflict. | |
1189 | * This function is called from rw_verify_area() and | |
1190 | * locks_verify_truncate(). | |
1191 | */ | |
1192 | int locks_mandatory_area(int read_write, struct inode *inode, | |
1193 | struct file *filp, loff_t offset, | |
1194 | size_t count) | |
1195 | { | |
1196 | struct file_lock fl; | |
1197 | int error; | |
29723ade | 1198 | bool sleep = false; |
1da177e4 LT |
1199 | |
1200 | locks_init_lock(&fl); | |
1da177e4 LT |
1201 | fl.fl_pid = current->tgid; |
1202 | fl.fl_file = filp; | |
1203 | fl.fl_flags = FL_POSIX | FL_ACCESS; | |
1204 | if (filp && !(filp->f_flags & O_NONBLOCK)) | |
29723ade | 1205 | sleep = true; |
1da177e4 LT |
1206 | fl.fl_type = (read_write == FLOCK_VERIFY_WRITE) ? F_WRLCK : F_RDLCK; |
1207 | fl.fl_start = offset; | |
1208 | fl.fl_end = offset + count - 1; | |
1209 | ||
1210 | for (;;) { | |
29723ade | 1211 | if (filp) { |
73a8f5f7 | 1212 | fl.fl_owner = filp; |
29723ade JL |
1213 | fl.fl_flags &= ~FL_SLEEP; |
1214 | error = __posix_lock_file(inode, &fl, NULL); | |
1215 | if (!error) | |
1216 | break; | |
1217 | } | |
1218 | ||
1219 | if (sleep) | |
1220 | fl.fl_flags |= FL_SLEEP; | |
1221 | fl.fl_owner = current->files; | |
150b3934 | 1222 | error = __posix_lock_file(inode, &fl, NULL); |
bde74e4b | 1223 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1224 | break; |
1225 | error = wait_event_interruptible(fl.fl_wait, !fl.fl_next); | |
1226 | if (!error) { | |
1227 | /* | |
1228 | * If we've been sleeping someone might have | |
1229 | * changed the permissions behind our back. | |
1230 | */ | |
a16877ca | 1231 | if (__mandatory_lock(inode)) |
1da177e4 LT |
1232 | continue; |
1233 | } | |
1234 | ||
1235 | locks_delete_block(&fl); | |
1236 | break; | |
1237 | } | |
1238 | ||
1239 | return error; | |
1240 | } | |
1241 | ||
1242 | EXPORT_SYMBOL(locks_mandatory_area); | |
1243 | ||
778fc546 BF |
1244 | static void lease_clear_pending(struct file_lock *fl, int arg) |
1245 | { | |
1246 | switch (arg) { | |
1247 | case F_UNLCK: | |
1248 | fl->fl_flags &= ~FL_UNLOCK_PENDING; | |
1249 | /* fall through: */ | |
1250 | case F_RDLCK: | |
1251 | fl->fl_flags &= ~FL_DOWNGRADE_PENDING; | |
1252 | } | |
1253 | } | |
1254 | ||
1da177e4 LT |
1255 | /* We already had a lease on this file; just change its type */ |
1256 | int lease_modify(struct file_lock **before, int arg) | |
1257 | { | |
1258 | struct file_lock *fl = *before; | |
1259 | int error = assign_type(fl, arg); | |
1260 | ||
1261 | if (error) | |
1262 | return error; | |
778fc546 | 1263 | lease_clear_pending(fl, arg); |
1da177e4 | 1264 | locks_wake_up_blocks(fl); |
3b6e2723 FB |
1265 | if (arg == F_UNLCK) { |
1266 | struct file *filp = fl->fl_file; | |
1267 | ||
1268 | f_delown(filp); | |
1269 | filp->f_owner.signum = 0; | |
96d6d59c BF |
1270 | fasync_helper(0, fl->fl_file, 0, &fl->fl_fasync); |
1271 | if (fl->fl_fasync != NULL) { | |
1272 | printk(KERN_ERR "locks_delete_lock: fasync == %p\n", fl->fl_fasync); | |
1273 | fl->fl_fasync = NULL; | |
1274 | } | |
1da177e4 | 1275 | locks_delete_lock(before); |
3b6e2723 | 1276 | } |
1da177e4 LT |
1277 | return 0; |
1278 | } | |
1279 | ||
1280 | EXPORT_SYMBOL(lease_modify); | |
1281 | ||
778fc546 BF |
1282 | static bool past_time(unsigned long then) |
1283 | { | |
1284 | if (!then) | |
1285 | /* 0 is a special value meaning "this never expires": */ | |
1286 | return false; | |
1287 | return time_after(jiffies, then); | |
1288 | } | |
1289 | ||
1da177e4 LT |
1290 | static void time_out_leases(struct inode *inode) |
1291 | { | |
1292 | struct file_lock **before; | |
1293 | struct file_lock *fl; | |
1294 | ||
1295 | before = &inode->i_flock; | |
ab83fa4b | 1296 | while ((fl = *before) && IS_LEASE(fl) && lease_breaking(fl)) { |
62af4f1f | 1297 | trace_time_out_leases(inode, fl); |
778fc546 BF |
1298 | if (past_time(fl->fl_downgrade_time)) |
1299 | lease_modify(before, F_RDLCK); | |
1300 | if (past_time(fl->fl_break_time)) | |
1301 | lease_modify(before, F_UNLCK); | |
1da177e4 LT |
1302 | if (fl == *before) /* lease_modify may have freed fl */ |
1303 | before = &fl->fl_next; | |
1304 | } | |
1305 | } | |
1306 | ||
df4e8d2c BF |
1307 | static bool leases_conflict(struct file_lock *lease, struct file_lock *breaker) |
1308 | { | |
1309 | if ((breaker->fl_flags & FL_DELEG) && (lease->fl_flags & FL_LEASE)) | |
1310 | return false; | |
1311 | return locks_conflict(breaker, lease); | |
1312 | } | |
1313 | ||
1da177e4 LT |
1314 | /** |
1315 | * __break_lease - revoke all outstanding leases on file | |
1316 | * @inode: the inode of the file to return | |
df4e8d2c BF |
1317 | * @mode: O_RDONLY: break only write leases; O_WRONLY or O_RDWR: |
1318 | * break all leases | |
1319 | * @type: FL_LEASE: break leases and delegations; FL_DELEG: break | |
1320 | * only delegations | |
1da177e4 | 1321 | * |
87250dd2 | 1322 | * break_lease (inlined for speed) has checked there already is at least |
1323 | * some kind of lock (maybe a lease) on this file. Leases are broken on | |
1324 | * a call to open() or truncate(). This function can sleep unless you | |
1da177e4 LT |
1325 | * specified %O_NONBLOCK to your open(). |
1326 | */ | |
df4e8d2c | 1327 | int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) |
1da177e4 | 1328 | { |
778fc546 | 1329 | int error = 0; |
1da177e4 LT |
1330 | struct file_lock *new_fl, *flock; |
1331 | struct file_lock *fl; | |
1da177e4 LT |
1332 | unsigned long break_time; |
1333 | int i_have_this_lease = 0; | |
df4e8d2c | 1334 | bool lease_conflict = false; |
8737c930 | 1335 | int want_write = (mode & O_ACCMODE) != O_RDONLY; |
1da177e4 | 1336 | |
8737c930 | 1337 | new_fl = lease_alloc(NULL, want_write ? F_WRLCK : F_RDLCK); |
6d4b9e38 LT |
1338 | if (IS_ERR(new_fl)) |
1339 | return PTR_ERR(new_fl); | |
df4e8d2c | 1340 | new_fl->fl_flags = type; |
1da177e4 | 1341 | |
1c8c601a | 1342 | spin_lock(&inode->i_lock); |
1da177e4 LT |
1343 | |
1344 | time_out_leases(inode); | |
1345 | ||
1346 | flock = inode->i_flock; | |
1347 | if ((flock == NULL) || !IS_LEASE(flock)) | |
1348 | goto out; | |
1349 | ||
df4e8d2c BF |
1350 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { |
1351 | if (leases_conflict(fl, new_fl)) { | |
1352 | lease_conflict = true; | |
1353 | if (fl->fl_owner == current->files) | |
1354 | i_have_this_lease = 1; | |
1355 | } | |
1356 | } | |
1357 | if (!lease_conflict) | |
778fc546 BF |
1358 | goto out; |
1359 | ||
1da177e4 LT |
1360 | break_time = 0; |
1361 | if (lease_break_time > 0) { | |
1362 | break_time = jiffies + lease_break_time * HZ; | |
1363 | if (break_time == 0) | |
1364 | break_time++; /* so that 0 means no break time */ | |
1365 | } | |
1366 | ||
1367 | for (fl = flock; fl && IS_LEASE(fl); fl = fl->fl_next) { | |
df4e8d2c BF |
1368 | if (!leases_conflict(fl, new_fl)) |
1369 | continue; | |
778fc546 BF |
1370 | if (want_write) { |
1371 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1372 | continue; | |
1373 | fl->fl_flags |= FL_UNLOCK_PENDING; | |
1da177e4 | 1374 | fl->fl_break_time = break_time; |
778fc546 BF |
1375 | } else { |
1376 | if (lease_breaking(flock)) | |
1377 | continue; | |
1378 | fl->fl_flags |= FL_DOWNGRADE_PENDING; | |
1379 | fl->fl_downgrade_time = break_time; | |
1da177e4 | 1380 | } |
778fc546 | 1381 | fl->fl_lmops->lm_break(fl); |
1da177e4 LT |
1382 | } |
1383 | ||
1384 | if (i_have_this_lease || (mode & O_NONBLOCK)) { | |
62af4f1f | 1385 | trace_break_lease_noblock(inode, new_fl); |
1da177e4 LT |
1386 | error = -EWOULDBLOCK; |
1387 | goto out; | |
1388 | } | |
1389 | ||
1390 | restart: | |
1391 | break_time = flock->fl_break_time; | |
f1c6bb2c | 1392 | if (break_time != 0) |
1da177e4 | 1393 | break_time -= jiffies; |
f1c6bb2c JL |
1394 | if (break_time == 0) |
1395 | break_time++; | |
4321e01e | 1396 | locks_insert_block(flock, new_fl); |
62af4f1f | 1397 | trace_break_lease_block(inode, new_fl); |
1c8c601a | 1398 | spin_unlock(&inode->i_lock); |
4321e01e MW |
1399 | error = wait_event_interruptible_timeout(new_fl->fl_wait, |
1400 | !new_fl->fl_next, break_time); | |
1c8c601a | 1401 | spin_lock(&inode->i_lock); |
62af4f1f | 1402 | trace_break_lease_unblock(inode, new_fl); |
1c8c601a | 1403 | locks_delete_block(new_fl); |
1da177e4 LT |
1404 | if (error >= 0) { |
1405 | if (error == 0) | |
1406 | time_out_leases(inode); | |
778fc546 BF |
1407 | /* |
1408 | * Wait for the next conflicting lease that has not been | |
1409 | * broken yet | |
1410 | */ | |
1da177e4 LT |
1411 | for (flock = inode->i_flock; flock && IS_LEASE(flock); |
1412 | flock = flock->fl_next) { | |
df4e8d2c | 1413 | if (leases_conflict(new_fl, flock)) |
1da177e4 LT |
1414 | goto restart; |
1415 | } | |
1416 | error = 0; | |
1417 | } | |
1418 | ||
1419 | out: | |
1c8c601a | 1420 | spin_unlock(&inode->i_lock); |
6d4b9e38 | 1421 | locks_free_lock(new_fl); |
1da177e4 LT |
1422 | return error; |
1423 | } | |
1424 | ||
1425 | EXPORT_SYMBOL(__break_lease); | |
1426 | ||
1427 | /** | |
a6b91919 | 1428 | * lease_get_mtime - get the last modified time of an inode |
1da177e4 LT |
1429 | * @inode: the inode |
1430 | * @time: pointer to a timespec which will contain the last modified time | |
1431 | * | |
1432 | * This is to force NFS clients to flush their caches for files with | |
1433 | * exclusive leases. The justification is that if someone has an | |
a6b91919 | 1434 | * exclusive lease, then they could be modifying it. |
1da177e4 LT |
1435 | */ |
1436 | void lease_get_mtime(struct inode *inode, struct timespec *time) | |
1437 | { | |
1438 | struct file_lock *flock = inode->i_flock; | |
0ee5c6d6 | 1439 | if (flock && IS_LEASE(flock) && (flock->fl_type == F_WRLCK)) |
1da177e4 LT |
1440 | *time = current_fs_time(inode->i_sb); |
1441 | else | |
1442 | *time = inode->i_mtime; | |
1443 | } | |
1444 | ||
1445 | EXPORT_SYMBOL(lease_get_mtime); | |
1446 | ||
1447 | /** | |
1448 | * fcntl_getlease - Enquire what lease is currently active | |
1449 | * @filp: the file | |
1450 | * | |
1451 | * The value returned by this function will be one of | |
1452 | * (if no lease break is pending): | |
1453 | * | |
1454 | * %F_RDLCK to indicate a shared lease is held. | |
1455 | * | |
1456 | * %F_WRLCK to indicate an exclusive lease is held. | |
1457 | * | |
1458 | * %F_UNLCK to indicate no lease is held. | |
1459 | * | |
1460 | * (if a lease break is pending): | |
1461 | * | |
1462 | * %F_RDLCK to indicate an exclusive lease needs to be | |
1463 | * changed to a shared lease (or removed). | |
1464 | * | |
1465 | * %F_UNLCK to indicate the lease needs to be removed. | |
1466 | * | |
1467 | * XXX: sfr & willy disagree over whether F_INPROGRESS | |
1468 | * should be returned to userspace. | |
1469 | */ | |
1470 | int fcntl_getlease(struct file *filp) | |
1471 | { | |
1472 | struct file_lock *fl; | |
1c8c601a | 1473 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1474 | int type = F_UNLCK; |
1475 | ||
1c8c601a | 1476 | spin_lock(&inode->i_lock); |
496ad9aa AV |
1477 | time_out_leases(file_inode(filp)); |
1478 | for (fl = file_inode(filp)->i_flock; fl && IS_LEASE(fl); | |
1da177e4 LT |
1479 | fl = fl->fl_next) { |
1480 | if (fl->fl_file == filp) { | |
778fc546 | 1481 | type = target_leasetype(fl); |
1da177e4 LT |
1482 | break; |
1483 | } | |
1484 | } | |
1c8c601a | 1485 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1486 | return type; |
1487 | } | |
1488 | ||
24cbe784 JL |
1489 | /** |
1490 | * check_conflicting_open - see if the given dentry points to a file that has | |
1491 | * an existing open that would conflict with the | |
1492 | * desired lease. | |
1493 | * @dentry: dentry to check | |
1494 | * @arg: type of lease that we're trying to acquire | |
1495 | * | |
1496 | * Check to see if there's an existing open fd on this file that would | |
1497 | * conflict with the lease we're trying to set. | |
1498 | */ | |
1499 | static int | |
1500 | check_conflicting_open(const struct dentry *dentry, const long arg) | |
1501 | { | |
1502 | int ret = 0; | |
1503 | struct inode *inode = dentry->d_inode; | |
1504 | ||
1505 | if ((arg == F_RDLCK) && (atomic_read(&inode->i_writecount) > 0)) | |
1506 | return -EAGAIN; | |
1507 | ||
1508 | if ((arg == F_WRLCK) && ((d_count(dentry) > 1) || | |
1509 | (atomic_read(&inode->i_count) > 1))) | |
1510 | ret = -EAGAIN; | |
1511 | ||
1512 | return ret; | |
1513 | } | |
1514 | ||
d4f22d19 | 1515 | static int generic_add_lease(struct file *filp, long arg, struct file_lock **flp) |
1da177e4 | 1516 | { |
7eaae282 | 1517 | struct file_lock *fl, **before, **my_before = NULL, *lease; |
0f7fc9e4 | 1518 | struct dentry *dentry = filp->f_path.dentry; |
1da177e4 | 1519 | struct inode *inode = dentry->d_inode; |
df4e8d2c | 1520 | bool is_deleg = (*flp)->fl_flags & FL_DELEG; |
c1f24ef4 | 1521 | int error; |
1da177e4 | 1522 | |
096657b6 | 1523 | lease = *flp; |
62af4f1f JL |
1524 | trace_generic_add_lease(inode, lease); |
1525 | ||
df4e8d2c BF |
1526 | /* |
1527 | * In the delegation case we need mutual exclusion with | |
1528 | * a number of operations that take the i_mutex. We trylock | |
1529 | * because delegations are an optional optimization, and if | |
1530 | * there's some chance of a conflict--we'd rather not | |
1531 | * bother, maybe that's a sign this just isn't a good file to | |
1532 | * hand out a delegation on. | |
1533 | */ | |
1534 | if (is_deleg && !mutex_trylock(&inode->i_mutex)) | |
1535 | return -EAGAIN; | |
1536 | ||
1537 | if (is_deleg && arg == F_WRLCK) { | |
1538 | /* Write delegations are not currently supported: */ | |
4fdb793f | 1539 | mutex_unlock(&inode->i_mutex); |
df4e8d2c BF |
1540 | WARN_ON_ONCE(1); |
1541 | return -EINVAL; | |
1542 | } | |
096657b6 | 1543 | |
24cbe784 JL |
1544 | error = check_conflicting_open(dentry, arg); |
1545 | if (error) | |
096657b6 | 1546 | goto out; |
6d5e8b05 | 1547 | |
1da177e4 LT |
1548 | /* |
1549 | * At this point, we know that if there is an exclusive | |
1550 | * lease on this file, then we hold it on this filp | |
1551 | * (otherwise our open of this file would have blocked). | |
1552 | * And if we are trying to acquire an exclusive lease, | |
1553 | * then the file is not open by anyone (including us) | |
1554 | * except for this filp. | |
1555 | */ | |
c1f24ef4 | 1556 | error = -EAGAIN; |
1da177e4 LT |
1557 | for (before = &inode->i_flock; |
1558 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1559 | before = &fl->fl_next) { | |
c1f24ef4 | 1560 | if (fl->fl_file == filp) { |
1da177e4 | 1561 | my_before = before; |
c1f24ef4 BF |
1562 | continue; |
1563 | } | |
1564 | /* | |
1565 | * No exclusive leases if someone else has a lease on | |
1566 | * this file: | |
1567 | */ | |
1568 | if (arg == F_WRLCK) | |
1569 | goto out; | |
1570 | /* | |
1571 | * Modifying our existing lease is OK, but no getting a | |
1572 | * new lease if someone else is opening for write: | |
1573 | */ | |
1574 | if (fl->fl_flags & FL_UNLOCK_PENDING) | |
1575 | goto out; | |
1da177e4 LT |
1576 | } |
1577 | ||
1da177e4 | 1578 | if (my_before != NULL) { |
8fb47a4f | 1579 | error = lease->fl_lmops->lm_change(my_before, arg); |
51ee4b84 CH |
1580 | if (!error) |
1581 | *flp = *my_before; | |
1da177e4 LT |
1582 | goto out; |
1583 | } | |
1584 | ||
1da177e4 LT |
1585 | error = -EINVAL; |
1586 | if (!leases_enable) | |
1587 | goto out; | |
1588 | ||
c5b1f0d9 | 1589 | locks_insert_lock(before, lease); |
24cbe784 JL |
1590 | /* |
1591 | * The check in break_lease() is lockless. It's possible for another | |
1592 | * open to race in after we did the earlier check for a conflicting | |
1593 | * open but before the lease was inserted. Check again for a | |
1594 | * conflicting open and cancel the lease if there is one. | |
1595 | * | |
1596 | * We also add a barrier here to ensure that the insertion of the lock | |
1597 | * precedes these checks. | |
1598 | */ | |
1599 | smp_mb(); | |
1600 | error = check_conflicting_open(dentry, arg); | |
1601 | if (error) | |
1602 | locks_unlink_lock(flp); | |
1da177e4 | 1603 | out: |
df4e8d2c BF |
1604 | if (is_deleg) |
1605 | mutex_unlock(&inode->i_mutex); | |
1da177e4 LT |
1606 | return error; |
1607 | } | |
8335ebd9 | 1608 | |
d4f22d19 | 1609 | static int generic_delete_lease(struct file *filp, struct file_lock **flp) |
8335ebd9 BF |
1610 | { |
1611 | struct file_lock *fl, **before; | |
1612 | struct dentry *dentry = filp->f_path.dentry; | |
1613 | struct inode *inode = dentry->d_inode; | |
1614 | ||
62af4f1f JL |
1615 | trace_generic_delete_lease(inode, *flp); |
1616 | ||
8335ebd9 BF |
1617 | for (before = &inode->i_flock; |
1618 | ((fl = *before) != NULL) && IS_LEASE(fl); | |
1619 | before = &fl->fl_next) { | |
1620 | if (fl->fl_file != filp) | |
1621 | continue; | |
1622 | return (*flp)->fl_lmops->lm_change(before, F_UNLCK); | |
1623 | } | |
1624 | return -EAGAIN; | |
1625 | } | |
1626 | ||
1627 | /** | |
1628 | * generic_setlease - sets a lease on an open file | |
1629 | * @filp: file pointer | |
1630 | * @arg: type of lease to obtain | |
1631 | * @flp: input - file_lock to use, output - file_lock inserted | |
1632 | * | |
1633 | * The (input) flp->fl_lmops->lm_break function is required | |
1634 | * by break_lease(). | |
1635 | * | |
1c8c601a | 1636 | * Called with inode->i_lock held. |
8335ebd9 BF |
1637 | */ |
1638 | int generic_setlease(struct file *filp, long arg, struct file_lock **flp) | |
1639 | { | |
1640 | struct dentry *dentry = filp->f_path.dentry; | |
1641 | struct inode *inode = dentry->d_inode; | |
1642 | int error; | |
1643 | ||
8e96e3b7 | 1644 | if ((!uid_eq(current_fsuid(), inode->i_uid)) && !capable(CAP_LEASE)) |
8335ebd9 BF |
1645 | return -EACCES; |
1646 | if (!S_ISREG(inode->i_mode)) | |
1647 | return -EINVAL; | |
1648 | error = security_file_lock(filp, arg); | |
1649 | if (error) | |
1650 | return error; | |
1651 | ||
1652 | time_out_leases(inode); | |
1653 | ||
1654 | BUG_ON(!(*flp)->fl_lmops->lm_break); | |
1655 | ||
1656 | switch (arg) { | |
1657 | case F_UNLCK: | |
1658 | return generic_delete_lease(filp, flp); | |
1659 | case F_RDLCK: | |
1660 | case F_WRLCK: | |
1661 | return generic_add_lease(filp, arg, flp); | |
1662 | default: | |
8d657eb3 | 1663 | return -EINVAL; |
8335ebd9 BF |
1664 | } |
1665 | } | |
0af1a450 | 1666 | EXPORT_SYMBOL(generic_setlease); |
1da177e4 | 1667 | |
b89f4321 AB |
1668 | static int __vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1669 | { | |
72c2d531 | 1670 | if (filp->f_op->setlease) |
b89f4321 AB |
1671 | return filp->f_op->setlease(filp, arg, lease); |
1672 | else | |
1673 | return generic_setlease(filp, arg, lease); | |
1674 | } | |
1675 | ||
1676 | /** | |
a9933cea | 1677 | * vfs_setlease - sets a lease on an open file |
1da177e4 LT |
1678 | * @filp: file pointer |
1679 | * @arg: type of lease to obtain | |
1680 | * @lease: file_lock to use | |
1681 | * | |
1682 | * Call this to establish a lease on the file. | |
8fb47a4f | 1683 | * The (*lease)->fl_lmops->lm_break operation must be set; if not, |
f9ffed26 BF |
1684 | * break_lease will oops! |
1685 | * | |
1686 | * This will call the filesystem's setlease file method, if | |
1687 | * defined. Note that there is no getlease method; instead, the | |
1688 | * filesystem setlease method should call back to setlease() to | |
1689 | * add a lease to the inode's lease list, where fcntl_getlease() can | |
1690 | * find it. Since fcntl_getlease() only reports whether the current | |
1691 | * task holds a lease, a cluster filesystem need only do this for | |
1692 | * leases held by processes on this node. | |
1693 | * | |
1694 | * There is also no break_lease method; filesystems that | |
c9404c9c | 1695 | * handle their own leases should break leases themselves from the |
f9ffed26 BF |
1696 | * filesystem's open, create, and (on truncate) setattr methods. |
1697 | * | |
1698 | * Warning: the only current setlease methods exist only to disable | |
1699 | * leases in certain cases. More vfs changes may be required to | |
1700 | * allow a full filesystem lease implementation. | |
1da177e4 LT |
1701 | */ |
1702 | ||
a9933cea | 1703 | int vfs_setlease(struct file *filp, long arg, struct file_lock **lease) |
1da177e4 | 1704 | { |
1c8c601a | 1705 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1706 | int error; |
1707 | ||
1c8c601a | 1708 | spin_lock(&inode->i_lock); |
b89f4321 | 1709 | error = __vfs_setlease(filp, arg, lease); |
1c8c601a | 1710 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
1711 | |
1712 | return error; | |
1713 | } | |
a9933cea | 1714 | EXPORT_SYMBOL_GPL(vfs_setlease); |
1da177e4 | 1715 | |
0ceaf6c7 BF |
1716 | static int do_fcntl_delete_lease(struct file *filp) |
1717 | { | |
1718 | struct file_lock fl, *flp = &fl; | |
1719 | ||
1720 | lease_init(filp, F_UNLCK, flp); | |
1721 | ||
1722 | return vfs_setlease(filp, F_UNLCK, &flp); | |
1723 | } | |
1724 | ||
1725 | static int do_fcntl_add_lease(unsigned int fd, struct file *filp, long arg) | |
1da177e4 | 1726 | { |
3df057ac | 1727 | struct file_lock *fl, *ret; |
1c8c601a | 1728 | struct inode *inode = file_inode(filp); |
f7347ce4 | 1729 | struct fasync_struct *new; |
1da177e4 LT |
1730 | int error; |
1731 | ||
c5b1f0d9 AB |
1732 | fl = lease_alloc(filp, arg); |
1733 | if (IS_ERR(fl)) | |
1734 | return PTR_ERR(fl); | |
1da177e4 | 1735 | |
f7347ce4 LT |
1736 | new = fasync_alloc(); |
1737 | if (!new) { | |
1738 | locks_free_lock(fl); | |
1739 | return -ENOMEM; | |
1740 | } | |
3df057ac | 1741 | ret = fl; |
1c8c601a | 1742 | spin_lock(&inode->i_lock); |
8896b93f | 1743 | error = __vfs_setlease(filp, arg, &ret); |
51ee4b84 | 1744 | if (error) { |
1c8c601a | 1745 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1746 | locks_free_lock(fl); |
1747 | goto out_free_fasync; | |
1748 | } | |
3df057ac BF |
1749 | if (ret != fl) |
1750 | locks_free_lock(fl); | |
1da177e4 | 1751 | |
f7347ce4 LT |
1752 | /* |
1753 | * fasync_insert_entry() returns the old entry if any. | |
1754 | * If there was no old entry, then it used 'new' and | |
1755 | * inserted it into the fasync list. Clear new so that | |
1756 | * we don't release it here. | |
1757 | */ | |
3df057ac | 1758 | if (!fasync_insert_entry(fd, filp, &ret->fl_fasync, new)) |
f7347ce4 LT |
1759 | new = NULL; |
1760 | ||
8896b93f | 1761 | error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0); |
1c8c601a | 1762 | spin_unlock(&inode->i_lock); |
51ee4b84 CH |
1763 | |
1764 | out_free_fasync: | |
f7347ce4 LT |
1765 | if (new) |
1766 | fasync_free(new); | |
1da177e4 LT |
1767 | return error; |
1768 | } | |
1769 | ||
0ceaf6c7 BF |
1770 | /** |
1771 | * fcntl_setlease - sets a lease on an open file | |
1772 | * @fd: open file descriptor | |
1773 | * @filp: file pointer | |
1774 | * @arg: type of lease to obtain | |
1775 | * | |
1776 | * Call this fcntl to establish a lease on the file. | |
1777 | * Note that you also need to call %F_SETSIG to | |
1778 | * receive a signal when the lease is broken. | |
1779 | */ | |
1780 | int fcntl_setlease(unsigned int fd, struct file *filp, long arg) | |
1781 | { | |
1782 | if (arg == F_UNLCK) | |
1783 | return do_fcntl_delete_lease(filp); | |
1784 | return do_fcntl_add_lease(fd, filp, arg); | |
1785 | } | |
1786 | ||
1da177e4 LT |
1787 | /** |
1788 | * flock_lock_file_wait - Apply a FLOCK-style lock to a file | |
1789 | * @filp: The file to apply the lock to | |
1790 | * @fl: The lock to be applied | |
1791 | * | |
1792 | * Add a FLOCK style lock to a file. | |
1793 | */ | |
1794 | int flock_lock_file_wait(struct file *filp, struct file_lock *fl) | |
1795 | { | |
1796 | int error; | |
1797 | might_sleep(); | |
1798 | for (;;) { | |
1799 | error = flock_lock_file(filp, fl); | |
bde74e4b | 1800 | if (error != FILE_LOCK_DEFERRED) |
1da177e4 LT |
1801 | break; |
1802 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); | |
1803 | if (!error) | |
1804 | continue; | |
1805 | ||
1806 | locks_delete_block(fl); | |
1807 | break; | |
1808 | } | |
1809 | return error; | |
1810 | } | |
1811 | ||
1812 | EXPORT_SYMBOL(flock_lock_file_wait); | |
1813 | ||
1814 | /** | |
1815 | * sys_flock: - flock() system call. | |
1816 | * @fd: the file descriptor to lock. | |
1817 | * @cmd: the type of lock to apply. | |
1818 | * | |
1819 | * Apply a %FL_FLOCK style lock to an open file descriptor. | |
1820 | * The @cmd can be one of | |
1821 | * | |
1822 | * %LOCK_SH -- a shared lock. | |
1823 | * | |
1824 | * %LOCK_EX -- an exclusive lock. | |
1825 | * | |
1826 | * %LOCK_UN -- remove an existing lock. | |
1827 | * | |
1828 | * %LOCK_MAND -- a `mandatory' flock. This exists to emulate Windows Share Modes. | |
1829 | * | |
1830 | * %LOCK_MAND can be combined with %LOCK_READ or %LOCK_WRITE to allow other | |
1831 | * processes read and write access respectively. | |
1832 | */ | |
002c8976 | 1833 | SYSCALL_DEFINE2(flock, unsigned int, fd, unsigned int, cmd) |
1da177e4 | 1834 | { |
2903ff01 | 1835 | struct fd f = fdget(fd); |
1da177e4 LT |
1836 | struct file_lock *lock; |
1837 | int can_sleep, unlock; | |
1838 | int error; | |
1839 | ||
1840 | error = -EBADF; | |
2903ff01 | 1841 | if (!f.file) |
1da177e4 LT |
1842 | goto out; |
1843 | ||
1844 | can_sleep = !(cmd & LOCK_NB); | |
1845 | cmd &= ~LOCK_NB; | |
1846 | unlock = (cmd == LOCK_UN); | |
1847 | ||
aeb5d727 | 1848 | if (!unlock && !(cmd & LOCK_MAND) && |
2903ff01 | 1849 | !(f.file->f_mode & (FMODE_READ|FMODE_WRITE))) |
1da177e4 LT |
1850 | goto out_putf; |
1851 | ||
2903ff01 | 1852 | error = flock_make_lock(f.file, &lock, cmd); |
1da177e4 LT |
1853 | if (error) |
1854 | goto out_putf; | |
1855 | if (can_sleep) | |
1856 | lock->fl_flags |= FL_SLEEP; | |
1857 | ||
2903ff01 | 1858 | error = security_file_lock(f.file, lock->fl_type); |
1da177e4 LT |
1859 | if (error) |
1860 | goto out_free; | |
1861 | ||
72c2d531 | 1862 | if (f.file->f_op->flock) |
2903ff01 | 1863 | error = f.file->f_op->flock(f.file, |
1da177e4 LT |
1864 | (can_sleep) ? F_SETLKW : F_SETLK, |
1865 | lock); | |
1866 | else | |
2903ff01 | 1867 | error = flock_lock_file_wait(f.file, lock); |
1da177e4 LT |
1868 | |
1869 | out_free: | |
993dfa87 | 1870 | locks_free_lock(lock); |
1da177e4 LT |
1871 | |
1872 | out_putf: | |
2903ff01 | 1873 | fdput(f); |
1da177e4 LT |
1874 | out: |
1875 | return error; | |
1876 | } | |
1877 | ||
3ee17abd BF |
1878 | /** |
1879 | * vfs_test_lock - test file byte range lock | |
1880 | * @filp: The file to test lock for | |
6924c554 | 1881 | * @fl: The lock to test; also used to hold result |
3ee17abd BF |
1882 | * |
1883 | * Returns -ERRNO on failure. Indicates presence of conflicting lock by | |
1884 | * setting conf->fl_type to something other than F_UNLCK. | |
1885 | */ | |
1886 | int vfs_test_lock(struct file *filp, struct file_lock *fl) | |
1887 | { | |
72c2d531 | 1888 | if (filp->f_op->lock) |
3ee17abd BF |
1889 | return filp->f_op->lock(filp, F_GETLK, fl); |
1890 | posix_test_lock(filp, fl); | |
1891 | return 0; | |
1892 | } | |
1893 | EXPORT_SYMBOL_GPL(vfs_test_lock); | |
1894 | ||
c2fa1b8a BF |
1895 | static int posix_lock_to_flock(struct flock *flock, struct file_lock *fl) |
1896 | { | |
cff2fce5 | 1897 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1898 | #if BITS_PER_LONG == 32 |
1899 | /* | |
1900 | * Make sure we can represent the posix lock via | |
1901 | * legacy 32bit flock. | |
1902 | */ | |
1903 | if (fl->fl_start > OFFT_OFFSET_MAX) | |
1904 | return -EOVERFLOW; | |
1905 | if (fl->fl_end != OFFSET_MAX && fl->fl_end > OFFT_OFFSET_MAX) | |
1906 | return -EOVERFLOW; | |
1907 | #endif | |
1908 | flock->l_start = fl->fl_start; | |
1909 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1910 | fl->fl_end - fl->fl_start + 1; | |
1911 | flock->l_whence = 0; | |
129a84de | 1912 | flock->l_type = fl->fl_type; |
c2fa1b8a BF |
1913 | return 0; |
1914 | } | |
1915 | ||
1916 | #if BITS_PER_LONG == 32 | |
1917 | static void posix_lock_to_flock64(struct flock64 *flock, struct file_lock *fl) | |
1918 | { | |
cff2fce5 | 1919 | flock->l_pid = IS_OFDLCK(fl) ? -1 : fl->fl_pid; |
c2fa1b8a BF |
1920 | flock->l_start = fl->fl_start; |
1921 | flock->l_len = fl->fl_end == OFFSET_MAX ? 0 : | |
1922 | fl->fl_end - fl->fl_start + 1; | |
1923 | flock->l_whence = 0; | |
1924 | flock->l_type = fl->fl_type; | |
1925 | } | |
1926 | #endif | |
1927 | ||
1da177e4 LT |
1928 | /* Report the first existing lock that would conflict with l. |
1929 | * This implements the F_GETLK command of fcntl(). | |
1930 | */ | |
c1e62b8f | 1931 | int fcntl_getlk(struct file *filp, unsigned int cmd, struct flock __user *l) |
1da177e4 | 1932 | { |
9d6a8c5c | 1933 | struct file_lock file_lock; |
1da177e4 LT |
1934 | struct flock flock; |
1935 | int error; | |
1936 | ||
1937 | error = -EFAULT; | |
1938 | if (copy_from_user(&flock, l, sizeof(flock))) | |
1939 | goto out; | |
1940 | error = -EINVAL; | |
1941 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
1942 | goto out; | |
1943 | ||
1944 | error = flock_to_posix_lock(filp, &file_lock, &flock); | |
1945 | if (error) | |
1946 | goto out; | |
1947 | ||
0d3f7a2d | 1948 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
1949 | error = -EINVAL; |
1950 | if (flock.l_pid != 0) | |
1951 | goto out; | |
1952 | ||
5d50ffd7 | 1953 | cmd = F_GETLK; |
cff2fce5 | 1954 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 1955 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
1956 | } |
1957 | ||
3ee17abd BF |
1958 | error = vfs_test_lock(filp, &file_lock); |
1959 | if (error) | |
1960 | goto out; | |
1da177e4 | 1961 | |
9d6a8c5c ME |
1962 | flock.l_type = file_lock.fl_type; |
1963 | if (file_lock.fl_type != F_UNLCK) { | |
1964 | error = posix_lock_to_flock(&flock, &file_lock); | |
c2fa1b8a | 1965 | if (error) |
1da177e4 | 1966 | goto out; |
1da177e4 LT |
1967 | } |
1968 | error = -EFAULT; | |
1969 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
1970 | error = 0; | |
1971 | out: | |
1972 | return error; | |
1973 | } | |
1974 | ||
7723ec97 ME |
1975 | /** |
1976 | * vfs_lock_file - file byte range lock | |
1977 | * @filp: The file to apply the lock to | |
1978 | * @cmd: type of locking operation (F_SETLK, F_GETLK, etc.) | |
1979 | * @fl: The lock to be applied | |
150b3934 ME |
1980 | * @conf: Place to return a copy of the conflicting lock, if found. |
1981 | * | |
1982 | * A caller that doesn't care about the conflicting lock may pass NULL | |
1983 | * as the final argument. | |
1984 | * | |
1985 | * If the filesystem defines a private ->lock() method, then @conf will | |
1986 | * be left unchanged; so a caller that cares should initialize it to | |
1987 | * some acceptable default. | |
2beb6614 ME |
1988 | * |
1989 | * To avoid blocking kernel daemons, such as lockd, that need to acquire POSIX | |
1990 | * locks, the ->lock() interface may return asynchronously, before the lock has | |
1991 | * been granted or denied by the underlying filesystem, if (and only if) | |
8fb47a4f | 1992 | * lm_grant is set. Callers expecting ->lock() to return asynchronously |
2beb6614 ME |
1993 | * will only use F_SETLK, not F_SETLKW; they will set FL_SLEEP if (and only if) |
1994 | * the request is for a blocking lock. When ->lock() does return asynchronously, | |
8fb47a4f | 1995 | * it must return FILE_LOCK_DEFERRED, and call ->lm_grant() when the lock |
2beb6614 ME |
1996 | * request completes. |
1997 | * If the request is for non-blocking lock the file system should return | |
bde74e4b MS |
1998 | * FILE_LOCK_DEFERRED then try to get the lock and call the callback routine |
1999 | * with the result. If the request timed out the callback routine will return a | |
2beb6614 ME |
2000 | * nonzero return code and the file system should release the lock. The file |
2001 | * system is also responsible to keep a corresponding posix lock when it | |
2002 | * grants a lock so the VFS can find out which locks are locally held and do | |
2003 | * the correct lock cleanup when required. | |
2004 | * The underlying filesystem must not drop the kernel lock or call | |
8fb47a4f | 2005 | * ->lm_grant() before returning to the caller with a FILE_LOCK_DEFERRED |
2beb6614 | 2006 | * return code. |
7723ec97 | 2007 | */ |
150b3934 | 2008 | int vfs_lock_file(struct file *filp, unsigned int cmd, struct file_lock *fl, struct file_lock *conf) |
7723ec97 | 2009 | { |
72c2d531 | 2010 | if (filp->f_op->lock) |
7723ec97 ME |
2011 | return filp->f_op->lock(filp, cmd, fl); |
2012 | else | |
150b3934 | 2013 | return posix_lock_file(filp, fl, conf); |
7723ec97 ME |
2014 | } |
2015 | EXPORT_SYMBOL_GPL(vfs_lock_file); | |
2016 | ||
b648a6de MS |
2017 | static int do_lock_file_wait(struct file *filp, unsigned int cmd, |
2018 | struct file_lock *fl) | |
2019 | { | |
2020 | int error; | |
2021 | ||
2022 | error = security_file_lock(filp, fl->fl_type); | |
2023 | if (error) | |
2024 | return error; | |
2025 | ||
764c76b3 MS |
2026 | for (;;) { |
2027 | error = vfs_lock_file(filp, cmd, fl, NULL); | |
2028 | if (error != FILE_LOCK_DEFERRED) | |
b648a6de | 2029 | break; |
764c76b3 MS |
2030 | error = wait_event_interruptible(fl->fl_wait, !fl->fl_next); |
2031 | if (!error) | |
2032 | continue; | |
2033 | ||
2034 | locks_delete_block(fl); | |
2035 | break; | |
b648a6de MS |
2036 | } |
2037 | ||
2038 | return error; | |
2039 | } | |
2040 | ||
cf01f4ee JL |
2041 | /* Ensure that fl->fl_filp has compatible f_mode for F_SETLK calls */ |
2042 | static int | |
2043 | check_fmode_for_setlk(struct file_lock *fl) | |
2044 | { | |
2045 | switch (fl->fl_type) { | |
2046 | case F_RDLCK: | |
2047 | if (!(fl->fl_file->f_mode & FMODE_READ)) | |
2048 | return -EBADF; | |
2049 | break; | |
2050 | case F_WRLCK: | |
2051 | if (!(fl->fl_file->f_mode & FMODE_WRITE)) | |
2052 | return -EBADF; | |
2053 | } | |
2054 | return 0; | |
2055 | } | |
2056 | ||
1da177e4 LT |
2057 | /* Apply the lock described by l to an open file descriptor. |
2058 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2059 | */ | |
c293621b PS |
2060 | int fcntl_setlk(unsigned int fd, struct file *filp, unsigned int cmd, |
2061 | struct flock __user *l) | |
1da177e4 LT |
2062 | { |
2063 | struct file_lock *file_lock = locks_alloc_lock(); | |
2064 | struct flock flock; | |
2065 | struct inode *inode; | |
0b2bac2f | 2066 | struct file *f; |
1da177e4 LT |
2067 | int error; |
2068 | ||
2069 | if (file_lock == NULL) | |
2070 | return -ENOLCK; | |
2071 | ||
2072 | /* | |
2073 | * This might block, so we do it before checking the inode. | |
2074 | */ | |
2075 | error = -EFAULT; | |
2076 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2077 | goto out; | |
2078 | ||
496ad9aa | 2079 | inode = file_inode(filp); |
1da177e4 LT |
2080 | |
2081 | /* Don't allow mandatory locks on files that may be memory mapped | |
2082 | * and shared. | |
2083 | */ | |
a16877ca | 2084 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2085 | error = -EAGAIN; |
2086 | goto out; | |
2087 | } | |
2088 | ||
c293621b | 2089 | again: |
1da177e4 LT |
2090 | error = flock_to_posix_lock(filp, file_lock, &flock); |
2091 | if (error) | |
2092 | goto out; | |
5d50ffd7 | 2093 | |
cf01f4ee JL |
2094 | error = check_fmode_for_setlk(file_lock); |
2095 | if (error) | |
2096 | goto out; | |
2097 | ||
5d50ffd7 JL |
2098 | /* |
2099 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2100 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2101 | */ |
2102 | switch (cmd) { | |
0d3f7a2d | 2103 | case F_OFD_SETLK: |
90478939 JL |
2104 | error = -EINVAL; |
2105 | if (flock.l_pid != 0) | |
2106 | goto out; | |
2107 | ||
5d50ffd7 | 2108 | cmd = F_SETLK; |
cff2fce5 | 2109 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2110 | file_lock->fl_owner = filp; |
5d50ffd7 | 2111 | break; |
0d3f7a2d | 2112 | case F_OFD_SETLKW: |
90478939 JL |
2113 | error = -EINVAL; |
2114 | if (flock.l_pid != 0) | |
2115 | goto out; | |
2116 | ||
5d50ffd7 | 2117 | cmd = F_SETLKW; |
cff2fce5 | 2118 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2119 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2120 | /* Fallthrough */ |
2121 | case F_SETLKW: | |
1da177e4 LT |
2122 | file_lock->fl_flags |= FL_SLEEP; |
2123 | } | |
5d50ffd7 | 2124 | |
b648a6de | 2125 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2126 | |
c293621b PS |
2127 | /* |
2128 | * Attempt to detect a close/fcntl race and recover by | |
2129 | * releasing the lock that was just acquired. | |
2130 | */ | |
0b2bac2f AV |
2131 | /* |
2132 | * we need that spin_lock here - it prevents reordering between | |
2133 | * update of inode->i_flock and check for it done in close(). | |
2134 | * rcu_read_lock() wouldn't do. | |
2135 | */ | |
2136 | spin_lock(¤t->files->file_lock); | |
2137 | f = fcheck(fd); | |
2138 | spin_unlock(¤t->files->file_lock); | |
2139 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2140 | flock.l_type = F_UNLCK; |
2141 | goto again; | |
1da177e4 LT |
2142 | } |
2143 | ||
c293621b | 2144 | out: |
1da177e4 LT |
2145 | locks_free_lock(file_lock); |
2146 | return error; | |
2147 | } | |
2148 | ||
2149 | #if BITS_PER_LONG == 32 | |
2150 | /* Report the first existing lock that would conflict with l. | |
2151 | * This implements the F_GETLK command of fcntl(). | |
2152 | */ | |
c1e62b8f | 2153 | int fcntl_getlk64(struct file *filp, unsigned int cmd, struct flock64 __user *l) |
1da177e4 | 2154 | { |
9d6a8c5c | 2155 | struct file_lock file_lock; |
1da177e4 LT |
2156 | struct flock64 flock; |
2157 | int error; | |
2158 | ||
2159 | error = -EFAULT; | |
2160 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2161 | goto out; | |
2162 | error = -EINVAL; | |
2163 | if ((flock.l_type != F_RDLCK) && (flock.l_type != F_WRLCK)) | |
2164 | goto out; | |
2165 | ||
2166 | error = flock64_to_posix_lock(filp, &file_lock, &flock); | |
2167 | if (error) | |
2168 | goto out; | |
2169 | ||
0d3f7a2d | 2170 | if (cmd == F_OFD_GETLK) { |
90478939 JL |
2171 | error = -EINVAL; |
2172 | if (flock.l_pid != 0) | |
2173 | goto out; | |
2174 | ||
5d50ffd7 | 2175 | cmd = F_GETLK64; |
cff2fce5 | 2176 | file_lock.fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2177 | file_lock.fl_owner = filp; |
5d50ffd7 JL |
2178 | } |
2179 | ||
3ee17abd BF |
2180 | error = vfs_test_lock(filp, &file_lock); |
2181 | if (error) | |
2182 | goto out; | |
2183 | ||
9d6a8c5c ME |
2184 | flock.l_type = file_lock.fl_type; |
2185 | if (file_lock.fl_type != F_UNLCK) | |
2186 | posix_lock_to_flock64(&flock, &file_lock); | |
2187 | ||
1da177e4 LT |
2188 | error = -EFAULT; |
2189 | if (!copy_to_user(l, &flock, sizeof(flock))) | |
2190 | error = 0; | |
2191 | ||
2192 | out: | |
2193 | return error; | |
2194 | } | |
2195 | ||
2196 | /* Apply the lock described by l to an open file descriptor. | |
2197 | * This implements both the F_SETLK and F_SETLKW commands of fcntl(). | |
2198 | */ | |
c293621b PS |
2199 | int fcntl_setlk64(unsigned int fd, struct file *filp, unsigned int cmd, |
2200 | struct flock64 __user *l) | |
1da177e4 LT |
2201 | { |
2202 | struct file_lock *file_lock = locks_alloc_lock(); | |
2203 | struct flock64 flock; | |
2204 | struct inode *inode; | |
0b2bac2f | 2205 | struct file *f; |
1da177e4 LT |
2206 | int error; |
2207 | ||
2208 | if (file_lock == NULL) | |
2209 | return -ENOLCK; | |
2210 | ||
2211 | /* | |
2212 | * This might block, so we do it before checking the inode. | |
2213 | */ | |
2214 | error = -EFAULT; | |
2215 | if (copy_from_user(&flock, l, sizeof(flock))) | |
2216 | goto out; | |
2217 | ||
496ad9aa | 2218 | inode = file_inode(filp); |
1da177e4 LT |
2219 | |
2220 | /* Don't allow mandatory locks on files that may be memory mapped | |
2221 | * and shared. | |
2222 | */ | |
a16877ca | 2223 | if (mandatory_lock(inode) && mapping_writably_mapped(filp->f_mapping)) { |
1da177e4 LT |
2224 | error = -EAGAIN; |
2225 | goto out; | |
2226 | } | |
2227 | ||
c293621b | 2228 | again: |
1da177e4 LT |
2229 | error = flock64_to_posix_lock(filp, file_lock, &flock); |
2230 | if (error) | |
2231 | goto out; | |
5d50ffd7 | 2232 | |
cf01f4ee JL |
2233 | error = check_fmode_for_setlk(file_lock); |
2234 | if (error) | |
2235 | goto out; | |
2236 | ||
5d50ffd7 JL |
2237 | /* |
2238 | * If the cmd is requesting file-private locks, then set the | |
cff2fce5 | 2239 | * FL_OFDLCK flag and override the owner. |
5d50ffd7 JL |
2240 | */ |
2241 | switch (cmd) { | |
0d3f7a2d | 2242 | case F_OFD_SETLK: |
90478939 JL |
2243 | error = -EINVAL; |
2244 | if (flock.l_pid != 0) | |
2245 | goto out; | |
2246 | ||
5d50ffd7 | 2247 | cmd = F_SETLK64; |
cff2fce5 | 2248 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2249 | file_lock->fl_owner = filp; |
5d50ffd7 | 2250 | break; |
0d3f7a2d | 2251 | case F_OFD_SETLKW: |
90478939 JL |
2252 | error = -EINVAL; |
2253 | if (flock.l_pid != 0) | |
2254 | goto out; | |
2255 | ||
5d50ffd7 | 2256 | cmd = F_SETLKW64; |
cff2fce5 | 2257 | file_lock->fl_flags |= FL_OFDLCK; |
73a8f5f7 | 2258 | file_lock->fl_owner = filp; |
5d50ffd7 JL |
2259 | /* Fallthrough */ |
2260 | case F_SETLKW64: | |
1da177e4 LT |
2261 | file_lock->fl_flags |= FL_SLEEP; |
2262 | } | |
5d50ffd7 | 2263 | |
b648a6de | 2264 | error = do_lock_file_wait(filp, cmd, file_lock); |
1da177e4 | 2265 | |
c293621b PS |
2266 | /* |
2267 | * Attempt to detect a close/fcntl race and recover by | |
2268 | * releasing the lock that was just acquired. | |
2269 | */ | |
0b2bac2f AV |
2270 | spin_lock(¤t->files->file_lock); |
2271 | f = fcheck(fd); | |
2272 | spin_unlock(¤t->files->file_lock); | |
2273 | if (!error && f != filp && flock.l_type != F_UNLCK) { | |
c293621b PS |
2274 | flock.l_type = F_UNLCK; |
2275 | goto again; | |
1da177e4 LT |
2276 | } |
2277 | ||
2278 | out: | |
2279 | locks_free_lock(file_lock); | |
2280 | return error; | |
2281 | } | |
2282 | #endif /* BITS_PER_LONG == 32 */ | |
2283 | ||
2284 | /* | |
2285 | * This function is called when the file is being removed | |
2286 | * from the task's fd array. POSIX locks belonging to this task | |
2287 | * are deleted at this time. | |
2288 | */ | |
2289 | void locks_remove_posix(struct file *filp, fl_owner_t owner) | |
2290 | { | |
ff7b86b8 | 2291 | struct file_lock lock; |
1da177e4 LT |
2292 | |
2293 | /* | |
2294 | * If there are no locks held on this file, we don't need to call | |
2295 | * posix_lock_file(). Another process could be setting a lock on this | |
2296 | * file at the same time, but we wouldn't remove that lock anyway. | |
2297 | */ | |
496ad9aa | 2298 | if (!file_inode(filp)->i_flock) |
1da177e4 LT |
2299 | return; |
2300 | ||
2301 | lock.fl_type = F_UNLCK; | |
75e1fcc0 | 2302 | lock.fl_flags = FL_POSIX | FL_CLOSE; |
1da177e4 LT |
2303 | lock.fl_start = 0; |
2304 | lock.fl_end = OFFSET_MAX; | |
2305 | lock.fl_owner = owner; | |
2306 | lock.fl_pid = current->tgid; | |
2307 | lock.fl_file = filp; | |
2308 | lock.fl_ops = NULL; | |
2309 | lock.fl_lmops = NULL; | |
2310 | ||
150b3934 | 2311 | vfs_lock_file(filp, F_SETLK, &lock, NULL); |
1da177e4 | 2312 | |
1da177e4 LT |
2313 | if (lock.fl_ops && lock.fl_ops->fl_release_private) |
2314 | lock.fl_ops->fl_release_private(&lock); | |
2315 | } | |
2316 | ||
2317 | EXPORT_SYMBOL(locks_remove_posix); | |
2318 | ||
2319 | /* | |
2320 | * This function is called on the last close of an open file. | |
2321 | */ | |
78ed8a13 | 2322 | void locks_remove_file(struct file *filp) |
1da177e4 | 2323 | { |
496ad9aa | 2324 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
2325 | struct file_lock *fl; |
2326 | struct file_lock **before; | |
2327 | ||
2328 | if (!inode->i_flock) | |
2329 | return; | |
2330 | ||
73a8f5f7 | 2331 | locks_remove_posix(filp, filp); |
5d50ffd7 | 2332 | |
72c2d531 | 2333 | if (filp->f_op->flock) { |
1da177e4 | 2334 | struct file_lock fl = { |
73a8f5f7 | 2335 | .fl_owner = filp, |
1da177e4 LT |
2336 | .fl_pid = current->tgid, |
2337 | .fl_file = filp, | |
2338 | .fl_flags = FL_FLOCK, | |
2339 | .fl_type = F_UNLCK, | |
2340 | .fl_end = OFFSET_MAX, | |
2341 | }; | |
2342 | filp->f_op->flock(filp, F_SETLKW, &fl); | |
80fec4c6 TM |
2343 | if (fl.fl_ops && fl.fl_ops->fl_release_private) |
2344 | fl.fl_ops->fl_release_private(&fl); | |
1da177e4 LT |
2345 | } |
2346 | ||
1c8c601a | 2347 | spin_lock(&inode->i_lock); |
1da177e4 LT |
2348 | before = &inode->i_flock; |
2349 | ||
2350 | while ((fl = *before) != NULL) { | |
2351 | if (fl->fl_file == filp) { | |
1da177e4 LT |
2352 | if (IS_LEASE(fl)) { |
2353 | lease_modify(before, F_UNLCK); | |
2354 | continue; | |
2355 | } | |
8c3cac5e JL |
2356 | |
2357 | /* | |
2358 | * There's a leftover lock on the list of a type that | |
2359 | * we didn't expect to see. Most likely a classic | |
2360 | * POSIX lock that ended up not getting released | |
2361 | * properly, or that raced onto the list somehow. Log | |
2362 | * some info about it and then just remove it from | |
2363 | * the list. | |
2364 | */ | |
2365 | WARN(!IS_FLOCK(fl), | |
2366 | "leftover lock: dev=%u:%u ino=%lu type=%hhd flags=0x%x start=%lld end=%lld\n", | |
2367 | MAJOR(inode->i_sb->s_dev), | |
2368 | MINOR(inode->i_sb->s_dev), inode->i_ino, | |
2369 | fl->fl_type, fl->fl_flags, | |
2370 | fl->fl_start, fl->fl_end); | |
2371 | ||
2372 | locks_delete_lock(before); | |
2373 | continue; | |
1da177e4 LT |
2374 | } |
2375 | before = &fl->fl_next; | |
2376 | } | |
1c8c601a | 2377 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2378 | } |
2379 | ||
1da177e4 LT |
2380 | /** |
2381 | * posix_unblock_lock - stop waiting for a file lock | |
1da177e4 LT |
2382 | * @waiter: the lock which was waiting |
2383 | * | |
2384 | * lockd needs to block waiting for locks. | |
2385 | */ | |
64a318ee | 2386 | int |
f891a29f | 2387 | posix_unblock_lock(struct file_lock *waiter) |
1da177e4 | 2388 | { |
64a318ee BF |
2389 | int status = 0; |
2390 | ||
7b2296af | 2391 | spin_lock(&blocked_lock_lock); |
5996a298 | 2392 | if (waiter->fl_next) |
1da177e4 | 2393 | __locks_delete_block(waiter); |
64a318ee BF |
2394 | else |
2395 | status = -ENOENT; | |
7b2296af | 2396 | spin_unlock(&blocked_lock_lock); |
64a318ee | 2397 | return status; |
1da177e4 | 2398 | } |
1da177e4 LT |
2399 | EXPORT_SYMBOL(posix_unblock_lock); |
2400 | ||
9b9d2ab4 ME |
2401 | /** |
2402 | * vfs_cancel_lock - file byte range unblock lock | |
2403 | * @filp: The file to apply the unblock to | |
2404 | * @fl: The lock to be unblocked | |
2405 | * | |
2406 | * Used by lock managers to cancel blocked requests | |
2407 | */ | |
2408 | int vfs_cancel_lock(struct file *filp, struct file_lock *fl) | |
2409 | { | |
72c2d531 | 2410 | if (filp->f_op->lock) |
9b9d2ab4 ME |
2411 | return filp->f_op->lock(filp, F_CANCELLK, fl); |
2412 | return 0; | |
2413 | } | |
2414 | ||
2415 | EXPORT_SYMBOL_GPL(vfs_cancel_lock); | |
2416 | ||
7f8ada98 | 2417 | #ifdef CONFIG_PROC_FS |
d8ba7a36 | 2418 | #include <linux/proc_fs.h> |
7f8ada98 PE |
2419 | #include <linux/seq_file.h> |
2420 | ||
7012b02a JL |
2421 | struct locks_iterator { |
2422 | int li_cpu; | |
2423 | loff_t li_pos; | |
2424 | }; | |
2425 | ||
7f8ada98 | 2426 | static void lock_get_status(struct seq_file *f, struct file_lock *fl, |
99dc8292 | 2427 | loff_t id, char *pfx) |
1da177e4 LT |
2428 | { |
2429 | struct inode *inode = NULL; | |
ab1f1611 VG |
2430 | unsigned int fl_pid; |
2431 | ||
2432 | if (fl->fl_nspid) | |
6c5f3e7b | 2433 | fl_pid = pid_vnr(fl->fl_nspid); |
ab1f1611 VG |
2434 | else |
2435 | fl_pid = fl->fl_pid; | |
1da177e4 LT |
2436 | |
2437 | if (fl->fl_file != NULL) | |
496ad9aa | 2438 | inode = file_inode(fl->fl_file); |
1da177e4 | 2439 | |
99dc8292 | 2440 | seq_printf(f, "%lld:%s ", id, pfx); |
1da177e4 | 2441 | if (IS_POSIX(fl)) { |
c918d42a | 2442 | if (fl->fl_flags & FL_ACCESS) |
5315c26a | 2443 | seq_puts(f, "ACCESS"); |
cff2fce5 | 2444 | else if (IS_OFDLCK(fl)) |
5315c26a | 2445 | seq_puts(f, "OFDLCK"); |
c918d42a | 2446 | else |
5315c26a | 2447 | seq_puts(f, "POSIX "); |
c918d42a JL |
2448 | |
2449 | seq_printf(f, " %s ", | |
1da177e4 | 2450 | (inode == NULL) ? "*NOINODE*" : |
a16877ca | 2451 | mandatory_lock(inode) ? "MANDATORY" : "ADVISORY "); |
1da177e4 LT |
2452 | } else if (IS_FLOCK(fl)) { |
2453 | if (fl->fl_type & LOCK_MAND) { | |
5315c26a | 2454 | seq_puts(f, "FLOCK MSNFS "); |
1da177e4 | 2455 | } else { |
5315c26a | 2456 | seq_puts(f, "FLOCK ADVISORY "); |
1da177e4 LT |
2457 | } |
2458 | } else if (IS_LEASE(fl)) { | |
8144f1f6 JL |
2459 | if (fl->fl_flags & FL_DELEG) |
2460 | seq_puts(f, "DELEG "); | |
2461 | else | |
2462 | seq_puts(f, "LEASE "); | |
2463 | ||
ab83fa4b | 2464 | if (lease_breaking(fl)) |
5315c26a | 2465 | seq_puts(f, "BREAKING "); |
1da177e4 | 2466 | else if (fl->fl_file) |
5315c26a | 2467 | seq_puts(f, "ACTIVE "); |
1da177e4 | 2468 | else |
5315c26a | 2469 | seq_puts(f, "BREAKER "); |
1da177e4 | 2470 | } else { |
5315c26a | 2471 | seq_puts(f, "UNKNOWN UNKNOWN "); |
1da177e4 LT |
2472 | } |
2473 | if (fl->fl_type & LOCK_MAND) { | |
7f8ada98 | 2474 | seq_printf(f, "%s ", |
1da177e4 LT |
2475 | (fl->fl_type & LOCK_READ) |
2476 | ? (fl->fl_type & LOCK_WRITE) ? "RW " : "READ " | |
2477 | : (fl->fl_type & LOCK_WRITE) ? "WRITE" : "NONE "); | |
2478 | } else { | |
7f8ada98 | 2479 | seq_printf(f, "%s ", |
ab83fa4b | 2480 | (lease_breaking(fl)) |
0ee5c6d6 JL |
2481 | ? (fl->fl_type == F_UNLCK) ? "UNLCK" : "READ " |
2482 | : (fl->fl_type == F_WRLCK) ? "WRITE" : "READ "); | |
1da177e4 LT |
2483 | } |
2484 | if (inode) { | |
2485 | #ifdef WE_CAN_BREAK_LSLK_NOW | |
ab1f1611 | 2486 | seq_printf(f, "%d %s:%ld ", fl_pid, |
1da177e4 LT |
2487 | inode->i_sb->s_id, inode->i_ino); |
2488 | #else | |
2489 | /* userspace relies on this representation of dev_t ;-( */ | |
ab1f1611 | 2490 | seq_printf(f, "%d %02x:%02x:%ld ", fl_pid, |
1da177e4 LT |
2491 | MAJOR(inode->i_sb->s_dev), |
2492 | MINOR(inode->i_sb->s_dev), inode->i_ino); | |
2493 | #endif | |
2494 | } else { | |
ab1f1611 | 2495 | seq_printf(f, "%d <none>:0 ", fl_pid); |
1da177e4 LT |
2496 | } |
2497 | if (IS_POSIX(fl)) { | |
2498 | if (fl->fl_end == OFFSET_MAX) | |
7f8ada98 | 2499 | seq_printf(f, "%Ld EOF\n", fl->fl_start); |
1da177e4 | 2500 | else |
7f8ada98 | 2501 | seq_printf(f, "%Ld %Ld\n", fl->fl_start, fl->fl_end); |
1da177e4 | 2502 | } else { |
5315c26a | 2503 | seq_puts(f, "0 EOF\n"); |
1da177e4 LT |
2504 | } |
2505 | } | |
2506 | ||
7f8ada98 | 2507 | static int locks_show(struct seq_file *f, void *v) |
1da177e4 | 2508 | { |
7012b02a | 2509 | struct locks_iterator *iter = f->private; |
7f8ada98 | 2510 | struct file_lock *fl, *bfl; |
1da177e4 | 2511 | |
139ca04e | 2512 | fl = hlist_entry(v, struct file_lock, fl_link); |
1da177e4 | 2513 | |
7012b02a | 2514 | lock_get_status(f, fl, iter->li_pos, ""); |
1da177e4 | 2515 | |
7f8ada98 | 2516 | list_for_each_entry(bfl, &fl->fl_block, fl_block) |
7012b02a | 2517 | lock_get_status(f, bfl, iter->li_pos, " ->"); |
094f2825 | 2518 | |
7f8ada98 PE |
2519 | return 0; |
2520 | } | |
1da177e4 | 2521 | |
7f8ada98 | 2522 | static void *locks_start(struct seq_file *f, loff_t *pos) |
b03dfdec | 2523 | __acquires(&blocked_lock_lock) |
7f8ada98 | 2524 | { |
7012b02a | 2525 | struct locks_iterator *iter = f->private; |
99dc8292 | 2526 | |
7012b02a JL |
2527 | iter->li_pos = *pos + 1; |
2528 | lg_global_lock(&file_lock_lglock); | |
7b2296af | 2529 | spin_lock(&blocked_lock_lock); |
7012b02a | 2530 | return seq_hlist_start_percpu(&file_lock_list, &iter->li_cpu, *pos); |
7f8ada98 | 2531 | } |
1da177e4 | 2532 | |
7f8ada98 PE |
2533 | static void *locks_next(struct seq_file *f, void *v, loff_t *pos) |
2534 | { | |
7012b02a JL |
2535 | struct locks_iterator *iter = f->private; |
2536 | ||
2537 | ++iter->li_pos; | |
2538 | return seq_hlist_next_percpu(v, &file_lock_list, &iter->li_cpu, pos); | |
7f8ada98 | 2539 | } |
1da177e4 | 2540 | |
7f8ada98 | 2541 | static void locks_stop(struct seq_file *f, void *v) |
b03dfdec | 2542 | __releases(&blocked_lock_lock) |
7f8ada98 | 2543 | { |
7b2296af | 2544 | spin_unlock(&blocked_lock_lock); |
7012b02a | 2545 | lg_global_unlock(&file_lock_lglock); |
1da177e4 LT |
2546 | } |
2547 | ||
d8ba7a36 | 2548 | static const struct seq_operations locks_seq_operations = { |
7f8ada98 PE |
2549 | .start = locks_start, |
2550 | .next = locks_next, | |
2551 | .stop = locks_stop, | |
2552 | .show = locks_show, | |
2553 | }; | |
d8ba7a36 AD |
2554 | |
2555 | static int locks_open(struct inode *inode, struct file *filp) | |
2556 | { | |
7012b02a JL |
2557 | return seq_open_private(filp, &locks_seq_operations, |
2558 | sizeof(struct locks_iterator)); | |
d8ba7a36 AD |
2559 | } |
2560 | ||
2561 | static const struct file_operations proc_locks_operations = { | |
2562 | .open = locks_open, | |
2563 | .read = seq_read, | |
2564 | .llseek = seq_lseek, | |
99dc8292 | 2565 | .release = seq_release_private, |
d8ba7a36 AD |
2566 | }; |
2567 | ||
2568 | static int __init proc_locks_init(void) | |
2569 | { | |
2570 | proc_create("locks", 0, NULL, &proc_locks_operations); | |
2571 | return 0; | |
2572 | } | |
2573 | module_init(proc_locks_init); | |
7f8ada98 PE |
2574 | #endif |
2575 | ||
1da177e4 LT |
2576 | /** |
2577 | * lock_may_read - checks that the region is free of locks | |
2578 | * @inode: the inode that is being read | |
2579 | * @start: the first byte to read | |
2580 | * @len: the number of bytes to read | |
2581 | * | |
2582 | * Emulates Windows locking requirements. Whole-file | |
2583 | * mandatory locks (share modes) can prohibit a read and | |
2584 | * byte-range POSIX locks can prohibit a read if they overlap. | |
2585 | * | |
2586 | * N.B. this function is only ever called | |
2587 | * from knfsd and ownership of locks is never checked. | |
2588 | */ | |
2589 | int lock_may_read(struct inode *inode, loff_t start, unsigned long len) | |
2590 | { | |
2591 | struct file_lock *fl; | |
2592 | int result = 1; | |
1c8c601a JL |
2593 | |
2594 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2595 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2596 | if (IS_POSIX(fl)) { | |
2597 | if (fl->fl_type == F_RDLCK) | |
2598 | continue; | |
2599 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2600 | continue; | |
2601 | } else if (IS_FLOCK(fl)) { | |
2602 | if (!(fl->fl_type & LOCK_MAND)) | |
2603 | continue; | |
2604 | if (fl->fl_type & LOCK_READ) | |
2605 | continue; | |
2606 | } else | |
2607 | continue; | |
2608 | result = 0; | |
2609 | break; | |
2610 | } | |
1c8c601a | 2611 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2612 | return result; |
2613 | } | |
2614 | ||
2615 | EXPORT_SYMBOL(lock_may_read); | |
2616 | ||
2617 | /** | |
2618 | * lock_may_write - checks that the region is free of locks | |
2619 | * @inode: the inode that is being written | |
2620 | * @start: the first byte to write | |
2621 | * @len: the number of bytes to write | |
2622 | * | |
2623 | * Emulates Windows locking requirements. Whole-file | |
2624 | * mandatory locks (share modes) can prohibit a write and | |
2625 | * byte-range POSIX locks can prohibit a write if they overlap. | |
2626 | * | |
2627 | * N.B. this function is only ever called | |
2628 | * from knfsd and ownership of locks is never checked. | |
2629 | */ | |
2630 | int lock_may_write(struct inode *inode, loff_t start, unsigned long len) | |
2631 | { | |
2632 | struct file_lock *fl; | |
2633 | int result = 1; | |
1c8c601a JL |
2634 | |
2635 | spin_lock(&inode->i_lock); | |
1da177e4 LT |
2636 | for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) { |
2637 | if (IS_POSIX(fl)) { | |
2638 | if ((fl->fl_end < start) || (fl->fl_start > (start + len))) | |
2639 | continue; | |
2640 | } else if (IS_FLOCK(fl)) { | |
2641 | if (!(fl->fl_type & LOCK_MAND)) | |
2642 | continue; | |
2643 | if (fl->fl_type & LOCK_WRITE) | |
2644 | continue; | |
2645 | } else | |
2646 | continue; | |
2647 | result = 0; | |
2648 | break; | |
2649 | } | |
1c8c601a | 2650 | spin_unlock(&inode->i_lock); |
1da177e4 LT |
2651 | return result; |
2652 | } | |
2653 | ||
2654 | EXPORT_SYMBOL(lock_may_write); | |
2655 | ||
1da177e4 LT |
2656 | static int __init filelock_init(void) |
2657 | { | |
7012b02a JL |
2658 | int i; |
2659 | ||
1da177e4 | 2660 | filelock_cache = kmem_cache_create("file_lock_cache", |
ee19cc40 MS |
2661 | sizeof(struct file_lock), 0, SLAB_PANIC, NULL); |
2662 | ||
7012b02a JL |
2663 | lg_lock_init(&file_lock_lglock, "file_lock_lglock"); |
2664 | ||
2665 | for_each_possible_cpu(i) | |
2666 | INIT_HLIST_HEAD(per_cpu_ptr(&file_lock_list, i)); | |
2667 | ||
1da177e4 LT |
2668 | return 0; |
2669 | } | |
2670 | ||
2671 | core_initcall(filelock_init); |